id
stringlengths 9
10
| text
stringlengths 1
18.1M
| source
stringclasses 1
value | created
timestamp[s] | added
stringlengths 26
26
| metadata
dict |
|---|---|---|---|---|---|
0806.2327 | # Lifetimes of tidally limited star clusters with different radii
M. Gieles1 and H. Baumgardt2
1 European Southern Observatory, Casilla 19001, Santiago 19, Chile
2 Argelander-Institut für Astronomie (Sternwarte), Universität Bonn, Auf dem
Hügel 71, D-53121 Bonn, Germany
(Released 2008 Xxxxx XX)
###### Abstract
We study the escape rate of stars, $\dot{N}$, from clusters with different
radii on circular orbits in a tidal field using analytical predictions and
direct $N$-body simulations. We find that $\dot{N}$ depends on the ratio
$\mbox{$\mathfrak{R}$}\equiv r_{\rm h}/r_{\rm J}$, where $r_{\rm h}$ is the
half-mass radius and $r_{\rm J}$ the radius of the zero-velocity surface
around the cluster. For $\mbox{$\mathfrak{R}$}\gtrsim 0.05$, the “tidal
regime”, there is almost no dependence of $\dot{N}$ on $\mathfrak{R}$. To
first order this is because the fraction of escapers per half-mass relaxation
time, $t_{\rm rh}$, scales approximately as $\mbox{$\mathfrak{R}$}^{3/2}$,
which cancels out the $r_{\rm h}^{3/2}$ term in $t_{\rm rh}$. For
$\mbox{$\mathfrak{R}$}\lesssim 0.05$, the “isolated regime”, $\dot{N}$ scales
as $\mbox{$\mathfrak{R}$}^{-3/2}$. The dissolution time-scale, $t_{\rm dis}$,
falls in three regimes. Clusters that start with their initial $\mathfrak{R}$,
$\mathfrak{R}_{i}$, in the tidal regime dissolve completely in this regime and
their $t_{\rm dis}$ is, therefore, insensitive to the initial $r_{\rm h}$. Our
model predicts that $\mathfrak{R}_{i}$ has to be $10^{-20}-10^{-10}$ for
clusters to dissolve completely in the isolated regime. This means that
realistic clusters that start with $\mathfrak{R}_{i}\lesssim 0.05$ always
expand to the tidal regime before final dissolution. Their $t_{\rm dis}$ has a
shallower dependence on $\mathfrak{R}_{i}$ than what would be expected when
$t_{\rm dis}$ is a constant times $t_{\rm rh}$. For realistic values of
$\mathfrak{R}_{i}$, the lifetime varies by less than a factor of 1.5 due to
changes in $\mathfrak{R}_{i}$. This implies that the “survival” or “vital”
diagram for globular clusters should allow for more small clusters to survive.
We note that with our result it is impossible to explain the universal peaked
mass function of globular cluster systems by dynamical evolution from a power-
law initial mass function, since the peak will be at lower masses in the outer
parts of galaxies. Our results finally show that in the tidal regime $t_{\rm
dis}$ scales as $N^{0.65}/\omega$, with $\omega$ the angular frequency of the
cluster in the host galaxy.
###### keywords:
stellar dynamics – methods: $N$-body simulations – globular clusters: general
– galaxies: star clusters
††pagerange: LABEL:firstpage–LABEL:lastpage††pubyear: 2006
## 1 Introduction
Stars escape from clusters due to internal two and three body encounters in
which stars get accelerated to velocities higher than the escape velocity. The
resulting dissolution time-scale, $t_{\rm dis}$, depends on the number of
stars, $N$, and the escape rate, $\dot{N}$, as $\mbox{$t_{\rm
dis}$}\equiv-N/\dot{N}$. For a constant $\dot{N}$ the instantaneous value of
$t_{\rm dis}$ is the remaining time to total dissolution. For clusters in
isolation $t_{\rm dis}$ scales linearly with the half-mass relaxation time,
$t_{\rm rh}$ (Ambartsumian, 1938; Spitzer, 1940). The presence of a tidal
field speeds up $\dot{N}$ by roughly an order of magnitude (Hénon, 1961;
Spitzer & Chevalier, 1973; Giersz & Heggie, 1997). When treating the tidal
field as a radial cut-off, as is often done in Fokker Planck and $N$-body
simulations (Chernoff & Weinberg, 1990; Gnedin & Ostriker, 1997; Takahashi &
Portegies Zwart, 2000) $t_{\rm dis}$ also scales with $t_{\rm rh}$ (Baumgardt
2001, hereafter B01).
Fukushige & Heggie (2000) demonstrated that for a realistic tidal field it is
important to consider the finite time it takes stars to escape through one of
the Lagrange points. B01 showed that then $\mbox{$t_{\rm dis}$}\propto t_{\rm
rh}^{3/4}$. This scaling was also found for more realistic $N$-body
simulations that include a stellar mass function and stellar evolution
(Vesperini & Heggie 1997; Baumgardt & Makino 2003, hereafter BM03). BM03 only
considered the dependence of $t_{\rm rh}$ on cluster mass, $M$, since in their
runs the initial half-mass radius, $r_{\rm h}$, and the initial tidal radius
are linked, so $r_{\rm h}\propto M^{1/3}$.
Observations of young (extra-galactic) clusters show that the scaling of
$r_{\rm h}$ with cluster mass, $M$, and galactocentric distance, $R_{\rm G}$,
is considerably shallower: $r_{\rm h}\propto M^{0.1}\,R^{\,0.1}_{\rm G}$ (Zepf
et al., 1999; Larsen, 2004; Scheepmaker et al., 2007) than the Roche-lobe
filling relation ($r_{\rm h}\propto M^{1/3}\,R_{\rm G}^{\,2/3}$), implying
that massive clusters and clusters at large $R_{\rm G}$ are initially under-
filling their Roche-lobe.
The relation between $t_{\rm dis}$ and the median cluster radius, $\bar{r}$,
was modelled by Wielen (1971) for clusters up to $N=250$. He found that
$t_{\rm dis}$ (in Myrs) scales with $\bar{r}^{3/2}$ for $\bar{r}\lesssim
0.01\,r_{\rm J}$, with $r_{\rm J}$ the Jabobi radius, being the radius of the
zero-velocity surface around the cluster imposed by the tidal field. This is
because for these clusters $\mbox{$t_{\rm dis}$}\propto t_{\rm rh}$. For
$\bar{r}\gtrsim 0.05\,r_{\rm J}$ this scaling is not followed any more and
$t_{\rm dis}$ is shorter. Large clusters are even more vulnerable to
disruption when the effect of passing molecular clouds is included (King,
1958; Wielen, 1985; Gieles et al., 2006).
Tanikawa & Fukushige (2005) modelled the evolution of clusters of larger $N$
that are initially Roche-lobe under-filling using collisionless $N$-body
simulations. They confirm that for Roche-lobe filling clusters $\mbox{$t_{\rm
dis}$}\propto t_{\rm rh}^{x}$, with $x=3/4$ as was found by B01, and $x$
somewhat larger for Roche-lobe under-filling clusters. They do not discuss the
effect of radius on $t_{\rm dis}$. More realistic $N$-body simulations,
including various stellar initial mass function and stellar evolution, of
Roche-lobe under-filling clusters were considered by Engle (1999). She
concludes that Roche-lobe under-filling clusters survive longer. However, this
is for a fixed $r_{\rm h,i}$ and different $r_{\rm J}$. Since her simulations
included stellar evolution it is not possible to scale these results to the
same $r_{\rm J}$.
Predictions for the survival probability of globular cluster, such as the
survival triangle (Fall & Rees, 1977; Gnedin & Ostriker, 1997), rely on the
assumption that $t_{\rm dis}$ is a constant times $t_{\rm rh}$. This is also
an important assumption in a recent attempt to explain the shape and the
universality of the turn-over of the globular cluster mass function
(McLaughlin & Fall, 2008). To be able to judge the applicability of such
models it is of importance that the relation between $t_{\rm dis}$ and $r_{\rm
h}$ is better understood for clusters with larger $N$.
The interplay between internal relaxation effects and external tidal effects
is the topic of this Letter. In § 2 we present a set of $N$-body simulations
of clusters with different initial Roche-lobe filling factors. We introduce a
simple analytical model for $\dot{N}$ that includes internal dynamics and the
external tides in § 3. In § 4 we confront our model with the simulations and
our conclusions are presented in § 5.
## 2 Description of the runs
We simulate the evolution of clusters containing between $N=1024$ and
$N=32768$ particles, without primordial binaries and mass-loss by stellar
evolution, orbiting with angular frequency $\omega$ in a steady point mass
tidal field to simulate circular orbits in a galactic potential. The stellar
masses are randomly drawn from a power-law mass function with index $-2.35$
with the maximum mass 30 times larger than the minimum mass.
For the density distribution of the clusters we use King (1966) models, with
$W_{0}=5$. We define an initial Roche-lobe filling factor as
$\mbox{$\mathfrak{F}$}\equiv\mbox{$r_{\rm t}$}/r_{\rm J,i}$, with $r_{\rm
J,i}$ the initial Jacobi radius and $r_{\rm t}$ the King tidal radius, that
is, the radius where the stellar density of the King (1966) model drops to
zero. We model four values of $\mathfrak{F}$, from 0.125 to 1. The $W_{0}$
parameter and $\mathfrak{F}$ set the ratio of the initial $r_{\rm h}$, $r_{\rm
h,i}$, and $r_{\rm J,i}$, which we denote by $\mathfrak{R}_{i}\,(\equiv r_{\rm
h,i}/r_{\rm J,i})$. Clusters with $N$=[1024, 2048, 4096, 8192] are run [16, 8,
4, 2] times to reduce statistical variations. We also run an $N=4096$
simulation in isolation up to $t_{1/2}$ to determine the mass loss parameters.
All clusters are scaled to $N$-body units, such that $G=M=1$ and $E=-0.25$
(Heggie & Mathieu, 1986). Here, $E$ is the total (potential and kinetic)
initial energy of the cluster. In these units the virial radius,
$\mbox{$r_{\rm v}$}\equiv GM/(-4\,E)$, equals unity and the crossing time at
$r_{\rm v}$ is $2\sqrt{2}$.
The half-time, $t_{1/2}$, is the when half of the initial number of stars have
become unbound, where bound is defined as the number of stars within $r_{\rm
J}$. We multiply $t_{1/2}$ (in $N$-body times) by $\omega$ to compare the
results of different $\mathfrak{R}_{i}$. The dimensionless $\omega t_{1/2}$
results are equivalent to physical times for clusters at the same $R_{\rm G}$.
A summary of the runs and the resulting $t_{1/2}$ and $\omega t_{1/2}$ values
is given in Table 1.
All $N$-body calculations were carried out with the kira (McMillan & Hut,
1996; Portegies Zwart et al., 2001) integrator on the special purpose GRAPE-6
boards (Makino et al., 2003) of the European Southern Observatory.
Table 1: Summary of the $N$-body simulations. $N$ | $\mbox{$\mathfrak{F}$}=\frac{\mbox{$\mbox{$r_{\rm t}$}$}}{\mbox{$r_{\rm J,i}$}}$ | $\mathfrak{R}_{i}=\frac{\mbox{$r_{\rm h,i}$}}{\mbox{$r_{\rm J,i}$}}$ | $t_{1/2}$ | $\omega t_{1/2}$ | $\frac{\mbox{$t_{1/2}$}}{\mbox{$t_{\rm rh,i}$}}$
---|---|---|---|---|---
1024 | 1 | 0.186 | 114 | 7.20 | 5.8
1024 | 0.5 | 0.093 | 306 | 6.84 | 15.7
1024 | 0.25 | 0.047 | 768 | 6.07 | 39.4
1024 | 0.125 | 0.023 | 1718 | 4.80 | 88.2
2048 | 1 | 0.186 | 174 | 11.0 | 5.1
2048 | 0.5 | 0.093 | 478 | 10.7 | 13.9
2048 | 0.25 | 0.047 | 1184 | 9.35 | 34.3
2048 | 0.125 | 0.023 | 2768 | 7.73 | 80.3
4096 | 1 | 0.186 | 269 | 17.0 | 4.4
4096 | 0.5 | 0.093 | 765 | 17.1 | 12.4
4096 | 0.25 | 0.047 | 1833 | 14.5 | 29.6
4096 | 0.125 | 0.023 | 4235 | 11.8 | 68.5
8192 | 1 | 0.186 | 435 | 27.5 | 3.9
8192 | 0.5 | 0.093 | 1188 | 26.5 | 10.6
8192 | 0.25 | 0.047 | 2925 | 23.1 | 26.1
8192 | 0.125 | 0.023 | 6614 | 18.5 | 59.0
16384 | 1 | 0.186 | 670 | 42.3 | 3.3
16384 | 0.5 | 0.093 | 1867 | 41.7 | 9.1
16384 | 0.25 | 0.047 | 4705 | 37.2 | 22.9
16384 | 0.125 | 0.023 | 10456 | 29.2 | 51.0
32768 | 1 | 0.186 | 1062 | 67.1 | 2.8
32768 | 0.5 | 0.093 | 3049 | 68.1 | 8.1
32768 | 0.25 | 0.047 | 8011 | 63.3 | 21.2
32768 | 0.125 | 0.023 | 17281 | 48.3 | 45.7
Table 2: Results of an isolated run. $N$ | $\nu$ | $t_{\rm cc}$ | $t_{\rm rh,i}$ | $n_{\rm rh}$ | $t_{1/2}$
---|---|---|---|---|---
4096 | 0.081 | 209 | 68 | 3.1 | 1.9$\times 10^{6}$
## 3 Analytical model for the escape rate of tidally limited clusters
### 3.1 The “classical” Ansatz
Lets first assume that a cluster, consisting of $N$ stars, loses a constant
fraction $\xi_{\rm e}$ of its stars each $t_{\rm rh}$, so that we can write
for $\dot{N}$ (for example Spitzer 1987, hereafter S87)
$\mbox{$\dot{N}$}=-\xi_{e}\frac{N}{t_{\rm rh}},$ (1)
where $t_{\rm rh}$ is conventionally expressed as (Spitzer & Hart, 1971)
$t_{\rm rh}=0.138\frac{N^{1/2}r_{\rm h}^{3/2}}{\sqrt{\bar{m}G}\ln\Lambda},$
(2)
where $\Lambda=\gamma N$ and $\gamma=0.11$ (Giersz & Heggie, 1994a), $G$ is
the gravitational constant and $\bar{m}$ is the mean stellar mass. The
crossing time at $r_{\rm h}$, $t_{\rm cr}$, is given by
$t_{\rm cr}=k\left(\frac{r_{\rm h}^{3}}{GM}\right)^{1/2},$ (3)
where $k$ is a constant of order unity depending on the cluster density
profile.
The escape energy of stars in an isolated cluster, $E_{\rm crit}^{\rm iso}$,
is four times the mean kinetic energy of stars in the clusters, so
$\mbox{$E_{\rm crit}^{\rm iso}$}=0.8GM/r_{\rm h}$ (S87) and the escape
velocity, $v_{\rm e}$, scales with the stellar root mean square velocity in
the cluster, $v_{\rm rms}$, as $v_{\rm e}^{2}=4\,v^{2}_{\rm rms}$. From
integration over a Maxwellian velocity distribution the fraction of stars with
$v^{2}>4\,v^{2}_{\rm rms}$ can be determined. We refer to this escape fraction
for isolated clusters as $\xi_{\rm e0}$ and S87 showed that $\mbox{$\xi_{\rm
e0}$}=0.0074$.
### 3.2 The escape fraction as a function of cluster radius
When a cluster evolves in a tidal field the critical energy for escape is
$\mbox{$E_{\rm crit}^{\rm tid}$}=-\frac{3GM}{2r_{\rm J}}.$ (4)
For a point-mass galaxy, $r_{\rm J}$ depends on $\omega$ and $M$ as
$r_{\rm J}=\left(\frac{G}{3\omega^{2}}\right)^{1/3}\,M^{1/3},$ (5)
where $\omega\equiv V_{\rm G}/R_{\rm G}$, with $V_{\rm G}$ the circular
velocity. A large $\omega$ means a strong tidal field, which results in a
small $r_{\rm J}$.
The ratio $E_{\rm crit}^{\rm tid}$ and $E_{\rm crit}^{\rm iso}$ gives the
relative reduction of the escape energy due to the tidal field (following S87)
$\Gamma=\frac{\mbox{$E_{\rm crit}^{\rm tid}$}}{\mbox{$E_{\rm crit}^{\rm
iso}$}}=-\frac{3GM}{2r_{\rm J}}{\mbox{\LARGE{/}}}\frac{0.8GM}{r_{\rm
h}}=-\frac{15}{8}\mbox{$\mathfrak{R}$},$ (6)
where we have used $\mbox{$\mathfrak{R}$}\equiv r_{\rm h}/r_{\rm J}$. Note
that this $\Gamma$ is a factor 1.5 higher than the original definition in S87,
since his result was based on $\mbox{$E_{\rm crit}^{\rm tid}$}=GM/r_{\rm J}$,
which he later refines to equation (4). We calculate $\xi_{\rm e}$ as a
function of $\mathfrak{R}$ by numerically integrating Maxwellian velocity
distributions for different $\mathfrak{R}$ to determine the fraction of stars
with velocities $v^{2}\geq 4\,[1-\Gamma(\mbox{$\mathfrak{R}$})]v^{2}_{\rm
rms}$.
In Fig. 1 we show that $\xi_{\rm e}$ increases exponentially for increasing
$\mathfrak{R}$ and can be well approximated by $\mbox{$\xi_{\rm
e0}$}\exp(10\mbox{$\mathfrak{R}$})$. For $\mbox{$\mathfrak{R}$}\gtrsim 0.05$,
the expression for $\xi_{\rm e}$ scales approximately as
$\mbox{$\mathfrak{R}$}^{3/2}$, which has important consequences for $\dot{N}$
(see equations (1) & (2)). We approximate $\xi_{\rm e}$ by
$\xi_{\rm e}$ $\displaystyle=$ $\displaystyle\mbox{$\xi_{\rm e0}$},\hskip
71.13188pt\mbox{$\mathfrak{R}$}<\mathfrak{R}_{1}$ (7) $\displaystyle=$
$\displaystyle\mbox{$\xi_{\rm
e0}$}\left(\frac{\mbox{$\mathfrak{R}$}}{\mathfrak{R}_{1}}\right)^{3/2},\hskip
28.45274pt\mbox{$\mathfrak{R}$}\geq\mathfrak{R}_{1},$
where $\mathfrak{R}_{1}=0.05$ is the boundary between the “isolated regime”
($\mbox{$\mathfrak{R}$}\leq\mathfrak{R}_{1}$) and the “tidal regime”
($\mbox{$\mathfrak{R}$}>\mathfrak{R}_{1}$).
Substituting equations (2) & (7) in equation (1) and using equation (5) we
find for $\dot{N}$ in the tidal regime
$\mbox{$\dot{N}$}=-\left(\frac{\sqrt{3}\mbox{$\xi_{\rm
e0}$}\ln\Lambda}{0.138\,\mathfrak{R}_{1}^{3/2}}\right)\,\omega.$ (8)
So $\dot{N}$ is independent of $r_{\rm h}$ in the regime where
$\mbox{$\xi_{\rm e}$}\propto\mbox{$\mathfrak{R}$}^{3/2}$. This is because for
a smaller(larger) star cluster the shorter(longer) $t_{\rm rh}$ is balanced by
the lower(higher) $\xi_{\rm e}$222Ivan King noticed a remarkable similarity
between this result and equation (53) in his 1966 paper. There he shows that
the escape rate of stars from a Roche-lobe filling cluster is independent of
position within the cluster. This is probably because of the same physical
reason, but he derived it in a different way. . Equation (8) also shows that
$\dot{N}$ depends only marginally on $N$ through $\ln\Lambda$.
Figure 1: The fraction of escapers, $\xi_{\rm e}$, for different
$\mathfrak{R}$ for clusters in a tidal field. The points show the result of a
numerical integration, the full line an exponential approximation and the
dashed line a double power-law approximation.
### 3.3 Including the escape time
Fukushige & Heggie (2000) consider the time-scale of escape for stars in a
cluster evolving in a tidal field. This time-scale is non-zero because stars
with energies (slightly) larger than the escape energy still need a finite
time to find one of the Lagrange points, where the escape energy is lowest, to
leave the cluster.
B01 derives an expression for $\dot{N}$ of stars in the potential escaper
regime, that is, with energies higher than the escape energy, but still
trapped in the potential, and shows that it scales as $Nt_{\rm
rh}^{-3/4}\,t_{\rm cr}^{-1/4}$, instead of $N\,t_{\rm rh}^{-1}$ (equation 1).
We include $\xi_{\rm e}$ and define $\dot{N}$ analogous to equation (1), as
$\mbox{$\dot{N}$}=-\mbox{$\xi_{\rm e}$}\,\frac{N}{t_{\rm rh}^{3/4}\,t_{\rm
cr}^{1/4}}.$ (9)
With the expressions for $\xi_{\rm e}$, $t_{\rm rh}$ and $t_{\rm cr}$ we then
find for $\dot{N}$ in the tidal regime, including the escape time,
$\mbox{$\dot{N}$}=-\left(\frac{\sqrt{3}\mbox{$\xi_{\rm
e0}$}}{k^{1/4}\,\mathfrak{R}_{1}^{3/2}}\right)\left(\frac{\ln\Lambda}{0.138}\right)^{3/4}\,N^{1/4}\,\omega.$
(10)
From a comparison between equation (8) and equation (10) we see that $\dot{N}$
becomes $N$ dependent when we include the escape time, but is still
independent of $r_{\rm h}$. The dissolution time-scale in the tidal regime,
which we define as $\mbox{$t^{\rm tid}_{\rm dis}$}\equiv-N/\mbox{$\dot{N}$}$,
is then
$\mbox{$t^{\rm tid}_{\rm
dis}$}=A\,\left(\frac{N}{\ln\Lambda}\right)^{3/4}\frac{1}{\omega},$ (11)
with $A=0.138^{3/4}\,\mathfrak{R}_{1}^{3/2}k^{1/4}/(\sqrt{3}\mbox{$\xi_{\rm
e0}$})$. For a $W_{0}=5$ cluster $k=3.85$, which together with the values for
$\mathfrak{R}_{1}$ and $\xi_{\rm e0}$ from § 3.2 results in $A=0.277$. The
term $(N/\ln\Lambda)^{3/4}$ can be well approximated by $B\,N^{\eta}$, with
$\eta\simeq 0.6$ (Lamers et al., 2005b). The values of $B$ and $\eta$ depend
slightly on the value of $\gamma$ in $\Lambda=\gamma N$. We find the best
agreement with the Roche-lobe filling simulations for $\gamma=0.2$, which
results in $B=0.5$ and $\eta=0.65$, so we can write $\mbox{$t^{\rm tid}_{\rm
dis}$}=0.138\,N^{0.65}/\omega$. This scaling of $t_{\rm dis}$ with $N$ was
also derived from observations (Boutloukos & Lamers, 2003; Lamers et al.,
2005b; Gieles et al., 2005). To compare the model to the results of the
simulations we derive $t_{1/2}$ in the tidal regime, $t_{1/2}^{\rm tid}$, from
$t^{\rm tid}_{\rm dis}$. Lamers et al. (2005a) show that when $\mbox{$t_{\rm
dis}$}=B\,N^{\eta}$, with $B$ a constant, then
$t_{1/2}=(B\,N^{\eta}/\eta)\,(1-[1/2]^{\eta})$, so
$\mbox{$t_{1/2}^{\rm tid}$}=0.077\,\frac{N^{0.65}}{\omega}.$ (12)
We assume that the effect of the escape time is the same for all
$\mbox{$\mathfrak{R}$}>\mathfrak{R}_{1}$, so that we can apply equation (12)
in this regime. With these relations we also assume that $\dot{N}$ in the pre-
collapse and post-collapse phase is the same.
### 3.4 Dissolution in the isolated regime
We assume that clusters with $\mbox{$\mathfrak{R}$}<\mathfrak{R}_{1}$ evolve
in the same way as clusters in isolation, up to the moment that $\mathfrak{R}$
becomes equal to $\mathfrak{R}_{1}$. This assumption is justified by equation
(6) from which we see that for a cluster with
$\mbox{$\mathfrak{R}$}=\mathfrak{R}_{1}$ the relative contribution of the
tidal field to the escape energy is less then 10%. In the absence of
primordial binaries and mass-loss by stellar evolution $r_{\rm h}$ and $N$
remain roughly constant up to the moment of core collapse, $t_{\rm cc}$
(Baumgardt et al. 2002, hereafter B02). After $t_{\rm cc}$, isolated clusters
evolve in a self-similar way ($t_{\rm rh}\propto t$), which results in two
fundamental relations for the evolution of $N$ and $r_{\rm h}$ (Goodman 1984;
S87; B02):
$N(t)=\mbox{$N_{\rm cc}$}\left(\frac{t}{\mbox{$t_{\rm cc}$}}\right)^{-\nu}$
(13) $r_{\rm h}(t)=\mbox{$r_{\rm h,cc}$}\left(\frac{t}{\mbox{$t_{\rm
cc}$}}\right)^{\frac{2+\nu}{3}},$ (14)
where $r_{\rm h,cc}$ and $N_{\rm cc}$ are $r_{\rm h}$ and $N$ at $t_{\rm cc}$.
More complicated relations, assuming a non-zero origin for these relations,
exist (Giersz & Heggie, 1994b). Core collapse happens after a multiple number
of the initial $t_{\rm rh}$, $t_{\rm rh,i}$: $\mbox{$t_{\rm
cc}$}=\mbox{$n_{\rm rh}$}t_{\rm rh,i}$.
From equation (13) we find that $t_{1/2}$ for clusters evolving completely in
the isolated regime, $t_{1/2}^{\rm iso}$, is
$\mbox{$t_{1/2}^{\rm iso}$}=\left(\frac{0.138\times 2^{1/\nu}\,\mbox{$n_{\rm
rh}$}}{\sqrt{3}}\right)\frac{N}{\ln\Lambda}\frac{\mathfrak{R}_{i}^{3/2}}{\omega}.$
(15)
Because clusters expand after $t_{\rm cc}$, not all clusters that start with
$\mathfrak{R}_{i}<\mathfrak{R}_{1}$ will reach $t_{1/2}^{\rm iso}$ before they
reach $\mathfrak{R}_{1}$.The maximum $\mathfrak{R}_{i}$ for which equation
(15) applies is found from equations (14) & (15) and depends on
$\mathfrak{R}_{1}$ and $\nu$ as
$\mathfrak{R}_{2}=\mathfrak{R}_{1}\,\left(\frac{1}{2}\right)^{\frac{2+2\nu}{3\nu}},$
(16)
which for $\mathfrak{R}_{1}=0.05$ (Fig. 1) and $\nu=0.05-0.1$ (B02) results in
$\mathfrak{R}_{2}\simeq 6\times 10^{-5}-6\times 10^{-3}$. If we define
complete dissolution as the moment where only 5% of the original number of
stars is still bound, then the corresponding value of $\mathfrak{R}_{2}$
reduces to $\sim 10^{-20}-10^{-10}$. This implies that realistic clusters
never dissolve completely in the isolated regime.
### 3.5 Combining the isolated and the tidal regime
Clusters that start with $\mathfrak{R}_{2}<\mathfrak{R}_{i}<\mathfrak{R}_{1}$
evolve partially in the isolated regime and partially in the tidal regime.
Though our model allows to numerically compute $t_{1/2}$ for clusters in this
regime, we simply connect $\log\left[\omega\mbox{$t_{1/2}^{\rm
iso}$}(\mathfrak{R}_{2})\right]$ and $\log\left[\omega\mbox{$t_{1/2}^{\rm
tid}$}(\mathfrak{R}_{1})\right]$ with a straight line. The slope of this line,
representing the $\mathfrak{R}_{i}$ dependence of $\omega t_{1/2}$, is
$\log\left[\mbox{$t_{1/2}^{\rm tid}$}(\mathfrak{R}_{1})/\mbox{$t_{1/2}^{\rm
iso}$}(\mathfrak{R}_{2})\right]/\log\left[\mathfrak{R}_{1}/\mathfrak{R}_{2}\right]$.
This slope is slightly $N$-dependent, and by using equations (12) & (15) and
the parameters from the isolated run (Table 2) we find that it decreases from
$0.55$ to $0.35$ between $N=1024$ and $N=32768$. For $N=10^{6}$ the slope
would be $0.15$.
In Fig. 2 we show the $\omega t_{1/2}$ following from our model for different
$N$ and a range of two orders of magnitude around $\mathfrak{R}_{1}$. The
results of the simulations are shown as dots and are discussed in the next
section.
Figure 2: Prediction for the dimensionless half-time, $\omega t_{1/2}$, for
clusters in the tidal regime (equation 12) and part of the intermediate regime
(§ 3.5) for different $N$. Results of the simulations are shown as dots.
## 4 Comparison to $N$-body simulations
The results for $\omega t_{1/2}$ of the simulations of clusters with different
$N$ and $\mathfrak{R}_{i}$ are presented in Fig. 2. The $\omega t_{1/2}$
results for $\mbox{$\mathfrak{F}$}=0.5$ clusters are nearly the same as those
for $\mbox{$\mathfrak{F}$}=1$. For $\mbox{$\mathfrak{F}$}=0.25(0.125)$ the
$\omega t_{1/2}$ values are approximately 10%(25%) shorter compared to the
Roche-lobe filling results, whereas a scaling with $t_{\rm rh}$ predicts a
difference of a factor of $4^{1.5}(8^{1.5})\simeq 8(23)$. From Table 1 we see
that clusters that start Roche-lobe under-filling, have evolved for a much
larger number of relaxation times than Roche-lobe filling clusters by the time
they reach $t_{1/2}$.
## 5 Conclusions
The dissolution time-scale of clusters evolving in a tidal field, in
dimensionless units ($\omega\mbox{$t_{\rm dis}$}$) or in physical units, is
almost independent of the initial half-mass radius, $r_{\rm h,i}$, when
$r_{\rm h,i}$ relative to the initial Jacobi (or tidal) radius, $r_{\rm J,i}$,
is larger than $\mathfrak{R}_{i}(\equiv r_{\rm h,i}/r_{\rm J,i})\gtrsim 0.05$.
For clusters that start with $\mathfrak{R}_{i}<0.05$, $t_{\rm dis}$ scales
mildly with $\mathfrak{R}_{i}$, between $\mathfrak{R}_{i}^{0.55}$ and
$\mathfrak{R}_{i}^{0.35}$ for the range of $N$ we consider and even flatter
for larger $N$. Only clusters that start with $\mathfrak{R}_{i}\lesssim
10^{-4}$ can lose half their stars before they reach the influence of the
tidal field. We find that in the tidal regime $t_{\rm dis}$ is mainly
determined by $N$ and the angular frequency: $\mbox{$t_{\rm dis}$}\propto
N^{0.65}/\omega$, that is, what was also found by BM03 for Roche-lobe filling,
multi-mass clusters dissolving in tidal fields.
Gnedin & Ostriker (1997) construct the “vital” diagram of globular clusters,
which is a triangle in $M$ vs. $r_{\rm h}$ space, outside which clusters
should have been destroyed. There are numerous globular clusters with small
$r_{\rm h}$ outside this triangle, which the authors denote as lucky
survivors. We show that small clusters can in fact survive.
Models that try to explain the evolution of the globular cluster mass function
(GCMF) from an initial power-law to a (universal) peaked distribution by
stellar dynamical processes are in difficulties, since such models will always
produce less dissolution in the outer parts of galaxies. McLaughlin & Fall
(2008) have recently proposed a solution to this problem by assuming that
$\mbox{$\mbox{${\rm d}$}M/\mbox{${\rm d}$}t$}\propto M/t_{\rm rh}$ and so
$\mbox{$\mbox{${\rm d}$}M/\mbox{${\rm d}$}t$}\propto\sqrt{\mbox{$\rho_{\rm
h}$}}$ (equations 1&2, with $\xi_{\rm e}$ constant), that is, their $t_{\rm
dis}$ is determined by internal relaxation effects only and is independent of
the strength of the tidal field. However, our results show that $\xi_{\rm e}$
is not constant and, therefore, $\mbox{${\rm d}$}M/\mbox{${\rm d}$}t$ does not
scale as $\sqrt{\mbox{$\rho_{\rm h}$}}$. This makes the universality of the
GCMF a problem (again), when trying to explain this by dynamical evolution
alone.
## Acknowledgement
We thank an anonymous referee for constructive comments. We are grateful to
Douglas Heggie and Ivan King for discussions. MG enjoyed discussions with
Henny Lamers during his stay in Santiago. The simulations were done on the
GRAPE-6 BLX64 boards of the European Southern Observatory in Garching. This
research was supported by the DFG cluster of excellence Origin and Structure
of the Universe (www.universe-cluster.de).
## References
* Aarseth (1999) Aarseth S. J., 1999, PASP, 111, 1333
* Ambartsumian (1938) Ambartsumian V. A., 1938, Sci. Mem. Leningrade State Univ. #22, ser. Math. Sci. (astronomy), 4, 19
* Baumgardt (2001) Baumgardt H., 2001, MNRAS, 325, 1323 (B01)
* Baumgardt et al. (2002) Baumgardt H., Hut P., Heggie D. C., 2002, MNRAS, 336, 1069
* Baumgardt & Makino (2003) Baumgardt H., Makino J., 2003, MNRAS, 340, 227
* Boutloukos & Lamers (2003) Boutloukos S. G., Lamers H. J. G. L. M., 2003, MNRAS, 338, 717
* Chernoff & Weinberg (1990) Chernoff D. F., Weinberg M. D., 1990, ApJ, 351, 121
* Engle (1999) Engle K. A., 1999, PhD thesis, AA(DREXEL UNIVERSITY)
* Fall & Rees (1977) Fall S. M., Rees M. J., 1977, MNRAS, 181, 37P
* Fukushige & Heggie (2000) Fukushige T., Heggie D. C., 2000, MNRAS, 318, 753
* Gieles et al. (2005) Gieles M., Bastian N., Lamers H. J. G. L. M., Mout J. N., 2005, A&A, 441, 949
* Gieles et al. (2006) Gieles M., Portegies Zwart S. F., Baumgardt H., Athanassoula E., Lamers H. J. G. L. M., Sipior M., Leenaarts J., 2006, MNRAS, 371, 793
* Giersz & Heggie (1994a) Giersz M., Heggie D. C., 1994a, MNRAS, 268, 257
* Giersz & Heggie (1994b) Giersz M., Heggie D. C., 1994b, MNRAS, 270, 298
* Giersz & Heggie (1997) Giersz M., Heggie D. C., 1997, MNRAS, 286, 709
* Gnedin & Ostriker (1997) Gnedin O. Y., Ostriker J. P., 1997, ApJ, 474, 223
* Goodman (1984) Goodman J., 1984, ApJ, 280, 298
* Heggie & Mathieu (1986) Heggie D. C., Mathieu R. D., 1986, in Hut P., McMillan S., eds, Lecture Notes in Physics Vol. 267, The Use of Supercomputers in Stellar Dynamics, Springer-Verlag, Berlin, p.233, 267, 233
* Hénon (1961) Hénon M., 1961, Annales d’Astrophysique, 24, 369
* King (1958) King I., 1958, AJ, 63, 465
* King (1966) King I. R., 1966, AJ, 71, 64
* Lamers et al. (2005a) Lamers H. J. G. L. M., Gieles M., Bastian N., Baumgardt H., Kharchenko N. V., Portegies Zwart S., 2005a, A&A, 441, 117
* Lamers et al. (2005b) Lamers H. J. G. L. M., Gieles M., Portegies Zwart S. F., 2005b, A&A, 429, 173
* Larsen (2004) Larsen S. S., 2004, A&A, 416, 537
* Makino et al. (2003) Makino J., Fukushige T., Koga M., Namura K., 2003, PASJ, 55, 1163
* McLaughlin & Fall (2008) McLaughlin D. E., Fall S. M., 2008, ApJ, 679, 1272
* McMillan & Hut (1996) McMillan S. L. W., Hut P., 1996, ApJ, 467, 348
* Portegies Zwart et al. (2001) Portegies Zwart S., McMillan S. L. W., Hut P., Makino J., 2001, MNRAS, 321, 199
* Scheepmaker et al. (2007) Scheepmaker R. A., Haas M. R., Gieles M., Bastian N., Larsen S. S., Lamers H. J. G. L. M., 2007, A&A, 469, 925
* Spitzer (1987) Spitzer L., 1987, Dynamical evolution of globular clusters. Princeton, NJ, Princeton University Press, 1987, 191 p. (S87)
* Spitzer (1940) Spitzer L. J., 1940, MNRAS, 100, 396
* Spitzer & Chevalier (1973) Spitzer L. J., Chevalier R. A., 1973, ApJ, 183, 565
* Spitzer & Hart (1971) Spitzer L. J., Hart M. H., 1971, ApJ, 164, 399
* Takahashi & Portegies Zwart (2000) Takahashi K., Portegies Zwart S. F., 2000, ApJ, 535, 759
* Tanikawa & Fukushige (2005) Tanikawa A., Fukushige T., 2005, PASJ, 57, 155
* Vesperini & Heggie (1997) Vesperini E., Heggie D. C., 1997, MNRAS, 289, 898
* Wielen (1971) Wielen R., 1971, Ap&SS, 13, 300
* Wielen (1985) Wielen R., 1985, in Goodman J., Hut P., eds, IAU Symp. 113: Dynamics of Star Clusters Dynamics of open star clusters. pp 449–460
* Zepf et al. (1999) Zepf S. E., Ashman K. M., English J., Freeman K. C., Sharples R. M., 1999, AJ, 118, 752
| arxiv-papers | 2008-06-15T20:12:15 | 2024-09-04T02:48:56.256334 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "M. Gieles (1) and H. Baumgardt (2) ((1) ESO/Santiago, (2) Bonn)",
"submitter": "Mark Gieles",
"url": "https://arxiv.org/abs/0806.2327"
} |
0806.2360 | # Existence of a polyhedron which does not have a non-overlapping pseudo-edge
unfolding
Alexey S Tarasov ISA RAS, Moscow, Russia,
###### Abstract
There exists a surface of a convex polyhedron $P$ and a partition $L$ of $P$
into intrinsically flat and convex geodesic polygons such that there are no
connected "edge" unfoldings of $P$ without self-intersections (whose spanning
tree is a subset of the edge skeleton of $L$).
###### keywords:
unfolding , polyhedron , partition , geodesic graph
††thanks: Work is supported by RFBR grants 08-01-00565, 08-01-91202, DFG
Research Unit ‘‘Polyhedral Surfaces’’
## 1 Introduction
Let $S$ be an abstract $2$-dimensional polyhedral surface with $n$ vertices.
We say that $S$ is intrinsically convex if the curvatures
$\omega_{1},\omega_{2},\ldots,\omega_{n}$ of all vertices satisfy
$0<\omega_{i}<2\pi$ for all $i\in[n]$. Of course the surface of every convex
polytope in $\mathbb{R}^{3}$ is intrinsically convex. The following result
shows that there are essentially no other examples:
(Alexandrov’s existence theorem). Every intrinsically convex $2$-dimensional
polyhedral surface homeomorphic to a sphere is isometric to the surface of a
convex polytope $P\in\mathbb{R}^{3}$ or to a doubly covered polygon.
There is a conjecture: every convex polyhedron has an edge net. This question
has a long story. The earliest reference to edge unfolding was made by Albreht
Dürer [2]. At first time this conjecutre was explicilty posted by Sheppard in
1975 She75 .
At first glance the conjecture seems to be true. If we take a polyhedron, it
is very easy to find its unfolding.
Even to find a polyhedron with at least one overlapping unfolding is not
trivial.
On fig. 1 we can see such example of overlapping unfolding to the thin
truncated pyramid. It is clear that the unfolding has overlapping because the
curvature (angle defect) of the upper base is very small and the left angle of
the upper base is noticeable less than $\pi/2$. We will use such idea later.
It is convenient to consider this conjecture by thinking about a polyhedral
surface as a metric space without embedding into $\mathbb{R}^{3}$.
According to Alexandrov’s theorem, such embedding always exists and we do not
need to care about it.
It is easy to find vertices on a surface – they have total sum of angles less
than $2\pi$. But points on edges and on faces do not have such distinctive
features.
There is no simple way without a lot of calculations to find edges skeleton.
Alexandrov’s theorem says that such skeleton exists and unique, but does not
give any way how to find it. There is a more constructive proof of this
theorem (BobIzm06 ). But still this proof need make a lot of calculations to
find edges.
Then we have the following question: does edge-skeleton a really important in
the Dürer conjecture.
We can formulate a similar conjecture, which does not depend from edge-
skeleton: Let $P$ be a polyhedral surface homeomorphic to a sphere. Let $L$ be
a partition of $P$ into geodesic polygons, flat and convex in terms of the
intrinsic metric (see def. 3) Does there exist a non-overlapping unfolding of
this polyhedral surface, whose spanning tree lies inside edge skeleton $E(L)$?
In this paper we construct a counterexample to this question.
A similar question about geodesic triangulated 1-skeletons was asked by Jeff
Erickson Eric06 . Let $T$ be an arbitray triangulation of a convex polytope
whose vertices are the vertices of the polytope and whose edges are geodesics.
Can the surface by unfolded without self-overlap by cutting it along edges of
$T$?
It appears that for Erickson’s question similar counterexample can also be
constructed.
This question was inspired by another very interesting instinsic conjecture on
the same conference:
Bobenko’s conjecture Bob06 Can the boundary of a convex polytope be unfolded
into the plane without self-overlap by cutting the surface along of the
Delaunay triangulation $T$ of the boundary? ($T$ can have loops and multiple
edges, but the faces are triangles). In particular, if all faces of a polytope
are acute triangles, can it be unfolded by cutting along the edges of the
polytope?
Here another two similar conjectures by O’Rourke ORED : Prove (or disprove)
that every convex polyhedron all of whose faces have no acute angles (i.e.,
all angles are $\geq 90^{\circ}$) has a non-overlapping edge-unfolding
Prove (or disprove) that every triangulated convex polyhedron, all of whose
angles are non-obtuse, i.e., $\leq 90^{\circ}$, has a non-overlapping edge-
unfolding. If the polyhedron has $F$ faces, what is the fewest nets into which
it may be cut along edges?
|
---|---
Figure 1: Overlapping unfolding.
Outline of the proof:
1. 1.
Criteria for non-existence of an unfolding for a special polyhedral surface
with border.
Here we will give a sufficient condition for non-existence of a one-piece
unfolding which is generalizing overlapping as on fig. 1. This condition is
defined on the plane in terms of the weighted center of subtree vertices with
respect to their corresponding root edges.
2. 2.
Construction a spiral-like gadget.
Here we will show the polygon $T$ and the partition $L_{T}$ with some special
properties. In particular, the polyhedral surface in some sense close enough
to this partition does not have an unfolding. does not have an unfoldable net.
3. 3.
Construction of a counterexample.
Here we will show a counterexample (in which additional vertices with no
curvature are used).
4. 4.
Proof that the counterexample does not have an unfolding. Construction of a
counterexample without additional zero-curvature vertices.
5. 5.
A simpler counterexample.
## 2 Definitions
Definition 1. An unfolding is a set of polygons with a rule for gluing the
boundaries.
Definition 2. The curvature $curv(v)$ of a vertex $v$ is the difference
between $2\pi$ and the sum of planar angles incident to this vertex.
Definition 3. A partition $L$ of a surface of a polyhedron is called convex
geodesic if the following conditions hold:
1\. Any vertex with non-zero curvature is a vertex of $L$;
2\. Any part $P_{i}$ of the partition $L$ unfolds into a flat convex polygon.
Definition 4. $L$-unfolding is a unfolding of $P$ spanning tree of which is a
subset of edge-skeleton $E(L)$ of the convex geodesic partition $L$.
Definition 5. An infinitesimally curved polyhedron is a convex polygon $P$ on
the plane $T$ together with finitely many internal points $p_{i}\in int(P)$ in
general position and weights $\alpha_{i}$ for each internal point
($\alpha_{i}\geq 0,\sum_{\forall i}\alpha_{i}=1$); Vertices $q_{i}$ are
vertices of polygon $P$.
Let us assume that diameter $max(x,y)\forall x,y\in P$ of $P$ is equal to $1$.
Definition 6. For a given infinitesimally curved surface $P$ and
$0<\beta<2\pi$ a cap $H$ is a polyhedral surface with border, if the following
conditions hold:
1\. $H$ above plane $T$ and curved downward.
2\. The projection of $H$ on the plane $T$ is $P$.
3\. Points $p_{i}$ are projections of vertices $h_{i}$ of $H$ on the plane
$T$.
4\. Curvature $curv(h_{i})=\alpha_{i}\cdot\beta$;
The existence of a cap follows this theorem:
Alexandrov’s theorem (Alex49 , Theorem 21, Part. II §5, p. 118) For any set of
points $A_{1},\ldots,A_{m}$ on the plane $T$ and any set of numbers
$\omega_{1},\ldots,\omega_{m}$, such that:
1\. $0<\omega_{i}<2\pi$ for all $1\leq i\leq m$
2\. $\sum_{i=1}^{m}\omega_{i}=2\pi$,
there exists an unique infinite convex polyhedral surface $Q$ with the
following properties:
1\. Points $A_{1},\ldots,A_{m}$ are projections of the vertices $h_{i}$ of
$Q$;
2\. Curvature $curv(h_{i})=\omega_{i}$.
Put to any point $q_{i}$ curvature $(2\pi-\angle
q_{i})(1-\frac{\beta}{2\pi})$, to point $p_{i}$ curvature $\alpha_{i}\beta$.
Total sum of curvature is $2\pi$. By Alexandrov’s theorem there exists an
unique infinite polyhedral surface $Q$ satisfying these conditions. Suppose
$H$ is a part of $Q$ and $H$ projects into $P$. The $H$ is a required cap.
Note that $H$ depends continuously on $\beta$. Indeed, the map of $P$ into a
set of points $A_{i}$ and numbers $\omega_{i}$ is continuous. Thus, the
reverse map not only unique but also continuous.
Definition 7. Let $L$ be any partition of the polygon $P$ into convex polygons
$P_{i}$, such that the vertices of $L$ are the vertices of $P$ and points
$p_{i}$. Denote by $E(L)$ the 1-skeleton of the partition $L$.
Note that the partition $L$ does not have any correspondence with faces of
$H$.
Definition 8. Define a map $f:P\to H\subset\mathbb{R}^{3}$. Let triangulation
$L^{\prime}$ be some subdivided partition $L$ if the following conditions
holds:
1. 1.
For every vertex $p_{i}$ and $q_{i}$ the projection of $f(x)$ into $T$ is $x$.
2. 2.
For every edge $xy$ of the triangulation $L^{\prime}$, an image $f(xy)$ is a
shortest path between $f(x)$ and $f(y)$ stretched evenly.
Hence points $p_{i}$ in general position, for small enough $\beta$ edges of
$f(L^{\prime})$ in $H$ are defined unique and continuously depending from
$\beta$.
3. 3.
For a triangle $\triangle xyz$ of $L^{\prime}$ we can unfold the corresponding
geodesic triangle $\triangle f(x)f(y)f(z)$ into a plane. Then $f$ is the
unique affine transformation of $\triangle xyz$ to $\triangle f(x)f(y)f(z)$.
It is clear that this map $f$ is one-one.
Denote by $h_{i}=f(p_{i})$ a vertices of the hat $H$.
When it is clear from context we will denote an objects (partition and cuts)
and its image similarly.
Definition 9. A subgraph $G\subset E(L)\setminus\partial P$ is called a cut if
the following conditions hold:
(i) $G$ contains all points with positive weights;
(ii) $G\cup\partial P$ is connected.
(iii) $G$ does not contain cycles.
Definition 10. Denote by $U_{G}$ the unfolding of $H$ corresponding to the cut
$G_{\beta}$.
We consider that the number of components of $U_{G}$ can be more than $1$ This
number is equal to the number of independent cycles of the graph
$G\cup\partial P$.
Definition 11. For a given cut (of graph) $G$ on the polyhedral surface $H$
let $g:H->U_{G}$ be a natural map. If the point $x\notin G$ then $g(x)$ is a
single point.
Definition 12. An edge $e$ of a cut $G$ is called an $A$-edge if $e$
disconnects one of its vertices from $\partial P$. Otherwise it is called a
$B$-edge (fig. 2).
Note that if a graph $G\cup\partial P$ contains only $2$ faces (inside and
outside $\partial P$), then $G$ does not contain $B$-edges.
Denote by $G_{A}$ the graph consisting of the $A-$edges of the cut $G$. Define
graph $G_{B}$ similarly.
Figure 2: Graphs $G_{A}$ and $G_{B}$
Note several facts about $G_{A}$ and $G_{B}$: Each component of $G_{A}$ is a
tree, which has one common point (outfall) with $\partial P$ or $G_{B}$. All
endpoints of $G_{B}$ lie in $\partial P$.
Definition 13. Given a cut $G$, we call a vertex $p_{1}$ upstream to a point
$p_{2}$ ($p_{1}>p_{2}$) if any path connecting $p_{1}$ with $\partial P$
passes through $p_{2}$. Similarly, we define the upstream partial order on
$A$-edges. Similarly we can define a downstream one.
From every vertex $v$ in $G_{A}$ goes only one downstream edge.
Definition 14. Consider a neighborhood $B(y,r)$ of a point $y$ on some
$A$-edge $e$ ($y$ is not a vertex) where $r$ is small enough that $B(y,r)\cap
G$ is a part of the edge $e$.
Image of $B(y,r)$ in the unfolding $U_{G}$ consists of two components (fig.
3). Each of these parts lies on one side of $e$ – left or right (we are
looking downstream $e$). Denote these parts respectively by $B_{L}(y,r)$ and
$B_{R}(y,r)$.
Denote by $y^{\prime}_{R}$ ($y^{\prime}_{L}$) an image of $y$ laying in
$B^{\prime}_{R}(y,r)$ ($B^{\prime}_{L}(y,r)$).
Similarly, we can define left and right images of any $A$-edge.
Vertex $p_{i}$ can have more than two images. If $p_{i}$ has an incident
downstream $A-$edge $e$, two images of $p_{i}$ are $u_{i,L}$ and $u_{i,R}$
corresponding to the images $e_{L}$ and $e_{R}$.
Figure 3: Left and right images of point $y$.
Definition 15. A stream of any point $x$ on an $A$-edge (including the
vertices) is the vector $\sum_{p_{i}>x}\alpha_{i}(x-p_{i}x)$. Another way to
define the stream is: A rotation center $c_{x}$ is
$\alpha_{x}^{-1}\sum_{p_{i}>x}\alpha_{i}p_{i}$ (the weighted barycenter). Then
the stream is equal to
$f_{x}=\alpha_{x}(x-c_{x}),$ $None$
where $\alpha_{x}=\sum_{p_{i}>x}\alpha_{i}$ (fig. 4).
Figure 4: Stream for point $p_{1}$
Term stream is called by analogy with a river. I imagine that graph $G$ is a
river net. The stream is the force and the direction there river wants to
flow. Later we will show that for a non-overlapping unfolding the angle
between the direction of the stream and the river (edges) should be no more
than $\pi/2$.
Definition 16. A cut $G$ is called admissible if for every point $xinG_{A}$
with $\alpha_{x}>0$
$<x-c_{x},e>/|e|\geq 0.$ $None$
We assume that the edge $e$ is directed from the upper to the lower vertex.
If $x$ is a vertex of $G$, the condition $(2)$ defines a half-plane where edge
$e$ should lie.
Denote by $l(G)$ the maximum of $-<x-c_{x},e>/|e|$. For a non-admissible cut
$G$ $l(G)>0$.
Note that if we draw away from the cut $G$ all edges $e$ with $\alpha_{e}=0$,
the obtained graph $G^{\prime}$ also is a cut without changing its
admissibility.
Definition 17. Denote by $l(L)=min_{\forall cutG,G_{B}=\emptyset}l(G)$.
A partition $L$ is called non-admissible if any admissible cut $G$ of $L$
contains non empty $G_{B}$.
As $L$ has finitely many vertices, it has finitely many subgraphs, Thus, the
number of possible cuts $G$ with $G_{B}=\emptyset$ is also finite. Hence, for
a non-admissible partition $L$ $l(L)>0$.
## 3 Criteria for non-existence of an unfolding in $H$
In this section we will show, that where is no $L$-unfolding for the
polyhedral surface $H$ to the infinitesimal surface with the non-admissible
partition $L$ when curvature $\beta$ of $mycap$ is small enough.
Unfolding is one or several polygons. We can arrange
Figure 5: Arrangement of $P$ and corresponding $U$.
Let us try to arrange these polygons on the plane $T$ to make any point $y\in
U_{G}$ be as close as possible to its pre-image (fig. 5).
Then $\beta=0$, $H$ is a flat polygon equal to $P$. And unfolding $U_{G}$ is
equal to $P$.
Since $H$ depends continuously on $\beta$ for a small enough $\beta$ we can
arrange $P$ and $U_{G}$ on the same plane in such way that:
1\. The distance
$|xx^{\prime}|<\varepsilon,$ $None$
for any point $x\in P$ and any image of it $x^{\prime}\in U_{G}$
2\. For any segment $xy\in P$ and its connected image $x^{\prime}y^{\prime}$,
$\left|\,|x-y|-|x^{\prime}-y^{\prime}|\,\right|<\varepsilon|x-y|$ $None$
Let $U_{G}$ be an unfolding combined with a polygon $P$ satisfying (3.1) and
(3.2). Consider some point $x$ on the cut $G$ and its two images
$x^{\prime}_{L},x^{\prime}_{R}$. Denote by $\tilde{c}$ a point satisfying
$x^{\prime}_{R}-x^{\prime}_{L}=\beta\alpha_{x}M_{rot}\cdot(x-\tilde{c}_{x}),$
$None$
where $M_{rot}$ is a $\pi/2$-clockwise rotation matrix.
###### Lemma 1
For a small enough $\beta$ and any point $x\in G$
$|\tilde{c}_{x}-c_{x}|<3\varepsilon.$
Proof.
1. 1.
Let $s$ be a Jordan curve on $P$ (with possible breaks), which intersects $G$
only at a point $x$, contains inside upstream part $G_{x}=\\{y\in G|y>x\\}$,
and touches $G$ at all vertices $p_{i}>x$ once (fig. 6). Let $f(s)$ be the
image of the curve $s$ on the polyhedral surface $H$, and let $g(f(s))$ be the
connected image of $f(s)$ on $U_{G}$.
Denote by $u_{i}$ the image of vertex $p_{i}$ touched by $g(f(s))$. Denote by
$c^{\prime}_{x}=\alpha_{x}^{-1}\sum_{p_{i}>x}\alpha_{i}u_{i}$
Endpoints of the curve $g(f(s))$ are $x^{\prime}_{L}$ and $x^{\prime}_{R}$.
Figure 6: $G_{x}$ and images of $s$ Figure 7: $G^{\prime}$ and $U^{\prime}$
2. 2.
The outer semi-neighborhood of the curve $s$ has a unique unfolding, which
does not depend from anything beyond this semi-neighborhood.
Let us replace $G_{x}$ in the graph $G$ by curves $s_{i}$, which connect each
vertex $p_{i}$ and $x$ without intersecting each other and curve $s$. Denote
by $U^{\prime}$ the unfolding corresponding to this constructed curved
spanning tree $G^{\prime}$ (fig. 7 b). This unfolding is not an $L$-unfolding.
It is possible. Since $s$ is a Jordan curve, it bounds a region homeomorphic
to a disk [SJ ].
Take the images of the vertex $p_{i}$ and the point $x$ on this disk and
connect them by the chord. Pre-image of this chord is $s_{i}$. Chords for
different $i$ do not intersect each other, so curves $s_{i}$ do also not
intersect each other.
Every vertex $p_{i}$ has a unique image $u_{i}$ in the unfolding $U^{\prime}$.
3. 3.
Consider the path connecting points $h_{i}$ and $x$ in the graph $G^{\prime}$.
There are two images of $x$ corresponding to this cut: $x^{\prime}_{i,L}$ and
$x^{\prime}_{i,R}$ (fig. 8 a). Vector $x^{\prime}_{i,R}-x^{\prime}_{i,L}$ does
not depend on the form of $s_{i}$ and is equal to
$M_{rot}\cdot(x^{\prime}_{i,c}-u_{i})2\sin(\beta\alpha_{i}/2)$, where
$x^{\prime}_{i,c}$ is the middle point of the segment
$x^{\prime}_{i,l}x^{\prime}_{i,r}$ (fig. 8 b).
The distance between $x^{\prime}_{i,c}$ and $x$ is less than $\varepsilon$.
Denote by $w_{i}$ the point such that
$(w_{i}-x^{\prime}_{i,c})=(u_{i}-x^{\prime}_{i,c})\frac{2\sin(\beta\alpha_{i}/2)}{\beta\alpha_{i}}$.
Then $|u_{i}w_{i}|<\frac{\beta\alpha_{i}^{2}}{24}<\beta$ when
$\beta\leq\varepsilon<\sqrt{1/5}$.
Let $\tilde{u}_{i}$ be given by the formula $\tilde{u}_{i}=w_{i}+(x-x_{i,c})$.
Distance $|\tilde{u}_{i}-u_{i}|<2\varepsilon$.
Figure 8: Images of $x$
Then
$x^{\prime}_{i,R}-x^{\prime}_{i,L}=M_{rot}\cdot(x^{\prime}_{i,c}-u_{i})2\sin(\beta\alpha_{i}/2)=M_{rot}\cdot(x^{\prime}_{i,c}-w_{i})\beta\alpha_{i}=\beta\alpha_{i}M_{rot}\cdot(x-\tilde{u}_{i})$
and $|\tilde{u}_{i}-u_{i}|<\varepsilon+\beta\leq 2\varepsilon$.
We get $x^{\prime}_{R}-x^{\prime}_{L}=\sum_{\forall
i:p_{i}>x}(x^{\prime}_{i,R}-x^{\prime}_{i,L})=\sum_{p_{i}>x}\beta\alpha_{i}M_{rot}\cdot(x-\tilde{u}_{i})$
By $(4)$
$|\tilde{c}_{x}=\alpha_{x}^{-1}\sum_{p_{i}>x}\alpha_{i}\tilde{u}_{i}$.
As $\tilde{c}_{x}$ is a weighted center of $\tilde{u}_{i}$, and
$c^{\prime}_{x}$ is a weighted center of $u_{i}$, we obtain:
$|\tilde{c}_{x}-c^{\prime}_{x}|<2\varepsilon$.
4. 4.
By (3.1) $|u_{i}-p_{i}|<\varepsilon$. Then
$|c^{\prime}_{x}-c_{x}|<\varepsilon$ and $|\tilde{c}_{x}-c_{x}|<3\varepsilon$.
###### Theorem 1
For a given infinitesimally curved surface $P$ and its partition $L$ there
exists an $\varepsilon$ such that for the cap $H$ with curvature
$\beta<\varepsilon$, for any non-admissible cut $G$ a $L$-unfolding $U_{G}$
overlaps.
Proof.
1. 1.
As $L$ is a non-admissible partition then $0<l(L)<1$. As $L$ is fixed, denote
in this theorem $l(L)$ by $l$.
Consider a cut $G$.
Let $A$ be a sufficiently small number, such that the disk with radius $A$ and
center $p_{i}$ for any $i$ intersects the skeleton $E(L)$ only at the edges
incident to the point $p_{i}$. Let $\gamma$ be the smallest angle between
incident edges of the skeleton $E(L)$.
Let the unfolding $U_{G}$ be arranged with the polygon $P$ satisfying (3.1)
and (3.2).
2. 2.
Since the cut $G$ is non-admissible, there exists an edge $p_{i}p_{j}\in G$,
$p_{i}>p_{j}$ and not satisfying the admissibility condition $(2)$:
$(f_{p_{i}},p_{i}p_{j})<-l|p_{j}-p_{i}|<0$. Let us take a point $a$ on the
edge $p_{i}p_{j}$, such that the distance between $p_{i}$ and this point $a$
is $r_{1}=\min(A/2,l/2)$. Since $(a-c_{a})=((p_{i}-c_{a})+(a-p_{i}))$. Using
$|a-p_{i}|<l/2<r_{1}$, we obtain $<a-p_{i},p_{j}-p_{i}><l/2|p_{j}-p_{i}|$.
Then
$<a-c_{a},p_{j}-p_{i}><-l/2|p_{j}-p_{i}|.$ $None$
Figure 9: The point $a$ and the disk $D$.
3. 3.
Let show that the unfolding $U_{G}$ has self-intersections in the point $a$,
i.e. the right image $a^{\prime}_{R}$ lies inside interior of $U_{G}$.
Denote by $b$ the point $a+\beta M_{rot}\cdot f_{a}$ (fig. 9).
Denote by $P_{L},P_{R}$ two polygons of the partition $L$ incident to the edge
$p_{i}p_{j}$, respectively to the left and to the right of the downstream
direction.
Let $D$ be a disk with the center $b$ and radius
$r_{D}=\beta\alpha_{a}\frac{l}{2}$.
Let us show that $d=|b^{\prime}-a^{\prime}_{R}|<r_{D}$, i.e. $a^{\prime}_{R}$
lies inside the image of $P_{L}$.
4. 4.
If $\beta$ is small enough ($\beta<\frac{A}{4}$), then we have
$|p_{i}-a|+|a-b|+r_{D}\leq r_{1}+\beta|f_{x}|+r_{D}$.
As $f_{a}\leq\alpha_{a}(c_{a}-a)\leq 1$, we obtain. $|p_{i}-a|+|a-b|+r_{D}\leq
A/2+\beta+\beta\frac{l}{2}<A/2+A/4(1+l/2)<A$.
The disk $D$ lies inside $P_{L}\cup P_{R}$. Indeed, for any point $x\in D$,
$\sin(\angle
ap_{i}x)<\frac{r_{D}+\beta|f_{x}|}{r_{1}-\beta|f_{x}|-r_{D}}<\frac{\beta(l/2+1)}{r_{1}-\beta(l/2+1}<\beta\frac{3}{r_{1}}$.
For $\beta<\frac{\sin(\gamma)r_{1}}{3}$. $\sin(\angle ap_{i}x)<\sin(\gamma)$.
Then $\angle ap_{i}x<\gamma$ and the disk $D$ lies inside $P_{L}\cup P_{R}$.
5. 5.
$\sin(\angle
xab)\leq\frac{r_{D}}{|b-a|}=\frac{\beta\alpha_{a}l/2}{\beta\alpha_{a}|a-c_{a}|}=\frac{l/2}{|a-c_{a}|}$.
Using $(5)$ we obtain $\sin(\angle bap_{j})=-\cos(\angle
c_{a}ap_{j})>\frac{l/2}{a-c_{a}}$. Then the disk $D$ does not intersect with
the edge $p_{i}p_{j}$. Hence, $D\subset int(P_{L})$.
6. 6.
An image $g(f(b))$ of the point $b$ is the single point $b^{\prime}$.
An image $g(f(a))$ consists of two points $a^{\prime}_{L},a^{\prime}_{R}$.
Taking into account second condition of Proposition $1$, we obtain
$|(b^{\prime}-a^{\prime}_{L})-(b-a)|<\varepsilon|(b-a)|=\varepsilon|f_{a}|$.
7. 7.
Consider the distance $d=|b^{\prime}-a^{\prime}_{R}|$.
$d=|b^{\prime}-a^{\prime}_{R}|=|(a^{\prime}_{R}-a^{\prime}_{L})-(b^{\prime}-a^{\prime}_{L})|$.
Using (3.1),
$b^{\prime}-a^{\prime}_{L}=b-a+\varepsilon|b-a|=b-a+\varepsilon\beta f_{a}$,
$d<|(a^{\prime}_{R}-a^{\prime}_{L})-(b-a)|+\varepsilon\beta|f_{a}|.$
Using Lemma 1,
$a^{\prime}_{R}-a^{\prime}_{L}=\beta\alpha_{a}M_{rot}\cdot(a-\tilde{c}_{a})$,
and $|\tilde{c}_{a}-c_{a}|<3\varepsilon$,
$d<\beta\alpha_{a}|(a-\tilde{c}_{a})-(a-c_{a})|+\varepsilon\beta\alpha_{a}|a-c_{a}|\leq\beta\alpha_{a}(3\varepsilon+\varepsilon|c_{a}-a|)<\beta\alpha_{a}4\varepsilon$.
Then when $\beta<\frac{l}{8}$, we obtain that
$d=|a^{\prime}_{R}-b^{\prime}|<r_{D}$.
Thus, if $\beta$ is small enough, the point $a^{\prime}_{R}$ lies inside the
image of the disk $D$ in the unfolding, i.e. an intersection has occurred.
## 4 Construction of the gadget
In this section we will construct a gadget $T$ – a special example of
infinitesimal polygon and its partition.
Our goal is the following theorem:
###### Theorem 2
There exists an infinitesimal surface $T$ and its partition $L_{T}$ that the
following conditions are hold:
1\. Polygon $T$ is a triangle.
2\. Partition $L_{T}$ is non-admissible, i.e. any admissible graph $G$ of
$L_{T}$ contains non-empty part $G_{B}$.
3\. A admissible graph $G$ connects all non-zero weighted vertices of $T$.
Outline of the proof:
1\. The basic idea of a spiral .
2\. Constructing the central ‘‘square’’ part of the gadget $T$.
3\. Constructing the whole gadget $T$.
4.–7. The proof the theorem
In spite of the fact that the gadget $T$ is very complicated, only a few
things are needed for the main proof: the combinatorial structure of $L_{T}$,
the fact that the partition $L_{T}$ is convex, and metric properties,
enumerated on fig. 17.
Proof.
1. 1.
At first we will the show basic idea how to construct a partition with very
few admissible spanning trees.
For a fixed vertex $a$ of the cut $G$, the rule $(2)$ defines a half-plane,
which the downstream edge from $a$ should point at. There exists at least one
such edge, because $L$ is a convex partition. But this rule allows this edge
not to be unique. It turns out, that one can construct special cases for which
such continuation is almost unique.
Here we construct a partition, for which, given a starting point, we can
define any admissible cut unambiguously.
For the center $C$, the number of points in period $n$, the growth parameter
$0<q\in\mathbb{R}$ and the starting point $s$ (or endpoint $e$ with index
$i_{e}$) take a logarithmic spiral
$f(x)=(C_{x}+r_{0}q^{x}\cos(\phi_{0}+2\pi x),C_{y}+r_{0}q^{x}\sin(\phi_{0}2\pi
x))$ $None$
where $x=i/n$, $r_{0},\phi_{0}$ are parameters to fit $f(0)=s$ (or to fit
$f(i_{e}/n)=e$).
Figure 10: Spiral Figure 11: Adjacent points to $f(0)$
Let $k,n\in\mathbb{N}$ be an integer numbers satisfying following conditions
(fig. 10 a):
$\angle Of(0)f(k/n)<\pi/2$ $None$ $f(0)\in int(\triangle Of(1/n)f(k/n)).$
$None$
Let $0<x<1$ be such value that $\angle Of(0)f(x)=\pi/2$. Then $k/n$ should be
less and close enough to the $x$. If $n$ is large enough such $k$ always
exists. Take $n_{max}>n+2k$.
Now take $\sf{a}_{0}=O$ $\sf{a}_{i}=f(i/n)$ for $1\leq i\leq n_{max}$.
$P_{spiral}=conv(\sf{a}_{0},\sf{a}_{1},..\sf{a}_{n_{max}-k})$. Weights
$\alpha_{0}=1$ and $\alpha_{i}=0$ for all $i>0$.
For any $i<n_{max}$ construct an edge between points $\sf{a}_{i}$ and
$\sf{a}_{i+1}$. For $0<i\leq k$ construct an edge between points $\sf{a}_{0}$
and $\sf{a}_{i}$. For $0<i\leq n_{max}-k$ construct an edge between points
$\sf{a}_{i}$ and $\sf{a}_{i+k}$. This set of edges intersecting with
$P_{spiral}$ defines the partition $L_{spiral}$. There are $3$ or $4$ angles
around any non-bound vertex $\sf{a}_{i}$. The biggest angle is equal to
$\angle f(1/n)f(0)f(k/n)$. This angle is less than $\pi$ because of $(7.2)$.
(fig. 10 b). Thus, all polygons of this partition are convex.
Consider some admissible graph $G$.
Suppose the vertex $\sf{a}_{i}$ is downstream to $\sf{a}_{0}$, then
$c_{\sf{a}_{i}}=\sf{a}_{0}$.
There are some possible ways from $\sf{a}_{i}$:
$O,\sf{a}_{i-1},\sf{a}_{i+1},\sf{a}_{i-k},\sf{a}_{i+k}$ (not all of them exist
at the same time).
There is only one way with supporting condition $(2)$ :
$\sf{a}_{i}\sf{a}_{i+1}$ (you see on fig. 11 only $f(1/n)$ is under the line).
Thus, any admissible cut of this partition contains the polyline
$\sf{a}_{k},\sf{a}_{k+1},\ldots,\sf{a}_{n_{3}}$, where $n_{3}$ is the minimal
index of the point $p_{i}$ on the boundary of the convex hull $P_{spiral}$
($\sf{a}_{71},\ldots,\sf{a}_{85}$ on fig. 10 b).
Such unambiguous continuation is the main idea in the construction of our
counterexample.
2. 2.
Now we construct the gadget $T$ (fig. 12, 13, 14, 15).
At first let us construct central part – the ‘‘square". This part is centrally
symmetrical.
Two vertices: ${\sf c}_{1}=(-70,0);{\sf c}_{2}=(70,0)$. These points have
equal weights $1/2$. All following vertices in the gadget have a zero weight.
Add points ${\sf h}_{1}=(-5,15);{\sf e}_{1}=(-5,10);$ ${\sf
j}_{1}=(-3.791,-5.006);{\sf f}_{1}=(-6.565,21.485);{\sf
f}^{\prime}_{1}=c_{1,199}=(-6.807,21.404);{\sf m}_{1}=(-11.002,42.812);{\sf
m}^{\prime}_{1}={\sf c}_{2,139}=(17.124,40.546);{\sf
n}_{1}=(213.886,53.695);{\sf n}^{\prime}_{1}={\sf
c}_{1,198}=(-53.695,218.886)$. Note that ${\sf f}^{\prime}_{1}$ is very close
to ${\sf f}_{1}$ (fig. 14 b).
Let us in a centrally symmetrical way take points with opposite index. Add
edges between ${\sf n}_{1},{\sf n}^{\prime}_{1},{\sf n}_{2},{\sf
n}^{\prime}_{2}$. This is our square.
Then add the following edges ${\sf h}_{1}{\sf h}_{2},{\sf e}_{1}{\sf
h}_{1},{\sf h}_{1}{\sf f}_{1},{\sf f}_{1}{\sf j}_{2},{\sf f}_{1}{\sf
m}^{\prime}_{1},{\sf f}_{1}{\sf f}^{\prime}_{1},{\sf f}^{\prime}_{1}{\sf
m}_{1},{\sf e}_{1}{\sf f}^{\prime}_{1},{\sf m}_{1}{\sf m}^{\prime}_{1},{\sf
m}^{\prime}_{1}{\sf n}^{\prime}_{1}$ and all centrally symmetrical them edges.
Using formula $(6)$, let us make two spirals around ${\sf c}_{1},{\sf c}_{2}$
ending in ${\sf e}_{1},{\sf e}_{2}$. Parameters for a spiral:
$q=2,n=83,k=70,n_{max}=199$ (we can take any value of $q$, but if $q$ is
small, then $n$ and $k$ must be very big, if $q$ is bigger – we can not draw a
picture).
Then we make a clock-wise spiral around ${\sf c}_{1}$ : ${\sf c}_{1,0},{\sf
c}_{1,1},\ldots,{\sf c}_{1,129}={\sf e}_{1},\ldots{\sf c}_{1,199}$.
Add the following edges: From ${\sf c}_{1}$ to ${\sf c}_{1,i}$ for all $i\leq
k$.
From ${\sf c}_{1,i}$ to ${\sf c}_{1,i+1}$ for all $i<k-1$.
From ${\sf c}_{i,i}$ to ${\sf c}_{1,i+k}$ for all $i<k$ (for $i=k$ the edge is
already added, ${\sf e}_{1}{\sf f}^{\prime}_{1}$).
We have to do something with tails: edges ${\sf c}_{1,i}{\sf c}_{1,i+k}$ where
$i>129$. We change position of the last vertex to the first intersection of a
segment ${\sf c}_{1,i}{\sf c}_{1,i+k}$ (from point ${\sf c}_{1,i}$) with a
line constructed above.
Now points ${\sf c}_{1,130},{\sf c}_{1,131}$ lie on ${\sf e}_{1}{\sf j}_{1}$,
${\sf c}_{1,132},\ldots{\sf c}_{1,138}$ lie on ${\sf j}_{1}{\sf
m}^{\prime}_{2}$ (${\sf m}^{\prime}_{2}={\sf c}_{1,139}$), ${\sf
c}_{1,140},\ldots{\sf c}_{1,150}$ lie on ${\sf m}_{2}{\sf n}^{\prime}_{2}$,
${\sf c}_{1,151},\ldots{\sf c}_{1,171}$ lie on ${\sf n}^{\prime}_{2}{\sf
n}_{2}$, ${\sf c}_{1,172},\ldots{\sf c}_{1,198}$ lie on ${\sf n}_{2}{\sf
n}^{\prime}_{1}$ (${\sf n}^{\prime}_{1}={\sf c}_{1,198})$.
We construct second spiral around ${\sf c}_{2}$ in the same way. At last, in
order to make all angles less than $\pi$, ley us slightly bend all semgnets
where ends of tails lie. Every ‘‘tail" become shorter and does not change its
direction.
The ‘‘square" is now slightly concave.
Figure 12: Square part of the gadget $T$ Figure 13: Zoom of the square part. |
---|---
Figure 14: Zoom of the square part and deep zoom around ${\sf f}_{1}$ with
${\sf f}^{\prime}_{1}$ Figure 15: Zoom around the edge ${\sf
m}^{\prime}_{1}{\sf c}_{2,69}$
3. 3.
Making a whole gadget (fig. 16).
Make a triangle with the vertices ${\sf t}_{i}=7500(\sin(2\pi i),\cos(2\pi
i))$, $i\in\\{1,2,3\\}$.
Make third spiral with center in origin. Parameters of spiral:
$q=2.4;n=83;k=68;n_{max}=87;$ (this spiral works for three centers: ${\sf
c}_{1},{\sf c}_{2},{\sf o}$ so we need bigger $q$).
Coordinates of starting point ${\sf o}_{0}=(0,-1198.4)$, of ending points
${\sf p}={\sf o}_{87}=(-894.6,-2863.5)$. Add edges ${\sf n}^{\prime}_{1}{\sf
o}_{42},{\sf n}^{\prime}_{2}{\sf o}_{0},{\sf n}_{1}{\sf o}_{21},{\sf
n}_{2}{\sf o}_{63}$.
Cut tails as in previous case. Vertices ${\sf o}_{88},\ldots,{\sf o}_{100}$
are on the edge ${\sf t}_{1}{\sf t}_{2}$. Vertices ${\sf o}_{101},\ldots,{\sf
o}_{127}$ are on the edge ${\sf t}_{2}{\sf t}_{3}$. Vertices ${\sf
o}_{128},\ldots,{\sf o}_{153}$ are on the edge ${\sf t}_{3}{\sf t}_{1}$. Bend
edges ${\sf t}_{1}{\sf t}_{2},{\sf t}_{2}{\sf t}_{3},{\sf t}_{3}{\sf t}_{1}$
to makes all angles strictly less than $\pi$. These edges became slightly
concave polylines.
The gadget is ready. The polygon $T$ is $\triangle{\sf t}_{1}{\sf t}_{2}{\sf
t}_{3}$.
Partition $L_{T}$ is defined as set of the vertices and edges.
To prove that any polygon of $L_{T}$ is convex we need to investigate every
vertex and find that any incident angle is less than $\pi$.
For the vertices on the spiral it was made by construction, also for the
vertices on the spiral tails. Other points ${\sf e}_{i},{\sf h}_{i},{\sf
f}_{i},{\sf f}^{\prime}_{i},{\sf j}_{i},{\sf m}_{i},{\sf m}^{\prime}_{i},{\sf
n}_{i},{\sf n}^{\prime}_{i}$ can be investigate manually. Also these
conditions were verified by the computer program.
Figure 16: The gadget $T$ (in the center the square part of $T$)
Coordinates of the gadget and the program one can downloaded from the web-page
http://dcs.isa.ru/taras/durer and explore it in detail.
Further we will need metric properties enumerated on fig.r̃eftable. This table
means that ${\sf y}$ is only one vertex adjacent to ${\sf x}$ that $\angle{\sf
y}{\sf x}c_{\sf x}$ is obtuse. For all other vertices corresponding angle is
acute.
Vertex (${\sf x}$) | Rotation center ($c_{\sf x}$) | Possible exit ${\sf y}$
---|---|---
${\sf c}_{i,j}$ $(j<91,j\neq 70)$ | ${\sf c}_{1}$ | ${\sf c}_{i,j+1}$
${\sf e}_{i}$ | ${\sf c}_{i}$ | ${\sf h}_{i}$
${\sf h}_{i}$ | ${\sf c}_{i}$ | ${\sf h}_{3-i}$
${\sf h}_{i}$ | ${\sf c}_{3-i}$ | ${\sf f}_{i}$
${\sf f}_{i}$ | ${\sf c}_{3-i}$ | ${\sf g}_{i}$
${\sf g}_{i}$ | ${\sf c}_{3-i}$ | ${\sf m}_{i}$
${\sf m}_{i}$ | ${\sf c}_{3-i}$ | ${\sf c}_{1,81}$
${\sf f}^{\prime}_{i}$ | ${\sf c}_{3-i}$ | ${\sf m}_{i}$
${\sf h}_{i}$ | ${\sf o}$ | ${\sf f}_{i}$
${\sf f}_{i}$ | ${\sf o}$ | ${\sf c}_{i,80}({\sf m}^{\prime}_{3-i})$
${\sf c}_{i,80}({\sf m}^{\prime}_{3-i})$ | ${\sf c}_{i}$ | ${\sf c}_{i,10}$
${\sf c}_{1,91}={\sf n}^{\prime}_{1}$ | ${\sf c}_{1},{\sf c}_{2},{\sf o}$ | ${\sf o}_{42}$
${\sf c}_{2,91}={\sf n}^{\prime}_{2}$ | ${\sf c}_{1},{\sf c}_{2},{\sf o}$ | ${\sf o}_{0}$
${\sf o}_{j}$ ($j\leq 99$) | ${\sf c}_{1},{\sf c}_{2},{\sf o}$ | ${\sf o}_{j+1}$
${\sf o}_{100}$ | ${\sf c}_{1},{\sf c}_{2},{\sf o}$ | ${\sf t}_{2}$
Figure 17: Necessary metric conditions of $T$
These conditions were verified by a simple computer program and can be
confirmed manually as well.
4. 4.
Divide $G_{A}$ into $4$ graphs $G_{1},G_{2},G_{1,2},G_{0}$ that do not share
edges. $G_{1}=\\{e\in G|{\sf c}_{1}>e{\rm\;and\;not\;}({\sf c}_{2}>e)\\}$,
$G_{2}=\\{e\in G|{\sf c}_{2}>e{\rm\;and\;not\;}({\sf c}_{1}>e)\\}$,
$G_{1}=\\{e\in G|{\sf c}_{1}>e{\rm\;and\;}{\sf c}_{2}>e\\}$ and $G_{0}$ all
the rest.
We adopt the convention that a graph containing an edge also contains its
vertices.
5. 5.
Consider the first case $G_{1,2}=\emptyset$.
Then the $G_{1},G_{2}$ polylines separately connecting with $\partial T$ or
$G_{B}$.
Using the properties of the gadget $T$ (fig. 17) and the condition $(2)$, we
can define the edges of $G_{1}$ and $G_{2}$ as follows:
The graph $G_{1}$ starts from the point $c_{1}$ and goes by any possible edge
to the point ${\sf c}_{1,i}$ ($i<k)$ of the spiral. Further $G_{1}$ continued
unambiguously along the follwing path: ${\sf c}_{1},{\sf c}_{1,i},{\sf
c}_{1,i+1},\ldots{\sf c}_{1,69},\ldots,{\sf c}_{1,129}={\sf e}_{1},{\sf
h}_{1},{\sf h}_{2},{\sf f}_{2},{\sf f}^{\prime}_{2},{\sf m}_{2},{\sf
c}_{1,140},{\sf c}_{1,141},\ldots,{\sf c}_{1,150},{\sf c}_{2,198}={\sf
n}^{\prime}_{2},{\sf o}_{0},{\sf o}_{1},\ldots,{\sf o}_{87}={\sf p},{\sf
o}_{88},\ldots,{\sf o}_{99},{\sf t}_{2}$.
The graph $G_{1}$ is a part of this path from ${\sf c}_{1}$ to the first point
of $G_{B}$ or $t_{2}$.
We can define the path for $G_{2}$ similarly: ${\sf c}_{2},{\sf c}_{2,i},{\sf
c}_{2,i+1},\ldots{\sf c}_{2,69},\ldots,{\sf c}_{1,129}={\sf e}_{2},{\sf
h}_{2},{\sf h}_{1},{\sf f}_{1},{\sf f}^{\prime}_{1},{\sf m}_{1},{\sf
c}_{2,140},{\sf c}_{2,141},\ldots,{\sf c}_{2,150},{\sf c}_{1,198}={\sf
n}^{\prime}_{1},{\sf o}_{42},{\sf o}_{43},\ldots,{\sf o}_{87}={\sf p},{\sf
o}_{88},\ldots,{\sf o}_{99},{\sf t}_{2}$. Again $G_{2}$ is a part of this path
from ${\sf c}_{2}$ to the first point in $G_{B}$ or ${\sf t}_{2}$.
We see that if $G_{B}=\emptyset$, both $G_{1}$ and $G_{2}$ cannot avoid the
bridge ${\sf h}_{1}{\sf h}_{2}$ and intersects each other. Then if $G_{1,2}$
is empty than $G_{B}$ is not empty.
6. 6.
Consider the case, that $G_{1}$ and $G_{2}$ are in different arc components of
$G$.
As $P_{T}$ has only $3$ vertices, and every component of $G_{B}$ should have
at least $2$ endpoints in $P_{T}$, $G_{B}$ is a connected graph. Then for some
$i\in[1,2]$ the graph $G_{i}$ falls into $\partial T$ and $G_{3-i}$ falls into
$G_{B}$. Because $G_{i}$ falls into ${\sf t}_{2}$, $G_{B}$ is a polyline
connecting vertices ${\sf t}_{1}$ and ${\sf t}_{3}$. But this polyline cannot
reach ‘‘square" part of $T$ without intersection with ${\sf
o}_{61},\ldots,{\sf o}_{85}$, the part of $G_{i}$.
Then if $G_{1,2}=\emptyset$, $G_{1}$ and $G_{2}$ falls separately into
$G_{B}$. Hence, $G$ connects vertices ${\sf c}_{1},{\sf c}_{2}$. We have
proved theorem for $G_{1,2}=\emptyset$.
7. 7.
Consider the case $G_{1,2}\neq\emptyset$. $G_{1}$ and $G_{2}$ go through the
bridge ${\sf h}_{1}{\sf h}_{2}$. $G_{2}$ does not intersect $G_{1}$ above
${\sf h}_{1}$, and $G_{1}$ does not intersect $G_{2}$ above ${\sf h}_{2}$.
Then $G_{1,2}$ can start only from one ${\sf h}_{i}$, $i\in\\{1,2\\}$.
We can unambiguously continue the path $G_{1,2}$ from each of these points
until it falls into $G_{B}$ or $\partial T$. For every $x\in G_{1,2}$ the
rotation center $c_{x}$ is the origin ${\sf o}$, midpoint between ${\sf
c}_{1}$ and ${\sf c}_{2}$.
From the vertex ${\sf h}_{i}$ the graph $G_{1,2}$ continued unambiguously:
${\sf h}_{i},{\sf f}_{i},{\sf m}^{\prime}_{i},{\sf o}_{3-i,69}$. Note that
${\sf o}_{i,69}\in G_{i}$ (${\sf o}_{i,69}$ is a last point which has common
edge with ${\sf c}_{i}$). $G_{1,2}$ cannot fall into $G_{i}$, so it falls into
$G_{B}$ in one of the points ${\sf f}_{1},{\sf f}_{2},{\sf
m}^{\prime}_{1},{\sf m}^{\prime}_{2}$.
As $G_{1,2}\neq\emptyset$, it is clear that $G$ connects two vertices ${\sf
c}_{1},{\sf c}_{2}$, because $G_{1},G_{2},G_{1,2}$ have a common point – the
outfall of $G_{1},G_{2}$ and origin of $G_{1,2}$.
Thus, if $G_{1,2}\neq\emptyset$, then $G_{B}\neq\emptyset$ and $G$ connects
${\sf c}_{1},{\sf c}_{2}$.
This completes the proof of the Theorem 2.
## 5 Proof of the main theorem
|
---|---
Figure 18: Cone-shaping
Definition 19. Let $Q$ be a polyhedral surface. Let $h_{1},h_{2}$ be two
vertices satisfying $curv(h_{1})+curv(h_{2})<2\pi$.
Take a geodesic $h_{1}h_{2}$. Cut the surface $Q$ by this geodesic. Take the
double triangle $a^{\prime}_{1}a^{\prime}_{2}a^{\prime}$ with
$a^{\prime}_{1}a^{\prime}_{2}=h_{1}h_{2};\angle
a^{\prime}a^{\prime}_{1}a^{\prime}_{2}=curv(h_{1})/2;\angle
a^{\prime}a^{\prime}_{2}a^{\prime}_{1}=curv(h_{2})/2$. Cut this triangle by
$u_{1}u_{2}$. Glue the triangle and the surface by the lines of the cut, so
that the new glued surface $Q^{\prime}$ is homeomorphic to sphere (disk).
The surface $Q^{\prime}$ is intrinsically convex and satisfies the conditions
of Alexandrov’s existence theorem.
Hence $Q^{\prime}$ is the surface of a convex polyhedron (or part of the
surface). This polyhedral surface contains one vertex $v$ with curvature
$curv(v)=curv(p_{1})+curv(p_{2})$ instead of the two vertices $p_{1},p_{2}$.
This procedure is called cone-shaping.
###### Lemma 2
Let $Q$ be the surface of a polyhedron, let $L_{Q}$ be a convex geodesic
partition. For a given vertex $v$ one can modify $Q$ by changing the
neighborhood of $v$ into a scaled copy of the gadget cap $T_{v}$ with
curvature $curv(v)$. The obtained polyhedral surface $Q^{\prime}$ has a
corresponding convex geodesic partition $L^{\prime}$.
Proof.
1. 1.
Denote by $e_{i}$ the edges incident to $v$ (fig. 20 a). Let $\gamma$ be the
biggest angle of the partition $L$ around the vertex $v$. Let $r$ be the
length of the smallest edge $e_{i}$. Put a point $z_{i}$ on each edge $e_{i}$
on distance $r/3$ from $v$.
For adjacent $e_{i}e_{j}$ add an edge between $z_{i}z_{j}$. Denote by $Z$ the
polyhedral surface bounded by the constructed cycle $z_{i}$ and containing
$v$.
Let us scale $T_{v}$ to make diameter $d(T_{v})<r/3\cos(\gamma/2)$.
|
---|---
Figure 19: Schematic $T_{v}$ with real edges and cone-shaped $T^{\prime}$
Let $T^{\prime}$ be a cone-shaped surface of $T_{v}$ (fig. 19). Then
$d(T^{\prime})<2d(T_{v})$ for small enough $curv(v)$. The polyhedral surface
$T^{\prime}$ has one vertex with curvature $curv(v)$. Then $T^{\prime}_{v}$ is
isometric to a fragment of the surface of $Q$ around the vertex $v$.
2. 2.
Cut from $Q$ the piece isometric to $T^{\prime}$ and replace it into $T_{v}$.
This gluing is made isometrically, because $T^{\prime}$ and $T_{v}$ has an
isometric neighborhood of the border.
$Q^{\prime}$ is a new surface. Surface $Q^{\prime}$ is an intrinsically
convex. Then it is a surface of some convex polyhedron.
|
---|---
Figure 20: Replacing neighborhood of $h_{i}$ into $T_{i}$
3. 3.
Let us define a partition $L_{Q^{\prime}}$ for $Q^{\prime}$ (fig. 20 b).
Inside $T_{i}$ take the partition $L_{v}$.
Outside $Z$ the partition $L_{Q^{\prime}}$ equals to $L_{Q}$. On $T_{i}$ the
partition $L_{Q^{\prime}}$ equals to $L_{i}$. The region outside $T_{i}$ and
inside $Z$ should be somehow triangulated.
It is clear that $L_{Q^{\prime}}$ is a geodesic convex partition.
Additional points $z_{i}$ only needed to make sure that all polygons of the
partition are convex.
###### Theorem 3
There exists a surface of a convex polyhedral surface $P$ and its geodesic
convex partition $L_{P}$ such that there are no connected non-overlapping
$L$-unfoldings of $P$.
Proof.
1. 1.
The technique of construction a counterexample is similar to the technique in
another counterexamples without non-overlapping unfolding (GlazTar08 ,Tar99 ):
we take a polyhedron $Q$ and replace its vertex with a special gadget. Any
spanning tree of the modified polyhedron $Q^{\prime}$ has a corresponding
spanning tree in the original polyhedron $Q$ (every time we have to show that
such correspondence exists).
Every spanning tree in $Q$ has at least one end. The unfoding of $Q^{\prime}$
has overlappings in a gadget corresponding to this end, because of the
properties of the gadget.
2. 2.
By Theorem 1, for the gadget $T$, there exists an $\varepsilon$ such that the
cap $T_{\beta}$ is not admissible for $\beta<\varepsilon$. Let $Q$ be a convex
polyhedral surface with the minimal distance between vertices at least $2d(T)$
and the curvature of any vertex less than $\varepsilon$.
For example, the prism based on the regular $N$-gon, where
$N>\frac{2\pi}{\varepsilon}$, satisfies this condition.
Using Lemma 2, change every vertex $h_{i}$ of $Q$ into a corresponding gadget
$T_{i}$ with the curvature $curv(h_{i})$.
We obtain a surface $P$ and its partition $L_{P}$. Denote by $L_{i}$ the part
of $L_{P}$ corresponding to the gadget $T_{i}$.
Every vertex $w$ of $P$ is a vertex $c_{1}$ or $c_{2}$ of $L_{i}$ by
construction.
3. 3.
Let $G$ be a spanning tree of the polyhedral surface $P$ such that $G\subset
L_{P}$ and the corresponding unfolding $U_{G}$ does not have any
intersections. Denote by $G^{\prime}$ the graph consisting of edges $e\in G$
which divide $G$ into two parts $G_{1},G_{2}$, where each part contains
vertices of non-zero curvature. It is easy to see that we can unfold $S$ by
cutting only $G^{\prime}$. Not cutted vertices have a zero curvature. The
obtained unfolding $U_{G^{\prime}}$ is equal to $U_{G}$.
Consider a fragment $L_{i}$, denote by $F_{i}=G^{\prime}\cap L_{i}$.!! As
$U_{G^{\prime}}$ is a non-overlapping unfolding, $F_{i}$ is a cut.
By the second condition of Theorem 2, $F_{i}$ has a non-empty subgraph
$F_{i,B}$ (which consists of $B$-edges of $F_{i}$).
Therefore $F_{i}$ is connected to $G^{\prime}\setminus F_{i}$ by at least two
points. By the third condition of Theorem 2, the graph $F_{i}$ is connected.
Since $F_{i,B}\subset G^{\prime}$, each endpoint of $F_{i,B}$ should be the
connection between $F_{i}$ and $G^{\prime}\setminus F_{i}$. Then $F_{i}$ and
$G^{\prime}\setminus F_{i}$ have at least $2$ connections.
Replace every fragment $F_{i}$ in $G^{\prime}$ into a single vertex. We obtain
a graph $G^{\prime\prime}$. This graph $G^{\prime\prime}$ has a cycle, because
the degree of any its vertex is at least $2$.
Since any $F_{i}$ is connected, $G^{\prime}$ also contains a cycle.
4. 4.
Note that there exists a polyhedral surface $P^{\prime}$ without zero-
curvature vertices, combinatorially equivalent and arbitrary near to $P$.
Let show this. Let any vertex $h_{i}$ with zero-curvature be lying inside some
face $f$. If some $h_{i}$ lies on the edge of $P$, we can slighlty change the
position of $h_{i}$ without changing the combinatorial structure and
properties of $P$.
For each vertex $h_{i}$ of the partition $L$ which is not a real vertex of $P$
take $h^{\prime}_{i}=h_{i}+\varepsilon(d^{2}-r_{i}^{2})n_{f}$, where $d$ is
the diameter of $P$ (as a body), $r_{i}$ is a distance from $h_{i}$ to the
center of the real facet $f$ of $P$ where this point is lying, $n_{f}$ is a
normal vector of the facet $f$. Then
$P^{\prime}=conv(P,conv_{i}(h^{\prime}_{i}))$. If $\varepsilon$ is small
enough all edges and vertices of $P$ exist in $P^{\prime}$.
Then $P^{\prime}=P\cup P_{f_{1}}\cup P_{f_{2}}\ldots P_{f_{n}}$, where
$P_{f_{j}}$ is a $conv(f_{j},conh_{h_{i}\in f_{j}}(v^{\prime}_{i}))$ for each
face $f_{j}$. Since all vertices $h^{\prime}_{i}$ of some $P(f_{j})$ are lying
on a paraboloid, any vertex $h^{\prime}_{i}$ cannot lie in the interior of
$P_{f_{j}}$ and have a non-zero curvature. Then every vertex of $P^{\prime}$
has a non-zero curvature.
Also, for a small enough $\varepsilon$ the surface $P^{\prime}$ has the
partition $L_{P^{\prime}}$ corresponding to $L_{P}$ without any combinatorial
changes.
The number of subtrees of $L_{P}$ is finite, so for a small enough
$\varepsilon$ any $L_{P^{\prime}}$-unfolding of $P^{\prime}$ overlaps as the
corresponding unfolding in $P^{\prime}$.
Thus, $P^{\prime}$ is also a counterexample.
## 6 A simpler counterexample
Let $P_{needle}$ be a polyhedron which holds the following conditions:
1\. The edge-skeleton of $P_{needle}$ is isomorphic to the Tutte graph (fig.
21) Tutte46 (this graph does not have a Hamiltonian cycle).
2\. $P_{needle}$ has $2$ adjacent vertices $h_{1},h_{2}$ with the sum of
curvature close to $4\pi$.
Such polyhedron exists. It can be constructed in the following way.
Using the Steinitz-Rademacher theorem SteinRad34 , take the polyhedron $P$
with its skeleton dual to the Tutte graph $G_{T}$
Take some edge $e$. Let the polyhedron $P^{*}$ be polar to $P$ with the center
$O$ of the polar transformation close to the center of the edge $e$.
The obtained $P^{*}$ has two vertices $h_{1},h_{2}$ incident to the edge
$e^{*}$. If $O$ is close enough to the edge $e$, the projection of every
vertex $h_{i},i>2$ into line $h_{1}h_{2}$ lies inside the edge $e^{*}$.
Squeeze affine $P^{*}$ in two directions perpendicular to $e^{*}$ to make
$P^{*}$ lie in the neighborhood of $e^{*}$.
Then the combinatorial structure of $E(P^{*})$ is equal to $G_{T}$ and $P^{*}$
has two adjacent vertices $h_{1},h_{2}$ with the sum of the curvatures near to
$4\pi$. The total curvature of all other vertices is arbitrary small.
Using Lemma 2, replace any vertex of $P^{*}$ except $h_{1},h_{2}$ into the
gadget. We obtain $P_{needle}$.
Repeat the argument from Theorem 2: let $G$ be any spanning tree of
$P_{needle}$ and $G^{\prime}$ its non-zero curvature part.
Any part of $G^{\prime}$ inside each $Z_{i}$ is connected. Indeed, any
component $F$ of $G^{\prime}\cap Z_{i}$ can have a vertex with degree $1$ only
in $z_{i}$ and $c_{1,i},c_{2,i}$. If this component $F$ contains non-zero
curvature vertices, then by third condition of theorem 2 it has two additional
ends.
Any component of $F$ of $G^{\prime}\cap Z_{i}$ should contain at least two
vertices from $z_{i,1},z_{i,2},z_{i,3}$. Hence, $G^{\prime}\cap Z_{i}$ is
connected graph.
Figure 21: Tutte graph $G_{T}$
For each $i$ replace $G^{\prime}\cap Z_{i}$ into single vertex. The obtained
graph $G^{\prime\prime}$ has degree $1$ only in vertices $h_{1},h_{2}$. As
$G^{\prime}$ does no have cycles, $G^{\prime\prime}$ is a Hamiltonian path
from $h_{1}$ to $h_{2}$. Adding to $G^{\prime}$ theedge $h_{1}h_{2}$ we obtain
a Hamiltonian cycle in $G_{T}$. This contradiction proves the statement.
$P_{needle}$ contains 43 gadgets and $2+43\cdot 442=19008$ vertices.
## References
* (1) Dürer A., The Painter’s Manual: A Manual of Measurement of Lines, Areas, and Solids my Mean of Compass and Rules assembled by Albrecht Dürer for the Use of All Lovers of Art with Appropriate Illustrations Arranged and Printed in the Year MDXXV. Abaris Books, Inc.
* (2) Sheppard, G.C., Convex polytops with convex nets, Math. Proc. Camb. Phil. Soc 78, 389-403, 1975.
* (3) Alexandov A. D., Convex polyhedra, Moscow, (in Russian) (1949).
* (4) Bobenko A.I., Izmestiev I., Alexandrov’s theorem, weighted Delaunay triangulations, and mixed volumes, To appear: Annales de l’Institut Fourier. Preprint: http://www.arxiv.org/math.DG/0609447.
* (5) Erickson J., Oberwolfach-Conference ‘‘Discrete Differential Geometry’’, problem 8, Berlin, March 2006.
* (6) Bobenko A.I., Oberwolfach-Conference ‘‘Discrete Differential Geometry’’, problem 7, Berlin, March 2006.
* (7) O’Rourke J., Demaine E., Geometric Folding Algorithms: Linkages, Origami, Polyhedra, Cambridge University Press,331-332, 2007.
* (8) Weisstein Eric W., ‘‘Schönflies Theorem.", From MathWorld–A Wolfram Web Resource. http://mathworld.wolfram.com/SchoenfliesTheorem.html
* (9) Tutte W. T., On Hamiltonian Circuits, J. London Math. Soc. 21, 98-101, 1946.
* (10) Steinitz E. and Rademacher H., Vorlesungen über 371 die Theorie der Polyeder, Springer, Berlin, 1934.
* (11) Glazyrin A., Tarasov. A.S., Anti-Dürer conjecture for unconvex polyhedra, To appear: Uspekhi Mat. Nauk (in Russian).
* (12) Tarasov A.S., Polyhedra that do not admit natural unfoldings. (In Russian) Uspekhi Mat. Nauk 54(1999), no. 3(327), pp. 185 – 186.
| arxiv-papers | 2008-06-14T06:57:50 | 2024-09-04T02:48:56.261277 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Alexey S Tarasov",
"submitter": "Alexey Tarasov S",
"url": "https://arxiv.org/abs/0806.2360"
} |
0806.2437 | # Phase diagram of Holstein-Kondo lattice model at half-filling
Reza Nourafkan Department of Physics, Sharif University of Technology,
P.O.Box: 11155-9161, Tehran, Iran Nasser Nafari Institute for Studies in
Theoretical Physics and Mathematics, P.O.Box: 19395-5531, Tehran, Iran
###### Abstract
We study the Kondo lattice model which is modified by the Holstein term,
involving both the Kondo exchange coupling and the electron-phonon coupling
constants, characterized by $J$ and $g$, respectively. The model is solved by
employing the dynamical mean-field theory in conjunction with exact
diagonalization technique. A zero temperature phase diagram of symmetry
unbroken states at half filling is mapped out which exhibits an interplay
between the two interactions and accounts for both spin and charge
fluctuations. When the Kondo exchange coupling is dominant the system is in
Kondo insulator state. Increasing $g$ for small values of $J$ leads to a Kondo
insulator-metal transition. Upon further enhancement of $g$ a transition to
the bipolaronic insulating phase takes place. Also a small region with non-
Fermi liquid behavior is found near the Kondo insulator-metal transition.
There has been a continued interest in a class of compounds called heavy
fermion semiconductors, which exhibit a spin and a charge gap at low
temperatures typically ranging between $1$ and $100$ meV Riseborough ; Misra .
In contrast to the ordinary band insulators, these two gaps are different,
indicating a separation of the spin and charge degrees of freedom brought
about by correlation effects. The gap formation in heavy fermion
semiconductors is attributed to the renormalized hybridization between a broad
band of conduction electrons and a nearly flat band of strongly correlated
$f$-electrons.
The Kondo lattice model (KLM) at half-filing is considered to be a good
starting point for investigating the properties of the heavy fermion
semiconductors. In this model, at each lattice site a local moment interacts
with the spin of a conduction electron, and thus, results in complex
correlation effects between them. In fact, a conduction and a localized
electron with antiparallel spins undergo a spin-flip process, causing
itinerant electrons to leave a trace of their spin exchange at each localized
spin site. As a result, the direction of a localized spin is affected by the
history of the electrons passing through it. There are similar correlation
effects in the periodic Anderson model due to the dynamical aspects of the
localized electrons.
Experiments involving the Kondo insulators at high magnetic fields indicate
the closure of the Kondo insulating gap, exemplifying a transition from the
Kondo insulator to a correlated metal Jaime ; Cooley . It is expected that the
electron-phonon (e-ph) interaction leads to similar results. Many experiments
suggest that the e-ph effects are important in describing a number of
observations such as the existence of an unusual phonon softening in the Kondo
lattice, CeCu2, which is indicative of coupling between electrons and phonons
Loewenhaupt . Furthermore, the lattice plays an important role in some heavy
fermion compounds, called 14-1-11, where various properties can be altered
through isoelectronic substitutions Burch . In fact, it is believed that the
coupling between phonon modes and the Kondo effect could manifest new material
properties, such as non-Fermi liquid behavior and unconventional
superconductivity Yotsuhashi ; Hotta ; Nayak .
Even less studies has been devoted to the role played by lattice vibrations in
these compounds. The role of the lattice vibrations is not trivial, but if, on
general grounds, the minimal effect of e-ph coupling is a phonon-retarded
attraction between conduction electrons with opposite spins, then the spin
excitation has a gap while the charge excitation, depending on the strength of
the e-ph coupling, can be either gap-full or gapless. Therefore, there arises
a competition between the spin- and the charge-fluctuations whose behavior is
determined, on the one hand, by the relative strength of the Kondo exchange
between the conduction electrons and the localized moments and, on the other
hand, by the conduction electron-phonon coupling leading to a complicated
phase diagram.
It is the goal of this paper to investigate the dynamical competition between
the e-ph and Kondo interactions. A natural way of incorporating the e-ph
coupling in the KLM is to add the Holstein coupling term to its hamiltonian.
In the Holstein coupling the phonon variables are coupled to the local density
of the conduction electrons. In this paper, we will present the zero
temperature phase diagram of the Holstein-Kondo lattice model (H-KLM) at half-
filling. The focus is on the transition between the unbroken symmetry ground
state as the e-ph and Kondo interactions parameters, $J$ and $g$, are varied.
The H-KLM Hamiltonian is defined by:
$\displaystyle H=$ $\displaystyle-$ $\displaystyle t\sum_{\langle
i,j\rangle\sigma}{\left(c^{\dagger}_{i\sigma}c_{j\sigma}+c.c.\right)}+\frac{J}{2}\sum_{i,\alpha\beta}{\bm{S}_{i}.\left(c^{\dagger}_{i\sigma}\bm{\sigma}_{\alpha\beta}c_{i\sigma}\right)}$
(1) $\displaystyle+$ $\displaystyle
g\sum_{i}{\left(n_{i}-1\right)\left(b^{\dagger}_{i}+b_{i}\right)}+\Omega_{0}\sum_{i}b^{\dagger}_{i}b_{i},$
(2)
where $c_{i\sigma}\left(c^{\dagger}_{i\sigma}\right)$ and
$b_{i}\left(b^{\dagger}_{i}\right)$ are, respectively, destruction (creation)
operators for itinerant electrons with spin $\sigma$ and local vibrons of
frequency $\Omega_{0}$ on site $i$, $n_{i}$ is the electron density on site
$i$, $\bm{S}_{i}$ is the spin operator for the localized spin on site $i$,
$\bm{\sigma}$ is a pseudo-vector represented by Pauli spin matrices, $t$
stands for the itinerant electrons hopping matrix elements between the
nearest-neighbor sites, $J$ is the coupling strength between itinerant
electrons and localized spins, and $g$ denote the electron-phonon coupling. We
do not consider the coulomb repulsion term between itinerant electrons,
because it tends to suppress the double occupation of sites and in our model
the exchange coupling, $J$, already does the same thing.
Our calculations are based on the dynamical mean field theory Georges , a
powerful, non-perturbative tool to study the properties of strongly correlated
systems, which allows us to treat, on equal footing, the two kinds of
interactions present in our model. This technique, which becomes exact in the
limit of infinite coordination number, reduces the full lattice many-body
problem to a local impurity embedded in a self-consistent effective bath of
free electrons, mimicing the effect of the full lattice on the local site. A
self consistency condition links the effective impurity model to the original
lattice problem. Adopting a semi-circular density of states (DOS)
$\rho_{0}(\epsilon)=(2/\pi D)\sqrt{D^{2}-\epsilon^{2}}$ of the noninteracting
system, corresponding to a Bethe lattice with the half bandwidth $D$, the
self-consistency relation imposed on the DMFT solution is given by
$\frac{D^{2}}{4}G(i\omega_{n})=\sum_{k}{\frac{V_{k}^{2}}{i\omega_{n}-\epsilon_{k}}},$
(3)
where $\epsilon_{k}$ and $V_{k}$ are the energies and the hybridization
parameters of the effective impurity model (bath parameters). We use exact
diagonalization (ED) technique to solve the effective impurity model Caffarel
. This solver allows us to access the ground state properties of the system
with a finite energy resolution. The ED technique consists of restricting the
sum in Eq. (3) to a small number of levels, and moreover, it truncates the
infinite phonon Hilbert space. The ground state and the Green’s function of
our discretized model are determined via the Lanczos procedure and the self-
consistency equation in turn allows us to derive a new set of bath parameters.
The process is iterated until convergence is reached. In the theory of Mott
transition, the investigation of the paramagnetic (PM) phase has been very
fruitful providing a lucid understanding of the finite temperature state,
above the magnetic order in many compounds. We pursue a similar approach in
our investigation and study the PM state. We force the system to be in a
paramagnetic state by averaging the spin up and spin down to study the
underlying normal state. In all our calculations presented here the
convergence of truncation has been checked.
Fig. 1 shows the $T=0$ phase diagram of the half-filled H-KLM in the parameter
space of $J$ and $g$ with $D=2t=2$ and $\Omega_{0}/t=0.2$. All types of long-
range order are excluded. Three different phases are distinguished: metallic
phase and the bipolaronic and Kondo insulating phases. In what follows, a
detailed discussion of the phase diagram of these systems will be presented.
The Kondo lattice model ($g=0$) and Holstein model ($J=0$), which are special
limiting cases of the H-KLM, have been extensively studied using the DMFT. The
ground state of KLM is the Kondo insulating phase with a spin and a charge gap
for all $J$ values Costi . For the Holstein model, the ground state is
metallic. The metallic phase is found to be a Fermi liquid, in the sense that,
the Luttinger sum rule $\rho(0)=\rho_{0}(0)$ for the spectral function
$\rho(\omega)=-ImG(\omega+i0^{+})/\pi$, or equally stated, the limit of
$ImG(i\omega_{n})\rightarrow-1$ as $\omega_{n}\rightarrow 0$, is satisfied
($\omega_{n}$ is the Matsubara frequency). Upon increasing $g$, the conduction
electrons lose their mobility, eventually acquiring polaronic character, in
which the presence of an electron is associated with a finite lattice
distortion. Also, the same e-ph coupling can cause any two polarons to attract
and form a bound pair in real space, called bipolaron Capone1 . In the absence
of pair hopping, the bipolaron formation would cause the system to undergo a
first order metal to bipolaronic insulating phase transition at the critical
coupling $g_{c}$ Koller1 ; Jeon . Meyer et al. have reported that there is a
coexistence region near $g_{c}$, which is reduced as the phonon frequency
$\Omega_{0}$ is decreased and disappears for $\Omega_{0}.leq.0.10D$ Meyer .
The bipolaron formation may be accommodated by reconstructuring the system
into a phase separated state Capone2 or a charge ordered state in which the
doubly occupied and empty sites alternate in real space Pietig .
At small fixed $J$-values, with increasing e-ph coupling, a continuous
transition to a metallic state occurs at a critical coupling $g_{1c}(J)$,
whose value increases with increasing $J$. This behavior is physically
expected. An increase in $J$ leads to a larger insulating gap, and this in
turn, leads to the suppression of the charge fluctuations which would
otherwise couple to phonons. As a result a transition to metallic state occurs
at larger e-ph coupling. We Also find that the metallic phase near $g_{1c}$
shows non-Fermi liquid character. Further increase of $g$ causes a metal-
bipolaronic phase transition taking place at a critical coupling $g_{2c}$. As
it can be distinguished, a Holstein coupling is weakly affected by exchange
coupling between conduction electrons with local spins. The metallic state
becomes more correlated as $g$ or $J$ is increased. This is reflected in the
decreasing behavior of the quasiparticle weight
$z=1/[1-Im\Sigma(i\omega_{0})/\omega_{0}]$ when $g$ or $J$ is increased (Fig.
2).
Figure 1: Zero temperature phase diagram of the unbroken symmetry Holstein-
Kondo lattice model at half-filling. The model shows three different phases:
metallic, bipolaronic and Kondo insulating phase. A narrow region with non-
Fermi liquid character is seen near the Kondo insulator-metal transition.
Figure 2: Behavior of quazi-particle weight for different values of $g$.
Fig. 3 shows the imaginary part of electron self-energy,
$Im\Sigma(i\omega_{n})$, for $J=0.1$ and several values of $g$ in the vicinity
of both phase transitions. For small $g$ values, the imaginary part of the
self-energies diverge as $\omega_{n}\rightarrow 0$, indicating the presence of
a charge gap (See panel a). Increasing $g$ causes the system to change its
phase from an insulator to a bad metal in the sense that its self-energy
extrapolates to a finite value $Im\Sigma(i0^{+})\equiv\Gamma(J)\neq 0$ for
$g\geq g_{1c}$. Hence, a finite lifetime is found at the Fermi level for a
narrow range of e-ph couplings near the $g_{1c}$, indicating that well defined
quasiparticles do not exist in this range. The violation of the Luttinger sum
rule in this region is also seen from $ImG(i\omega_{n})$, which tends to a
negative constant $c<0$ in the limit of $\omega_{n}\rightarrow 0$, with
$c<\pi\rho_{0}(0)=1$. Although the discretness of the spectra obtained in the
exact diagonalization techniqe does not allow us to unambiguously identify the
non-Fermi liquid region, we believe that the spectral function at $g=0$
displays a narrow insulating gap, whose width is proportional to the value of
$J$, with two peaks on each side. For a fixed $J$, increasing $g$ causes the
low-energy spectrum widen and are also suppressed. If these peaks overlap
before being damped completely, a narrow pseudogap forms near the Fermi level,
$E_{F}$. With further enhancement of $g$, there is a rapid shallowing of the
pseudogap till finally a quasiparticle peak forms at $E_{F}$. At this stage,
the system will have a Fermi-liquid character. Upon increasing $g$ further,
there is a weakly narrowing of quasiparticle peak until it disappears at the
second critical value of e-ph coupling $g_{2c}$ where a gap opens. A more
detailed results on the spectra might be obtained by the numerical
renormalization group technique. The inset of panel (b) shows double occupancy
$d=<n_{\uparrow}n_{\downarrow}>$ as a function of $g$. There is no signature
of the Kondo insulator-metal transition in the double occupancy, but at
$g_{2c}$ the double occupancy jumps suddenly to $d\approx 1/2$, indicating a
discontinuous transition to bipolaronic phase.
Figure 3: Imaginary part of electron self-energy, $Im\Sigma(i\omega_{n})$,
obtained at different values of $g$ in the vicinity of both phase transitions,
with fixed $J=0.1$. Panel (a): $Im\Sigma(i\omega_{n})$ in the vicinity of the
transition from the Kondo insulator state to the metallic state. Changing the
$Im\Sigma(i\omega_{n})$ behavior as $\omega_{n}\rightarrow 0$ from diverging
to extrapolating to zero shows the insulator-matallic phase transition. Panel
(b): $Im\Sigma(i\omega_{n})$ in the vicinity of the transition from the
metallic state to bipolaronic state. Inset: the double occupancy
$d=<n_{\uparrow}n_{\downarrow}>$ as a function of $g$. The transition from the
metallic state to bipolaronic state is clearly visible by observing when the
double occupancy’s jump to $\approx 1/2$ begins to set in.
Fig. 4(a) shows the phonon spectral function,
$\rho_{ph}(\omega)=-Imd(\omega+i0^{+})/\pi$, for $J=0.1$ as a function of e-ph
interaction strengths. The phonon Green’s function is defined by
$d(\omega)=\ll b_{i};b^{\dagger}_{i}\gg_{\omega}$. The figure illustrates how
the phonon mode is softened with increasing $g$. The softening phonon mode is
a manifestation of a lattice instability as in structural phase transitions. A
stability is restored by the condensation of the unstable mode. It results in
a nonzero expectation value of the phonon operator ($<b>\neq 0$) or in large
average number of excited phonons in the ground state. The appearance of
negative spectral function for $\omega<0$, when the bipolaronic state is
approached implies that there is a large increase in the lattice displacement.
In the bipolaronic state, the phonon mode hardens back to the bare mode as $g$
assumes values greater than $g_{c2}$. This is due to the fact that screening
is not effective in an insulating state. This is the same behavior which had
already been seen for pure Holstein model Koller2 . Panel (b) of Fig. 4 shows
the phonon spectral function for $g=0.2$ and various values of $J$. The phonon
mode gradually hardens back to $\Omega_{0}$, as the $J$-values increase. We
observe no signature of a transition to Kondo insulator in the phonon
spectrum. The effect of increasing $J$ is to suppress contiuously the charge
fluctuations which results in a decoupling of electrons and phonons causing
the phonon peak to exhibit hardening. In contrast to the Holstein-Hubbard
model results, where softening is absent in Mott insulator phase and phonons
are effectively decoupled from electrons Koller2 , here the hardening of the
phonon peak takes place very slowly.
Figure 4: Phonon spectral function for different values of $g$. The bare
phonon frequency is $\Omega_{0}=0.2$ and a Lorentzian broadening with the full
width at half maximum of $0.02$ has been implemented. Panel (a): Spectral
function for $J=0.1$ and various values of $g$. A cosiderable phonon softening
is seen upon approaching the transition to the bipolaronic insulator. Panel
(b): Spectral function for $g=0.2$ and various values of $J$. The transition
to Kondo insulator does not obviously affect the phonon spectral function.
In concluson, we have studied the Holstein-Kondo lattice model at half-
filling. We find that the model presents the physics of the Kondo insulator
when the exchange coupling, $J$, plays a dominant role and a transition to
correlated metal takes place for small $J$ and intermediate e-ph coupling,
$g$. Moreover, a bipolaronic-metal insulator takes place for small $J$ and
large $g$. We also find a small region with non-Fermi liquid character near
the Kondo insulator-metal transition. The remaining interesting questions will
be how the phase diagram and nature of transitions will change as $\Omega_{0}$
or electron density is changed. It is also interesting to study the symmetry
breaking states such as the antiferromagnetic and superconducting states.
Works in this direction are in progress and will be reported in a separate
publication.
###### Acknowledgements.
## References
* (1) P. S. Riseborough, Adv. Phys, 49, 257 (2000).
* (2) P. Mira, Heavy-Fermion Systems (Elusive, 2008).
* (3) M. Jaime et al., Nature (London) 405, 160 (2000).
* (4) J. C. Cooley et al., J. Superfund. 12, 171 (1999).
* (5) M. Loewenhaupta, U. Wittea, S. Krampa, M. Bradenc, and P. Svoboda, Physica B 312–313, 181 (2002).
* (6) K. S. Burch et al., Phys. Rev. Lett. 100, 026409 (2008).
* (7) S. Yotsuhashi et al., J. Phys. Soc. Jpn. 74, 49 (2005).
* (8) T. Hotta, Phys. Rev. Lett. 96, 197201 (2006).
* (9) P. Nayak et al., Int. J. Mod. Phys. B 16, 3595 (2002).
* (10) A. Georges, G. Kotlier, W. Kraut, and M. J. Rosenberg, Rev. Mod. Phys. 68, 13 (1996).
* (11) M. Caffarel and W. Kraut, Phys. Rev. Lett. 72, 1545 (1994).
* (12) T. A. Costi and N. Manini, J. of Low Temp. Physics, 126, 835-866 (2002).
* (13) P. Werner and A. J. Millis, Phys. Rev. B 74, 155107 (2006).
* (14) M. Capone, and S. Chichi, Phys. Rev. Lett. 91, 186405 (2003).
* (15) W. Koller, D. Meyer, Y. Ono, and A. C. Hewson, Europhys. Lett. 66, 559 (2004).
* (16) G. S. Jeon, T. Park, J. H. Han, H. C. Lee, and H. Y. Choi, Phys. Rev. B 70, 125114 (2004).
* (17) D. Meyer, A. C. Howsen, and R. Bull, Phys. Rev. Lett. 89, 196401 (2002).
* (18) M. Capone, G. Sangiovanni, C. Castellani, C. Di Castro, and M. Grilli, Phys. Rev. Lett. 92, 106401 (2004).
* (19) R. Poetic, R. Bull, and S. Bland, Phys. Rev. Lett. 82, 4046 (1999).
* (20) W. Koller, D. Meyer, and A. C. Hewson, Phys. Rev. B 70, 155103 (2004).
| arxiv-papers | 2008-06-15T11:52:29 | 2024-09-04T02:48:56.268242 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Reza Nourafkan, Nasser Nafari",
"submitter": "Nasser Nafari",
"url": "https://arxiv.org/abs/0806.2437"
} |
0806.2476 | # Finite-Temperature Scaling of Magnetic Susceptibility and Geometric Phase
in the XY Spin Chain
H. T. Quan Theoretical Division, MS B213, Los Alamos National Laboratory, Los
Alamos, NM, 87545, U.S.A.
###### Abstract
We study the magnetic susceptibility of 1D quantum XY model, and show that
when the temperature approaches zero, the magnetic susceptibility exhibits the
finite-temperature scaling behavior. This scaling behavior of the magnetic
susceptibility in 1D quantum XY model, due to the quantum-classical mapping,
can be easily experimentally tested. Furthermore, the universality in the
critical properties of the magnetic susceptibility in quantum XY model is
verified. Our study also reveals the close relation between the magnetic
susceptibility and the geometric phase in some spin systems, where the quantum
phase transitions are driven by an external magnetic field.
###### pacs:
75.10.Jm, 64.70.Tg, 75.40.Cx, 03.65.Vf
_Introduction:_ Quantum Phase Transitions (QPTs), which occur at absolute zero
and are driven by zero-point quantum fluctuations, are one of the most
fascinating aspects of many-body systems. QPTs and related quantum critical
phenomena have been a topic of tremendous interest in condensed matter physics
and have been extensively studied in the past decade sachdev . In recent work,
quantum criticality has been characterized by using the methods and notions
borrowed from quantum information science, such as the concurrence concurrence
, the entanglement entropy entanglemententropy , geometric phase (GP) pachos ,
Loschmidt Echo LE , and quantum fidelity fidelity in the place of traditional
criteria, such as specific heat or magnetic susceptibility (MS). Most of these
studies focus on the zero-temperature properties of the critical systems. In
recent years, the finite-temperature properties of QPTs nature05 ; nonzero ,
such as, thermal entanglement thermal have begun to attract more attention.
This is because, firstly, all experiments are confined to finite temperature.
Thus, to experimentally verify the theoretical results, knowing only the zero-
temperature properties of the quantum system is not sufficient. Secondly,
though genuine QPTs occur only at absolute zero, quantum criticality has
profound influence on system properties up to a surprisingly high temperature
nature05 . Interesting phenomena at finite temperature related to QPTs have
been experimentally observed in various systems, such as the heavy fermion
system and the BEC review .
On the other hand, it has been shown that a QPT in $d$ space dimensions is
related to a classical transition in $d+z$ space dimensions sachdev ; map ,
where $z$ is the dynamical critical exponent. Under this quantum-classical
mapping, the temperature $T$ of quantum system maps onto an imaginary time
direction: $\tau=-it/\hbar\in(0,1/k_{B}T)$, where $\tau$ and $t$ are imaginary
and real time map . Accordingly accessing the QPT by reducing the temperature
amounts to increasing the size of imaginary time dimension towards infinity,
and leads to a divergence of the spatial correlation length $\xi$. This one-
to-one mapping motivates us to study the finite-temperature properties of QPTs
through its higher dimensional classical counterpart. Studies of these QPTs
and the quantum-classical mapping rely heavily on the exactly solvable models.
One of the most common examples is the one-dimensional quantum Transverse
Ising Model (1D TIM) free energy , which exhibits a second-order QPT at the
critical point $\lambda_{c}=1$, and its classical counterpart - the two-
dimensional classical Ising model suzuki , which exhibits a second-order
thermal phase transitions at the Curie point.
Historically, scaling has played a central role in the study of classical
criticality. It is well known that the 2D classical Ising model obeys finite-
size scaling behavior fishier . A straightforward idea is to study $T\neq 0$
scaling laws of 1D TIM. In Refs. thermal and heavyfermion , the authors use
Gruneisen Parameter and concurrence to characterize finite-temperature
properties induced by QPT at zero temperature. In this paper, instead we will
use a classical macroscopic thermodynamic obsevable - the MS - to study the
finite-temperature properties of the generalized 1D TIM - the quantum XY
chain. The MS has the advantage of being easily experimentally accessible and
has been used as a witness of macroscopic quantum entanglement explain ;
verdral . We will show how the finite-temperature scaling is manifested when
the temperature approaches zero, in analogy with finite-size scaling in the
imaginary time direction of the 2D classical Ising model. We will also verify
the universality in the properties of the MS in quantum XY chain. Finally we
will elucidate the close relation between the MS and another well studied
observable - the GP pachos ; zhu ; yi ; hamma .
_Magnetic susceptibility of quantum XY chain at finite temperature:_ The
Hamiltonian of quantum XY chain can be written as free energy
$H(\gamma,\lambda)=J\sum_{i=1}^{N}\left[\frac{1+\gamma}{2}\sigma_{i}^{x}\sigma_{i+1}^{x}+\frac{1-\gamma}{2}\sigma_{i}^{y}\sigma_{i+1}^{y}+\lambda\sigma_{i}^{z}\right],$
(1)
where $N$ is the number of spins in the chain; $J$ is the coupling strength
(for simplicity we choose $J=1$ hereafter); $\lambda$ is external magnetic
field, and $\gamma$ describes the anisotropy of the system;
$\sigma_{i}^{\alpha},\alpha=x,y,z$ are the Pauli matrix on the $i$th site of
the chain. After a standard procedure free energy , this Hamiltonian can be
diagonalized as
$H(\gamma,\lambda)=\sum_{k}2\Lambda_{k}(\eta_{k}^{\dagger}\eta_{k}-1/2)$,
where $\eta_{k}$ is the Fermionic annihilation operator of the $k$-th mode
quasi particle; $\Lambda_{k}=\sqrt{(\lambda-\cos k)^{2}+\gamma^{2}\sin^{2}k}$
are one half of the excitation energy for modes
$k=2\pi(i-0.5)/N,i=1,2,\cdots,N/2$. The partition function of the system can
be obtained as
$Z=\prod_{k}\left(e^{-\beta\Lambda_{k}}+e^{\beta\Lambda_{k}}\right)=\prod_{k}2\cosh\left(\beta\Lambda_{k}\right)$,
where $\beta=1/k_{B}T$ is the inverse temperature and $k_{B}$ is the Boltzmann
constant. Accordingly, the free energy per spin of the system can be
calculated as $F=-k_{B}T\ln
Z/N=-k_{B}T\sum_{k}\ln\left[2\cosh\left(\beta\Lambda_{k}\right)\right]/N$. In
the thermodynamic limit, $N\rightarrow\infty$, we use an integral to replace
the sum and obtain the exact expression of the free energy per spin at
temperature $T$ free energy
$F=-k_{B}T\ln
2-k_{B}T\times\frac{1}{\pi}\int_{0}^{\pi}dk\ln\left[\cosh\left(\beta\Lambda_{k}\right)\right].$
(2)
The magnetization per spin along the direction of the external magnetic field
$\lambda$ at temperature $T$ can be obtained
$M_{z}(T)=-\frac{\partial
F}{\partial\lambda}=\frac{1}{\pi}\int_{0}^{\pi}\tanh\left(\beta\Lambda_{k}\right)\frac{\lambda-\cos
k}{\Lambda_{k}}dk,$ (3)
and then the MS along z direction
$\chi_{z}=-\partial^{2}F/\partial\lambda^{2}$ as a function of the temperature
$T$ and the magnetic field $\lambda$ of the system can also be obtained
$\chi_{z}(\lambda,T)=\frac{1}{\pi}\int_{0}^{\pi}\left[\frac{\beta}{\cosh^{2}{(\beta\Lambda_{k})}}\frac{(\lambda-\cos
k)^{2}}{\Lambda_{k}^{2}}+\tanh(\beta\Lambda_{k})\frac{\gamma^{2}\sin^{2}k}{\Lambda_{k}^{3}}\right]dk$
(4)
We plot the MS $\chi_{z}$ of 1D TIM ($\gamma=1$) as a function of external
magnetic field $\lambda$ and the temperature $T$ in Fig. 1. Clearly it can be
seen that the logarithmic divergence of the MS at zero temperature indicates
the second-order QPT at the QCP $\lambda_{c}=1$. We would like to point it out
that at zero temperature, the magnetization is reduced to
$M_{z}(T=0)=\int_{0}^{\pi}(\lambda-\cos k)/(\pi\Lambda_{k})dk$.
For the convenience of later study, we introduce another observable – the GP,
which is a fundamental concept in quantum mechanics berry . To obtain a
geometric phase, we rotate the Hamiltonian (1) around the z axis for an angle
$\phi$. The effective Hamiltonian after the rotation is
$H_{\phi}=U_{\phi}HU^{\dagger}_{\phi},\\\ \\\ \\\ \\\ \\\ \\\ \
U_{\phi}=\prod_{j=1}^{N}e^{i\phi\sigma_{j}^{z}/2}.$ (5)
The periodicity of the Hamiltonian in $\phi$ is $\pi$. After we rotate the
Hamiltonian back to its initial form ($\phi=\pi$), the GP of the ground state
accumulated by varying the angle $\phi$ from 0 to $\pi$ is given by
$\beta_{g}=-i\frac{2}{N}\int_{0}^{\pi}\left(\left\langle
GS\right|U^{\dagger}_{\phi}\right)\frac{\partial}{\partial\phi}\left(U_{\phi}\left|GS\right\rangle\right)d\phi,$
(6)
which is an extra phase in addition to the usual dynamic phase. From Refs
pachos ; zhu ; yi ; hamma we know that the ground-state GP studied there can
be expressed as
$\beta_{g}=\pi+\int_{0}^{\pi}\frac{\lambda-\cos k}{\Lambda_{k}}dk=\pi+\pi
M_{z}(T=0).$ (7)
Hence, the derivative $\partial\beta_{g}/\partial\lambda$ of the ground-state
GP over the external field is $\pi$ times of the zero-temperature MS
$\chi_{z}=\partial M_{z}(T=0)/\partial\lambda$. We can understand this
relation in the following way: the ground-state GP studied in Refs. pachos ;
zhu ; yi ; hamma is a function of the derivative of the ground state energy
with respect to the external magnetic field zhu ; yi ; hamma , and at zero
temperature, the free energy is equal to the ground state energy. Thus, at
zero temperature, the GP is a function of the magnetization.
As is well known, at zero temperature, the MS of 1D TIM shows logarithmic
singularity at the QCP and exhibits finite-size scaling behavior in the
proximity of the QPT point $\lambda_{c}=1$. Thus, it is not surprising that
the GP exhibits singularity and finite-size scaling behavior near the QCP zhu
. Instead of studying the finite-size scaling of the GP (MS at zero
temperature), in this letter, we will study finite-temperature scaling of the
quantum XY chain. We will see that when the temperature approaches zero, in
analogy with the imaginary time direction approaching the infinity in the
finite-size scaling, the MS obeys $T\neq 0$ scaling behavior in the proximity
of the QPT.
Figure 1: MS $\chi_{z}$ of 1D TIM model as a function of external magnetic
field $\lambda$ and temperature $T$. It can be seen that the MS at zero
temperature show logarithmic divergence at the QCP $\lambda_{c}=1$. At nonzero
temperature the MS is analytical. This agrees with the known result that 1D
TIM model does not exhibits thermal phase transition at nonzero temerature.
Figure 2: (color online). (above) The MS for the 1D TIM ($\gamma=1$) as a
function of the controlling parameter $\lambda$. The curve corresponds to
different temperatures $k_{B}T=0\mathrm{J}$, $0.02\mathrm{J}$,
$0.06\mathrm{J}$, $0.21\mathrm{J}$, $0.5\mathrm{J}$, and $1.01\mathrm{J}$.
With the decrease of the temperature, the maximum gets pronounced, and (below)
the pseudopoint $\lambda_{m}$ changes and tends as $T^{1.706}$ towards the QCP
$\lambda_{c}=1$.
_Scaling of the magnetic susceptibility of the quantum XY chain:_ In order to
further understand the relation between the 1D TIM and 2D classical Ising
model, we investigate the finite-temperature scaling behavior of the MS by the
finite-size scaling ansatz barber83 . For simplicity, we first look at 1D TIM
($\gamma=1$), and we will discuss the properties of the family of $\gamma\neq
1$ later. The MSs as a function of the external magnetic field $\lambda$ at
different temperatures $T$ (including zero temperature) are presented in Fig.
2. At zero temperature the MS shows a singularity at $\lambda_{c}=1$, but at
nonzero temperature, there are no real divergence of $\chi_{z}$. Nevertheless,
there are clear anomalies at low temperature, and the height of which
increases with the decrease of the temperature. This can be regarded as the
precursors of the QPT. What is more, the position $\lambda_{m}$ of the maximum
susceptibility (pseudocritical point) barber83 changes and tends as
$T^{1.704}$ towards the QCP and clearly approaches $\lambda_{c}$ when
$T\rightarrow 0$ (see Fig. 2b). Meanwhile, the maximum value
$\chi_{z}|_{\lambda_{m}}$ of the MS diverges logarithmically with the decrease
of the temeprature
$\chi_{z}|_{\lambda_{m}}\approx\kappa_{1}\ln{T}+\mathrm{const}.$ (8)
Our numerical results (see Fig. 3a) give $\kappa_{1}=0.320$. On the other
hand, when $T=0$, from Ref. barber81 we know that the MS in the proximity of
the QCP exhibits logarithmic singularity
$\chi_{z}\approx\kappa_{2}\ln{|\lambda-\lambda_{c}|}+\mathrm{const}.$ (9)
Our numericals in Fig. 3b give the result $\kappa_{2}\approx 0.317$, while the
exact result barber81 gives $\kappa_{2}=1/\pi\approx 0.3183$. We would like
to point it out that the coefficient $\kappa_{2}$ here is the same as that in
Ref. zhu , where the author gives $\kappa_{2}\approx 0.3123$ and our numerical
result is closer to the exact result $\kappa_{2}=1/\pi$.
Figure 3: (color online). (above) The maximum value of the MS at the
pseudocritical point $\lambda_{m}$ of the 1D quantum XY chain as a function of
temperature $T$. The slope of the line is $0.317$ ($0.394$) for $\gamma=1$
(blue) ($\gamma=0.8$ (red)). (below) The MS at zero temperature diverges
logarithmically in the proximity of the QCP $\lambda_{c}=1$. The slope of the
line is $0.320$ ($0.401$) for $\gamma=1$ (blue) ($\gamma=0.8$ (red)). The
ratio of the two slopes (below and above) for a fixed parameter $\gamma$ is
equal to the critical exponent $\nu$. Here $\nu\approx 1.009$ ($\nu\approx
1.017$) for $\gamma=1$ ($\gamma=0.8$) is obtained. The numerical results agree
with the scaling ansatz and the universality of the XY model.
According to the scaling ansatz in the logarithmic singularities, the ratio
$|\kappa_{2}/\kappa_{1}|$ gives the critical exponent $\nu$ that governs the
divergence of the correlation length $\xi\sim|\lambda-\lambda_{c}|^{-\nu}$. In
our case, $\nu\approx 1.009\sim 1$ is obtained in the numerical calculation
for the 1D TIM, which agrees well with the known result about 1D TIM free
energy . Furthermore, by proper scaling and taking into account the distance
of the maximum $\chi_{z}$ from the QCP, it is possible to make all the data
for the value of
$F=1-\exp\left[\chi_{z}(\lambda)-\chi_{z}|_{\lambda_{m}}\right]$ as a function
of $(\lambda-\lambda_{m})/T$ for different temperatures $T$ to collapse onto a
single curve (see Fig. 4). This figure contains the data for temperatures
ranging from $k_{B}T=e^{-3}\mathrm{J}$, $e^{-4}\mathrm{J}$,
$e^{-5}\mathrm{J}$, $e^{-5.5}\mathrm{J}$. These results demonstrate that the
MS does obey the scaling behavior as the temperature decrease to zero, in
analogy to the lattice size approaching the infinity in the finite-size
scaling cases.
In the following we will study the universality of the critical behavior of
the MS. It is well known that the anisotropic XY chain ($\gamma\ni(0,1]$)
belongs to the 1D TIM universality, while isotropic XY chain ($\gamma=0$)
belongs to the XX universality. For the 1D TIM universality, $\nu=1$, while
for the XX universality, $\nu=1/2$. We will show that the finite-temperature
scaling behavior of $\chi_{z}$ also manifests the universality principle - the
critical properties depends only on the dimensionality of the system and the
broken symmetry in ordered phase. To verify the universality principle of the
XY model, we consider the case for $\gamma\neq 1$. The asymptotic behavior is
also described by Eqs. (8) and (9). From Fig. 3 we see that for $\gamma=0.8$
numerical simulation gives $\kappa_{1}\approx 0.394$ and $\kappa_{2}\approx
0.401$, while the exact result barber81 should be
$\kappa_{2}=(\gamma\pi)^{-1}\approx 0.398$. As a result the critical exponent
for $\gamma=0.8$ is $\nu=|\kappa_{2}/\kappa_{1}|\approx 1.017$, very close to
the exact value $\nu=1$. Moreover we also verify that by proper scaling, all
data for different temperatures $T$ but a specific $\gamma$ will collapse onto
the same curve. The data for $\gamma=0.8$ are shown in Fig. 4.
Figure 4: (color online). The value of
$F=1-\exp{[\chi_{z}(\lambda)-\chi_{z}|\lambda_{m}]}$ as a function of
$(\lambda-\lambda_{m})/T$ for different temperatures (dots of different
colors) $k_{B}T=e^{-3}\mathrm{J},e^{-4}\mathrm{J},e^{-5}\mathrm{J}$, and
$e^{-5.5}\mathrm{J}$. For fixed $\gamma$ (here we choose $\gamma=1$ and
$\gamma=0.8$), all data collapse on a single curve, which agrees with the
finite-size scaling behavior. The critical exponent $\nu=1$ can be obtained
from this figure.
What is more, through a similar analysis to that in Ref. zhu , we can directly
extract the finite-temperature scaling behavior of the XX ($\gamma=0$)
universality class. It can be found that, at zero temperature $T=0$, for the
XX universality, the magnetization can be written in the following compact
form.
$M_{z}=\left\\{\begin{array}[]{c}1-\frac{2}{\pi}\arccos\lambda,(0\leq\lambda\leq
1)\\\ 1,(\lambda>1)\end{array}\right..$ (10)
Accordingly, the critical exponent $\nu=1/2$ and $z=2$ can be extracted from
the MS $\chi_{z}=\sqrt{2}(1-\lambda)^{-\frac{1}{2}}$, $(\lambda\to 1^{-})$ zhu
, which is different from the TIM universality ($\nu=1$ and $z=1$). When we
change the anisotropy $\gamma$ from 1 to 0, we find the range of the validity
of the quantum scaling ansatz in $\lambda$ (Eq. (9)) shrink gradually. The
leading term of the MS crossover from $\frac{1}{\pi\gamma}\ln{(1-\lambda)}$ to
$\sqrt{2}(1-\lambda)^{-\frac{1}{2}}$. Hence, when $0<\gamma\leq 1$, the
scaling belongs to XY universality, while when $\gamma=0$, the scaling belongs
to XX universality. Finally, we also would like to point it out that our
numerical result shows that the scaling behavior of Eq. (8) can persist up to
a temperature $k_{B}T\approx 0.5\mathrm{J}$. This result agrees well with that
of Ref. nature05 . In addition, the crossover line in the region $0<\lambda<1$
given by the MS $\chi_{z}$ is roughly
$T_{c}\sim\left|\lambda-\lambda_{c}\right|^{\nu z}$, which agrees well with
the result obtained in the analysis elsewhere thermal . Hence the boundary of
quantum critical scaling region can be confirmed by the behavior of the MS
$\chi_{z}$.
_Magnetic susceptibility and geometric phase:_ As we have mentioned before,
the derivative of the ground-state GP discussed in Refs. pachos ; zhu ; yi ;
hamma is equal to $\pi$ times of the MS, and the finite-size scaling of the
GP zhu actually represents the finite-size scaling of the MS. Based on these
studies, we would like to further study the relation between the thermal-state
GP and the MS at a finite temperature. Similar to the definition of the
ground-state GP in Refs. pachos ; zhu ; yi ; hamma , we define the thermal-
state GP in the following way: four eigenstates of the modes $(k,-k)$ of
$H_{\phi}$ (see Refs. pachos ; zhu ; yi ; hamma ) can be expressed as
$\left|00\right\rangle_{k}=\cos{(\theta_{k}/2)}\left|0\right\rangle_{k}\left|0\right\rangle_{-k}+ie^{i2\phi}\sin{(\theta_{k}/2)}\left|1\right\rangle_{k}\left|1\right\rangle_{-k}$,
$\left|11\right\rangle_{k}=ie^{-i2\phi}\sin{(\theta_{k}/2)}\left|0\right\rangle_{k}\left|0\right\rangle_{-k}+\cos{(\theta_{k}/2)}\left|1\right\rangle_{k}\left|1\right\rangle_{-k}$,
$\left|01\right\rangle_{k}=\left|0\right\rangle_{k}\left|1\right\rangle_{-k}$
, and
$\left|10\right\rangle_{k}=\left|1\right\rangle_{k}\left|0\right\rangle_{-k}$
with the angle $\theta_{k}$ defined by $\theta_{k}=\arctan[-\sin k/(\cos
k-\lambda)]$. The GP of the thermal state at temperature $T$ accumulated by
varying the angle $\phi$ from $0$ to $\pi$ is described by
$\beta_{T}=\frac{-2i}{N}\sum_{k=1}^{N/2}\sum_{n}\int e^{-\beta
E_{n}^{k}}\left\langle
n\right|_{k}\frac{\partial}{\partial\phi}\left|n\right\rangle_{k}d\phi,$ (11)
where
$\left|n\right\rangle_{k}=\left|00\right\rangle_{k},\left|01\right\rangle_{k},\left|10\right\rangle_{k}$,
and $\left|11\right\rangle_{k}$. After a straightforward calculation, we
obtain the same relation between the magnetization and the GP as that of zero
temperature $\beta_{T}=\pi+\int_{0}^{\pi}(\lambda-\cos
k)/\Lambda_{k}\tanh{(\beta\Lambda_{k})}dk=\pi[1+M_{z}(T)]$ (3). Thus we prove
that at both zero temperature and nonzero temperature, the GP of the quantum
XY chain is a linear function of the magnetization and the derivative of the
GP is proportional to the MS. The discussions of the finite-temperature
scaling of the MS in this letter can be alternatively regarded as the finite-
temperature scaling of the GP in the proximity of the QPT point. Finally the
close relation between the GP and the MS does not confined to 1D quantum XY
chain. In Ref. plaslina the ground-state GP of the Dicke model and its
relation to quantum criticality are studied. We would like to point it out
that, similar to the discussions about the 1D XY chain, the ground-state GP of
the Dicke model is a linear function of the ground state magnetization
$\beta_{g}=\pi(1+\left\langle S_{x}\right\rangle/N)$, where $\left\langle
S_{x}\right\rangle/N$ is the magnetization $M_{x}$ per spin in Dicke model.
Hence the derivative of ground-state GP of the Dicke model is also equal to
$\pi$ times of the MS. Besides the above two examples, it can be proved that
for any QPTs driven by an external magnetic field, such as the Lipkin-Meshkov-
Glick model yi and 1D XXZ model xxz , the relation between the GP and the
magnetization still holds true. The proof is give as follows. For those QPTs
driven by an external magnetic field, we apply a $\pi$-rotation along the z
axis for every spin $U_{\phi}=\prod_{j=1}^{N}e^{i\phi\sigma_{j}^{z}/2}$ to
obtain the GP. The GP of the ground state can be expressed as (6)
$\beta_{g}=-i\frac{2}{N}\int_{0}^{\pi}\left(\left\langle
GS\right|U^{\dagger}_{\phi}\right)\left(i\sum_{j}\frac{\sigma_{j}^{z}}{2}\right)\left(U_{\phi}\left|GS\right\rangle\right)d\phi=\frac{2\pi}{N}\sum_{j}\left\langle
GS\right|\frac{\sigma_{j}^{z}}{2}\left|GS\right\rangle=\pi M_{z}(T=0),$ (12)
where $\left|GS\right\rangle$ is the ground state of the Hamiltonian (1)
before the rotation, and $U_{\phi}\left|GS\right\rangle$ is the instantaneous
ground state after the rotation for an angle $\phi$. We know that
$\frac{2}{N}\sum_{j}\left\langle
GS\right|\frac{\sigma_{j}^{z}}{2}\left|GS\right\rangle$ is the definition of
the GP of the ground state. Thus we prove that the GP obtained by applying a
rotation around the z axis to each spin, is proportional to the magnetization
along z axis (similarly if we rotate along x axis, the GP will be proportional
to the magnetization along x axis). We also would like to point out that the
GP in Eq. (12) differs from that in Eq. (7) by a constant $\pi$. This is
because the ground state of $U_{\phi}HU_{\phi}^{\dagger}$ has an uncertainty
of the global phase. When we choose a proper global phase, we can eliminate
the difference between Eq. (7) and Eq. (12). When we study the scaling of
$\frac{d\beta_{g}}{d\lambda}$, the difference does not affect. In addition, we
can generalize the above discussions to eigenstates other than the ground
state. We find the same proportional factor between the GP and the
magnetization for all eigenstates. Thus the relation between the GP and the
magnetization can be straightforwardly generalized to a thermal state at
finite temperature. The above XY model is a good example.
In summary, we study the finite-temperature scaling of the MS of the quantum
XY chain. All key features of the quantum criticality, such as scaling,
critical exponent, the universality, etc. are presented in the MS of the XY
spin chain. Though the nature of the QPT and the $T\neq 0$ scaling is purely
quantum mechanical, the classical macroscopic thermodynamic observable MS,
which can be easily accessed experimentally, can be used to witness and
characterize the quantum features of the system explain ; verdral . Our
studies shed light on the mechanism of bring quantum criticality up to a
finite temperature, and opens the possibility of observing the footprint of
quantum criticality experimentally. We also would like to point it out that
the results obtained in this paper does not depend on the model and
thermodynamic observable used here and can be generalized to other QPT models
with the only change of MS to a “controlling parameter-dependent
susceptibility”. For example, in a QPT driven by the pressure instead of the
external magnetic field, the observable $\chi_{p}=-\partial^{2}F/\partial
p^{2}$ is expected to exhibit the finite-temperature scaling behavior, and the
critical exponent can be extracted through a similar analysis. Finally, our
study establishes the connection between the MS and the GP at both zero
temperature and nonzero temperature in a family of spin systems, where the
QPTs are driven by an external magnetic field.
The author thanks F. M. Cucchietti and Rishi Sharma for stimulating
discussions and gratefully acknowledges the support of the U.S. Department of
Energy through the LANL/LDRD Program for this work.
## References
* (1) S. Sachdev, Quantum Phase Transitions (Cambridge University Press, Cambridge, England, 1999).
* (2) T. J. Osborne, et al, Phys. Rev. A, 66, 032110 (2002); A. Osterloh, et al, Nature 416, 608 (2002).
* (3) G. Vidal, et al, Phys. Rev. Lett. 90,227902 (2003); Y. Chen, et al, New J. Phys. 8, 97 (2006); S. J. Gu, et al, Phys. Rev. Lett. 93, 086402 (2004).
* (4) A. C. M. Carollo et al, Phys. Rev. Lett. 95, 157203 (2005); Phil. Trans. R. Soc. A, 364, 3463 (2006).
* (5) H.T. Quan, et al, Phys. Rev. Lett. 96, 140604 (2006); P. Zanardi, et al, Phys. Rev. A 75, 032109 (2007).
* (6) P. Zanardi, et al, Phys. Rev. E 74, 031123 (2006);
* (7) A. Kopp, et al, Nature Phys. 1, 53 (2005); Zhihua Yang, et al, Phys. Rev. Lett. 100, 067203 (2008).
* (8) P. Coleman, et al, Nature 433, 226 (2005).
* (9) L. Amico, et al, Phys. Rev. A 74, 022322 (2006); L. Amico, et al, Europhys. Lett. 77, 17001 (2007); A. Cuccoli, et, al, Phys. Rev. B 76, 064405 (2007); H. T. Quan and F. M. Cucchietti, Phys. Rev. B 79, 031101 (2009).
* (10) S. Sachdev, Nature Phys. 4, 173 (2008); T. Giamarchi, et al, Nature Phys. 4, 198 (2008).
* (11) S. L. Sondhi et al, Rev. Mod. Phys., 69, 315 (1997); M. Vojta, Rep. Prog. Phys., 66, 2069 (2003).
* (12) E. Lieb, Ann. Phys. (1961); S. Katsura, Phys. Rev. 127, 1508 (1962); P. Pfeuty, Ann. Phys., 57, 79 (1970).
* (13) M. Suzuki, Prog. Theor. Phys., 46, 1337 (1971); 56, 1454 (1976).
* (14) A. E. Ferdinand, et al, Phys. Rev., 185, 832 (1969.
* (15) P. Gegenwart, et al, Nature Phys. 4, 186 (2008); L. Zhu, et al, Phys. Rev. Lett. 91, 066404 (2003); R. Kuchler, et al, Phys. Rev. Lett. 91, 066405 (2003).
* (16) the use of thermodynamics observables to witness quantum entanglement is studied in Ref. verdral . For XY chain, the separable bound of the MS is given by the sum of three susceptibilities along three orthogonal axes $\chi_{x}+\chi_{y}+\chi_{z}\geqslant N(2k_{B}T)^{-1}$. We know that when $0\leqslant\gamma\leqslant 1$ and at $T=0$, $\chi_{x}\approx(1-\lambda)^{-7/4}$, $\chi_{y}=0$, and $\chi_{z}\approx(\gamma\pi)^{-1}\ln{|1-\lambda|}$. Hence, except for $\lambda=1$, the MS cannot be explained without entanglement at $T=0$. The above result agrees with the analysis using the specific heat as an entanglement witness verdral .
* (17) V. Verdral, New J. Phys. 6, 102 (2004); Nature, 453, 1004 (2008); J. Anders, et ak, Open Sys. Inform. Dyn. 14, 1 (2007); M. Wiesniak, et al, New J. Phys. 7, 258 (2005); C. Brukner, et al, Phys. Rev. A 73, 012110 (2006); M. Wiesniak, et al, Phys. Rev. B, 78, 064108 (2008); C. Brukner, et al. arXiv: quant-ph/0406040; B. C. Hiesmayr, et al, arXiv: quant-ph/501015.
* (18) S. L. Zhu, Phys. Rev. Lett. 96, 077206 (2006); Int. J. Mod. Phys. B, 22, 561 (2008);
* (19) H. T. Cui, et al, Phys. Lett. A, 360, 243 (2006).
* (20) A. Hamma, arXiv: quant-ph/0602091.
* (21) M. V. Berry, Proc. R. Soc. London Ser. A, 392, 45 (1984).
* (22) M. N. Barber in Phase Transition and Critical Phenomena, Edited by C. Domb, and J. L. Lebowits, Vol.8, P154 (Academic Press, London, 1983).
* (23) Th. Niemejer, Physica, 36, 377 (1967); C. J. Hamer, et al, J. Phys. A, 14, 241 (1981).
* (24) F. Plastina, et al, Europhys. Lett. 76, 182 (2006).
* (25) The Hamiltonian of XXZ model is $H=\sum_{i=1}^{N}\left[\sigma_{i}^{x}\sigma_{i+1}^{x}+\sigma_{i}^{y}\sigma_{i+1}^{y}+\Delta\sigma_{i}^{z}\sigma_{i+1}^{z}+\lambda\sigma_{i}^{z}\right]$. Here both the anisotropy $\Delta$ and the external magnetic field $\lambda$ can induce QPTs takahashi . However, only the QPT induced by $\lambda$ can be characterized by the GP induced by a rotation $U_{\phi}=\prod_{j=1}^{N}e^{-i\phi\sigma_{j}^{z}/2}$. Because $\frac{\partial F}{\partial\Delta}$ is not proportional to the magnetization in z direction $M_{z}$, and hence is not proportional to $\left\langle\sigma^{z}\right\rangle$.
* (26) M. Takahashi, Thermodynamics of One-Dimensional Solvable Models (Cambridge University Press, Cambridge, England, 1999).
| arxiv-papers | 2008-06-15T23:47:45 | 2024-09-04T02:48:56.273645 | {
"license": "Public Domain",
"authors": "H. T. Quan",
"submitter": "Haitao Quan",
"url": "https://arxiv.org/abs/0806.2476"
} |
0806.2478 | # The effect of localized new Higgs doublet on the radiative lepton flavor
violating decays in the Randall Sundrum background
E. O. Iltan
Middle East Technical University, Northern Cyprus Campus, Guzelyurt,
Mersin 10, TURKEY
E-mail address: eiltan@newton.physics.metu.edu.tr
###### Abstract
We study the radiative lepton flavor violating $l_{1}\rightarrow l_{2}\gamma$
decays in the two Higgs doublet model with localized new Higgs doublet in the
Randall Sundrum background. We estimate the contributions of the KK modes of
new Higgs bosons and left (right) handed charged lepton doublets (singlets) on
the branching ratios of the decays considered. We observe that there is an
enhancement in the branching ratios with the addition of new Higgs boson and
lepton KK modes.
## 1 Introduction
The processes with flavor violation (FV) are worthwhile to study since they
exist at least in the one loop level in the standard model (SM) and,
therefore, they are rich from the theoretical point of view. The lepton flavor
violating (LFV) interactions are among the most exciting candidates of these
processes, since they are clean in the sense that they are free from strong
interactions. Furthermore, the small numerical values of branching ratios
(BRs) of LFV decays stimulate one to search beyond and to study the more
fundamental models in order to enhance these numerical values to reach the
current experimental upper limits. Among the LFV decays the radiative
$l_{1}\rightarrow l_{2}\gamma$ processes reach great interest and their
current experimental upper limits of the BRs are: BR $(\mu\rightarrow
e\gamma)=1.2\times 10^{-11}$ [1], BR $(\tau\rightarrow e\gamma)=3.9\times
10^{-7}$ [2] and BR $(\tau\rightarrow\mu\gamma)=1.1\times 10^{-6}\,(9.0\times
10^{-8}\,\mathbf{;}\,6.8\times 10^{-8}\mathbf{,}\,\,90\%CL)$ [3] ([4]; [5]),
respectively. Furthermore, in order to search the $\mu\rightarrow e\gamma$
decay, a new experiment, aiming to reach a sensitivity of BR$\sim 10^{-14}$,
at PSI has been described [6]. At present, this experiment (PSI-R-99-05
Experiment) is still running in the MEG [7].
The theoretical values of the BRs of the radiative LFV decays in the framework
of the SM are negligible compared to the experimental upper limits and the
addition of one more Higgs doublet, which drives the flavor changing neutral
currents (FCNCs) and the LFV interactions at tree level, may cause to pull the
theoretical values of the BRs near to the experimental upper limits. This is
the case that the lepton FV is induced by the internal new neutral Higgs
bosons, $h^{0}$ and $A^{0}$, and the strength of this violation is regulated
by the Yukawa couplings, appearing as free parameters which should be
restricted by using the experimental data. These decays were examined in the
framework of the SM with one more Higgs doublet [8]-[11], the so called two
Higgs doublet model (2HDM), in [12, 13, 14, 15, 16, 17]. Besides the
theoretical calculations based on the 2HDM, they were studied in the
supersymmetric models [18]-[24], in a model independent way [25], in the
framework of 2HDM and the supersymmetric model [26] and, recently, in the SM
including effective operators coming from the possible unparticle effects
[27]-[28].
Another possibility to enhance the numerical values of the BRs of these
processes is to consider the extra dimension which results in the additional
effects of the KK modes of the particles in the loops, after the
compactification. In the present work, we consider the extended Higgs sector,
the 2HDM, in the the RS1 background [29, 30]. The RS1 model is based on the
curved extra dimension and the corresponding metric reads
$\displaystyle
ds^{2}=e^{-2\,\sigma}\,\eta_{\mu\nu}\,dx^{\mu}\,dx^{\nu}-dy^{2}\,,$ (1)
where $\sigma=k\,|y|$, $k$ is the bulk curvature constant, the exponential
$e^{-\sigma}$, with $y=R\,|\theta|$, is the warp factor. Here, the extra
dimension is compactified onto $S^{1}/Z_{2}$ orbifold and $R$ is the
compactification radius. The extra dimension has two boundaries, the hidden
(Planck) brane and the visible (TeV) brane, with opposite and equal tensions.
This choice leads to the fact that the low energy effective theory has flat 4D
spacetime, even if the 5D cosmological constant is non vanishing. In the RS1
background, the gravity is taken to be localized on the hidden brane and to be
extended into the bulk with varying strength and the SM fields live in the
visible brane. If some of the SM fields are accessible to the extra dimension,
the phenomenology becomes richer and there are various work done in the
literature respecting such scenarios [31]-[49]. If fermions are accessible to
the extra dimension and there is a Dirac mass term in the Lagrangian, the
fermion mass hierarchy can be explained. In this case the fermion mass
hierarchy comes from the possible fermion field locations [35, 38, 39, 40].
The quark and lepton FV, which is based on the different locations of the
fermion fields in the extra dimension, is extensively studied in [45, 46]. In
these works, it is considered that the FV is carried by the Yukawa
interactions, coming from the SM Higgs-fermion-fermion vertices. In [47], the
high precision measurements of top pair production at the ILC is addressed by
considering that the fermions are localized in the bulk of RS1 background. In
recent works [48, 49], the various experimental FCNC constraints and the
electro weak precision tests for the location parameters of the fermions in
the extra dimension are discussed. The other possibility is to consider the
localization of Higgs field in the extra dimension. The brane localized mass
terms for scalar fields are considered in order to get small couplings of KK
modes with the boundaries [35] and these mass terms result in that the zero
mode localized solution is obtained. In [41] the hierarchy of fermion masses
is analyzed by taking that the Higgs field has an exponential profile around
the TeV brane. [42] is devoted to an extensive work on the bulk fields in
various multi-brane models.
In our work, we assume that the new Higgs doublet is accessible to the extra
dimension of RS1 background. First, we study the case that the charged leptons
are restricted to the 4D brane and, second, we consider that the charged
leptons are also localized in the extra dimension. Notice that, in both cases,
the gauge bosons are necessarily accessible to the extra dimension.
The paper is organized as follows: In Section 2, we present the BRs of LFV
interactions $l_{1}\rightarrow l_{2}\gamma$ in the 2HDM, by considering that
the new Higgs doublet is localized in the extra dimension of RS1 background.
Section 3 is devoted to discussion and our conclusions. In Appendix A, we
study the construction of new Higgs boson mass matrix. In Appendix B, we
present the amplitudes appearing in the calculation of the decay widths of the
radiative decays under consideration Appendix C is devoted to calculation of
the zero mode lepton fields and their KK modes.
## 2 LFV $l_{1}\rightarrow l_{2}\gamma$ decays in the Randall Sundrum
background with localized new Higgs boson
We start with the action for the new Higgs doublet $\phi$ (see for example
[31], [42] for a massive bulk scalar field case),
$\displaystyle{\cal{S}}_{\phi}=\frac{1}{2}\,\int d^{4}x\int
dy\,\sqrt{g}\,\Big{(}g^{MN}\,(\partial_{M}\phi)^{\dagger}\,\partial_{N}\,\phi+m^{2}_{\phi}\,\phi^{\dagger}\,\phi\Big{)}\,,$
(2)
where $g=Det[g_{MN}]=e^{-8\,\sigma}$, $M,N=0,1,...,4$. The decomposition of
the scalar doublet into KK modes
$\displaystyle\phi(x,y)=\sum_{n=0}^{\infty}\,\phi^{(n)}(x)\,f_{n}(y)$ (3)
brings the action eq.(2) into form
$\displaystyle{\cal{S}}_{S}=\frac{1}{2}\,\sum_{n=0}^{\infty}\,\int
d^{4}x\,\Big{(}\eta^{\mu\nu}\,(\partial_{\mu}\phi_{n}(x))^{\dagger}\,\partial_{\nu}\,\phi_{n}(x)+m_{n\,S}^{2}\,(\phi_{n}(x))^{\dagger}\,\phi_{n}(x)\Big{)}\,,$
(4)
with the second order differential equation
$\displaystyle-e^{4\,\sigma}\,\frac{d}{dy}\,\Big{(}e^{-4\,\sigma}\,\frac{d\,f_{n}(y)}{dy}\Big{)}+m_{\phi}^{2}\,f_{n}(y)=m_{n\,S}^{2}\,e^{2\,\sigma}\,f_{n}(y)\,,$
(5)
and the orthogonality relation
$\displaystyle\int_{-\pi\,R}^{\pi\,R}dye^{-2\,\sigma}\,f^{*}_{n}(y)\,f_{m}(y)=\delta_{nm}\,.$
(6)
The choice of the mass term in eq.(2)
$\displaystyle
m^{2}_{\phi}=a\,(\frac{d\,\sigma}{dy})^{2}+b\,\frac{d^{2}\,\sigma}{dy^{2}}\,,$
(7)
results in the differential equation in the bulk
$\displaystyle-e^{4\,\sigma}\,\frac{d}{dy}\,\Big{(}e^{-4\,\sigma}\,\frac{d\,f_{n}(y)}{dy}\Big{)}+a\,k^{2}\,f_{n}(y)=m_{n\,S}^{2}\,e^{2\,\sigma}\,f_{n}(y)\,,$
(8)
where $n=1,2,....$. Now, the the boundary mass term111Here the boundary mass
terms have the same magnitude and the opposite sign on the branes. The idea of
brane localized mass terms has been considered for scalar fields in [31],
[35].
$\displaystyle m^{2}_{\phi,bound}=b\,\frac{d^{2}\,\sigma}{dy^{2}}\,,$ (9)
is considered in order to obtain zero mode Higgs doublet and this term induces
the boundary condition
$\displaystyle\Bigg{(}\frac{\partial\phi(x,y)}{\partial
y}-b\,k\,\phi(x,y)\Bigg{)}|_{y=0,\pi\,R}=0\,\,.$ (10)
Notice that the non-vanishing zero mode can be obtained with the fine tuning
of the parameters222There is another possibility of fine tuning of the
parameters $b$ and $a$ for the non-vanishing zero mode, namely
$b=2-\sqrt{4+a}$. However we ignore this choice since it is not appropriate
for the brane localized fermion scenario and bulk fermion scenario with the
parameter set used (see discussion section for details). $b$ and $a$,
$\displaystyle b=2+\sqrt{4+a}\,,$ (11)
and it reads
$\displaystyle
f_{0}(y)=\frac{e^{b\,k\,y}}{\sqrt{\frac{e^{2\,(b-1)\,k\,\pi\,R}-1}{(b-1)\,k}}}\,\,.$
(12)
On the other hand, the KK mode Higgs doublet is obtained as
$\displaystyle
f_{n}(y)=\frac{e^{2\,\sigma}}{N_{S\,n}}\,\Bigg{(}J_{\sqrt{4+a}}(e^{\sigma}\,x_{nS})+\alpha_{n}\,Y_{\sqrt{4+a}}(e^{\sigma}\,x_{nS})\Bigg{)}\,,$
(13)
where $N_{S\,n}$ is the normalization constant, $x_{nS}=\frac{m_{n\,S}}{k}$
and $\alpha_{n}$ reads
$\displaystyle\alpha_{n}$ $\displaystyle=$
$\displaystyle\frac{(2-b)\,J_{\sqrt{4+a}}(x_{nS})+x_{nS}\,J^{\prime}_{\sqrt{4+a}}(x_{nS})}{(b-2)\,Y_{\sqrt{4+a}}(x_{nS})-x_{nS}\,Y^{\prime}_{\sqrt{4+a}}(x_{nS})}\,.$
(14)
Here, the functions $J_{\beta}(w)$ and $Y_{\beta}(w)$ are the Bessel function
of the first kind and of the second kind, respectively. Finally, the mass
spectrum of KK modes ($n=1,2,...$) are obtained by using the boundary
conditions at $y=0$ and $y=\pi\,R$ (see eq.( 10)),
$\displaystyle
m_{n\,S}\simeq(n+\frac{1}{2}\,(b-2)-\frac{3}{4})\,\pi\,k\,e^{-k\,\pi\,R}\,\,,$
(15)
for $k\,e^{-k\,\pi\,R}\ll m_{n\,S}\ll k$.
At this stage, we consider two possibilities for the charged leptons :
* •
they are restricted to the 4D brane
* •
they are localized in the extra dimension.
### 2.1 The charged leptons restricted to the brane
The LFV interactions are driven by the part of the action
$\displaystyle{\cal{S}}_{Y}=\int
d^{5}x\sqrt{g}\,\Bigg{(}\xi^{E}_{5\,ij}\,\bar{l}_{iL}\phi_{2}E_{jR}+h.c.\Bigg{)}\,\delta(y-\pi
R)\,\,\,,$ (16)
where $L$ and $R$ denote chiral projections $L(R)=1/2(1\mp\gamma_{5})$,
$\phi_{2}$ is the new scalar doublet, $l_{iL}$ ($E_{jR}$) are lepton doublets
(singlets), $\xi^{E}_{5\,ij}$, with family indices $i,j$ , are the Yukawa
couplings in five dimensions, which are responsible for the flavor violating
interactions in the leptonic sector. Here, we assume that the Higgs doublet
$\phi_{1}$ lives on the visible brane and it has non-zero vacuum expectation
value in order to ensure the ordinary masses of the gauge fields and the
fermions. On the other hand the second doublet, which is accessible to the
extra dimension, has no vacuum expectation value333 Here we take the Higgs
potential as $\displaystyle
V(\phi_{1},\phi_{2})=c_{1}(\phi_{1}^{+}\phi_{1}-v^{2}/2)^{2}+c_{2}(\phi_{2}^{+}\phi_{2})^{2}+c_{3}[(\phi_{1}^{+}\phi_{1})(\phi_{2}^{+}\phi_{2})-(\phi_{1}^{+}\phi_{2})(\phi_{2}^{+}\phi_{1})]+c_{4}[Re(\phi_{1}^{+}\phi_{2})]^{2}+c_{5}[Im(\phi_{1}^{+}\phi_{2})]^{2}\,.$
This choice leads to no tree level mixing between the CP even neutral Higgs
bosons, namely $H^{0}$ and $H^{1}$. Therefore, the SM particles (new
particles) are collected in the first (second) doublet and $H^{1}$, $H^{2}$
are obtained as the mass eigenstates $h^{0}$ and $A^{0}$ respectively. Notice
that, in general, the mixing between the CP even neutral Higgs bosons can
exist in the loop level when one considers the quantum corrections.:
$\displaystyle\phi_{1}=\frac{1}{\sqrt{2}}\left[\left(\begin{array}[]{c c}0\\\
v+H^{0}\end{array}\right)\;+\left(\begin{array}[]{c c}\sqrt{2}\chi^{+}\\\
i\chi^{0}\end{array}\right)\right]\,;\phi_{2}=\frac{1}{\sqrt{2}}\left(\begin{array}[]{c
c}\sqrt{2}H^{+}\\\ H^{1}+iH^{2}\end{array}\right)\,\,,$ (23)
and
$\displaystyle<\phi_{1}>=\frac{1}{\sqrt{2}}\left(\begin{array}[]{c c}0\\\
v\end{array}\right)\,\,;<\phi_{2}>=0\,\,.$ (26)
The new Higgs doublet $\phi_{2}$ is expanded into KK modes after the
compactification of the extra dimension as given in eq.(3) and the zero (KK)
mode Higgs fields are obtained by imposing the fine tuning condition in
eq.(11). On the other hand, after the electro weak breaking, the SM Higgs
acquires the vacuum expectation value eq.(26) and there appears mixing between
zero mode and KK mode new Higgs bosons. However, we do not take into account
the additional effects coming from this mixing since they are suppressed (see
Appendix A for detail).
For the effective Yukawa coupling $\xi^{E}_{ij}$ we integrate out the Yukawa
interaction eq.(16) over the fifth dimension by taking the zero mode neutral
Higgs fields $S=h^{0},A^{0}$ :
$\displaystyle\xi^{E}_{ij}=V^{0}_{ij}\,\xi^{E}_{5\,ij}\,,$ (27)
where
$\displaystyle V^{0}_{ij}$ $\displaystyle=$
$\displaystyle\int_{-\pi\,R}^{\pi\,R}\,dy\,e^{-4\,\sigma}\,f_{0}(y)\,\delta(y-\pi
R)$ (28) $\displaystyle=$
$\displaystyle\frac{e^{(b-4)\,k\,\pi\,R}}{\sqrt{\frac{e^{2\,(b-1)\,k\,\pi\,R}-1}{k\,(b-1)}}}\,.$
Here it is assumed that the coupling $\xi^{E}_{5\,ij}$ is flavor dependent and
it is regulated in such a way that the overall quantity
$V^{0}_{ij}\,\xi^{E}_{5\,ij}$ is pointed to the chosen numerical value of
$\xi^{E}_{ij}$.
The effective Yukawa coupling inducing the tree level interaction among the KK
mode Higgs and charged leptons reads
$\displaystyle\xi^{E\,n}_{ij}=V^{S\,n}_{ij}\,\xi^{E}_{5\,ij}\,,$ (29)
where
$\displaystyle
V^{S\,n}_{ij}=\frac{e^{-2\,k\,\pi\,R}}{N_{S\,n}}\,\Big{(}J_{b-2}(e^{k\,\pi\,R}\,x_{nS})+\alpha_{n}\,Y_{b-2}(e^{k\,\pi\,R}\,x_{nS})\Big{)}\,.$
(30)
Here $N_{S\,n}$ is the normalization constant (see eq.(13)) and $b$
($\alpha_{n}$) is defined in eqs.(11) ((14)). Finally, the effective Yukawa
coupling $\xi^{E\,n}_{ij}$ is obtained as
$\displaystyle\xi^{E\,n}_{ij}=\frac{V^{n}_{S\,ij}}{V^{0}_{ij}}\,\xi^{E}_{ij}=\frac{e^{(2-b)\,k\,\pi\,R}\,\sqrt{\frac{e^{2\,(b-1)\,k\,\pi\,R}-1)}{k\,(b-1)}}}{N_{S\,n}}\,\Big{(}J_{b-2}(e^{k\,\pi\,R}\,x_{nS})+\alpha_{n}\,Y_{b-2}(e^{k\,\pi\,R}\,x_{nS})\Big{)}\,\xi^{E}_{ij}\,.$
(31)
Now, we present the decay widths of the LFV $l_{1}\rightarrow l_{2}\gamma$
decays, including the KK modes of new neutral Higgs fields. Since these decays
exist at least in the one loop level, there appear the logarithmic divergences
in the calculations. In order to eliminate these divergences, we follow the
on-shell renormalization scheme. In this scheme, the self energy diagrams can
be written in the form
$\sum(p)=(\hat{p}-m_{l_{1}})\bar{\sum}(p)(\hat{p}-m_{l_{2}})\,,$ which results
in that these diagrams do not contribute for on-shell leptons and, only, the
vertex diagrams (see Fig.1) contribute444This is the case that the divergences
can be eliminated by introducing a counter term $V^{C}_{\mu}$ with the
relation $V^{Ren}_{\mu}=V^{0}_{\mu}+V^{C}_{\mu}\,,$ where $V^{Ren}_{\mu}$
($V^{0}_{\mu}$) is the renormalized (bare) vertex and by using the gauge
invariance $k^{\mu}V^{Ren}_{\mu}=0$. Here, $k^{\mu}$ is the four momentum
vector of the outgoing photon.. Taking only tau lepton for the internal
line555We take into account only the internal tau lepton contribution since we
respect the idea that the couplings $\bar{\xi}^{E}_{N,ij}$ ($i,j=e,\mu$), are
small compared to $\bar{\xi}^{E}_{N,\tau\,i}$ $(i=e,\mu,\tau)$, due to the
possible proportionality of them to the masses of leptons under consideration
in the vertices. Here, we use the dimensionful coupling $\bar{\xi}^{E}_{N,ij}$
with the definition
$\xi^{E}_{N,ij}=\sqrt{\frac{4\,G_{F}}{\sqrt{2}}}\,\bar{\xi}^{E}_{N,ij}$ where
N denotes the word ”neutral”., the decay width $\Gamma$ reads
$\displaystyle\Gamma(l_{1}\rightarrow
l_{2}\gamma)=c_{1}(|A_{1}|^{2}+|A_{2}|^{2})\,\,,$ (32)
where
$\displaystyle A_{1}$ $\displaystyle=$ $\displaystyle
A^{0}_{1}+A^{S\,KK}_{1}\,,$ $\displaystyle A_{2}$ $\displaystyle=$
$\displaystyle A^{0}_{2}+A^{S\,KK}_{2}\,.$ (33)
For the explicit expression of these amplitudes see Appendix B.
### 2.2 The charged leptons localized in the extra dimension
The part of the action which drives the LFV interactions in this case reads
$\displaystyle{\cal{S}}_{Y}=\int
d^{5}x\sqrt{g}\,\Bigg{(}\xi^{E}_{5\,ij}\,\bar{l}_{iL}\phi_{2}E_{jR}+h.c.\Bigg{)}\,,$
(34)
where $l_{iL}$ ($E_{jR}$) are lepton doublets (singlets) which are localized
in the extra dimension. The addition of Dirac mass term to the lagrangian of
bulk fermions causes this localization [32, 34, 35, 37, 38, 40, 41]. Since the
combination $\bar{\psi}\psi$ is odd due to the two possible transformation
properties of fermions under the orbifold $Z_{2}$ symmetry,
$Z_{2}\psi=\pm\gamma_{5}\psi$, in order to construct the $Z_{2}$ invariant
mass term, one needs $Z_{2}$ odd scalar field to be coupled. This discussion
leads to the mass term
$\displaystyle{\cal{S}}_{m}=-\int d^{4}x\int
dy\,\sqrt{g}\,m(y)\,\bar{\psi}\psi\,,$ (35)
where $m(y)=m\frac{\sigma^{\prime}(y)}{k}$ with
$\sigma^{\prime}(y)=\frac{d\sigma}{dy}$. With the help of the given mass term
the localized zero mode leptons are obtained. We present the construction of
the zero mode and KK mode leptons in the Appendix C extensively.
For the effective Yukawa coupling $\xi^{E}_{ij}$, similar to the previous
case, we integrate out the Yukawa interaction eq.(34) over the fifth
dimension. By taking the zero mode lepton doublets, singlets (see eq. (65))
and neutral Higgs fields $S=h^{0},A^{0}$ (see eq.(12)), we get
$\displaystyle\xi^{E}_{ij}\,((\xi^{E}_{ij})^{\dagger})=V^{00}_{S\,RL(LR)\,ij}\,\xi^{E}_{5\,ij}\,((\xi^{E}_{5\,ij})^{\dagger})\,,$
(36)
where
$\displaystyle V^{00}_{S\,RL\,ij}$ $\displaystyle=$
$\displaystyle\frac{1}{2\,\pi\,R}\,\int_{-\pi\,R}^{\pi\,R}\,dy\,\chi_{iR0}(y)\,\chi_{jL0}(y)\,f_{0}(y)$
(37) $\displaystyle=$
$\displaystyle\frac{\Big{(}1-e^{(b-r_{iR}-r_{jL})\,k\,\pi\,R}\Big{)}}{(r_{iR}+r_{jL}-b)\,\sqrt{\frac{1-e^{(1-2\,r_{iR})\,k\,\pi\,R}}{(2\,r_{iR}-1)}}\,\sqrt{\frac{1-e^{(1-2\,r_{jL})\,k\,\pi\,R}}{(2\,r_{jL}-1)}}\,\sqrt{\frac{1-e^{2\,(b-1)\,k\,\pi\,R}}{k\,(1-b)}}}\,.$
Here, similar to the previous scenario, the coupling $\xi^{E}_{5\,ij}$ in five
dimension is flavor dependent and it is regulated in such a way that the
overall quantity $V^{0}_{RL\,(LR)\,ij}\,\xi^{E}_{5\,ij}$ is pointed to the
chosen numerical value of
$\xi^{E}_{ij}\,\Big{(}(\xi^{E}_{ij})^{\dagger}\Big{)}$. This is the case that
the hierarchy of new Yukawa couplings, describing the tree level Higgs zero
mode($S^{(0)}$)-lepton zero mode ($l^{(0)}$)-lepton zero mode ($l^{(0)}$)
interaction, is not related to the Higgs field and lepton field locations.
The effective Yukawa coupling which drives the tree level KK mode Higgs
($S^{(n)}$)-$l^{(0)}$-$l^{(0)}$ interaction is
$\displaystyle\xi^{E\,n\,0}_{ij}\,((\xi^{E\,n\,0}_{ij})^{\dagger})=V^{n0}_{S\,RL(LR)\,ij}\,\xi^{E}_{5\,ij}\,((\xi^{E}_{5\,ij})^{\dagger})\,,$
(38)
where
$\displaystyle V^{n0}_{S\,RL\,ij}$ $\displaystyle=$
$\displaystyle\frac{\int_{-\pi\,R}^{\pi\,R}\,dy\,e^{(2-r_{iR}-r_{jL})\,\sigma}\,\Big{(}J_{b-2}(e^{\sigma}\,x_{nS})+\alpha_{n}\,Y_{b-2}(e^{\sigma}\,x_{nS})\Big{)}}{N_{S\,n}\,\sqrt{\frac{1-e^{(1-2\,r_{iR})\,k\,\pi\,R}}{k\,(2\,r_{iR}-1)}}\,\sqrt{\frac{1-e^{(1-2\,r_{jL})\,k\,\pi\,R}}{k\,(2\,r_{jL}-1)}}}\,.$
(39)
Using the eqs. (38) and (39), the effective Yukawa coupling
$\xi^{E\,n\,0}_{ij}$ is obtained as
$\displaystyle\xi^{E\,n\,0}_{ij}$ $\displaystyle=$
$\displaystyle\frac{V^{n0}_{S\,RL\,ij}}{V^{00}_{S\,RL\,ij}}\,\xi^{E}_{ij}=\frac{(r_{iR}+r_{jL}-b)\,\sqrt{\frac{k\,(e^{2\,(b-1)\,k\,\pi\,R}-1)}{(b-1)}}}{N_{S\,n}\,\Big{(}1-e^{(b-r_{iR}-r_{jL})\,k\,\pi\,R}\Big{)}}$
(40) $\displaystyle\times$
$\displaystyle\int_{-\pi\,R}^{\pi\,R}\,dy\,e^{(2-r_{iR}-r_{jL})\,\sigma}\,\Big{(}J_{b-2}(e^{\sigma}\,x_{nS})+\alpha_{n}\,Y_{b-2}(e^{\sigma}\,x_{nS})\Big{)}\,\xi^{E}_{ij}\,.$
$S^{(0)}-l^{(0)}-l^{(n)}$ vertex drives another possible tree level
interaction appearing in the loop calculations and the corresponding is
effective Yukawa coupling reads
$\displaystyle\xi^{E\,0\,n}_{ij}\,\Big{(}(\xi^{E\,0\,n}_{ij})^{\dagger}\Big{)}=V^{0n}_{S\,RL\,(LR)\,ij}\,\xi^{E}_{5\,ij}\,\Big{(}(\xi^{E}_{5\,ij})^{\dagger}\Big{)}\,,$
(41)
where
$\displaystyle
V^{0n}_{S\,RL\,(LR)\,ij}=\frac{N_{Ln\,(Rn)}\,\int_{-\pi\,R}^{\pi\,R}\,dy\,e^{(b-r_{iR\,(iL)}+\frac{1}{2})\,\sigma}\,\Bigg{(}J_{\frac{1}{2}\mp
r_{jL\,(jR)}}(e^{\sigma}\,x_{nL(R)})+c_{L\,(R)}\,Y_{\frac{1}{2}\mp
r_{jL\,(jR)}}(e^{\sigma}\,x_{nL(R)})\Bigg{)}}{\pi\,R\,\sqrt{\frac{1-e^{(1-2\,r_{iR\,(iL)})\,k\,\pi\,R}}{k\,\pi\,R\,(2\,r_{iR\,(iL)}-1)}}\,\sqrt{\frac{e^{2\,(b-1)\,k\,\pi\,R}-1}{(b-1)\,k}}}\,.$
(42)
Here the parameters $x_{nR(L)}$, $c_{R}(L)$, the lepton localization
parameters $r_{iR\,(iL)}$ and the normalization constant $N_{R(L)n}$ are given
in Appendix C. By using the eqs. (41) and (42) we get the effective Yukawa
coupling $\xi^{E\,0\,n}_{ij}\,\Big{(}(\xi^{E\,0\,n}_{ij})^{\dagger}\Big{)}$ as
$\displaystyle\xi^{E\,0\,n}_{ij}\\!\\!\\!\\!\\!\\!\\!\\!\\!\\!\\!\\!\\!\\!\\!$
$\displaystyle\Big{(}(\xi^{E\,0\,n}_{ij})^{\dagger}\Big{)}=\frac{V^{0n}_{S\,RL\,(LR)\,ij}}{V^{00}_{S\,RL(LR)\,ij}}\,\xi^{E}_{ij}\,\Big{(}(\xi^{E}_{ij})^{\dagger}\Big{)}=N_{Ln\,(Rn)}\,\,\sqrt{\frac{k\,(1-e^{(1-2\,r_{jL\,(jR)})\,k\,\pi\,R})}{\pi\,R\,(2\,r_{jL\,(iR)}-1)}}(r_{iR\,(iL)}+r_{jL\,(jR)}-b)$
$\displaystyle\times$
$\displaystyle\frac{\int_{-\pi\,R}^{\pi\,R}\,dy\,e^{(b-r_{iR\,(iL)}+\frac{1}{2})\,\sigma}\,\Bigg{(}J_{\frac{1}{2}\mp
r_{jL\,(jR)}}(e^{\sigma}\,x_{nL(R)})+c_{L\,(R)}\,Y_{\frac{1}{2}\mp
r_{jL\,(jR)}}(e^{\sigma}\,x_{nL(R)})\Bigg{)}}{\Big{(}1-e^{(b-r_{iR\,(iL)}-r_{jL\,(jR)})\,k\,\pi\,R}\Big{)}}\,\xi^{E}_{ij}\,\Big{(}(\xi^{E}_{ij})^{\dagger}\Big{)}\,.$
Finally, the tree level $S^{(m)}-l^{(0)}-l^{(n)}$ interaction is carried by
the effective Yukawa coupling $\xi^{E\,m\,n}_{ij}$ and it reads
$\displaystyle\xi^{E\,m\,n}_{ij}\,\Big{(}(\xi^{E\,m\,n}_{ij})^{\dagger}\Big{)}=V^{mn}_{S\,RL\,(LR)\,ij}\,\xi^{E}_{5\,ij}\,\Big{(}(\xi^{E}_{5\,ij})^{\dagger}\Big{)}\,,$
(44)
with
$\displaystyle
V^{mn}_{S\,RL\,(LR)\,ij}=\frac{N_{Ln\,(Rn)}}{N_{S\,m}\,\pi\,R\,\sqrt{\frac{1-e^{(1-2\,r_{iR\,(iL)})\,k\,\pi\,R}}{k\,\pi\,R\,(2\,r_{iR\,(iL)}-1)}}}\\!\\!\\!\\!\\!\\!\\!\\!\\!\\!\\!\\!$
$\displaystyle\int_{-\pi\,R}^{\pi\,R}\,dy\,e^{(\frac{5}{2}-r_{iR\,(iL)})\,\sigma}\,\Bigg{(}J_{b-2}(e^{\sigma}\,x_{mS})+\alpha_{n}\,Y_{b-2}(e^{\sigma}\,x_{mS})\Bigg{)}$
$\displaystyle\times$ $\displaystyle\Bigg{(}J_{\frac{1}{2}\mp
r_{jL\,(jR)}}(e^{\sigma}\,x_{nL(R)})+c_{L\,(R)}\,Y_{\frac{1}{2}\mp
r_{jL\,(jR)}}(e^{\sigma}\,x_{nL(R)})\Bigg{)}\,,$
and, therefore, we get
$\displaystyle\xi^{E\,m\,n}_{ij}\\!\\!\\!\\!\\!\\!\\!\\!\\!\\!\\!\\!\\!\\!\\!$
$\displaystyle\Big{(}(\xi^{E\,m\,n}_{ij})^{\dagger}\Big{)}=\frac{V^{mn}_{S\,RL\,(LR)\,ij}}{V^{00}_{S\,RL(LR)\,ij}}\,\xi^{E}_{ij}\,\Big{(}(\xi^{E}_{ij})^{\dagger}\Big{)}$
$\displaystyle=$
$\displaystyle\frac{N_{Ln\,(Rn)}\,\sqrt{\frac{1-e^{(1-2\,r_{jL\,(jR)})\,k\,\pi\,R}}{\pi\,R\,(2\,r_{jL\,(jR)}-1)}}\,\sqrt{\frac{e^{2\,(b-1)\,k\,\pi\,R}-1}{(b-1)}}(r_{iR\,(iL)}+r_{jL\,(jR)}-b)}{N_{S\,m}\,\Big{(}1-e^{(b-r_{iR\,(iL)}-r_{jL\,(jR)})\,k\,\pi\,R}\Big{)}}$
$\displaystyle\times$
$\displaystyle\int_{-\pi\,R}^{\pi\,R}\,dy\,e^{(\frac{5}{2}-r_{iR\,(iL)})\,\sigma}\,\Bigg{(}J_{b-2}(e^{\sigma}\,x_{mS})+\alpha_{n}\,Y_{b-2}(e^{\sigma}\,x_{mS})\Bigg{)}$
$\displaystyle\times$ $\displaystyle\Bigg{(}J_{\frac{1}{2}\mp
r_{jL\,(jR)}}(e^{\sigma}\,x_{nL(R)})+c_{L\,(R)}\,Y_{\frac{1}{2}\mp
r_{jL\,(jR)}}(e^{\sigma}\,x_{nL(R)})\Bigg{)}\,\xi^{E}_{ij}\,\Big{(}(\xi^{E}_{ij})^{\dagger}\Big{)}\,.$
(46)
The decay widths of the LFV $\mu\rightarrow e\gamma$, $\tau\rightarrow
e\gamma$ and $\tau\rightarrow\mu\gamma$ decays are calculated by using on-
shell renormalization scheme (see the section 2.1) and we have
$\displaystyle\Gamma(l_{1}\rightarrow
l_{2}\gamma)=c_{1}(|A_{1}|^{2}+|A_{2}|^{2})\,,$ (47)
with
$\displaystyle A_{1}$ $\displaystyle=$ $\displaystyle
A^{0}_{1}+A^{S\,KK}_{1}+A^{l\,KK}_{1}+A^{S,\,l\,KK}_{1}\,,$ $\displaystyle
A_{2}$ $\displaystyle=$ $\displaystyle
A^{0}_{2}+A^{S\,KK}_{2}+A^{l\,KK}_{2}+A^{S,\,l\,KK}_{2}\,.$ (48)
Notice that we present the explicit expression of the amplitudes
$A^{0}_{1\,(2)}$, $A^{S\,KK}_{1\,(2)}$ and $A^{S,\,l\,KK}_{1\,(2)}$ in
Appendix B.
## 3 Discussion
The Yukawa interactions coming from lepton-lepton-$S$ vertices drive the
radiative LFV $l_{1}\rightarrow l_{2}\gamma$ decays666Here, we do not take
into account the internal neutrino mediation due to their weak contribution to
the $BRs$ of the processes we study and, therefore, we assume that the lepton
FV comes from the internal new neutral Higgs bosons, $h^{0}$ and $A^{0}$.
Notice that we ignored the possible restrictions coming from the hadronic
decays., and their strengths are regulated by the Yukawa couplings which are
free parameters of the model used. In the present work, we study these LFV
decays in the RS1 background and we assume that the new Higgs doublet and the
gauge fields are accessible to the extra dimension. Here, in order to obtain
zero mode Higgs doublet, one considers the boundary mass term (see eq.(9)) and
impose the fine tuning $b=2+\sqrt{4+a}$ (eq.(11)) of the parameters $b$ and
$a$ which regulates to the boundary and bulk mass terms. Finally, the zero
mode new Higgs doublet is obtained as an exponential function of the parameter
$b$ (eq.(12)) and it is highly localized around the visible brane. The choice
$b=2-\sqrt{4+a}$ is also possible for the non-vanishing zero mode. However, we
do not take this possibility into account because of the following reason: If
the fermions are localized on the 4D brane, the overall quantity
$V^{0}_{ij}\,\xi^{E}_{5\,ij}$ is fixed to the chosen numerical value of
$\xi^{E}_{ij}$ with the assumption that the coupling $\xi^{E}_{5\,ij}$ is
flavor dependent and appropriately regulated. For $b=2-\sqrt{4+a}$ the
coupling $V^{0}_{ij}$ is a number of orders smaller compared to
$b=2+\sqrt{4+a}$ and, since this term appear in the denominator of the
coupling $\xi^{E\,n}$ according to our definition, $\xi^{E\,n}$ exceeds the
range of perturbative calculation. Notice that the coupling $V^{S\,n}_{ij}$ is
not so much sensitive to the parameter $b$. In the case of bulk fermions, the
choice $b=2-\sqrt{4+a}$ results in extremely large coupling which breaks the
perturbative upper limit (negligible coupling which causes weak sensitivity to
KK mode contributions) for set II (set I) .
As a first attempt we assume that the leptons are restricted to the 4D brane.
In this case the contribution of the extra dimension is due to the new Higgs
KK modes which appear in the internal line of the loop with the modified
Yukawa couplings (eq.(31)). Second, we consider that the leptons are also
localized in the extra dimension. We follow the idea that the localization of
the lepton fields in the extra dimension occurs with the addition of a Dirac
mass term $m_{l}=r\sigma^{\prime}$ with $\sigma=k\,|y|$ (eq.(35)). In this
case, the right and left handed lepton zero modes (eq.(65)) are chosen to
locate at different positions in the extra dimension in order to explain
different flavor mass hierarchy. In the scenario we choose the contribution of
the extra dimension is coming from the new Higgs KK modes and the lepton KK
modes appearing in the internal line of the loop. The FV is carried by the new
Yukawa couplings which are fixed to an appropriate number, respecting the
current measurements and the location parameters of leptons are responsible
for the lepton mass hierarchy. This choice makes the constraints coming from
various LFV processes to be more relaxed. Here, we consider two different set
of locations of charged leptons in order to obtain the masses of different
flavors777The gauge sector is necessarily lives in the extra dimension and
their KK modes appear after the compactification of the extra dimension. The
different fermion locations can induce additional FCNC effects at tree level
due to the couplings of neutral gauge KK modes-leptons and they should be
suppressed even for low KK masses, by choosing the location parameters $r_{L}$
($r_{R}$) appropriately. In the set of location parameters we use (Table 2),
we verify the various experimental FCNC constraints with KK neutral gauge
boson masses as low as few TeVs (see the similar the set of location
parameters and the discussion given in [48, 49].). In the first set (Table 2),
we consider the left and right handed fields having the same location in the
extra dimension. In the second, we choose the left handed charged lepton
locations as the same for each flavor, and we estimate the right handed ones
by respecting the current charged lepton masses. For the second set, we
observe that the BRs of the decays under consideration enhance since the KK
mode couplings to the new Higgs scalars, which are highly localized near the
visible brane, become stronger if the left handed lepton field is near to this
brane. For the effective Yukawa couplings in four dimension we choose that
$\bar{\xi}^{E}_{N,ij},\,i,j=e,\mu$ are smaller compared to
$\bar{\xi}^{E}_{N,\tau\,i}\,i=e,\mu,\tau$, since latter ones contain heavy
flavor and we assume that, in four dimensions, the couplings
$\bar{\xi}^{E}_{N,ij}$ is symmetric with respect to the indices $i$ and $j$.
Furthermore, the curvature parameter $k$ and the compactification radius $R$
are among the free parameters of the theory. Here, we take $k\,R=10.83$ and
consider in the region $10^{17}\,(GeV)\leq k\leq 10^{18}\,(GeV)$ (see the
discussion in Appendix C and [40]). Throughout our calculations we use the
input values given in Table (1).
Parameter | Value
---|---
$m_{\mu}$ | $0.106$ (GeV)
$m_{\tau}$ | $1.78$ (GeV)
$m_{h^{0}}$ | $100$ (GeV)
$m_{A^{0}}$ | $200$ (GeV)
$G_{F}$ | $1.1663710^{-5}(GeV^{-2})$
Table 1: The values of the input parameters used in the numerical
calculations.
In the case that the leptons live on the 4D brane the contribution of the
Higgs boson KK modes is negligible for the decays under consideration The
weakness of the new contribution is due to the tiny ratio $z_{Sn}$ appearing
in the expression (eq.(55)) which represents the additional effects to the
amplitudes.
Now we analyze the case that the leptons are also accessible to the extra
dimension.
Fig.2 represents the parameter $k$ dependence of the BR of the LFV
$\mu\rightarrow e\gamma$ decay for $\bar{\xi}^{E}_{N,\tau e}=0.01\,GeV$,
$\bar{\xi}^{E}_{N,\tau\mu}=1.0\,GeV$. Here the solid (dashed, short dashed)
line represents the BR without KK modes of leptons and new Higgs bosons (with
KK modes of leptons and new Higgs bosons for lepton location set II, set I),
for $a=0.01$ and $0.1$888For $a=0.01$ and $a=0.1$ the curves almost coincide.
It is observed that the BR ($\mu\rightarrow e\gamma$) is of the order of
$10^{-11}$ without the internal lepton and new Higgs boson KK mode
contributions. The addition of these KK modes result in that the BR enhances
almost $2\times$ one order, for the lepton location set II, especially for the
small values of the parameter $k$. For the set I, the enhancement of the BR is
negligible. On the other hand, the BRs are weakly sensitive to the parameter
$a$ which plays a crucial role in the localization of new Higgs bosons. We
present the parameter $a$ dependence of the BR ($\mu\rightarrow e\gamma$) for
the lepton location set II, in Fig.3 for $k=10^{18}\,GeV$. This figure shows
that the enhancement of the BR ($\mu\rightarrow e\gamma$) is of the order of
$\sim 0.1\%$ in the range of $a$, $0.01\leq a\leq 1.0$. This is a negligible
enhancement which can not be determined. On the other hand the enhancement in
the case of set II is due to the fact that the left handed leptons (KK modes)
are near to the visible brane and their couplings to the new Higgs bosons,
which are localized near the visible brane, become stronger.
Fig.4 is devoted to the parameter $k$ dependence of the BR of the LFV
$\tau\rightarrow e\gamma$ decay for $\bar{\xi}^{E}_{N,\tau e}=0.1\,GeV$,
$\bar{\xi}^{E}_{N,\tau\tau}=50\,GeV$. Here the solid (dashed, short dashed)
line represents the BR without KK modes of leptons and new Higgs bosons (with
KK modes of leptons and new Higgs bosons for lepton location set II, set I),
for $a=0.01$ and $0.1$. This figure shows that the BR ($\tau\rightarrow
e\gamma$) is of the order of $10^{-12}$ without the internal lepton and new
Higgs boson KK mode contributions. The addition of these KK modes results in
that the BR enhances almost three orders for the small values of the parameter
$k$ and the lepton location set II. For the set I, the BR enhances to the
value almost two times larger compared to the one without KK modes. Fig.5
represents the parameter $a$ dependence of the BR ($\tau\rightarrow e\gamma$)
for $k=10^{17}\,GeV$, for the lepton location set II. It is observed that the
enhancement of the BR ($\tau\rightarrow e\gamma$) is greater than $\sim 2.0\%$
in the range of $a$, $0.01\leq a\leq 1.0$. Similar to the previous decay, this
is a small enhancement which can not be determined.
Fig.6 shows is the parameter $k$ dependence of the BR of the LFV
$\tau\rightarrow\mu\gamma$ decay for $\bar{\xi}^{E}_{N,\tau\mu}=1.0\,GeV$,
$\bar{\xi}^{E}_{N,\tau\tau}=50\,GeV$. Here the solid (dashed, short dashed)
line represents the BR without KK modes of leptons and new Higgs bosons (with
KK modes of leptons and new Higgs bosons for lepton location set II, set I),
for $a=0.01$ and $0.1$. The BR ($\tau\rightarrow\mu\gamma$) is at the order of
the magnitude of $10^{-10}$ without the internal lepton and new Higgs boson KK
mode contributions. The lepton and Higgs boson KK modes cause more than three
order enhancement in the BR for the small values of the parameter $k$ and for
the lepton location set II. For the set I, this enhancement is almost two
times of the BR without KK modes. In Fig.7 we present the parameter $a$
dependence of the BR ($\tau\rightarrow\mu\gamma$) for $k=10^{17}\,GeV$ and the
lepton location set II. We observe that the enhancement of the BR
($\tau\rightarrow\mu\gamma$) is more than $\sim 2.0\%$ in the range of $a$,
$0.01\leq a\leq 1.0$.
At this stage, we would like to summarize our results. For the brane leptons,
the contribution of the Higgs boson KK modes to the BRs of the radiative LFV
decays is too small to be detected. However, if one considers that the leptons
are also accessible to the extra dimension, there exists a considerable
enhancement in the BRs, especially for the small values of the parameter $k$.
This enhancement occurs for the lepton location set II and it is due to the
fact that the left handed leptons (KK modes), which are near to the visible
brane, have enhanced couplings to the new Higgs bosons, which are also
localized near the visible brane. Finally, we observe that, the BRs are weakly
sensitive to the parameter $a$ which regulates the localization of the new
Higgs doublet in the extra dimension. With the more accurate forthcoming
measurements of the BRs of the LFV decays it would be possible to test the
existence of the warped extra dimensions and, to get a considerable
information which fields are accessible to the extra dimension.
Appendix
## Appendix A The mass matrix of new Higgs boson
The Higgs potential which creates the masses of neutral CP even and CP odd
Higgs bosons, $S=h^{0}$, $A^{0}$ reads
$\displaystyle V_{m_{S}}$ $\displaystyle=$ $\displaystyle
c^{\prime}_{h}[Re(\phi_{1}^{+}\phi_{2})]^{2}+c^{\prime}_{A}[Im(\phi_{1}^{+}\phi_{2})]^{2}\,,$
(49)
where $\phi_{1}$ and $\phi_{2}$ are given in (eq. (23)). After the electrowek
breaking the SM Higgs acquires a vacuum expectation value (see (eq.(26)) and
the mass lagrangian of new CP even Higgs boson ($h^{0}$) 999The similar mass
lagrangian appears for the CP odd Higgs boson $A^{0}$ with the replacement
$c^{\prime}_{h}\rightarrow c^{\prime}_{A}$. becomes
$\displaystyle{\cal{L}}_{S}=\frac{1}{2}\,\sum_{n=1}^{\infty}\,m_{n\,S}^{2}\,(S^{(n)}(x))^{2}\,+c^{\prime}_{h}\,\frac{v^{2}}{2}\,\Bigg{(}S^{(0)}(x)+\sum_{n=1}^{\infty}\,S^{(n)}(x)\,\alpha_{n}\Bigg{)}^{2}\,,$
(50)
with $c_{h}=c^{\prime}_{h}\,f^{2}_{0}(\pi\,R)$,
$\alpha_{n}=\frac{f_{n}(\pi\,R)}{f_{0}(\pi\,R)}$ and $m_{S}^{2}=c_{h}\,v^{2}$,
$S=h^{0}$. By using the mass lagrangian the $S$ boson mass matrix is obtained
as (see [50] and [51] for boson mass matrix in the one and two non-universal
extra dimensions, [36] for $U(1)_{Y}$ gauge boson mass matrix.):
$M_{S}^{2}\,=\,\left(\begin{array}[]{cccc}m_{S}^{2}&\alpha_{1}\,m_{S}^{2}&\alpha_{2}\,m_{S}^{2}&\cdots\\\
\alpha_{1}\,m_{S}^{2}&m_{1\,S}^{2}+m_{S}^{2}\,\alpha_{1}^{2}&m_{S}^{2}\,\alpha_{1}\,\alpha_{2}&\cdots\\\
\alpha_{2}\,m_{S}^{2}&\,\alpha_{2}\,\alpha_{1}\,m_{S}^{2}&m_{2\,S}^{2}+m_{S}^{2}\,\alpha_{2}^{2}&\cdots\\\
\vdots&\vdots&\vdots&\ddots\end{array}\right)\quad,$ (51)
and the determinant equation reads
$\det\Big{(}M_{S}^{2}-\lambda\,I\Big{)}=\Big{(}\prod_{n=1}^{\infty}(m_{n\,S}^{2}-\lambda)\Big{)}\,\Big{(}m_{S}^{2}-\lambda-\lambda\,m_{S}^{2}\sum^{\infty}_{n=1}\,\frac{\alpha_{n}}{m_{n\,S}^{2}-\lambda}\Big{)}=0\,.$
(52)
This equation is used to calculate the physical masses of zero and KK modes of
$S$ bosons and their eigenstates. To leading order, the physical $S$ boson
mass (the zero mode one) reads
$(m^{phys}_{S})^{2}=m_{S}^{2}\,\Bigg{(}1+\sum_{n=1}^{\infty}\,\frac{m_{S}^{2}\,\alpha_{n}^{2}}{m_{n\,S}^{2}}\Bigg{)}\,,$
(53)
since $m_{n\,S}>>m_{S}$. Notice that in our numerical calculations we do not
take into account the additional effects coming from the mixing because they
are suppressed due to the fact that the KK mode masses are considerably larger
compared to the zero mode one.
## Appendix B The amplitudes appearing in the text
Here, we present the amplitudes which appear in the calculation of the decay
widths of the radiative decays under consideration.
In the case that the charged leptons are restricted to the brane, the
amplitudes $A^{0}_{1}$, $A^{0}_{2}$, $A^{S\,KK}_{1}$ and $A^{S\,KK}_{2}$ in
eq.(33) read
$\displaystyle A^{0}_{1}$ $\displaystyle=$ $\displaystyle
Q_{\tau}\frac{1}{4\,m_{\tau}^{2}}\Bigg{\\{}m_{l_{1}}\,\bar{\xi}^{E}_{N,\tau
l_{2}}\,\bar{\xi}^{E}_{N,\tau
l_{1}}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,\,x\,(x+y-1)\,(\frac{z_{h^{0}}}{L_{h^{0}}}+\frac{z_{A^{0}}}{L_{A^{0}}})$
$\displaystyle-$ $\displaystyle
m_{l_{2}}\,\bar{\xi}^{E}_{N,l_{2}\tau}\,\bar{\xi}^{E}_{N,l_{1}\tau}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,x\,y\,(\frac{z_{h^{0}}}{L_{h^{0}}}+\frac{z_{A^{0}}}{L_{A^{0}}})$
$\displaystyle+$ $\displaystyle m_{\tau}\,\bar{\xi}^{E}_{N,\tau
l_{2}}\,\bar{\xi}^{E}_{N,l_{1}\tau}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,(x-1)\,(\frac{z_{h^{0}}}{L_{h^{0}}}-\frac{z_{A^{0}}}{L_{A^{0}}})\Bigg{\\}}\,\,,$
$\displaystyle A^{0}_{2}$ $\displaystyle=$ $\displaystyle
Q_{\tau}\frac{1}{4\,m_{\tau}^{2}}\Bigg{\\{}-m_{l_{1}}\,\bar{\xi}^{E}_{N,l_{2}\tau}\,\bar{\xi}^{E}_{N,l_{1}\tau}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,\,x\,(x+y-1)\,(\frac{z_{h^{0}}}{L_{h^{0}}}+\frac{z_{A^{0}}}{L_{A^{0}}})$
(54) $\displaystyle+$ $\displaystyle m_{l_{2}}\,\bar{\xi}^{E}_{N,\tau
l_{2}}\,\bar{\xi}^{E}_{N,\tau
l_{1}}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,x\,y\,(\frac{z_{h^{0}}}{L_{h^{0}}}+\frac{z_{A^{0}}}{L_{A^{0}}})$
$\displaystyle-$ $\displaystyle
m_{\tau}\,\bar{\xi}^{E}_{N,l_{2}\tau}\,\bar{\xi}^{E}_{N,\tau
l_{1}}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,(x-1)\,(\frac{z_{h^{0}}}{L_{h^{0}}}-\frac{z_{A^{0}}}{L_{A^{0}}})\Bigg{\\}}\,\,,$
$\displaystyle A^{S\,KK}_{1}$ $\displaystyle=$ $\displaystyle
Q_{\tau}\frac{1}{4\,m_{\tau}^{2}}\,\sum_{n=1}^{\infty}\,\Bigg{\\{}m_{l_{1}}\,\bar{\xi}^{E\,n}_{N,\tau
l_{2}}\,\bar{\xi}^{E\,n}_{N,\tau
l_{1}}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,\,x\,(x+y-1)\,(\frac{z_{h^{0}\,n}}{L_{h^{0}\,n}}+\frac{z_{A^{0}\,n}}{L_{A^{0}\,n}})$
$\displaystyle-$ $\displaystyle
m_{l_{2}}\,\bar{\xi}^{E\,n}_{N,l_{2}\tau}\,\bar{\xi}^{E\,n}_{N,l_{1}\tau}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,x\,y\,(\frac{z_{h^{0}\,n}}{L_{h^{0}\,n}}+\frac{z_{A^{0}\,n}}{L_{A^{0}\,n}})$
$\displaystyle+$ $\displaystyle m_{\tau}\,\bar{\xi}^{E\,n}_{N,\tau
l_{2}}\,\bar{\xi}^{E\,n}_{N,l_{1}\tau}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,(x-1)\,(\frac{z_{h^{0}\,n}}{L_{h^{0}\,n}}-\frac{z_{A^{0}\,n}}{L_{A^{0}\,n}})\Bigg{\\}}\,\,,$
$\displaystyle A^{S\,KK}_{2}$ $\displaystyle=$ $\displaystyle
Q_{\tau}\frac{1}{4\,m_{\tau}^{2}}\,\sum_{n=1}^{\infty}\,\Bigg{\\{}-m_{l_{1}}\,\bar{\xi}^{E\,n}_{N,l_{2}\tau}\,\bar{\xi}^{E\,n}_{N,l_{1}\tau}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,\,x\,(x+y-1)\,(\frac{z_{h^{0}\,n}}{L_{h^{0}\,n}}+\frac{z_{A^{0}\,n}}{L_{A^{0}\,n}})$
(55) $\displaystyle+$ $\displaystyle m_{l_{2}}\,\bar{\xi}^{E\,n}_{N,\tau
l_{2}}\,\bar{\xi}^{E\,n}_{N,\tau
l_{1}}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,x\,y\,(\frac{z_{h^{0}\,n}}{L_{h^{0}\,n}}+\frac{z_{A^{0}\,n}}{L_{A^{0}\,n}})$
$\displaystyle-$ $\displaystyle
m_{\tau}\,\bar{\xi}^{E\,n}_{N,l_{2}\tau}\,\bar{\xi}^{E\,n}_{N,\tau
l_{1}}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,(x-1)\,(\frac{z_{h^{0}\,n}}{L_{h^{0}\,n}}-\frac{z_{A^{0}\,n}}{L_{A^{0}\,n}})\Bigg{\\}}\,\,,$
where
$\displaystyle L_{S}$ $\displaystyle=$ $\displaystyle
z_{S}+x^{2}\,z_{S}+x\Big{(}1+(y-2)\,z_{S}\Big{)}\,,$ $\displaystyle L_{S\,n}$
$\displaystyle=$ $\displaystyle
z_{S\,n}+x^{2}\,z_{S\,n}+x\Big{(}1+(y-2)\,z_{S\,n}\Big{)}\,\,,$ (56)
with $z_{S}=\frac{m^{2}_{\tau}}{m^{2}_{S}}$,
$z_{S\,n}=\frac{m^{2}_{\tau}}{m^{2}_{n\,S}}$. Here ,
$l_{1}\,(l_{2})=\tau;\mu\,(\mu$ or $e;e)$,
$c_{1}=\frac{G_{F}^{2}\alpha_{em}m^{3}_{l_{1}}}{32\pi^{4}}$, $A_{1}$ ($A_{2}$)
is the left (right) chiral amplitude, $Q_{\tau}$ is the charge of tau lepton
and $m_{n\,S}$ is the internal Higgs KK mode mass (see eq.(15)). Notice that
we take the Yukawa couplings real.
If the charged leptons are also accessible to the extra dimension the
amplitudes $A^{0}_{1\,(2)}$, $A^{S\,KK}_{1\,(2)}$ and $A^{S,\,l\,KK}_{1\,(2)}$
(see eq.(48)) are
$\displaystyle A^{0}_{1}$ $\displaystyle=$ $\displaystyle
Q_{\tau}\frac{1}{4\,m_{\tau}^{2}}\Bigg{\\{}m_{l_{1}}\,\bar{\xi}^{E}_{N,\tau
l_{2}}\,\bar{\xi}^{E}_{N,\tau
l_{1}}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,\,x\,(x+y-1)\,(\frac{z_{h^{0}}}{L_{h^{0}}}+\frac{z_{A^{0}}}{L_{A^{0}}})$
$\displaystyle-$ $\displaystyle
m_{l_{2}}\,\bar{\xi}^{E}_{N,l_{2}\tau}\,\bar{\xi}^{E}_{N,l_{1}\tau}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,x\,y\,(\frac{z_{h^{0}}}{L_{h^{0}}}+\frac{z_{A^{0}}}{L_{A^{0}}})$
$\displaystyle+$ $\displaystyle m_{\tau}\,\bar{\xi}^{E}_{N,\tau
l_{2}}\,\bar{\xi}^{E}_{N,l_{1}\tau}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,(x-1)\,(\frac{z_{h^{0}}}{L_{h^{0}}}-\frac{z_{A^{0}}}{L_{A^{0}}})\Bigg{\\}}\,\,,$
$\displaystyle A^{0}_{2}$ $\displaystyle=$ $\displaystyle
Q_{\tau}\frac{1}{4\,m_{\tau}^{2}}\Bigg{\\{}-m_{l_{1}}\,\bar{\xi}^{E}_{N,l_{2}\tau}\,\bar{\xi}^{E}_{N,l_{1}\tau}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,\,x\,(x+y-1)\,(\frac{z_{h^{0}}}{L_{h^{0}}}+\frac{z_{A^{0}}}{L_{A^{0}}})$
(57) $\displaystyle+$ $\displaystyle m_{l_{2}}\,\bar{\xi}^{E}_{N,\tau
l_{2}}\,\bar{\xi}^{E}_{N,\tau
l_{1}}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,x\,y\,(\frac{z_{h^{0}}}{L_{h^{0}}}+\frac{z_{A^{0}}}{L_{A^{0}}})$
$\displaystyle-$ $\displaystyle
m_{\tau}\,\bar{\xi}^{E}_{N,l_{2}\tau}\,\bar{\xi}^{E}_{N,\tau
l_{1}}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,(x-1)\,(\frac{z_{h^{0}}}{L_{h^{0}}}-\frac{z_{A^{0}}}{L_{A^{0}}})\Bigg{\\}}\,\,,$
$\displaystyle A^{S\,KK}_{1}$ $\displaystyle=$ $\displaystyle
Q_{\tau}\frac{1}{4\,m_{\tau}^{2}}\,\sum_{n=1}^{\infty}\,\Bigg{\\{}m_{l_{1}}\,\bar{\xi}^{E\,n0}_{N,\tau
l_{2}}\,\bar{\xi}^{E\,n0}_{N,\tau
l_{1}}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,\,x\,(x+y-1)\,(\frac{z_{h^{0}\,n}}{L_{h^{0}\,n}}+\frac{z_{A^{0}\,n}}{L_{A^{0}\,n}})$
$\displaystyle-$ $\displaystyle
m_{l_{2}}\,\bar{\xi}^{E\,n0}_{N,l_{2}\tau}\,\bar{\xi}^{E\,n0}_{N,l_{1}\tau}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,x\,y\,(\frac{z_{h^{0}\,n}}{L_{h^{0}\,n}}+\frac{z_{A^{0}\,n}}{L_{A^{0}\,n}})$
$\displaystyle+$ $\displaystyle m_{\tau}\,\bar{\xi}^{E\,n0}_{N,\tau
l_{2}}\,\bar{\xi}^{E\,n0}_{N,l_{1}\tau}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,(x-1)\,(\frac{z_{h^{0}\,n}}{L_{h^{0}\,n}}-\frac{z_{A^{0}\,n}}{L_{A^{0}\,n}})\Bigg{\\}}\,\,,$
$\displaystyle A^{S\,KK}_{2}$ $\displaystyle=$ $\displaystyle
Q_{\tau}\frac{1}{4\,m_{\tau}^{2}}\,\sum_{n=1}^{\infty}\,\Bigg{\\{}-m_{l_{1}}\,\bar{\xi}^{E\,n0}_{N,l_{2}\tau}\,\bar{\xi}^{E\,n0}_{N,l_{1}\tau}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,\,x\,(x+y-1)\,(\frac{z_{h^{0}\,n}}{L_{h^{0}\,n}}+\frac{z_{A^{0}\,n}}{L_{A^{0}\,n}})$
(58) $\displaystyle+$ $\displaystyle m_{l_{2}}\,\bar{\xi}^{E\,n0}_{N,\tau
l_{2}}\,\bar{\xi}^{E\,n0}_{N,\tau
l_{1}}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,x\,y\,(\frac{z_{h^{0}\,n}}{L_{h^{0}\,n}}+\frac{z_{A^{0}\,n}}{L_{A^{0}\,n}})$
$\displaystyle-$ $\displaystyle
m_{\tau}\,\bar{\xi}^{E\,n0}_{N,l_{2}\tau}\,\bar{\xi}^{E\,n0}_{N,\tau
l_{1}}\,\int_{0}^{1}\,dx\,\int_{0}^{1-x}\,dy\,(x-1)\,(\frac{z_{h^{0}\,n}}{L_{h^{0}\,n}}-\frac{z_{A^{0}\,n}}{L_{A^{0}\,n}})\Bigg{\\}}\,\,,$
$\displaystyle A^{l\,KK}_{1}$ $\displaystyle=$
$\displaystyle\frac{Q_{\tau}}{48\,m_{\tau}^{2}}\,\sum_{n=1}^{\infty}\,\Bigg{\\{}\frac{m_{\tau}^{2}}{m^{2}_{nR}}\,\,m_{l_{1}}\,(\bar{\xi}^{E\,0n}_{N,l_{2}\tau})^{\dagger}\,(\bar{\xi}^{E\,0n}_{N,l_{1}\tau})^{\dagger}\,\Big{(}G(z_{nR,h^{0}})+G(z_{nR,A^{0}})\Big{)}$
$\displaystyle+$
$\displaystyle\frac{m_{\tau}^{2}}{m^{2}_{nL}}\,\,m_{l_{2}}\,\bar{\xi}^{E\,0n}_{N,l_{2}\tau}\,\bar{\xi}^{E\,0n}_{N,l_{1}\tau}\,\Big{(}G(z_{nL,h^{0}})+G(z_{nL,A^{0}})\Big{)}\Bigg{\\}}\,\,,$
$\displaystyle A^{l\,KK}_{2}$ $\displaystyle=$
$\displaystyle\frac{-Q_{\tau}}{48\,m_{\tau}^{2}}\,\sum_{n=1}^{\infty}\,\Bigg{\\{}\frac{m_{\tau}^{2}}{m^{2}_{nR}}\,m_{l_{2}}\,(\bar{\xi}^{E\,0n}_{N,l_{2}\tau})^{\dagger}\,(\bar{\xi}^{E\,0n}_{N,l_{1}\tau})^{\dagger}\,\Big{(}G(z_{nR,h^{0}})+G(z_{nR,A^{0}})\Big{)}$
(59) $\displaystyle+$
$\displaystyle\frac{m_{\tau}^{2}}{m^{2}_{nL}}\,m_{l_{1}}\,\bar{\xi}^{E\,0n}_{N,l_{2}\tau}\,\bar{\xi}^{E\,0n}_{N,l_{1}\tau}\,\Big{(}G(z_{nL,h^{0}})+G(z_{nL,A^{0}})\Big{)}\Bigg{\\}}\,\,,$
$\displaystyle A^{S,l\,KK}_{1}$ $\displaystyle=$
$\displaystyle\frac{Q_{\tau}}{48\,m_{\tau}^{2}}\,\sum_{n,m=1}^{\infty}\,\Bigg{\\{}\frac{m_{\tau}^{2}}{m^{2}_{nR}}\,\,m_{l_{1}}\,(\bar{\xi}^{E\,mn}_{N,l_{2}\tau})^{\dagger}\,(\bar{\xi}^{E\,mn}_{N,l_{1}\tau})^{\dagger}\,\Big{(}G(z_{nR,h^{0}\,m})+G(z_{nR,A^{0}\,m})\Big{)}$
$\displaystyle+$
$\displaystyle\frac{m_{\tau}^{2}}{m^{2}_{nL}}\,\,m_{l_{2}}\,\bar{\xi}^{E\,mn}_{N,l_{2}\tau}\,\bar{\xi}^{E\,mn}_{N,l_{1}\tau}\,\Big{(}G(z_{nL,h^{0}\,m})+G(z_{nL,A^{0}\,m})\Big{)}\Bigg{\\}}\,\,,$
$\displaystyle A^{S,l\,KK}_{2}$ $\displaystyle=$
$\displaystyle\frac{-Q_{\tau}}{48\,m_{\tau}^{2}}\,\sum_{n,m=1}^{\infty}\,\Bigg{\\{}\frac{m_{\tau}^{2}}{m^{2}_{nR}}\,m_{l_{2}}\,(\bar{\xi}^{E\,mn}_{N,l_{2}\tau})^{\dagger}\,(\bar{\xi}^{E\,mn}_{N,l_{1}\tau})^{\dagger}\,\Big{(}G(z_{nR,h^{0}\,m})+G(z_{nR,A^{0}\,m})\Big{)}$
(60) $\displaystyle+$
$\displaystyle\frac{m_{\tau}^{2}}{m^{2}_{nL}}\,m_{l_{1}}\,\bar{\xi}^{E\,mn}_{N,l_{2}\tau}\,\bar{\xi}^{E\,mn}_{N,l_{1}\tau}\,\Big{(}G(z_{nL,h^{0}\,m})+G(z_{nL,A^{0}\,m})\Big{)}\Bigg{\\}}\,.$
Here $A_{1}$ ($A_{2}$) is the left (right) chiral amplitude,
$l_{1}\,(l_{2})=\tau;\mu\,(\mu$ or $e;e)$, the functions $F(w)$, $G(w)$ are
$\displaystyle F(w)$ $\displaystyle=$
$\displaystyle\frac{w\,(3-4\,w+w^{2}+2\,ln\,w)}{(1-w)^{3}}\,,$ $\displaystyle
G(w)$ $\displaystyle=$
$\displaystyle-\frac{w\,(2+3\,w-6\,w^{2}+w^{3}+6\,w\,ln\,w)}{(1-w)^{4}}\,\,,$
(61)
$c_{1}=\frac{G_{F}^{2}\alpha_{em}m^{3}_{l_{1}}}{32\pi^{4}}$, ,
$z_{S}=\frac{m^{2}_{\tau}}{m^{2}_{S}}$,
$z_{S\,n}=\frac{m^{2}_{\tau}}{m^{2}_{n\,S}}$,
$z_{nL(nR),S}=\frac{m^{2}_{nL\,(nR)}}{m^{2}_{S}}$,
$z_{nL(nR),S\,m}=\frac{m^{2}_{nL\,(nR)}}{m^{2}_{m\,S}}$ with left (right)
handed internal lepton KK mode mass $m_{nL\,(nR)}$ (eq.(76)). In eqs. (57) and
(58) the functions $L_{S}$ and $L_{S\,n}$ are given in (eq.(56)). In the case
that the incoming and outgoing lepton masses are ignored in the functions
$L_{S}$ and $L_{S\,n}$ one gets the integrated form of $A^{0}_{1(2)}$ as
$\displaystyle A^{0}_{1}$ $\displaystyle=$ $\displaystyle
Q_{\tau}\frac{1}{48\,m_{\tau}^{2}}\Bigg{\\{}6\,m_{\tau}\,\bar{\xi}^{E}_{N,\tau
l_{2}}\,\bar{\xi}^{E}_{N,l_{1}\tau}\,\Big{(}F(z_{h^{0}})-F(z_{A^{0}})\Big{)}+m_{l_{1}}\,\bar{\xi}^{E}_{N,\tau
l_{2}}\,\bar{\xi}^{E}_{N,\tau
l_{1}}\,\Big{(}G(z_{h^{0}})+G(z_{A^{0}})\Big{)}\Bigg{\\}}\,\,,$ $\displaystyle
A^{0}_{2}$ $\displaystyle=$ $\displaystyle-
Q_{\tau}\frac{1}{48\,m_{\tau}^{2}}\Bigg{\\{}6\,m_{\tau}\,\bar{\xi}^{E}_{N,l_{2}\tau}\,\bar{\xi}^{E}_{N,\tau
l_{1}}\,\Big{(}F(z_{h^{0}})-F(z_{A^{0}})\Big{)}+m_{l_{1}}\,\bar{\xi}^{E}_{N,l_{2}\tau}\,\bar{\xi}^{E}_{N,l_{1}\tau}\,\Big{(}G(z_{h^{0}})+G(z_{A^{0}})\Big{)}\Bigg{\\}}\,.$
Notice that, for the amplitudes $A^{l\,KK}_{1(2)}$ and $A^{S,l\,KK}_{1(2)}$,
the incoming and outgoing lepton masses are ignored in the functions $L_{S}$
and $L_{S\,n}$ since the internal KK leptons are heavy.
## Appendix C The construction of zero mode and KK mode leptons
This Appendix is devoted to the construction of the zero mode and KK mode
leptons in the case that the leptons are localized in the extra dimension with
the help of the Dirac mass term given in (eq.(35)). We start with the
expansion of the bulk fermion as
$\displaystyle\psi(x^{\mu},y)=\frac{1}{\sqrt{2\,\pi\,R}}\,\sum_{n=0}^{\infty}\,\psi^{(n)}(x^{\mu})\,e^{2\,\sigma}\,\chi_{n}(y)\,.$
(63)
By using the normalization
$\displaystyle\frac{1}{2\,\pi\,R}\,\int_{-\pi\,R}^{\pi\,R}\,dy\,e^{\sigma}\,\chi_{n}(y)\,\chi_{m}(y)=\delta_{nm}\,,$
(64)
and the Dirac equation the zero mode fermion is obtained as
$\displaystyle\chi_{0}(y)=N_{0}\,e^{-r\,\sigma}\,,$ (65)
where $r=m/k$ and $N_{0}$ is the normalization constant:
$\displaystyle
N_{0}=\sqrt{\frac{k\,\pi\,R\,(1-2\,r)}{e^{k\,\pi\,R\,(1-2\,r)}-1}}\,.$ (66)
The appropriately normalized solution
$\displaystyle\chi^{\prime}_{0}(y)=e^{-\frac{\sigma}{2}}\,\chi_{0}(y)$ (67)
is localized in the extra dimension where the localization is regulated by the
parameter $r$. For $r>\frac{1}{2}$ ($r<\frac{1}{2}$) this solution is
localized near the hidden (visible) brane and it has a constant profile for
$r=\frac{1}{2}$.
Now, we are ready construct the SM leptons. What we need is to consider
$SU(2)_{L}$ doublet $\psi_{L}$ and singlet $\psi_{R}$ with separate $Z_{2}$
projection conditions: $Z_{2}\psi_{R}=\gamma_{5}\psi_{R}$ and
$Z_{2}\psi_{L}=-\gamma_{5}\psi_{L}$ (see for example [32]). Finally we get the
leptons accessible to the extra dimension as
$\displaystyle l_{iL}(x^{\mu},y)$ $\displaystyle=$
$\displaystyle\frac{1}{\sqrt{2\,\pi\,R}}\,e^{2\,\sigma}\,l_{iL}^{(0)}(x^{\mu})\,\chi_{i\,L0}(y)$
$\displaystyle+$
$\displaystyle\frac{1}{\sqrt{2\,\pi\,R}}\,\sum_{n=1}^{\infty}\,e^{2\,\sigma}\,\Bigg{(}l_{iL}^{(n)}(x^{\mu})\,\chi^{l}_{i\,Ln}(y)+l_{iR}^{(n)}(x^{\mu})\,\chi^{l}_{i\,Rn}(y)\Bigg{)}\,,$
$\displaystyle E_{jR}(x^{\mu},y)$ $\displaystyle=$
$\displaystyle\frac{1}{\sqrt{2\,\pi\,R}}\,e^{2\,\sigma}\,E_{jR}^{(0)}(x^{\mu})\,\chi_{j\,R0}(y)$
(68) $\displaystyle+$
$\displaystyle\frac{1}{\sqrt{2\,\pi\,R}}\,\sum_{n=0}^{\infty}\,e^{2\,\sigma}\,\Bigg{(}E_{jR}^{(n)}(x^{\mu})\,\chi^{E}_{j\,Rn}(y)+E_{jL}^{(n)}(x^{\mu})\,\chi^{E}_{j\,Ln}(y)\Bigg{)}\,.$
Here the zero mode leptons $\chi_{i\,L0}(y)$ and $\chi_{j\,R0}(y)$ are given
in eq.(65) with the replacements $r\rightarrow r_{iL}$ and $r\rightarrow
r_{jR}$, respectively.
The zero mode fermions can get mass through the $Z_{2}$ invariant left handed
fermion-right handed fermion-Higgs interaction,
$\bar{\psi}_{R}\,\psi_{L}\,H$101010Here, we consider different location
parameters $r$ for each left handed and right handed part of different
flavors. The location parameters for fermion fields are chosen so that this
interaction creates the current masses of fermions.. If the SM Higgs field
lives on the visible brane as in our choice, the masses of fermions are
calculated by using the integral
$\displaystyle
m_{i}=\frac{1}{2\,\pi\,R}\,\int_{-\pi\,R}^{\pi\,R}\,dy\,\lambda_{5}\,\chi_{iL0}(y)\,\chi_{iR0}(y)\,<H>\,\delta(y-\pi\,R)\,,$
(69)
where $\lambda_{5}$ is the coupling in five dimensions and it can be
parametrized in terms of the one in four dimensions, the dimensionless
coupling $\lambda$, $\lambda_{5}=\lambda/\sqrt{k}$. Here the expectation value
of the Higgs field $<H>$ reads $<H>=v/\sqrt{k}$ where $v$ is the vacuum
expectation value111111We take $v=0.043\,M_{Pl}$ to provide the measured gauge
boson masses [40] and choose $k\,R=10.83$ in order to get the correct
effective scale on the visible brane, i.e., $M_{W}=e^{-\pi\,k\,R}\,M_{pl}$ is
of the order of TeV.. Now, we choose two different sets of location of charged
lepton fields in order to obtain the masses of different flavors.
| SET I | SET II
---|---|---
| $r_{L}$ $r_{R}$ | $r_{L}$ $r_{R}$
e | 0.6710 0.6710 | -0.4900 0.8800
$\mu$ | 0.5826 0.5826 | -0.4900 0.7160
$\tau$ | 0.5273 0.5273 | -0.4900 0.6249
Table 2: Two possible locations of charged lepton fields. Here $r_{L}$ and
$r_{R}$ are left handed and right handed lepton field location parameters,
respectively.
In Set I, the left and right handed fields of the same flavor have the same
location, however, in set II, we choose the left handed charged lepton
locations the same for each flavor. For both cases we estimate the left-right
handed charged lepton locations by respecting their current masses.
The $Z_{2}$ the projection condition $Z_{2}\psi=-\gamma_{5}\psi$, used in
constructing the left handed fields on the branes, results in that the left
handed zero mode appears, the left (right) handed KK modes appear (disappear)
on the branes, with the boundary conditions due to the Dirac mass term in the
action eq.(35):
$\displaystyle\Big{(}\frac{d}{dy}-m\Big{)}\,\chi^{l}_{iLn}(y_{0})=0\,,$
$\displaystyle\chi^{l}_{iRn}(y_{0})=0\,,$ (70)
where $y_{0}=0$ or $\pi\,R$. Using the Dirac equation for KK mode leptons one
gets the left handed lepton $\chi^{l}_{i\,Ln}(y)$ that lives on the visible
brane as
$\displaystyle\chi^{l}_{iLn}(y)=N_{Ln}\,e^{\sigma/2}\Bigg{(}J_{\frac{1}{2}-r_{iL}}(e^{\sigma}\,x_{nL})+c_{L}\,Y_{\frac{1}{2}-r_{iL}}(e^{\sigma}\,x_{nL})\Bigg{)}\,,$
(71)
with the constant
$\displaystyle
c_{L}=-\frac{J_{-r_{iL}-\frac{1}{2}}(x_{nL})}{Y_{-r_{iL}-\frac{1}{2}}(x_{nL})}\,.$
(72)
Here, $N_{Ln}$ is the normalization constant and $x_{nL}=\frac{m_{Ln}}{k}$.
The functions $J_{\beta}(w)$ and $Y_{\beta}(w)$ appearing in eq.(71) are the
Bessel function of the first kind and of the second kind, respectively. On the
other hand, the $Z_{2}$ projection condition $Z_{2}\psi=\gamma_{5}\psi$ is
used in order to construct the right handed fields on the branes and this
ensures that the right handed zero mode appears, the right (left) handed KK
modes appear (disappear) on the branes with the boundary conditions:
$\displaystyle\Big{(}\frac{d}{dy}+m\Big{)}\,\chi^{E}_{iRn}(y_{0})=0\,,$
$\displaystyle\chi^{E}_{iLn}(y_{0})=0\,.$ (73)
Again, using the Dirac equation for KK mode leptons, one gets the right handed
lepton $\chi^{E}_{i\,Rn}(y)$ that lives on the visible brane as
$\displaystyle\chi^{E}_{iRn}(y)=N_{Rn}\,e^{\sigma/2}\Bigg{(}J_{\frac{1}{2}+r_{iR}}(e^{\sigma}\,x_{nR})+c_{R}\,Y_{\frac{1}{2}+r_{iR}}(e^{\sigma}\,x_{nR})\Bigg{)}\,,$
(74)
with
$\displaystyle
c_{R}=-\frac{J_{r_{iR}-\frac{1}{2}}(x_{nR})}{Y_{r_{iR}-\frac{1}{2}}(x_{nR})}\,,$
(75)
where $N_{Rn}$ is the normalization constant and $x_{nR}=\frac{m_{Rn}}{k}$.
Notice that the constant $c_{L}$, the $n^{th}$ KK mode mass $m_{Ln}$ in
eq.(71) and the constant $c_{R}$, the $n^{th}$ KK mode mass $m_{Rn}$ in
eq.(74) are obtained by using the boundary conditions eq.(70) and eq.(73),
respectively. For $m_{L(R)n}\ll k$ and $kR\gg 1$ they are approximated as:
$\displaystyle m_{Ln}$ $\displaystyle\simeq$ $\displaystyle
k\,\pi\,\Big{(}n-\frac{\frac{1}{2}-r}{2}+\frac{1}{4}\Big{)}\,e^{-\pi\,k\,R}\,,$
$\displaystyle m_{Rn}$ $\displaystyle\simeq$ $\displaystyle
k\,\pi\,\Big{(}n-\frac{\frac{1}{2}+r}{2}+\frac{1}{4}\Big{)}\,e^{-\pi\,k\,R}\,\,\,\,\,\,\,\,\mbox{for
$r<0.5$}\,,$ $\displaystyle m_{Rn}$ $\displaystyle\simeq$ $\displaystyle
k\,\pi\,\Big{(}n+\frac{\frac{1}{2}+r}{2}-\frac{3}{4}\Big{)}\,e^{-\pi\,k\,R}\,\,\,\,\,\,\,\,\mbox{for
$r>0.5$}\,.$ (76)
## References
* [1] M. L. Brooks et. al., MEGA Collaboration, Phys. Rev. Lett. 83 1521, (1999).
* [2] K. Hayasaka et al.., Phys.Lett. D63 20, (2005).
* [3] S. Ahmed et.al., CLEO Collaboration, Phys. Rev. D61 071101, (2000).
* [4] J.M. Roney and the BABAR Collaboration, Nucl. Phys. Proc. Suppl. 144 155, (2005).
* [5] B. Aubert et. al., BABAR Collaboration, SLAC-PUB-11028, BABAR-PUB-04-049, Feb. 2005, 7. pp, Phys. Rev. Lett. 95 041802, (2005).
* [6] Donato Nicolo, MUEGAMMA Collaboration, Nucl. Instrum. Meth A503 287, (2003).
* [7] S. Yamada, Nucl. Phys. Proc. Suppl. 144 185, (2005).
* [8] T. D. Lee, Phys. Rev. D8 1226, (1973).
* [9] S. Glashow and S. Weinberg, Phys. Rev. D15 1958, (1977).
* [10] For a review see J. Gunion, H. Haber, G. Kane, and S. Dawson, The Higgs Hunter s Guide Addison-Wesley, New York, (1990).
* [11] D. Atwood, L. Reina and A. Soni, Phys. Rev. D55 3156, (1997).
* [12] E. O. Iltan, Phys. Rev. D64 115005, (2001).
* [13] E. O. Iltan,Phys. Rev. D64 013013, (2001)
* [14] R. Diaz, R. Martinez and J. A. Rodriguez, Phys. Rev. D63 095007, (2001).
* [15] E. O. Iltan, JHEP 0402 20, (2004).
* [16] E. O. Iltan, Int. J. Mod. Phys. A23 1055, (2008).
* [17] R. Diaz, R. Martinez, J. A. Rodriguez, Phys. Rev. D67 075011, (2003).
* [18] R. Barbieri and L. J. Hall, Phys. Lett. B338 212, (1994).
* [19] R. Barbieri, L. J. Hall and A. Strumia, Nucl. Phys. B445 219, (1995).
* [20] R. Barbieri, L. J. Hall and A. Strumia, Nucl. Phys. B449 437, (1995).
* [21] P. Ciafaloni, A. Romanino and A. Strumia, IFUP-YH-42-95.
* [22] T. V. Duong, B. Dutta and E. Keith, Phys. Lett. B378 128, (1996).
* [23] G. Couture, et. al., Eur. Phys. J. C7 135, (1999).
* [24] Y. Okada, K. Okumara and Y. Shimizu, Phys. Rev. D61 094001, (2000).
* [25] D. Chang, W. S. Hou and W. Y. Keung, Phys. Rev. D48 217, (1993).
* [26] P. Paradisi, JHEP 0602 050, (2006).
* [27] G. J. Ding, M. L. Yan, Phys. Rev. D77 014005, (2008).
* [28] A. Hektor, Y. Kajiyama, K. Kannike, hep-ph/0802.4015, (2008).
* [29] L. Randall, R.Sundrum, Phys. Rev. Lett. 83 3370, (1999).
* [30] L. Randall, R.Sundrum, Phys. Rev. Lett. 83 4690 (1999);
* [31] W. D. Goldberger, M. B. Wise, Phys. Rev. Lett. D 83 4922, (1999).
* [32] S. Chang, J. Hisano, H. Nakano, N. Okada, M. Yamaguchi, Phys. Rev. D62 084025, (2000).
* [33] A. Pomarol, Phys. Lett. B486 153, (2000).
* [34] H. Davoudias, J. L. Hewett, T. G. Rizzo, Phys. Lett. B473 43, (2000).
* [35] T. Gherghetta, A. Pomarol, Nucl. Phys. B586 141, (2000).
* [36] B. Batell, T. Gherghetta, Phys. Rev. D73, 045016 (2006).
* [37] S. J. Huber,C. A. Lee, Q. Shafi, Phys. Lett. B531 112, (2002).
* [38] Y. Grossman, M. Neubert, Phys. Lett. B474 361, (2000).
* [39] S. J. Huber, hep-ph/0211056, (2002).
* [40] S. J. Huber, Nucl. Phys. B666 269, (2003).
* [41] S. J. Huber, Q. Shafi, Phys. Rev. D63 045010, (2001).
* [42] I. I. Kogan, S. Mouslopolous, A. Papazoglou, G. G. Ross, Nucl. Phys. B615 191, (2001).
* [43] T. Gherghetta, A. Pomarol, Nucl. Phys. B602 3, (2001).
* [44] K. Ghoroku, A. Nakamura, Phys. Rev. D65 084017, (2002).
* [45] K. Agashe, G. Perez and A. Soni, Phys. Rev. D71 016002, (2005).
* [46] K.Agashe, A. E. Blechman and F. Petriello, Phys. Rev. D74 053011, (2006).
* [47] E. D. Pree, M. Sher., Phys. Rev. D73, 0955006 (2006).
* [48] G. Moreau, J. I. S. Marcos, JHEP 0603, 090 (2006)
* [49] F. Ledroit, G. Moreau, J. Morel, JHEP 09 071 (2007).
* [50] A. Muck, A. Pilaftsis, and R. Ruckl, Phys. Rev. D65, 085037 (2002).
* [51] E. Iltan, Eur. Phys. J. C41, 233 (2005).
Figure 1: One loop diagrams contribute to $l_{1}\rightarrow l_{2}\gamma$ decay
due to the zero mode (KK mode) leptons and Higgs fields in the 2HDM.
Figure 2: $k$ dependence of the BR($\mu\rightarrow e\gamma$) for
$\bar{\xi}^{E}_{N,\tau e}=0.01\,GeV$, $\bar{\xi}^{E}_{N,\tau\mu}=1.0\,GeV$.
Here the solid (dashed, short dashed) line represents the BR without KK modes
of leptons and new Higgs bosons (with KK modes of leptons and new Higgs bosons
for lepton location set II, set I), for $a=0.01$ and $0.1$. Figure 3: $a$
dependence of the BR($\mu\rightarrow e\gamma$) for $\bar{\xi}^{E}_{N,\tau
e}=0.01\,GeV$, $\bar{\xi}^{E}_{N,\tau\mu}=1.0\,GeV$, for the lepton location
set II and $k=10^{18}\,GeV$. Figure 4: The same as Fig.2 but for
$\tau\rightarrow e\gamma$ decay and for $\bar{\xi}^{E}_{N,\tau e}=0.1\,GeV$,
$\bar{\xi}^{E}_{N,\tau\tau}=50\,GeV$. Figure 5: The same as Fig.3 but for
$\tau\rightarrow e\gamma$ decay, for $\bar{\xi}^{E}_{N,\tau e}=0.1\,GeV$,
$\bar{\xi}^{E}_{N,\tau\tau}=50\,GeV$ and $k=10^{17}\,GeV$. Figure 6: The same
as Fig.2 but for $\tau\rightarrow\mu\gamma$ decay and for
$\bar{\xi}^{E}_{N,\tau\mu}=1.0\,GeV$, $\bar{\xi}^{E}_{N,\tau\tau}=50\,GeV$.
Figure 7: The same as Fig.3 but for $\tau\rightarrow\mu\gamma$ decay, for
$\bar{\xi}^{E}_{N,\tau\mu}=1.0\,GeV$, $\bar{\xi}^{E}_{N,\tau\tau}=50\,GeV$ and
$k=10^{17}\,GeV$.
| arxiv-papers | 2008-06-16T00:21:16 | 2024-09-04T02:48:56.279033 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "E. Iltan",
"submitter": "Erhan Iltan",
"url": "https://arxiv.org/abs/0806.2478"
} |
0806.2583 | # Avoided crossings between bound states of ultracold Cesium dimers
Jeremy M. Hutson Department of Chemistry, University of Durham, South Road,
Durham, DH1 3LE, United Kingdom Eite Tiesinga and Paul S. Julienne Joint
Quantum Institute, National Institute of Standards and Technology and The
University of Maryland, Gaithersburg, Maryland 20899-8423, USA
###### Abstract
We present an efficient new computational method for calculating the binding
energies of the bound states of ultracold alkali-metal dimers in the presence
of magnetic fields. The method is based on propagation of coupled differential
equations and does not use a basis set for the interatomic distance
coordinate. It is much more efficient than the previous method based on a
radial basis set and allows many more spin channels to be included. This is
particularly important in the vicinity of avoided crossings between bound
states. We characterize a number of different avoided crossings in Cs2 and
compare our converged calculations with experimental results. Small but
significant discrepancies are observed in both crossing strengths and level
positions, especially for levels with $l$ symmetry (rotational angular
momentum $L=8$). The discrepancies should allow the development of improved
potential models in the future.
## I Introduction
Ultracold Cs atoms are of great interest for a number of experiments, which
have produced a Bose-Einstein condensate of such atoms Weber et al. (2003),
formed a cold cloud of Cs2 dimer molecules Herbig et al. (2003), probed three-
body Efimov physics Kraemer et al. (2006), studied collisional shifts
Szymaniec et al. (2007) or quantum scattering Hart et al. (2007) of atomic
clock states, carried out high-resolution molecular spectroscopy Vanhaecke et
al. (2004) or used magnetic fields to switch among a variety of very weakly
bound molecular states of the Cs2 dimer Mark et al. (2007a, b). These
experiments all depend upon and take advantage of the collisional interactions
between two Cs atoms. Consequently, accurate theoretical and computational
models of near-threshold Cs atom scattering and bound states are necessary for
maximum understanding of existing experiments and for making quantitative
predictions for new experimental domains.
Because of the complex spin structure of two ground-state Cs atoms, many
different near-threshold bound states exist and have different magnetic
moments. They thus tune differently with magnetic field. When one of these
bound states crosses a collision threshold, a low-energy scattering resonance
occurs, commonly known as a Feshbach resonance. Extensive study of such
resonances has allowed the construction of quite accurate coupled-channel
models for calculating the magnetic field-dependent scattering and bound-state
properties near collision thresholds Chin et al. (2000); Leo et al. (2000);
Chin et al. (2003, 2004); Mark et al. (2007a, b). These models incorporate the
electron and nuclear spins, their mutual interactions, and the adiabatic Born-
Oppenheimer potentials for the X${}^{1}\Sigma_{g}^{+}$ and
a${}^{3}\Sigma_{u}^{+}$ molecular states that correlate with two 2S1/2 ground-
state Cs atoms. By adjusting the model parameters to fit the measured magnetic
fields for resonances in different scattering channels, the model quite
accurately predicts near-threshold scattering properties and the binding
energies of weakly bound states within a few GHz of threshold. Such threshold
models can also be adapted to treat three-body interactions, for which an
accurate knowledge of the threshold two-body bound states is necessary Lee et
al. (2007). The models are sensitive to relatively few parameters, and may or
may not be adequate when extended into new experimental domains.
Recently, Mark et al. Mark et al. (2007a, b) have characterized a number of
avoided crossings between levels bound by only $E/h\approx 5$ MHz with respect
to the energy of two separated Cs atoms in their lowest-energy Zeeman
sublevels. Using time-dependent magnetic field ramping, they were able to
convert two Cs atoms into a number of different molecular states with
different rotational quantum numbers and magnetic moments. Most of the bound
states are well described by the existing coupled-channel model in regions far
from avoided crossings. However, characterizing the avoided crossings
themselves presents problems for the existing computational methods. In
particular, Ref. Chin et al. (2004) calculated bound states using a method
based on a basis set expansion of the radial wavefunctions in a discrete
variable representation (DVR). This method can use only a restricted spin
basis in determining the molecular bound states because of the large number of
grid points required.
The present paper develops an improved computational method that is necessary
to calculate and understand the avoided crossings in Cs2. This method uses a
propagator approach Hutson (1994) in place of a radial basis set to represent
the molecular bound states. It can readily be adapted to threshold states of
other molecules Köhler et al. (2006); Lang et al. (2008). The propagator
approach is computationally much cheaper than the DVR approach and as a result
can include many more coupled spin channels. The new approach is used to
compare the calculated and observed properties of the avoided crossings, in
order to identify aspects of the ground-state coupled-channel model for Cs2
that are still in need of improvement.
## II Computational methods
The present work solves the bound-state Schrödinger equation for Cs2 using two
independent methods. In either case the Hamiltonian may be written
$\frac{\hbar^{2}}{2\mu}\left[-R^{-1}\frac{d}{dR^{2}}R+\frac{\hat{L}^{2}}{R^{2}}\right]+\hat{h}_{1}+\hat{h}_{2}+\hat{V}(R),$
(1)
where $\mu$ is the reduced mass and $\hat{L}^{2}$ is the operator for the end-
over-end angular momentum of the two atoms about one another. The monomer
Hamiltonians including Zeeman terms are
$\hat{h}_{j}=\zeta\hat{\imath}_{j}\cdot\hat{s}_{j}+g_{e}\mu_{\rm
B}B\,\hat{s}_{zj}+g_{n}\mu_{\rm B}B\,\hat{\imath}_{zj},$ (2)
where $\hat{s}_{1}$ and $\hat{s}_{2}$ represent the electron spins of the two
atoms and $\hat{\imath}_{1}$ and $\hat{\imath}_{2}$ represent nuclear spins.
$g_{e}$ and $g_{n}$ are the electron and nuclear $g$-factors, $\mu_{\rm B}$ is
the Bohr magneton, and $\hat{s}_{z}$ and $\hat{\imath}_{z}$ represent the
$z$-components of $\hat{s}$ and $\hat{\imath}$ along a space-fixed $Z$ axis
whose direction is defined by the external magnetic field $B$. The interaction
between the two atoms $\hat{V}(R)$ is given by Stoof et al. Stoof et al.
(1988) as the sum of two terms,
${\hat{V}}(R)=\hat{V}^{\rm c}(R)+\hat{V}^{\rm d}(R)\,.$ (3)
Here $\hat{V}^{\rm
c}(R)=V_{0}(R)\hat{\cal{P}}^{(0)}+V_{1}(R)\hat{\cal{P}}^{(1)}$ is an isotropic
potential operator that depends on the potential energy curves $V_{0}(R)$ and
$V_{1}(R)$ for the respective X${}^{1}\Sigma_{g}^{+}$ singlet and
a${}^{3}\Sigma_{u}^{+}$ triplet states of the diatomic molecule. The singlet
and triplet projectors $\hat{\cal{P}}^{(0)}$ and $\hat{\cal{P}}^{(1)}$ project
onto subspaces with total electron spin quantum numbers 0 and 1 respectively.
Figure 1 shows the two potential energy curves for Cs2. The $\hat{V}^{\rm
d}(R)$ term represents small, anisotropic spin-dependent couplings that are
responsible for the avoided crossings discussed in this paper and are
discussed further in Section III below.
Figure 1: Molecular potential energy curves $V_{0}(R)$ and $V_{1}(R)$ for the
respective singlet and triplet states of Cs2 correlating with two separated
2S1/2 ground-state atoms. The inset shows an expanded view of the long-range
potentials separating to the two different $f=$ 3 and 4 hyperfine states of
the 2S1/2 atom with nuclear spin $i$=$7/2$ and magnetic field $B=0$. The inset
shows the adiabatic potentials obtained from diagonalizing the matrix form of
the operator $\hat{h}_{1}+\hat{h}_{2}+\hat{V}(R)$ at each $R$ for the case of
$L=0$, $M_{F}=+6$. There are 5 channels, and the $3+4$ and $4+4$ separated-
atom limits are doubly degenerate at $B=0$. All 5 channels have the same long-
range variation as $-C_{6}/R^{6}$, with $C_{6}=6860$ $E_{\rm h}a_{0}^{6}$ Chin
et al. (2004) ($E_{\rm h}=4.3597\times 10^{-18}$ J is the Hartree and
$a_{0}$=0.0529177 nm is the Bohr radius). The level crossings discussed in
this paper are for very weakly bound levels that lie within about $E/h\approx
5$ MHz of the dissociation limit to two $\\{fm_{f}\\}=\\{3,+3\\}$ atoms in the
magnetic field range from 0 mT to 5 mT.
The first method for finding eigenvalues is a conventional full matrix
diagonalization in a discrete variable representation (DVR) Colbert and Miller
(1992). It uses a basis set made up of products of internal and radial
functions. The internal Bose-symmetrized basis set is made up of functions in
which the operators $\hat{L}^{2}$ and $\hat{h}_{j}$ are diagonal, that is,
$|\alpha_{1}m_{f1}\rangle|\alpha_{2}m_{f2}\rangle|LM_{L}\rangle,$ (4)
where $|LM_{L}\rangle$ and $|\alpha_{j}m_{fj}\rangle$ respectively represent
the eigenstates of $\hat{L}^{2}$ and the $B$-dependent monomer Hamiltonian
$\hat{h}_{j}$, and where $M_{L}$ and $m_{fj}$ are projection quantum numbers
along the magnetic field direction. When $B=0$,
$|\alpha_{j}m_{fj}\rangle=|(s_{j}i_{j})f_{j}m_{fj}\rangle$, where $f_{j}$ is
the total spin of atom $j$ and $m_{fj}$ is its space-fixed projection. As $B$
increases from zero, different $f_{j}$ values become mixed. The DVR radial
functions are unevenly spaced collocation points obtained from a nonlinear
coordinate transformation Tiesinga et al. (1998).
This DVR method requires diagonalizing a large $N\times N$ matrix, the
dimension of which is given by the product of the number of spatial
collocation points $N_{\rm c}$ and the number of spin basis functions $N_{\rm
s}$. We use the LAPACK subroutine DSPEVX to find a selected range of
eigenvalues and eigenvectors Anderson et al. (1992). In order to use a direct
diagonalization procedure to calculate the bound-state energies Not shown in
Refs. Chin et al. (2004); Mark et al. (2007b), the magnitude of $N=N_{\rm
c}N_{\rm s}$ was limited to around 25000 using a processor with 4 GB of
memory. With $N_{\rm c}\approx 800$, in order to give 5 points per node with
about 150 nodes for threshold wave functions, the number of spin basis
functions is thus restricted to be about $N_{\rm s}\approx 35$. When this is
fewer than is needed for a complete calculation, an approximation scheme
becomes necessary, as described in Section III.
The second method avoids the use of a basis set for the interatomic distance
$R$ and instead relies on propagation of coupled differential equations Hutson
(1994). In this case the Bose symmetrized basis set used is a fully decoupled
set,
$\Phi_{k}=|s_{1}m_{s1}\rangle|i_{1}m_{i1}\rangle|s_{2}m_{s2}\rangle|i_{2}m_{i2}\rangle|LM_{L}\rangle.$
(5)
The compound channel index $k$ is used to simplify notation and implies values
of all the quantum numbers in the basis set. While the choice of the basis
sets in Eqs. (4) and (5) represent different approaches, they are equivalent
for representing molecular energy levels when the two basis sets span the same
space. There is a simple unitary transformation between the two basis sets.
The matrix elements of the different terms in the Hamiltonian in basis set (5)
are given in the Appendix.
In the propagation method, we expand the total wavefunction for state $n$ as
$\Psi_{n}=R^{-1}\sum_{k}\Phi_{k}\psi_{kn}(R).$ (6)
Substituting into the Schrödinger equation and projecting onto each channel
function in turn gives a set of coupled equations for the radial channel
functions $\psi_{kn}(R)$,
$\frac{d^{2}\psi_{jn}}{dR^{2}}=\sum_{k}\left[W_{jk}(R)-\varepsilon\delta_{jk}\right]\psi_{kn}(R),$
(7)
where $\delta_{jk}$ is the Kronecker delta, $\varepsilon$ is the energy $E$
scaled by $2\mu/\hbar^{2}$, and
$W_{jk}(R)=\int\Phi_{j}^{*}\left[\frac{\hat{L}^{2}}{R^{2}}+\frac{2\mu}{\hbar^{2}}\left(\hat{h}_{1}+\hat{h}_{2}+\hat{V}(R)\right)\right]\Phi_{k}\,d\tau,$
(8)
where $d\tau$ indicates integration over all coordinates except $R$. If there
are $N_{\rm s}$ basis functions, the required solution $\psi_{n}(R)$ is a
column vector of order $N_{\rm s}$ with elements $\psi_{kn}(R)$. However, Eq.
7 has $N_{\rm s}$ independent solution vectors at any energy, so that until
the boundary conditions are applied $\psi_{n}(R)$ is an $N_{\rm s}\times
N_{\rm s}$ wavefunction matrix.
The Schrödinger equation can be solved to find an $N_{\rm s}\times N_{\rm s}$
wavefunction matrix at any energy $E$. In practice it is numerically stabler
to propagate the log-derivative matrix
$Y(R)=[d\psi_{n}/dR][\psi_{n}(R)]^{-1}$. However, a solution that satisfies
bound-state boundary conditions can be found only at the eigenvalues $E_{n}$.
Solutions are propagated outwards from a point $R_{\rm min}$ in the inner
classically forbidden region and inwards from a point $R_{\rm max}$ at long
range to a matching point $R_{\rm mid}$. The outwards and inwards solutions
are designated $Y^{+}(R)$ and $Y^{-}(R)$. If $E$ is an eigenvalue of the
coupled equations, there must exist a wavefunction vector $\psi_{n}(R_{\rm
mid})=\psi_{n}^{+}(R_{\rm mid})=\psi_{n}^{-}(R_{\rm mid}$) for which the
derivatives also match,
$\left.\frac{d\psi_{n}^{+}}{dR}\right|_{R_{\rm
mid}}=\left.\frac{d\psi_{n}^{-}}{dR}\right|_{R_{\rm mid}},$ (9)
so that
$Y^{+}(R_{\rm mid})\psi_{n}(R_{\rm mid})=Y^{-}(R_{\rm mid})\psi_{n}(R_{\rm
mid}).$ (10)
Thus $\psi_{n}(R_{\rm mid})$ is an eigenvector of $Y^{+}(R_{\rm
mid})-Y^{-}(R_{\rm mid})$ with eigenvalue 0. It is thus possible to locate
eigenvalues of the Schrödinger equation by propagating solutions of the
coupled equations and searching for zeroes in the eigenvalues of the log-
derivative matching matrix $Y^{+}(R_{\rm mid})-Y^{-}(R_{\rm mid})$ as a
function of energy. This approach is much stabler for large multichannel
problems than the older approach Johnson (1978) of searching for zeroes of the
determinant of the matching matrix.
The major advantage of the propagator method is that the matrices handled are
only of dimension $N_{\rm s}\times N_{\rm s}$, where $N_{\rm s}$ is the number
of internal basis functions. The computational cost is proportional to $N_{\rm
s}^{3}$ but only linear in the number of propagation steps. By contrast, a
full diagonalization with $N_{\rm c}$ radial basis functions (collocation
points) involves matrices of dimension $N_{\rm s}N_{\rm c}\times N_{\rm
s}N_{\rm c}$. The computational cost is proportional to $N_{\rm s}^{3}N_{\rm
c}^{3}$. Since $N_{\rm c}$ typically needs to be greater than 500 for the
present application, the propagator approach is much cheaper.
The BOUND program Hutson (1993) is a general-purpose package to solve the
bound-state Schrödinger equation using propagator methods. The algorithms used
are described in more detail in Ref. Hutson, 1994. For the purpose of the
present work we have generalised the BOUND package in three significant
respects:
1. 1.
We have generalised the structure of the code so that it can handle coupled
equations in basis sets that are not diagonal at $R=\infty$;
2. 2.
We have implemented the specific set of coupled equations required for Cs2
with the basis set of Eq. 5;
3. 3.
We have added an option to use the log-derivative propagator of Alexander and
Manolopoulos Alexander and Manolopoulos (1987), which is based on Airy
functions and allows very large step sizes at long range.
In the presence of a magnetic field, the only rigorously conserved quantum
numbers are $M_{\rm
tot}=m_{f1}+m_{f2}+M_{L}=m_{s1}+m_{s2}+m_{i1}+m_{i2}+M_{L}$ and the total
parity $(-1)^{L}$. This leads to an infinite number of channels. However, $L$
and $M_{F}=m_{f1}+m_{f2}$ are very good approximate quantum numbers because
the only term in the Hamiltonian that is off-diagonal in them is the small
anisotropic coupling term $\hat{V}^{\rm d}$. In either computational approach
it is possible to restrict the number of channels by selecting only one or a
few values of $L$ and all or a subset of possible $M_{F}$ values. Here we
consider the case studied experimentally by Mark et al. Mark et al. (2007b),
who used Cs atoms is their lowest energy hyperfine state with $m_{f}=+3$ to
make Cs2 molecules with $M_{\rm tot}=+6$. The number of channels with $L=0$,
2, 4, 6, 8, including all allowed $M_{F}$ values, are 5, 23, 46, 76, 103,
respectively. Thus, for example, a full calculation including all channels
with $L=4$, 6 and 8 requires 225 channels.
In practical terms, for example, a run with the DVR method to find 28 bound
states within 3 GHz of the $E=0$ threshold for Cs2 for a single magnetic field
with 30 channels and 720 collocation points took about 7 hours on an 2.4 GHz
processor. With the propagator approach we were able to find selected near-
dissociation levels for 30 channels in about 40 seconds per level with a 2.0
GHz processor. The great advantage of the propagator approach was demonstrated
by our ability to find levels with 225 channels in about 45 minutes per level.
A calculation with 225 channels would not be possible at all using the DVR
method with a direct eigenvalue solver.
## III Comparison of computational results
The DVR and propagator calculations described here both use the same potential
model, with the parameters given by Chin et al. Chin et al. (2004). The
potential energy curves are based on the ab initio calculations of Krauss and
Stevens Krauss and Stevens (1990). The singlet and triplet scattering lengths
$a_{\rm S}$ and $a_{\rm T}$, the long-range coefficients $C_{6}$ and $C_{8}$,
and a scaling factor $S_{C}$ for the second-order spin-orbit coupling were
adjusted by Chin et al. to reproduce a substantial number of Feshbach
resonances with $L\leq 4$.
Figure 2 shows an example of weakly bound levels of the Cs2 molecule with
$M_{\mathrm{tot}}=+6$ in the 0 mT to 6 mT range of $B$. Many of these levels
have been probed in the experiment of Mark et al. Mark et al. (2007b). The
figure also shows the bound-state classification scheme of Chin et al. Chin et
al. (2004), namely $FL(M_{F})$, where $F$ is the resultant of the separated-
atom spins $f_{1}$ and $f_{2}$ and $M_{F}$ is its projection defined above.
Like $f_{1}$ and $f_{2}$, $F$ is a good approximate quantum number for
labeling near-threshold levels at low $B$. Quantum numbers $L=0$, 2, 4, 6, 8
are represented by labels $s$, $d$, $g$, $i$, $l$, respectively. $M_{L}$ need
not be specified since $M_{L}=M_{\mathrm{tot}}-M_{F}$. Fig. 2 shows levels
with $L\leq 4$ obtained from a DVR calculation that included only basis
functions for a single $L$ and $M_{F}$. This neglects the small off-diagonal
couplings between levels with different $L$ and $M_{F}$ quantum numbers due to
$\hat{V}^{\rm d}$, so that levels of different symmetry show crossings rather
than avoided crossings in Fig. 2.
Figure 2: Bound state energy $E/h$ as a function of $B$ for levels of the Cs2
molecules with even $L\leq 4$ and $M_{\mathrm{tot}}=+6$. Energies are given
relative to the energy of two Cs atoms in their ground Zeeman sublevel ($f=3$,
$m_{f}=+3$). The $FL(M_{F})$ labeling scheme is shown for each level. Off-
diagonal coupling between levels with different $FL(M_{L})$ quantum numbers is
neglected in this calculation.
For ground-state alkali-metal atom interactions, the $\hat{V}^{\rm d}$
operator has the form of spin-dipolar coupling
$\hat{V}^{\rm
d}(R)=\lambda(R)\left(\hat{s}_{1}\cdot\hat{s}_{2}-3(\hat{s}_{1}\cdot\vec{e}_{R})(\hat{s}_{2}\cdot\vec{e}_{R})\right)\,,$
(11)
where $\vec{e}_{R}$ is a unit vector along the internuclear axis and $\lambda$
is an $R$-dependent coupling constant, which for our model is
$\lambda(R)=E_{\rm
h}\alpha^{2}\left(\frac{1}{(R/a_{0})^{3}}-0.071968e^{-0.83[(R/a_{0})-10]}\right)\,,$
(12)
where $\alpha\approx 1/137$ is the fine structure constant. At large $R$ the
coupling becomes the long-range dipolar interaction between the spins on the
separated atoms that varies as $1/R^{3}$ Stoof et al. (1988); Moerdijk et al.
(1995). In the short-range region of chemical bonding the magnitude of
$\lambda(R)$ is primarily determined by the second-order spin-orbit coupling
term represented by the exponential term Mies et al. (1996); Leo et al.
(1998); Kotochigova et al. (2001); Chin et al. (2004).
The crossings in Fig. 2 become avoided crossings when the small interactions
due to $\hat{V}^{\rm d}$ are taken into account. The energy splitting at the
crossing varies greatly, depending on the quantum numbers of the two levels.
In first order, the $\hat{V}^{\rm d}$ operator couples states $FL(M_{F})$ and
$F^{\prime}L^{\prime}(M_{F}^{\prime})$ according to the selection rules
$|L-L^{\prime}|=$ 0 or 2, $|F-F^{\prime}|=$ 0, 1, or 2, and
$|M_{F}-M_{F}^{\prime}|=$ 0, 1, or 2. These selection rules immediately follow
from the tensor form of the operator in Eq. (11), as given by Stoof et al.
Stoof et al. (1988), who write Eq. (11) as a sum of products of $L=2$
spherical harmonic components $Y_{LM_{L}}({\vec{e}_{R}})$ and rank 2 spin
tensor components. We refer to a crossing as direct when there is a first
order coupling of the two states involved and indirect when there is not.
The success of a calculation of the Cs2 energy levels and their avoided
crossings depends on the sufficiency of the basis set expansion of the wave
function. Suppose we wish to calculate the energy of one $FL(M_{F})$ state
that crosses a different $F^{\prime}L^{\prime}(M_{F}^{\prime})$ state. It is
necessary to include sufficient basis functions to represent each state
adequately, and to represent their interaction. This is simplified by taking
advantage of the selection rules described above. In order to represent a
level with a given $FL(M_{F})$, it is necessary to include all basis functions
with the same set of three quantum numbers, since such levels are coupled by
terms due to the strong central potential $\hat{V}^{\rm c}$. A level
calculated with such a basis is coupled through the $\hat{V}^{\rm d}$ operator
to other levels in which one or more of the three quantum numbers are
different. Such off-diagonal coupling causes shifts in level positions and
also induces avoided crossings.
In the propagator calculations, the basis set usually includes all functions
with $L$ and $L^{\prime}$ of the levels in question consistent with
$M_{\mathrm{tot}}$. Additional basis functions with different quantum numbers
$L_{i}$ are added to account for shifts and crossings due to coupling of $L$
or $L^{\prime}$ with $L_{i}$. The propagator basis is specified by giving $L$,
$L^{\prime}$, and a list of additional values $L_{i}$ needed to account for
higher-order coupling. In the DVR calculations, the basis sets are
additionally limited by restricting the calculation to functions with
$L(M_{F})$, $L^{\prime}(M_{F}^{\prime})$ and additional quantum numbers
$L_{i}(M_{F,i})$ as needed. Thus the basis set is specified by giving the list
$L(M_{F})L^{\prime}(M_{F}^{\prime})[L_{i}(M_{F,i})]$. Some propagator
calculations were done with a similarly restricted list to verify that the two
methods gave exactly equivalent results. Neither the propagator nor DVR
calculations make any additional restrictions by $F$, although this could be
done.
Figure 3: Example of coupling between different $L(M_{F})$ symmetry blocks
with the symmetry of the dipole-dipole interaction of $\hat{V}^{\rm d}$. Each
block represents a Hamiltonian matrix for spin states with the $L(M_{F})$
values indicated. The labels “x” and “y” indicate the existence of
nonvanishing coupling due to $\hat{V}^{\rm d}$; a “0” indicates no coupling.
The case shown is for a $g(6)$ and a $g(3)$ level, which have no direct
coupling. The left panel shows the symmetries that give rise to second-order
interactions between the two levels and thus contribute to the strength of the
avoided crossing between them. The right panel shows a truncated set of
interactions through intermediate $d(4)$ and $d(5)$ levels. Figure 4:
Calculated energy levels $E/h$ with $M_{\mathrm{tot}}=+6$ as a function of
magnetic field $B$ near the crossing of the $4g(3)$ level with the 6g(6) level
near 1.0 mT. The points and solid line show the propagator calculation with a
$sdgi$ basis set. The dashed lines show the crossing levels from two uncoupled
DVR calculations with $g(3)$ or $g(6)$ basis functions only. The dash-dot and
dotted lines show the crossing levels from DVR calculations with added
$g(4,5)$ and $d(4,5)$ functions respectively. The DVR calculation with
$i(4,5)$ basis functions is not shown, but lies near the uncoupled crossing
and has a very small splitting, indicating very weak second-order coupling
through distant $i$ states.
Figure 3 illustrates the size of the basis set needed, as governed by the
selection rules on $\hat{V}^{\rm d}$ coupling. Since the matrix elements of
$\hat{V}^{\rm d}$ are relatively small, they are normally of practical
significance only through second order. Thus it is necessary to include only
intermediate levels with $L_{i}$ and $M_{F,i}$ that differ from $L$ or
$L^{\prime}$ and $M_{F}$ or $M_{F}^{\prime}$ by at most 2 units. Any higher-
order couplings would be much smaller than those discussed here. Thus, in
order to represent the crossing of a $6g(6)$ and a $4g(3)$ level, for which
there is no first-order direct coupling, $d$-, $g$-, or $i$-basis functions
with $M_{F}=$ 4 and 5 need to be included in the basis, as shown in Fig. 3. To
represent additional second-order shifts of the two $g$ levels, $d$-, $g$\-
and $i$-basis functions with $M_{F}=$ 1, 2, 7 and 8 also need to be added.
Figure 4 illustrates calculations with different basis sets, comparing
energies calculated with the propagator and DVR methods for the crossing of
the $4g(3)$ and $6g(6)$ levels near 1.0 mT. Table 1 tabulates the positions
and strengths of this crossing, as well as a number of others. The position
$B_{0}$ of the crossing is defined as the field at which the two levels are
closest together and the strength $2V$ is the minimum of the difference
between the two energies as a function of $B$; $2V$ is used since the
splitting is twice the effective coupling matrix element $V$ in a 2-level
representation of the crossing Mark et al. (2007b). We have verified that the
two methods give identical results within numerical accuracy when exactly
equivalent basis sets are used. Since there is no direct interaction between
the two crossing levels in this case, the splitting at the crossing originates
principally in second-order interactions mediated through distant levels of
$d$, $g$, or $i$ symmetry with $M_{F}=$ 4 or 5. However, as mentioned above,
second-order couplings to levels with other $M_{F}$ values can cause
additional shifts. Both bound and scattering states can contribute, and the
contribution from any given distant state varies inversely with the its
separation in energy from the crossing. Intermediate $g$ levels are the
closest in energy to the crossing, whereas intermediate $i$ levels are the
most distant. In Figure 4, the $sdgi$ basis set used in the propagator
calculation is effectively complete. It may be seen that a calculation
including only the $g(4,5)$ intermediate states captures most of the crossing
strength but does not reproduce the level shifts well. Conversely, a
calculation including only the $d(4,5)$ states gives a crossing strength that
is much too small but overestimates the level shifts. The contributions to the
crossing strength from different intermediate states are far from additive.
There are no experimental results for this crossing.
Figure 5: Calculated energy levels $E/h$ with $M_{\mathrm{tot}}=+6$ as a
function of magnetic field $B$ near the crossing of the $4g(4)$ level with the
6g(6) level near 1.0 mT. The points and solid line show the propagator
calculation with a $dgi$ basis set. The dashed lines shows the crossing levels
from two uncoupled DVR calculations with $g(4)$ or $g(6)$ basis functions
only. The dash-dot lines show the avoided crossing from a DVR calculation with
only the direct coupling in the $g(4,6)$ basis set included. The doubled-
headed arrow at the position of the propagator crossing shows the measured
splitting Mark et al. (2007b). The actual experimental crossing was observed
$0.046$ mT lower in $B$ value than the propagator crossing. Figure 6:
Calculated energy difference $\Delta/h$ between the $6g(6)$ and $4d(4)$ levels
with $M_{\mathrm{tot}}=+6$ as a function of magnetic field $B$. The solid line
is from a propagator calculation with the $sdgi$ basis. The diamonds show the
experimental results obtained by Ferlaino et al. using their more accurate
field modulation method Ferlaino et al. (2008). The dashed line shows the
calculated points shifted by $+0.034$ mT. The DVR calculation (not shown) with
direct coupling included in the $d(4)g(6)$ basis is virtually identical to the
dashed line when the DVR results are shifted by $+.062$ mT.
Figure 5 illustrates a different case, a $4g(4)-6g(6)$ crossing with a
splitting that is about 8 times larger at the crossing. This is a case where
the two states involved have a direct coupling to one another through
$\hat{V}^{\rm d}$. While additional second-order coupling can change the
position and strength of the crossing slightly, the direct coupling is
dominant and the restricted-basis DVR calculation agrees much better with the
full propagator calculations. Both are in reasonable agreement with the
measured splitting of the crossing Mark et al. (2007b). However, the
calculated position in $B$ needs to be shifted by $-0.046$ mT to agree with
the measured position Mark et al. (2007b). This remaining discrepancy reflects
a real deficiency in the parameters of our potential model as discussed below.
Figure 6 shows the difference between the upper and lower branches of the
crossing for the case of the $4d(4)-6g(6)$ crossing near 4.5 mT. This is
another case of direct coupling, where the measured and calculated crossings
agree well in coupling strength, although the calculated position needs to be
shifted by $+0.034$ mT to agree with the measured one.
Table 1 show comparisons between the propagator and DVR calculations for a
number of other crossings in Figure 2. Crossing positions generally agree
within about 0.01 mT among the different basis sets. Relatively good agreement
between propagator and DVR coupling strengths $2V$ is seen in the cases where
there is direct coupling between the two crossing levels, or where the DVR
method includes all second-order intermediate states allowed by the symmetry
of the $\hat{V}^{\rm d}$ operator. However, for higher-$L$ crossings it was
usually necessary to select a subset of the allowed intermediate states to
make DVR calculations feasible. In such cases the DVR method can give
unreliable results, depending on the choice of restricted basis set.
Figure 7 shows calculated bound states for $s$ and $d$ levels on a broader
energy scale. (Levels with other symmetries are not shown). A DVR calculation
with a full $sd$ basis is possible in this case. The $6s(6)$ and $4d(4)$
uncoupled levels show two crossings. The low-field crossing around 0.24 mT
occurs near the observed location (0.72 mT) of a three-body Borromean state of
the Cs3 trimer associated with the exotic Efimov physics of this species
Kraemer et al. (2006). Lee et al. Lee et al. (2007) used the last two $6s(6)$
two-body states of the Cs2 dimer to construct the parameters for full three-
body calculations of bound states and recombination coefficients in the 0 mT
to 3 mT range. While their method was able to give semi-quantitative agreement
with the measurements, the avoided crossing of the $6s(6)$ level with the
$4d(4)$ level needs to be taken into account in subsequent calculations
because of the mixed spin character of the target molecular state produced by
the three-body recombination in this region of $B$. The strong $s-d$
interactions modify the $s$-wave scattering length at small $B$, but this is
easy to take into account by including $s$ and $d$ basis functions in
scattering calculations.
The higher-field $6s(6)-4d(4)$ crossing near 4.8 mT has been studied in Refs.
Mark et al. (2007b); Julienne et al. (2004). Figure 9 shows an expanded view
of the very-near-threshold region of this crossing and the additional
$6s(6)-2g(2)$ crossing near 5.4 mT. The interaction between $s$ and $d$ states
results in an overall shift in the binding energy of the $6s(6)$ level, where
the uncoupled level is too high in energy. This case illustrates one advantage
of the propagator method over the DVR method. The latter has to use a finite
range of spatial points and is restricted by the length of the “box” in which
the calculation is carried out. When this length is too large, the number of
the spatial collocation points can become too large for practical
calculations. A 5000 $a_{0}$ “box” is sufficient for levels with binding
energies on the order of 40 kHz, since the scattering length, which gives an
indication of the “size” of the weakly bound molecular levels Köhler et al.
(2006), is on the order of 1000 $a_{0}\ll$ 5000 $a_{0}$. Such restrictions on
spatial grid do not apply to the propagator method, which is capable of
calculating levels arbitrarily close to $E=0$, as long as the propagation is
to sufficiently large distances. Since the propagator used can take very large
steps at long range, this presents no difficulty.
Figure 7: Energy levels $E/h$ as a function of $B$ for the Cs2 molecule for $L=0$ and 2 only with $M_{\mathrm{tot}}=+6$ ($g$ and $l$ levels are not shown). The dashed lines show the DVR levels calculated with the uncoupled $s(6)$ and $d(M_{F})$ basis sets, where $M_{F}=4$, 5, or 6. The diamonds show the results of Mark et al. Mark et al. (2007b). The $4d(4)$ level crosses the $6s(6)$ level twice, near 0.24 mT and 4.77 mT. The closed circles and dotted lines show the levels obtained from a DVR calculation with an $sd$ basis. A propagator calculation with a full $sdg$ basis shows negligible differences for this case. Figure 8: Energy levels $E/h$ as a function of $B$ for the Cs2 molecule for $L=8$ only with $M_{\mathrm{tot}}=+6$. The solid lines show the DVR levels calculated with the uncoupled $l(M_{F})$ basis sets, where $M_{F}=0,\ldots,6$. The dashed line shows the $6g(6)$ level for which avoided crossings have been calculated (see Table 1) and measured Mark et al. (2007b). Figure 9: Expanded view of the crossing in Fig. 7 of the $4d(4)$ and $6s(6)$ levels near 4.8 mT. The long dashed line shows the uncoupled calculation with the $s(6)$ and $d(4)$ basis sets. The solid lines show the propagator calculations with an $sdg$ basis. The upper crossing near 5.4 mT is due to a $2g(2)$ level. The open circles show DVR calculations with a full $sd$ basis in a finite box of $5000$ $a_{0}$. The diamonds show experimental results of Lange et al. Lange et al. (2008). Table 1: Levels crossing the $6g(6)$ level of the Cs2 molecule with a binding energy near $-5$ MHz relative to the energy of two atoms in their lowest energy hyperfine state at each $B$. The columns label the symmetry of the crossing state, the computational method (propagator or DVR), the $L$ functions in the basis set used for the calculation (only intermediate basis states are listed for the DVR method, since basis states for the two crossing states are automatically included), and the position $B_{0}$, energy $E/h$, and splitting $2V/h$ of each crossing. State | Method | Basis | $B_{0}$ | $E/h$ | $2V/h$
---|---|---|---|---|---
| | | [mT] | [MHz] | [kHz]
4g(2) | propagator | $g$ | 0.7615 | -5.180 | 3.6
| propagator | $dg$ | 0.7670 | -5.192 | 1.0
| propagator | $sdgi$ | 0.7617 | -5.152 | 1.5
| DVR | $g(2,6)[d(4)]$ | 0.7745 | -5.152 | 2.3
6l(3) | propagator | $gil$ | 0.9181 | -5.135 | 21.2
| DVR | $g(6)l(3)[i(4,5)]$ | 0.9112 | -5.138 | 21.7
4g(3) | propagator | $sdg$ | 1.0105 | -5.169 | 37.3
| propagator | $sdgi$ | 1.0024 | -5.138 | 33.5
| DVR | $g(3,6)[d(4,5)]$ | 1.0097 | -5.134 | 12.1
| DVR | $g(3,6)[g(4,5)]$ | 0.9935 | -5.134 | 32.9
| DVR | $g(3,6)[i(4,5)]$ | 0.9937 | -5.1314 | 1.1
6l(4) | propagator | $gil$ | 1.2715 | -5.088 | 43.6
4g(4) | propagator | $g$ | 1.368 | -5.14 | 264.
| propagator | $dgi$ | 1.375 | -5.11 | 277.
| DVR | $g(4,6)$ | 1.367 | -5.10 | 265.
6s(6) | propagator | $sdg$ | 1.8648 | -5.114 | 44.8
| propagator | $sdgi$ | 1.8664 | -5.079 | 44.8
6l(5) | propagator | $gil$ | 2.0089 | -5.056 | 66.6
6l(6) | propagator | $gil$ | 4.3211 | -4.890 | 77.6
4d(4) | propagator | $dg$ | 4.4403 | -4.937 | 55.5
| propagator | $sdgi$ | 4.4766 | -4.885 | 57.6
| DVR | $d(4)g(6)$ | 4.4485 | -4.905 | 53.0
2g(2) | propagator | $g$ | 5.1906 | -4.887 | $<$ 0.1
| propagator | $dgi$ | 5.1176 | -4.842 | 9.5
## IV Comparison with experiment
When the basis set is sufficiently large, there is good overall agreement
between our calculations and the experimental measurements, as already noted
in relation to Figs. 5 and 6. Table 2 lists other examples, including the
crossings of the $6g(6)$ level with the $6l(M_{F})$ levels shown in Fig. 8.
Since the potentials and second-order spin-orbit coupling of the model were
originally adjusted to reproduce Feshbach resonances due to zero-energy bound
states of $d$ and $g$ symmetry, the positions of crossings between $s$, $d$,
and $g$ levels tend to be accurate to within the model uncertainties, which
are on the order of 0.05 mT or less Leo et al. (2000); Chin et al. (2004). On
the other hand, the levels of $l$ symmetry, corresponding to $L=8$, are off by
up to 0.5 mT, a much larger amount. One plausible reason for this has to do
with the large rotational energy of the $6l(M_{F})$ levels that cross the
$6g(6)$ level. The $6g(6)$ level has the vibrational character of the second
$6s(6)$ vibrational level below the lowest separated-atom limit, with about
110 MHz of $l=4$ rotational energy added. The crossing $6l(M_{F})$ levels, by
contrast, have the vibrational character of the third vibrational level below
the limit, with about 740 MHz of rotational energy added to bring them near
threshold. More deeply bound levels with more rotational energy can have
larger errors due to deficiencies in the model potentials. An error of only a
few parts per 1000 in the rotational energy can lead to a 0.5 mT error in the
crossing positions for $6l(M_{F})$ levels.
Additional information is contained in the coupling strengths that govern the
closest approach $2V$ between levels at avoided crossings. In the
calculations, this quantity is determined largely by the second-order spin-
orbit contribution to $\hat{V}^{\mathrm{d}}$. This is a relatively poorly
determined parameter in our model and is uncertain to about 15% Leo et al.
(2000).
Several different experimental methods have been used to determine coupling
strengths. Mark et al. Mark et al. (2007a) used a method based on Stückelberg
interferometry, which gives precise measurements of the energy difference
between the two states. Mark et al. Mark et al. (2007b) used a different
method based on integrating magnetic moment values. This gives absolute
energies for the two states (rather than just the difference between them) but
is now believed to overestimate the coupling strengths in some cases Ferlaino
and Grimm (2008), especially for crossings between states with very different
magnetic moments. Some crossing strengths were also estimated from a Landau-
Zener approach. Lastly, Ferlaino et al. Ferlaino et al. (2008) have used a
method in which transitions are induced by modulating the magnetic field
Thompson et al. (2005). This is the most precise of the different methods.
Table 2: Comparison of results from the best propagator calculation with the
experimental results for selected level crossings with the 6g(6) state. The
columns label the symmetry of the crossing state, the origin of the value, the
$L$ functions in the basis set used for the calculation, and the position
$B_{0}$, energy $E/h$, and the energy splitting $2V/h$ for each crossing. The
lines labeled “Exp” show the experimental values.
State | method | basis | $B_{0}$ [mT] | $E/h$ [MHz] | $2V/h$ [kHz]
---|---|---|---|---|---
6l(3) | propagator | $gil$ | 0.9181 | -5.135 | 21.2
| Expa | | 1.122(2) | | 32(6)
| Expb | | 1.1339(1) | | 28(2)
6l(4) | propagator | $gil$ | 1.2715 | -5.088 | 43.6
| Expa | | 1.550(3) | | 128(26)
4g(4) | propagator | $dgi$ | 1.375 | -5.11 | 277.
| Expa | | 1.329(4) | | 328(60)
| Expc | | 1.357(1) | | 291.4(8)
6s(6) | propagator | $sdgi$ | 1.8664 | -5.079 | 44.8
| Expc | | 1.8651(3) | | 58(17)
6l(5) | propagator | $gil$ | 2.0089 | -5.056 | 66.6
| Expa | | 2.53(1) | | 126(44)
4d(4) | propagator | $sdgi$ | 4.4766 | -4.885 | 57.6
| Expa | | 4.515(4) | | 240(42)
| Expc | | 4.5106(3) | | 78(9)
a. Reference Mark et al. (2007b)
b. Reference Mark et al. (2007a)
c. Reference Ferlaino et al. (2008).
The crossing strengths for various different levels crossing the $6g(6)$ level
near 5 MHz are compared with the available experimental values in Table 2. The
most reliable experimental results are those from Stückelberg oscillations
Mark et al. (2007a) and magnetic field modulation Ferlaino and Grimm (2008)
for the $6l(3)$, $4g(4)$, $6s(6)$ and $4d(4)$ levels. The $6l(3)$ and $6s(6)$
levels are indirectly coupled to $6g(6)$, and for both these the calculated
crossing strength is about 25% lower than the best experimental value. The
$4g(4)$ and $4d(4)$ levels are directly coupled to $6g(6)$; for the $4g(4)$
level the calculated crossing strength is about 5% lower than experiment,
while for the $6s(6)$ level the discrepancy is larger but is within the
experimental error bars. This suggests that the strength of the coupling term
$V^{\rm d}(R)$ is underestimated but within the error range of Leo et al. Leo
et al. (2000).
Some of the other crossings in Table 2 show larger differences between
experiment and theory, but in all these cases the experimental value was
obtained using the less reliable magnetic moment method. The possible
experimental errors for the magnetic moment approach are illustrated by the
$4d(4)$ crossing, where it gives a crossing strength a factor of 3 larger than
the more accurate magnetic field modulation method. It would be very
interesting to remeasure the $6l(4)$, $6l(5)$ and other crossings in order to
establish whether there is a consistent relative error between experiment and
theory.
Errors in the level positions can result from deficiencies in either the long-
range or the short-range part of the model potentials. As discussed above,
there are remaining discrepancies in level positions of up to 0.05 mT for $s$,
$d$, and $g$ levels, and up to 0.5 mT for $l$ levels. Further improvements in
the potential model are thus needed for this important prototype system. This
is particularly important for predicting the resonances and crossings in the
80 mT region, where interesting Efimov physics is predicted Lee et al. (2007)
and even greater sensitivity to model errors is expected. A major advantage of
the propagator method introduced here is that it is inexpensive enough to be
used to determine model parameters by least-squares fitting to level energies
and locations and strengths of level crossings.
## V Conclusions
We have presented a new computational method for calculating bound states of
molecules such as Cs2. The method is based on solving a set of coupled
differential equations by propagation, without relying on a basis set for the
interatomic coordinate. This is much more efficient than using a radial basis
set and allows the use of much larger basis sets of spin functions. It also
eliminates problems with calculating bound states very near to dissociation,
because the propagation can be extended to very large separations at very
little expense. The new method makes it possible for the first time to carry
out fully converged calculations on bound states of Cs2, including anisotropic
couplings due to spin-spin and second-order spin-orbit interactions, and to
characterize avoided crossings between pairs of levels.
We have compared the results of converged calculations using the current best
Cs2 model potentials with experimental measurements on the near-dissociation
states of Cs2 in a magnetic field. The model generally performs well for $s$,
$d$ and $g$ states (with $L=0$, 2 and 4), though even there there are
quantitative discrepancies of up to 0.05 mT in the magnetic fields at which
levels cross. The discrepancies are much larger (0.5 mT) for $l$ states
($L=8$). The strengths of the avoided crossings also appear to be
systematically underestimated by the current model. These discrepancies should
in future allow the development of improved models for the potential curves
and couplings in the Cs2 dimer. Such model improvement is both desirable and
possible, not only for near-threshold levels but also to provide an improved
representation of more deeply bound states such as those measured by Vanhaecke
et al. Vanhaecke et al. (2004). High-quality models are also important for
proposals to use precision measurements on Cs$2$ for fundamental physics
studies Chin and Flambaum (2006); DeMille et al. (2008).
## VI Acknowledgements
P.S. Julienne acknowledges the Office of Naval Research for partial support.
J. M. Hutson is grateful to EPSRC for support under the ESF EUROCORES
Programme EuroQUAM.
## Appendix A Matrix elements
In the decoupled basis set (5), the matrix elements of the isotropic potential
operator $\hat{V}^{\rm c}(R)$ between primitive (unsymmetrized) basis
functions are
$\displaystyle\langle
s_{1}m_{s1}i_{1}m_{i1}s_{2}m_{s2}i_{2}m_{i2}LM_{L}|\hat{V}^{\rm
c}(R)|s_{1}m_{s1}^{\prime}i_{1}m_{i1}^{\prime}s_{2}m_{s2}^{\prime}i_{2}m_{i2}^{\prime}L^{\prime}M_{L}^{\prime}\rangle=\delta_{LL^{\prime}}\delta_{M_{L}M_{L}^{\prime}}\delta_{m_{i1}m_{i1}^{\prime}}\delta_{m_{i2}m_{i2}^{\prime}}$
$\displaystyle\sum_{S}V_{S}(R)(-1)^{2s_{1}-2s_{2}+m_{s1}+m_{s2}+m_{s1}^{\prime}+m_{s2}^{\prime}}(2S+1)\left(\matrix{s_{1}&s_{2}&S\cr
m_{s1}&m_{s2}&-m_{s1}-m_{s2}}\right)\left(\matrix{s_{1}&s_{2}&S\cr
m_{s1}^{\prime}&m_{s2}^{\prime}&-m_{s1}^{\prime}-m_{s2}^{\prime}}\right).$
(13)
The corresponding matrix elements of the spin-spin operator are
$\displaystyle\langle
s_{1}m_{s1}i_{1}m_{i1}s_{2}m_{s2}i_{2}m_{i2}LM_{L}|\hat{V}^{\rm
d}(R)|s_{1}m_{s1}^{\prime}i_{1}m_{i1}^{\prime}s_{2}m_{s2}^{\prime}i_{2}m_{i2}^{\prime}L^{\prime}M_{L}^{\prime}\rangle=\delta_{m_{i1}m_{i1}^{\prime}}\delta_{m_{i2}m_{i2}^{\prime}}\lambda(R)$
$\displaystyle(-1)^{s_{1}+s_{2}-m_{s1}-m_{s2}-M_{L}}\left[s_{1}(s_{1}+1)(2s_{1}+1)s_{2}(s_{2}+1)(2s_{2}+1)(2L+1)(2L^{\prime}+1)\right]^{1/2}\left(\matrix{L&2&L^{\prime}\cr
0&0&0}\right)$ $\displaystyle\sum_{q_{1}q_{2}}\left(\matrix{L&2&L^{\prime}\cr-
M_{L}&-q_{1}-q_{2}&M_{L}^{\prime}}\right)\left(\matrix{1&1&2\cr
q_{1}&q_{2}&-q_{1}-q_{2}}\right)\left(\matrix{s_{1}&1&s_{1}\cr-
m_{s1}&q_{1}&m_{s1}^{\prime}}\right)\left(\matrix{s_{2}&1&s_{2}\cr-
m_{s2}&q_{2}&m_{s2}^{\prime}}\right),$ (14)
where for any individual matrix element the sums over $q_{1}$ and $q_{2}$
collapse because of the selection rules imposed by the last two 3-$j$ symbols.
The matrix elements of the atomic nuclear spin operators are particularly
simple in this basis set,
$\displaystyle\langle
s_{1}m_{s1}i_{1}m_{i1}s_{2}m_{s2}i_{2}m_{i2}LM_{L}|\hat{\imath}_{1}\cdot\hat{s}_{1}|s_{1}m_{s1}^{\prime}i_{1}m_{i1}^{\prime}s_{2}m_{s2}^{\prime}i_{2}m_{i2}^{\prime}L^{\prime}M_{L}^{\prime}\rangle=$
$\displaystyle\delta_{LL^{\prime}}\delta_{m_{s2}m_{s2}^{\prime}}\delta_{m_{i2}m_{i2}^{\prime}}\langle
s_{1}m_{s1}i_{1}m_{i1}|\hat{\imath}_{1}\cdot\hat{s}_{1}|s_{1}m_{s1}^{\prime}i_{1}m_{i1}^{\prime}\rangle$
(15)
where
$\displaystyle\langle
s_{1}m_{s1}i_{1}m_{i1}|\hat{\imath}_{1}\cdot\hat{s}_{1}|s_{1}m_{s1}i_{1}m_{i1}\rangle$
$\displaystyle=$ $\displaystyle m_{i1}m_{s1};$ (16) $\displaystyle\langle
s_{1}m_{s1}i_{1}m_{i1}|\hat{\imath}_{1}\cdot\hat{s}_{1}|s_{1}m_{s1}\pm
1i_{1}m_{i1}\mp 1\rangle$ $\displaystyle=$
$\displaystyle\left[s_{1}(s_{1}+1)-m_{s1}(m_{s1}\pm
1)\right]^{1/2}\left[i_{1}(i_{1}+1)-m_{i1}(m_{i1}\mp 1)\right]^{1/2},$ (17)
and similarly for $\hat{\imath}_{2}\cdot\hat{s}_{2}$. The matrix elements of
$\hat{L}^{2}$ are simply
$\displaystyle\langle
s_{1}m_{s1}i_{1}m_{i1}s_{2}m_{s2}i_{2}m_{i2}LM_{L}|\hat{L}^{2}|s_{1}m_{s1}^{\prime}i_{1}m_{i1}^{\prime}s_{2}m_{s2}^{\prime}i_{2}m_{i2}^{\prime}L^{\prime}M_{L}^{\prime}\rangle=$
$\displaystyle\delta_{LL^{\prime}}\delta_{M_{L}M_{L}^{\prime}}\delta_{m_{s1}m_{s1}^{\prime}}\delta_{m_{i1}m_{i1}^{\prime}}\delta_{m_{s2}m_{s2}^{\prime}}\delta_{m_{i2}m_{i2}^{\prime}}L(L+1)\,,$
(18)
and those of the Zeeman operator are
$\displaystyle\langle
s_{1}m_{s1}i_{1}m_{i1}s_{2}m_{s2}i_{2}m_{i2}LM_{L}|\hat{g}_{e}\mu_{\rm
B}B\,\hat{s}_{zj}+g_{n}\mu_{\rm
B}B\,\hat{\imath}_{zj}|s_{1}m_{s1}^{\prime}i_{1}m_{i1}^{\prime}s_{2}m_{s2}^{\prime}i_{2}m_{i2}^{\prime}L^{\prime}M_{L}^{\prime}\rangle=$
$\displaystyle\delta_{LL^{\prime}}\delta_{M_{L}M_{L}^{\prime}}\delta_{m_{s1}m_{s1}^{\prime}}\delta_{m_{i1}m_{i1}^{\prime}}\delta_{m_{s2}m_{s2}^{\prime}}\delta_{m_{i2}m_{i2}^{\prime}}(g_{e}\mu_{\rm
B}B\,\hat{m}_{sj}+g_{n}\mu_{\rm B}B\,m_{ij}).$ (19)
All the calculations in the present paper used basis functions symmetrized for
exchange of two identical particles with $s_{1}=s_{2}=s$ and $i_{1}=i_{2}=i$.
For $m_{s1}=m_{s2}$ or $m_{i1}=m_{i2}$ the symmetrized functions are identical
to the unsymmetrized ones, except that only even $L$ is allowed for bosons and
only odd $L$ for fermions. For $m_{s1}\neq m_{s2}$ or $m_{i1}\neq m_{i2}$, the
symmetrized functions are
$\left[|sm_{s1}im_{i1}sm_{s2}im_{i2}LM_{L}\rangle\pm(-1)^{L}|sm_{s2}im_{i2}sm_{s1}im_{i1}LM_{L}\rangle\right]/\sqrt{2}\,,$
(20)
with the $+$ sign for bosons and the $-$ sign for fermions.
The Hamiltonian in the basis set, Eq. (4), used in the DVR calculations can be
derived from the Hamiltonian in the uncoupled basis by performing a unitary
transformation, namely, the transformation $|\alpha_{j}m_{fj}\rangle$ to
$|s_{j}m_{sj}\rangle|i_{j}m_{ij}\rangle$ for each of the two atoms ($j$=1 or
2). The transformation depends on the magnetic field strength Breit and Rabi
(1931). In practice, the eigenvectors for the monomer $h_{j}$ must be
evaluated. As $m_{fj}$ is conserved at most a 2$\times$2 matrix needs to be
diagonalized. Bose/Fermi symmetrization is ensured by
$[|\alpha_{1}m_{f1}\alpha_{2}m_{f2},LM_{L}\rangle\pm(-1)^{L}|\alpha_{2}m_{f2}\alpha_{1}m_{f1},LM_{L}\rangle]/\sqrt{2}$
when $\alpha_{1}\neq\alpha_{2}$ or $m_{f1}\neq m_{f2}$. The state with
$\alpha_{1}=\alpha_{2}$ and $m_{f1}=m_{f2}$ exists only for even (odd) $L$ for
bosonic (fermionic) atoms respectively.
## References
* Weber et al. (2003) T. Weber, J. Herbig, M. Mark, H.-C. Nägerl, and R. Grimm, Science 299, 232 (2003).
* Herbig et al. (2003) J. Herbig, T. Kraemer, M. Mark, T. Weber, C. Chin, H.-C. Nägerl, and R. Grimm, Science 301, 1510 (2003).
* Kraemer et al. (2006) T. Kraemer, M. Mark, P. Waldburger, J. G. Danzl, C. Chin, B. Engeser, A. D. Lange, K. Pilch, A. Jaakkola, H.-C. Nägerl, et al., Nature 440, 315 (2006).
* Szymaniec et al. (2007) K. Szymaniec, W. Chalupczak, E. Tiesinga, C. J. Williams, S. Weyers, and R. Wynands, Phys. Rev. Lett. 98, 153002 (2007).
* Hart et al. (2007) R. A. Hart, X. Xu, R. Legere, and K. Gibble, Nature 446, 892 (2007).
* Vanhaecke et al. (2004) N. Vanhaecke, C. Lisdat, B. T Jampens, D. Comparat, A. Crubellier, and P. Pillet, Eur. Phys. J. D 28, 351 (2004).
* Mark et al. (2007a) M. Mark, T. Kraemer, P. Waldburger, J. Herbig, C. Chin, H.-C. Nägerl, and R. Grimm, Phys. Rev. Lett. 99, 113201 (2007a).
* Mark et al. (2007b) M. Mark, F. Ferlaino, S. Knoop, J. G. Danzl, T. Kraemer, C. Chin, H.-C. Nägerl, and R. Grimm, Phys. Rev. A 76, 042514 (2007b).
* Chin et al. (2000) C. Chin, V. Vuletić, A. J. Kerman, and Chu, Phys. Rev. Lett. 85, 2717 (2000).
* Leo et al. (2000) P. J. Leo, C. J. Williams, and P. S. Julienne, Phys. Rev. Lett. 85, 2721 (2000).
* Chin et al. (2003) C. Chin, A. J. Kerman, V. Vuletić, and S. Chu, Phys. Rev. Lett. 90, 033201 (2003).
* Chin et al. (2004) C. Chin, V. Vuletić, A. J. Kerman, S. Chu, E. Tiesinga, P. J. Leo, and C. J. Williams, Phys. Rev. A 70, 032701 (2004).
* Lee et al. (2007) M. D. Lee, T. Köhler, and P. S. Julienne, Phys. Rev. A 76, 012720 (2007).
* Hutson (1994) J. M. Hutson, Comput. Phys. Commun. 84, 1 (1994).
* Köhler et al. (2006) T. Köhler, K. Góral, and P. S. Julienne, Rev. Mod. Phys. 78, 1311 (2006).
* Lang et al. (2008) F. Lang, P. van der Straten, B. Brandstätter, G. Thalhammer, K. Winkler, P. S. Julienne, R. Grimm, and J. Hecker Denschlag, Nature Phys. 4, 223 (2008).
* Stoof et al. (1988) H. T. C. Stoof, J. M. V. A. Koelman, and B. J. Verhaar, Phys. Rev. B 38, 4688 (1988).
* Colbert and Miller (1992) D. T. Colbert and W. H. Miller, J. Comp. Phys. 96, 1982 (1992).
* Tiesinga et al. (1998) E. Tiesinga, C. J. Williams, and P. S. Julienne, Phys. Rev. A 57, 4257 (1998).
* Anderson et al. (1992) E. Anderson, Z. Bai, C. Bischof, J. W. Demmell, J. J. Dongarra, J. Du Croz, A. Greenbaum, S. Hammarling, A. McKenney, S. Ostrouchov, et al., _LAPACK User’s Guide_ (SIAM, Philadelphia, 1992).
* (21) The scattering resonance positions given in Ref. Chin et al. (2004) were calculated by a propagator method, and are not subject to the basis set restrictions that apply to bound states.
* Johnson (1978) B. R. Johnson, J. Chem. Phys. 69, 4678 (1978).
* Hutson (1993) J. M. Hutson, _Bound computer program, version 5_ , distributed by Collaborative Computational Project No. 6 of the UK Engineering and Physical Sciences Research Council (1993).
* Alexander and Manolopoulos (1987) M. H. Alexander and D. E. Manolopoulos, J. Comp. Phys. 86, 2044 (1987).
* Krauss and Stevens (1990) M. Krauss and W. J. Stevens, J. Comp. Phys. 93, 4236 (1990).
* Moerdijk et al. (1995) A. J. Moerdijk, B. J. Verhaar, and A. Axelsson, Phys. Rev. A 51, 4852 (1995).
* Mies et al. (1996) F. H. Mies, C. J. Williams, P. S. Julienne, and M. Krauss, [J. Res. Natl. Inst. Stand. Technol 101, 521 (1996).
* Leo et al. (1998) P. J. Leo, E. Tiesinga, P. S. Julienne, and D. K. Walker, Phys. Rev. Lett. 81, 1389 (1998).
* Kotochigova et al. (2001) S. Kotochigova, E. Tiesinga, and P. S. Julienne, Phys. Rev. A 63, 012517 (2001).
* Ferlaino et al. (2008) F. Ferlaino, S. Knoop, M. Berninger, M. Mark, H.-C. Nägerl, and R. Grimm, _private communication_ (2008).
* Julienne et al. (2004) P. S. Julienne, E. Tiesinga, and T. Köhler, J. Mod. Optics 51, 1787 (2004).
* Lange et al. (2008) A. D. Lange, K. Pilch, A. Prantner, F. Ferlaino, B. Engeser, H. C. Nägerl, R. Grimm, and C. Chin, arXiv p. in preparation (2008).
* Ferlaino and Grimm (2008) F. Ferlaino and R. Grimm, _private communication_ (2008).
* Thompson et al. (2005) S. T. Thompson, E. Hodby, and C. E. Wieman, Phys. Rev. Lett. 95, 190404 (2005).
* Chin and Flambaum (2006) C. Chin and V. V. Flambaum, Phys. Rev. Lett. 96, 230801 (2006).
* DeMille et al. (2008) D. DeMille, S. Sainis, J. Sage, T. Bergeman, S. Kotochigova, and E. Tiesinga, Phys. Rev. Lett. 100, 043202 (2008).
* Breit and Rabi (1931) G. Breit and I. I. Rabi, Phys. Rev. A 38, 2082 (1931).
| arxiv-papers | 2008-06-16T13:56:08 | 2024-09-04T02:48:56.286234 | {
"license": "Public Domain",
"authors": "Jeremy M. Hutson, Eite Tiesinga, Paul S. Julienne",
"submitter": "Paul Julienne",
"url": "https://arxiv.org/abs/0806.2583"
} |
0806.2685 | RAId_aPS: MS/MS analysis with multiple scoring functions and
spectrum-specific statistics
Gelio Alves1, Aleksey Y. Ogurtsov1 and Yi-Kuo Yu1,∗
1 National Center for Biotechnology Information, National Library of Medicine,
National Institutes of Health, Bethesda, MD 20894
$\ast$ E-mail: yyu@ncbi.nlm.nih.gov
## Abstract
Statistically meaningful comparison/combination of peptide identification
results from various search methods is impeded by the lack of a universal
statistical standard. Providing an $E$-value calibration protocol, we
demonstrated earlier the feasibility of translating either the score or
heuristic $E$-value reported by any method into the textbook-defined
$E$-value, which may serve as the universal statistical standard. This
protocol, although robust, may lose spectrum-specific statistics and might
require a new calibration when changes in experimental setup occur. To
mitigate these issues, we developed a new MS/MS search tool, RAId_aPS, that is
able to provide spectrum-specific $E$-values for additive scoring functions.
Given a selection of scoring functions out of RAId score, K-score, Hyperscore
and XCorr, RAId_aPS generates the corresponding score histograms of all
possible peptides using dynamic programming. Using these score histograms to
assign $E$-values enables a calibration-free protocol for accurate
significance assignment for each scoring function. RAId_aPS features four
different modes: (i) compute the total number of possible peptides for a given
molecular mass range, (ii) generate the score histogram given a MS/MS spectrum
and a scoring function, (iii) reassign $E$-values for a list of candidate
peptides given a MS/MS spectrum and the scoring functions chosen, and (iv)
perform database searches using selected scoring functions. In modes (iii) and
(iv), RAId_aPS is also capable of combining results from different scoring
functions using spectrum-specific statistics.
The web link is
http://www.ncbi.nlm.nih.gov/CBBresearch/Yu/raid_aps/index.html.
Relevant binaries for Linux, Windows, and Mac OS X are available from the same
page.
## Introduction
Gaining popularity in biology over the last decade, mass spectrometry (MS) has
become the core technology in the field of proteomics. Although this
technology holds the promise to identity and quantify proteins in complex
biological mixtures/samples, such a goal has not yet been achieved due to the
presence of a number of difficulties ranging from experimental design and
experimental protocol standardization to data analysis [1, 2, 3]. This paper
mainly focuses on the data analysis, especially providing accurate statistical
significance assignments for peptide candidates in peptide identifications.
There are many peptide identification methods that are available to the
proteomics community. Because different identification methods process
(filter) the MS/MS spectra differently and also have different scoring
functions, it is natural for users to wish to compare search results from
different search methods or to combine these results to enhance identification
confidence. Nevertheless, there are important issues to be addressed prior to
successfully reaching this goal.
Due to intrinsic experimental variability, differences in the peptide
chemistry, peptide-peptide interactions, ionization sources, and mass
analyzers used, it is natural to expect among tandem mass spectra variations
in signal to noise ratios even when each peptide in the mixture has equal
molar concentration. That said, one anticipates the noise in a mass spectrum
to be spectrum-specific and the meaning of a search score depends on its
context, i.e., the spectrum used. That is, although search score can be used
to compare candidate peptides associated with the same query spectrum, it is
no longer a valid measure when one wishes to compare peptides identified
across spectra. Not only posing a challenge for ranking identified peptides
within a single experiment, this also raise a serious problem when one wishes
to compare or combine search results from different scoring functions (or
search methods).
If one knows how to translate the score or reported $E$-value of one method to
that of another method, or to a universal standard, it helps significantly the
task of comparing/combining search results. This is particularly true when one
wishes to combine search results from multiple scoring functions. We showed in
an earlier publication [4] that it is possible to use the textbook-defined
$E$-value as that universal standard. Providing an $E$-value calibration
protocol, we demonstrated the feasibility of translating either the score or
heuristic $E$-value reported by any method to the textbook-defined $E$-value,
the proposed universal statistical standard. This protocol, although robust,
may (a) lose spectrum-specific statistics, and may (b) require a new
calibration when changes in experimental set up occur.
Without attempting a universal statistical standard, several machine-learning
based approaches have been developed to either re-rank identified candidate
peptides [5, 6] or to combine search results from several search methods [7,
8]. These approaches require for their analyses training data set(s), either
pre-constructed or obtained on-the-fly, to aid the parameter selections for
their discriminant functions. For methods with feature vector (allowed to
contain some spectrum-specific quantities) updated on-the-fly [6, 8], the
spectrum-specific bias may be partially compensated, but not giving rise to
spectrum-specific statistics. This is because the feature vector, although may
be trained with spectrum-specific quantities, aims to categorize the whole
training set into finite number of classes but does not solely reflect the
properties of any individual spectrum.
To address the issue of spectrum-specific statistics, we developed a new MS/MS
search tool, RAId_aPS (a new module of the RAId suite), that is able to
provide spectrum-specific $E$-values for additive scoring functions that do
not have known theoretical score distributions. RAId_aPS provides the users
with four different modes to choose from: (i) compute the total number of
possible peptides (TNPP), (ii) generate score histogram, (iii) reassign
$E$-values, and (iv) database search. In modes (iii) and (iv), RAId_aPS is
also capable of combining results [9] from different scoring functions.
Founded on the algorithm published earlier [10], mode (i) is a straight
implementation of an existing idea. However, modes (ii) to (iv) are novel,
albeit at different levels. Mode (ii) uses the algorithm published earlier
[10], nevertheless, generating the all-possible-peptide (APP) score histograms
of different scoring functions was never done. Mode (iii) is novel from the
concept to its implementation. Modes (i-iii) do not have counter-parts in
other components of RAId suite. Mode (iv) is similar to RAId_DbS [11] in the
sense that it performs database searches. However, the difference between mode
(iv) of RAId_aPS and RAId_DbS lies in the use of statistics. The theoretical
score distribution of RAId_DbS fits score histogram of database peptides per
spectrum, while mode (iv) RAId_aPS uses score distributions of APP and is able
to provide statistics for multiple scoring functions.
The term “all possible peptides” (or APP) deserves some deliberation. The pool
of APP includes any linear arrangement of amino acids. Therefore, when
considering peptides of $L$ amino acids without modification, the APP pool
includes all the $(20)^{L}$ combinations. For the purpose of mass spectrometry
data analysis, instead of peptides with a fixed length one is more interested
in APP within a specified molecular mass range. The number of possible
peptides (PP) within a molecular mass range is much larger than the number of
database peptides within the same molecular mass range. For example, for the
molecular mass range $[2208{\rm Da},2304{\rm Da}]$, there are approximately
$10,000$ peptides in the Bos Taurus database, while there are in total
$1.385\times 10^{26}$ PP with lengths (number of amino acids) ranging from
$13$ to $39$.
Using dynamic programming, RAId_aPS generates the score histograms from
scoring APP. These score histograms are then used to assign accurate,
spectrum-specific $E$-values. Since RAId_aPS uses the score histograms, or the
(weighted) rank of each candidate peptide considered among APP, it is already
in conformity to the textbook defined $P$-value and thus there is no need to
translate the score or heuristic $E$-value into the universal standard.
Consequently, RAId_aPS is able to provide a calibration-free protocol for
accurate significance assignment and for combining search results.
In order to provide a clear exposition, it is necessary for us to go into some
technical details. Readers not interested in the details, however, may want to
read the results section first and then come back to read other sections. To
make the paper easier to read and more modular, we outline below the
organization of this paper. In the Technical Background section, we will
review the similarities and differences between two major approaches in
dealing with peptide identification statistics, describe how one may achieve
calibration-free, spectrum-specific statistics. In the Method section, we
first describe the dynamic programming algorithm needed to generate the score
distribution of APP, followed by spectral filtering procedures each associated
with a scoring function implemented. The incorporation of the four scoring
functions are then reported since some of them are nontrivial to encode via
dynamic programming. We then describe how the APP statistics are implemented
in practice, how to include modified amino acids in APP statistics, and how to
combine search results from different scoring functions. In the Results
section, we describe several tests performed using various modes of RAId_aPS,
as well as the $E$-value accuracy assessment. The paper is then concluded by
the Discussion section. All the technical aspects that are not most essential
in understanding the basic idea are provided either as supplementary texts or
supplementary figures. The most important message is that RAId_aPS serves as a
calibration-free, statistically sound method for comparing or combining search
results from different scoring functions.
## Technical Background
Since this paper is focused on the statistical aspect of peptide
identifications, we will start with such an example. In general, it is rather
easy to rank candidate peptides given a tandem mass spectrum. Once a scoring
function is selected to score peptides, qualified database peptides (those
within a molecular mass range and with correct enzymatic cleavages) can be
ranked based on their scores. However, it becomes difficult to rank candidate
peptides across all spectra. Although a number of publications have proposed
different ways tailored to deal with various aspects of this difficulty [4,
12], this problem remains very challenging. Should one take the best candidate
peptide per spectrum and then postprocess to globally re-rank those best hits
or should one devise something different to achieve the maximum robustness?
Instead of discussing the differences between these two possibilities, we
first wish to point out a common theme that is often unnoticed: spectrum-
specificity.
### Spectrum Specificity
As mentioned in the Introduction section, spectrum-specificity has not been
emphasized enough. However, there does exist evidence of community’s
recognition of this point. For example, by picking the best hit out of each
spectrum, one is acknowledging spectrum-specificity, because one has chosen to
keep the best candidate per spectrum regardless of the fact that the best hit
in one spectrum might have lower score than the second best hit in some other
spectrum. In other words, by picking only the best hits one has endorsed the
view that the score should not be used as an objective measure of
identification confidence across all candidate peptides; or more precisely,
the meaning of score depends on its context, i.e., the spectrum used.
There exists another route to apply the concept of spectrum-specificity. That
is to use a spectrum-specific score distribution to assign an $E$-value to
each candidate peptide of a spectrum. Although the term spectrum-specific
statistics was not explicitly mentioned, the proposal of Fenyo and Beavis [13]
to fit per spectrum the tail of score distribution to an exponential
represents the first attempt, to the best of our knowledge, in this direction.
The concept of spectrum-specific statistics was formally introduced by Alves
and Yu [14]. The same group also developed RAId_DbS [11], so far the only
database search tool with a theoretically derived spectrum-specific score
distribution. The importance of spectrum-specific statistics is then
emphasized through a series of publications [4, 9, 11, 15]. The key point of
this type of approach is to exemplify spectrum-specificity via spectrum-
specific score statistics. After describing the common theme, spectrum
specificity, we now turn to features associated with different types of
approaches to elucidate the usefulness of an even more general statistical
framework.
### Best hit per spectrum versus Accurate $E$-value
When keeping only the best hit per spectrum, a global re-ranking among those
best hits becomes necessary in order to decide which best hits to trust over
the others. This is usually achieved in one of the two ways to be described.
The first possible choice is to use the original score in conjunction with
either false discovery rate (FDR) or $q$-value analysis through introduction
of a decoy database. The second choice is to use some kind of refined score in
conjunction with an empirical expectation-maximization-based Bayesian approach
[5]. This global re-ranking type of strategies, unfortunately, makes
assumptions contradicting spectrum-specificity, a fundamental fact that is
respected when only the best hit per spectrum is retained.
In the FDR (be it global or local) or $q$-value analyses, one pools together
the best hits across spectra and order the hits by their scores. This
contradicts the idea of picking best hit per spectrum, which essentially
endorses the notion that the meaning of a peptide score is spectrum-dependent
and can’t be used to rank peptides globally across spectra. For the Bayesian
type of analyses [5], one assumes the existence of two score distributions:
one for the score of correctly identified spectra, in terms of best hit, and
another for the score of incorrectly identified spectra. This means that all
correctly identified spectra –in terms of best hit– should be ranked according
to the best hit’s refined score, implying that one may use the refined score
to assign relative identification confidence across spectra. This again
contradicts the idea that the meaning of a peptide score is spectrum-
dependent. Furthermore, to perform the expectation maximization procedure, one
often needs to assume the parametric forms of the two distribution functions,
which might not be applicable to all scoring functions.
When the reported spectrum-specific $E$-value (assigned to each of the
candidate peptides per spectrum) is in agreement with its definition, it can
serve as an objective measure of identification confidence. For a given
spectrum and a score threshold, the $E$-value associated with that score
threshold is defined to be the expected number of false hits that have score
better than or equal to that threshold. In simple terms, the $E$-value
associated with a candidate peptide in the database may be viewed as the
number of false positive hits anticipated, from querying a spectrum, before
calling the peptide at hand a true positive hit. However, a previous study
[11] showed that most $E$-value reporting methods investigated report
inaccurate $E$-values. To rectify this problem, we provided a protocol [4] to
calibrate $E$-values reported by other search methods, including search tools
that don’t report $E$-values such as ProbID [16] and SEQUEST [17]. However,
the calibration procedure cannot restore/recreate spectrum-specificity for
methods not reporting $E$-values or reporting $E$-values that are not obtained
via characterizing the score histogram for each spectrum (spectrum-specific
score modelling).
Nevertheless, spectrum-specific statistics can be obtained provided that one
extracts statistical significance from the score histogram for each spectrum
[4]. A recent reimplementation [18, 19, 20] of the SEQUEST XCorr follows
exactly this idea. To avoid possible confusion, however, we must first note
that the $p^{*}$-value in reference [18] is actually the $E$-value. Authors of
reference [18] assume that the XCorr from every spectrum can be fitted by a
stretched exponential without providing, like most other methods, a measure on
the agreement between the best fitted parametric form and the score
distribution per spectrum. To ensure the accuracy of statistics, a measure of
the goodness of the model [21, 11] is actually necessary even for scoring
systems that have a theoretically characterized distribution. This is because
very biased sampling might lead to a discrepancy between the theoretical
distribution and the score distribution, not to mention a discrepancy between
a fitted parametric form and the score distribution.
One way to circumvent the aforementioned problem is to apply a target-decoy
strategy at the per spectrum level. This means that one uses the hits from
decoy database to estimate the identification confidence of peptides from the
target database. This approach, unfortunately, is not computationally
efficient because one will need a decoy database that is much larger than the
target database in order to have a good estimate of the $E$-value for each hit
in the target database. For example, if the number of qualified peptides in
the decoy database is $1,000$ times that in the target database, and if a
peptide in the target database scores between the third and the fourth decoy
hits, then that peptide will acquire an $E$-value between $3\times 10^{-3}$
and $4\times 10^{-3}$. And if there are target hits that score better than the
best decoy hit, all one can say is that they all have $E$-values smaller than
$10^{-3}$. If one keeps increasing the size of the decoy database, one will
eventually be able to globally rank the candidate peptides from all spectra
using $E$-value. However, computational efficiency prevents us from using this
strategy.
These aforementioned problems associated with obtaining spectrum-specific
statistics can be avoided provided that one uses a search method that has a
theoretically derived score distribution [11]. However, restricting to methods
that have theoretically derived statistics is not necessarily the best
strategy since each search method does have different strengths [9, 22]. It
can be advantageous to combine different types of search scores. Therefore,
for assigning peptides’ identification confidence, it is desirable to have a
unified framework which we now turn to.
### APP Statistics (calibration-free)
Alves and Yu in 2005 proposed [14] using the de novo rank as the statistical
significance measure. Despite the simplicity of this idea, it was never fully
carried out. Since it is this idea that inspired the development of RAId_aPS,
we need to describe the basic concept to some detail so that various
extensions employed in RAId_aPS can be properly explained.
The fundamental idea is as follows. For a given MS/MS spectrum $\sigma$ with
parent molecular mass $MW$ and a given mass error tolerance $\delta$, we
denote by $\Pi(\sigma,\delta)$ the set of APP subjected to enzymatic cleavage
condition in the mass range $[MW-\delta,MW+\delta]$. We also denote by
$\Delta(\sigma,\delta,C)$ the set of peptides in the (target) database,
subjected to a set of conditions $C$, in the mass range
$[MW-\delta,MW+\delta]$. The set of conditions $C$ may contain, for example,
the enzymatic cleavage constraints, number of miscleavage sites per peptide
allowed, and others [23]. The following argument is also applicable to the
case when one wishes to weight each peptide in the APP set by its elemental
composition. This may be used to form a background model mimicking the amino
acid composition in the target database [10, 24].
Let $N(S,\sigma)$ be the (weighted) number of peptides out of
$\Pi(\sigma,\delta)$ that have scores greater than or equal to $S$. We then
define the APP $P$-value corresponding to score $S$ by
$N(S,\sigma)/|\Pi(\sigma,\delta)|$, with $|\Pi(\sigma,\delta)|$ representing
the total (weighted) number of peptides in the set $\Pi(\sigma,\delta)$. In
general, for a given spectrum $\sigma$ and a score cutoff $S$, the $P$-value
$P(S|\sigma)$ refers to the probability for a qualified random peptide to
attain a score greater than or equal to $S$ when using spectrum $\sigma$ as a
query. If a database contains $N_{d}$ qualified, unrelated random peptides,
one will expect to have $E(S|\sigma)=N_{d}P(S|\sigma)$ number of random
peptides to have quality score greater than or equal to $S$. This expectation
value $E(S|\sigma)$ is by definition the $E$-value associated with score
cutoff $S$.
The $E$-value associated with a peptide of score $S$ using the APP $P$-value
will therefore be
$E(S|\sigma)=|\Delta(\sigma,\delta,C)|\frac{N(S,\sigma)}{|\Pi(\sigma,\delta)|}$
where the spectrum-specific $E(S|\sigma)$ represents the $E$-value for a hit
with score $S$ when the spectrum $\sigma$ is used as the query and
$|\Delta(\sigma,\delta,C)|$ represents the total number of peptides in the set
$\Delta(\sigma,\delta,C)$. When cast in the aspect of per spectrum target-
decoy approach, $\Pi(\sigma,\delta)\setminus\Delta(\sigma,\delta,C)$
represents the largest possible decoy database, which is supposed to provide
numerically the finest $E$-values for candidate peptides in the target
database. (The symbol $\setminus$ is called “setminus”. $A\setminus B$ can be
called $A$ minus $B$ in the set sense or called complement of $B$ provided
that set $A$ is the largest set considered and every set is a subset of $A$.)
Let $N^{\prime}(S|\sigma)$ be the (weighted) number of peptide hits in the
target database with score greater than $S$. The per spectrum target-decoy
approach will have
$E(S|\sigma)=|\Delta(\sigma,\delta,C)|\frac{N(S,\sigma)-N^{\prime}(S,\sigma)}{|\Pi(\sigma,\delta)\setminus\Delta(\sigma,\delta,C)|}\approx|\Delta(\sigma,\delta,C)|\frac{N(S,\sigma)}{|\Pi(\sigma,\delta)|}$
where the last result comes from $N^{\prime}(S,\sigma)\ll N(S,\sigma)$ and
$|\Pi(\sigma,\delta)\setminus\Delta(\sigma,\delta,C)|\approx|\Pi(\sigma,\delta)|$
for any practical applications.
For a typical molecular mass of $1500$ Dalton (Da) and in the absence of
weighting, $|\Pi(\sigma,\pm 1Da)|\approx 5\times 10^{15}$. For a typical
organismal database, such as that of Homo sapiens, the total number of
peptides within the molecular mass range without any condition is only
$|\Delta(\sigma,\pm 1Da)|\approx 3\times 10^{3}$. Therefore, $5\times
10^{15}\geq|\Pi(\sigma,\pm 1Da)\setminus\Delta(\sigma,\pm 1Da,C)|\geq 5\times
10^{15}-3\times 10^{3}$, and $|\Pi(\sigma,\pm 1Da)\setminus\Delta(\sigma,\pm
1Da,C)|\approx 5\times 10^{15}$. In the presence of peptide weighting, one
still has $|\Pi(\sigma,\pm 1Da)|/|\Pi(\sigma,\pm
1Da)\setminus\Delta(\sigma,\pm 1Da,C)|\approx 1$. Therefore,
$|\Pi(\sigma,\delta)\setminus\Delta(\sigma,\delta,C)|\approx|\Pi(\sigma,\delta)|$.
As for $N^{\prime}(S,\sigma)$ versus $N(S,\sigma)$, by definition
$N^{\prime}=0$ for best target hit and $N(S,\sigma)$ typically increases much
faster than $N^{\prime}(S,\sigma)$ when $S$ is lowered, thus
$N^{\prime}(S,\sigma)\ll N(S,\sigma)$, a fact also observed in reference [24].
Consequently, $N(S,\sigma)-N^{\prime}(S,\sigma)\approx N(S,\sigma)$ is a very
good approximation. Therefore, the APP statistics also serve as the best per
spectrum target-decoy statistics. The only question now is how does one get
the score distribution of APP?
It turns out that if the score of a peptide is the sum of local contributions,
meaning each term in the sum is uniquely determined by specifying a fragment’s
m/z value, then it is possible to construct the score histogram of APP via
dynamic programming [10, 24]. When there exists intrinsically nonlocal
contribution in peptide scoring, it is no longer possible to obtain the full
histogram by dynamic programming. However, it is still possible to estimate
the de novo rank via a scaling approach [15] similar to that used in
statistical physics. The key point, as will be shown later, is that for the
four scoring functions implemented in RAId_aPS, by using the APP statistics,
it is no longer critical to theoretically characterize the score distribution
obtained from the database search. This is because the $E$-value obtained via
RAId_aPS does agree well with the textbook definition. The APP statistics
employed by RAId_aPS may be extended to provide robust spectrum-specific
statistics for scoring functions that do not have theoretically characterized
score distributions. One advantage to having a method that can provide robust
spectrum-specific statistics for different scoring functions is that if the
$E$-value reported by each method agrees with its definition, one can compare
and combine search results from different search methods [9].
## Methods
### Basic Dynamic Programming Algorithm
To generate the score histogram of APP in a speedy manner, RAId_aPS does not
score every possible peptide individually. As a matter of fact, it is
impossible to score every possible peptide individually. For example, consider
a typical parent ion molecular mass of $1,500$ Da. It can be shown that the
TNPP within $1$ Da of this molecular mass is more than $10^{15}$. Even if one
has a simple scoring function and a fast computer that can score one hundred
millions peptides per second, it will take more than $116$ days of computer
time to generate the score histogram for a single spectrum.
In real application, one needs to analyze a spectrum in a short time. How
could one achieve this? One may use a 1-dimensional (1D) mass grid to
encode/score APP [10, 24]. At each mass index of the grid, the local score
contribution associated with all partial peptides reaching that location is
computed only once and this information may be propagated forward to other
mass entries via dynamic programming, making it possible to generate the score
histogram of APP without individually scoring all peptides. In the score
histogram, instead of counting number of peptides associated with a certain
score, it is also possible to weight each peptide sequence according to its
elemental composition. For a peptide sequence $[a_{1},a_{2},\ldots,a_{M}]$,
one may assign it a weight [10, 24] $p(a_{1})p(a_{2})\ldots p(a_{M})$ with
$p(a_{i})$ being the emitting probability of amino acid $a_{i}$.
For illustration purposes, the mass grid of 1Da resolution is used in Figure
1. Each mass index contains a score histogram, with each entry in the left
column indicating a score and the corresponding entry at the right column
recording the number of partial peptides reaching that mass index with that
score. The score histogram is obtained using a backtracking update rule. For
example, at the mass grid $558$, the local score contribution from evidence
peaks in the spectrum is assumed to contribute $\Delta$ amount of score.
Looking back to mass grid $501$ ($57$ Da less than $558$ Da), one knows that
by attaching a glycine residue to the partial peptides reaching mass index
$501$ one will then advance these peptides to index $558$. Similarly, any
partial peptides reaching mass index $487$ will move to mass index $558$ by
adding an alanine residue. Therefore, at mass index $558$ the score histogram
is the superposition of score histograms associated with the other twenty
lighter mass grids corresponding respectively to the twenty amino acids. For
simplicity, the illustration is drawn as if there are only two amino acids,
glycine and alanine. When one weights each peptide by its elemental
composition, the counts next to the scores in the histogram are weighted and
no longer integers. For example, the weighted count $n(558)$ at mass index
$558$ will be given by $n(558)=\sum_{a=1}^{20}p_{a}\;n(558-m_{a})$ where
$m_{a}$ is the mass of amino acid $a$ rounded to the nearest Da and $p_{a}$ is
the emitting probability associated with amino acid $a$. In addition to
attaching a score histogram to each mass grid, one may also include other
internal structures such as peptide lengths, peak counts, etc. as shown in the
caption of Figure 1. When one suppresses the score and only counts number of
partial peptides reaching a certain mass index, the update rule readily
provides the total number of peptides within a given mass range.
### Spectral Filtering
Before describing the scoring functions, the major component of peptide
database search tools, we first mention spectral filtering, an often under-
emphasized but equally important ingredient. Starting with a raw tandem mass
spectrum, spectral filtering produces a processed spectrum that is used to
score candidate peptides in the database. Apparently, information kept in the
processed spectrum plays an important role in the effectiveness of a tool’s
performance in database searches. Customized for different scoring functions,
different filtering strategies are employed by different search tools. In
order for RAId_aPS to capture the essence of a scoring function, it is very
important for RAId_aPS to produce, for every input raw spectrum, a filtered
spectrum that is as close as possible to the one produced by other search
tool’s filtering protocol. For most search tools, the filtering heuristics are
not clearly documented. For that reason, it becomes necessary to delve into
the source code of the search program to find out each method’s spectral
filtering protocol. We are thus limited to search tools whose source programs
are available or those with filtering strategies clearly documented.
For RAId score, the spectral filtering strategy was described in an earlier
publication [11]. For Hyperscore [25], XCorr [17], and K-score [26, 27], the
details of spectral filtering will be described in Text S1. Since the SEQUEST
source code is not available, for XCorr score we attempt to replicate the
filtering of Crux [20], a search method that has been shown to reproduce
SEQUEST XCorr [20]. That the filtering strategies extracted are accurate can
be seen from Figure S1. The spectral correlation histograms between the
filtered spectra produced by RAId_aPS’s Hyperscore/XCorr/K-score with the
filtered spectra from X!Tandem/Crux/X!Tandem(with K-score plug-in) show that
RAId_aPS is able to produce filtered spectra identical to those generated by
the canonical programs. Although the spectral filtering strategies associated
with various search tools investigated seem stable, it is still possible that
the developers may change their filtering strategies in the future. When that
happens, one should be able to update RAId_aPS to reflect the filtering
changes provided that the source programs are still accessible and clearly
documented.
Instead of elaborating on various filtering strategies, let us first use a
experimentally obtained spectrum to demonstrate the effect of spectral
filtering employed by different methods. Figure 2 shows the raw spectrum, and
the filtered spectra processed by the four scoring methods mentioned. The
general trend is as follows: RAId score usually produces the filtered spectrum
that resembles the original spectrum the most; Hyperscore filtering also
produces a processed spectrum that is similar to the original spectrum; for
XCorr and K-score the filtered spectra in general look quite different from
the original spectrum. The differences in the filtered spectra might be a
major factor contributing to the fact that different search methods have
different and often complementary strengths. The correlation between any pair
of filtering strategies can be quantified. Starting with a large set of raw
spectra, one may process these spectra with a pair of different methods. For
each raw spectrum, one obtains two different filtered spectra and can compute
their correlation. The correlation between every pair of filtered spectra can
then be collected to form the correlation histogram, reflecting the
correlation between a pair of filtering strategies. Figure 3 and Figure S2
exhibit the correlation histograms between each pair of filtering strategies
using different data types: centroid (A1-A4 of ISB data set [28], Figure 3)
and profile (NHLBI data set [4], Figure S2). The large correlation between
XCorr and K-score may be the cause of their significant scoring correlation
observed.
### Scoring Functions
To better express the scoring functions, let us first define the following
notations. For a given peptide $\pi$, the set of corresponding theoretical
mass over charge (m/z) ratios taken into consideration by a scoring function
is called $T(\pi)$, which is also used to indicate the number of elements in
the set $T(\pi)$ whenever no confusion arises. The set $T(\pi)$ varies from
software to software. However, the fragmentation series
$(a_{n},b_{n},b_{n}\\!\\!-\\!\\!18,b_{n}\\!\\!-\\!\\!17,c_{n},x_{n},y_{n},y_{n}\\!\\!-\\!\\!18,y_{n}\\!\\!-\\!\\!17,z_{n})$
include what most methods consider. The Heaviside step function $\theta(x)$ is
defined by $\theta(x<0)=0$ and $\theta(x>0)=1$. We introduce $I_{i}$ as a
shorthand notation for $I(m_{i})$, the peak intensity associated with
theoretical mass $m_{i}$ in the processed spectrum. In an experimental
spectrum, the mass giving rise to $I_{i}$ usually does not coincide with
$m_{i}$. The absolute difference between the experimental mass (giving rise to
$I_{i}$) and the theoretical mass $m_{i}$ is denoted by $\Delta m_{i}$. The
notation $I^{\prime}_{i}$ is used in place of $I_{i}$ when the preprocessing
of the spectrum involves a nonlinear transformation of the peak intensity or
involves generation of additional peaks. We now list the four different
scoring function implemented:
$\displaystyle{\rm RAId~{}}S(\pi)$ $\displaystyle=$
$\displaystyle\frac{1}{T(\pi)}\sum_{i=1}^{T(\pi)}\ln(I_{i})\;e^{-\Delta
m_{i}}\theta(1-\Delta m_{i})$ (1) $\displaystyle{\rm Hyperscore~{}}S(\pi)$
$\displaystyle=$ $\displaystyle
4\log_{10}\left[\left(\sum_{i=1}^{T(\pi)}\;I^{\prime}_{i}\right)b\,!\;y!\right]$
(2) $\displaystyle{\rm XCorr~{}}S(\pi)$ $\displaystyle=$ $\displaystyle{1\over
10000}\sum_{i=1}^{T(\pi)}\;w_{i}I^{\prime}_{i}$ (3) $\displaystyle{\rm
K\\!\\!-\\!score~{}}S(\pi)$ $\displaystyle=$ $\displaystyle{1000\,\ln(l)\over
3\sqrt{l}}\sum_{i=1}^{T(\pi)}\;w_{i}I^{\prime}_{i}$ (4)
The first scoring function listed is employed by RAId_DbS [11]; the second one
mimicks the Hyperscore (${\mathrm{X}}_{II}$) of X!Tandem [13]; the third one
mimicks the XCorr score used in SEQUEST and is similar to what was implemented
in Crux [20, 19]; the last one mimicks K-score [26], a plug-in for X!Tandem.
For the RAId score, the set $T(\pi)$ includes only the $b$\- and $y$-series
peaks. For the Hyperscore, $T(\pi)$ includes $\\{b_{n},~{}y_{n}\\}$. For
XCorr, $T(\pi)$ includes
$\\{b_{n},~{}y_{n},~{}b_{n}-1,~{}b_{n}+1,~{}y_{n}-1,~{}y_{n}+1,~{}b_{n}-18,~{}b_{n}-17,y_{n}-17,~{}a_{n}\\}$
with the corresponding weights given by
$\\{50,~{}50,~{}25,~{}25,~{}25,~{}25,~{}10,~{}10,~{}10,~{}10\\}$. For K-score,
$T(\pi)$ includes
$\\{b_{n},~{}y_{n},~{}b_{n}-1,~{}b_{n}+1,~{}y_{n}-1,~{}y_{n}+1\\}$ with the
corresponding weights given by $\\{1,~{}1,~{}0.5,~{}0.5,~{}0.5,~{}0.5\\}$. To
speed up the code, we have chosen to rescale the weights for XCorr (see the
“Crux Filtering and XCorr” section of Text S1 for detail).
Very often it is useful to include the peptide length in the scoring of a
peptide. Using RAId score as a simple example, two peptides of length $11$ and
$16$ may achieve the same raw score $S^{\prime}_{11}=S^{\prime}_{16}=10$, sum
of the logarithm of evidence peak intensity. A longer peptide consists of a
longer list of theoretical peaks to look for and may thus score higher by
chance. RAId_DbS scoring function [11] deals with this issue by dividing the
raw score by the length of the theoretical peak list. Upon doing so, one has
$S_{11}=S^{\prime}_{11}/(2\times(11-1))=1/2$ and
$S_{16}=S^{\prime}_{16}/(2\times(16-1))=1/3$. This score normalization may
help in discriminating true positives from false positives. The other scoring
function utilizing the peptide length information is the K-score. Hyperscore,
employed by X!Tandem, uses a slightly different score renormalization
strategy. Inside the logarithm, the Hyperscore contains two factorials, $b!$
and $y!$. For each candidate peptide, $b$ ($y$) represents the total number of
$b$-series ($y$-series) evidence peaks found in the spectrum. At any specified
mass index in the mass grid, unlike the peak intensity associated with that
index, neither the peptide length nor the total number of the b (y) peaks has
a unique corresponding value. Therefore, one needs to extend the basic
algorithm outlined in the previous subsection to accommodate these additional
information needed for scoring.
As documented in reference [10], it is possible to introduce additional
structures in the score histogram associated with each mass index. The
flexibility to introduce additional structures of various dimensions makes
RAId_aPS a versatile tool: it can accommodate the scoring functions that
utilize length information or the number of $b$-series ($y$-series) peaks to
compute the final peptide score. Using peptide length as an example, Figure 1
demonstrates the inclusion of additional structures. More detailed exposition
about the inclusion of internal structures can be found in reference [10].
Although the spectral filtering parts of various scoring functions are
replicated exactly, a candidate peptide may receive different scores from
RAId_aPS and the original programs. This phenomenon can be seen in Figure 4:
the ordinate of each data point displays the search score of the best hit of a
centroid spectrum using the original programs, while the abscissa of the same
data point shows the score reported by RAId_aPS. The corresponding plots for
profile data are shown in Figure S3.
The major source of score difference is due to RAId_aPS’s omission of
heuristics while implementing a published scoring function. For each scoring
function, many scoring heuristics are present in the source code. While some
of the heuristics cannot be included via dynamic programming, all these
heuristics are either not described or not justified in the original papers.
For these reasons, RAId_aPS does not include those unpublished heuristics.
Therefore, the Hyperscore/XCorr/K-score scoring functions implemented in
RAId_aPS should be regarded as our attempt to mimick the original
Hyperscore/XCorr/K-score scoring functions. Although the scoring functions we
implemented are not exact replicas of the original ones, due to omission of
heuristics, we can see from Figure 4 (and also Figure S3 when tested on
profile data) that there exist strong correlation between each scoring
function implemented in RAId_aPS and the original, corresponding scoring
function. In other words, the scoring functions implemented in RAId_aPS do
capture the essence of these original scoring functions.
### APP Statistics: practical implementation
In the APP statistics section, we described how to use APP statistics to
obtain $P$-values and $E$-values with or without weighting each peptide by its
elemental composition. In this subsection, we will complement the theoretical
presentation by describing some pragmatic aspects of the implementation.
In order to build the score histogram quickly, it is necessary to discretize
the score, thereby compromising to some degree the score precision. However,
this rounding of scores does not affect peptide scoring when using RAId_aPS as
a database search tool or a tool to provide statistical significance for a
list of peptides. Specifically, the evidence score collected at each mass
index is stored in two formats: one with much higher precision and the other
rounded to nearest integer. The rounded values are used in dynamic programming
to propagate the score histogram forward, facilitating a speedy construction
of the score histogram. The slight error introduced in individual peptide
scoring does not influence the accuracy of the score histogram much since
these errors largely cancel each other when lumping the scores into a
histogram. In the database search mode, RAId_aPS will sum the high precision
evidence scores in the mass indices traversed by the candidate peptide being
scored. Therefore the score associated with each candidate peptide in the
database search mode has a better resolution than that in the score histogram.
To obtain the statistical significance associated with each candidate peptide,
RAId_aPS performs an interpolation procedure to obtain the $P$-value,
$P(S,\sigma)=\frac{N(S,\sigma)}{|\Pi(\sigma,\delta)|}\;.$
Multiplying the $P$-value by the number of qualified peptides
$|\Delta(\sigma,\delta,C)|$ in the target database provides the $E$-value
$E(S,\sigma)=|\Delta(\sigma,\delta,C)|~{}P(S,\sigma)\;.$
### APP Statistics including PTM amino acids
Since proteins do contain PTM amino acids, it is important for peptide
identification tools to consider amino acid modifications in the statistical
analysis. By scoring only qualified peptides, database search methods have
little problem including PTM amino acids provided that the score distribution
is theoretically characterizable. For APP based statistics, even though the
score distribution is not always characterizable, information from qualified
peptides in database search may be used to generate the emission probabilities
of all the amino acids, PTMs included, needed for APP based statistics.
Given a parent ion mass and a database, once the allowable PTMs are specified,
the number of peptides along with possible types of modifications are fixed.
This renders a parent-ion-mass specific and database specific emission
probabilities for PTMs. Nevertheless, the list of qualified peptides may vary
with molecular mass error tolerance while the allowable PTMs may also vary
with users’ specification for a search. Once the list of qualified peptides
for a spectrum is given, the emission probabilities of each amino acid
(including PTMs) are computed as follows: for each amino acid $B$, RAId_aPS
first counts the number of occurrences of the unmodified amino acids $n(B)$
and the number of occurrences $n(B_{i})$ of $B$ modified into a different form
$B_{i}$, with $i=1,\ldots,k$. RAId_aPS then proportionally distributes the
emission probability $p_{0}(B)$ associated with amino acid $B$ to all the
possible modified forms using the following formulas
$\displaystyle p(B)$ $\displaystyle=$
$\displaystyle\frac{n(B)+1}{n(B)+1+\sum_{i=1}^{k}n(B_{k})}\;p_{0}(B)$ (5)
$\displaystyle p(B_{i})$ $\displaystyle=$
$\displaystyle\frac{n(B_{i})}{n(B)+1+\sum_{i=1}^{k}n(B_{k})}\;p_{0}(B)\;.$ (6)
Effectively, one pseudocount is always given to each unmodified amino acid.
Therefore, for a given list of peptides, RAId_aPS will count the total number
of distinct amino acids modifications. In principle, RAId_aPS can incorporate
all those modified amino acids in the score histogram construction. However,
for reasons to be described below, RAId_aPS retains no more than the ten most
abundant PTMs in calculating the new emission probabilities. First, the
estimated emission probabilities of PTMs become less trustworthy when the
occurrences of those PTMs are rare. Second, inclusion of many PTMs can slow
down the process, although not very much. Assume that one incorporates
${\mathcal{M}}$ modified amino acids in the score histogram construction, the
number of trace backs per mass index becomes $20+{\mathcal{M}}$ instead of
$20$. This introduces a factor of $(20+{\mathcal{M}})/20$ compared to the
original construction. Further, the size of score array associated with each
mass index needs to be larger than before and thus require more time to
compound the score histogram. This approximately introduce another factor of
$(20+{\mathcal{M}})/20$ to the computation speed. Thus, introducing
${\mathcal{M}}$ modifications will introduce a multiplicative factor of
$(1+\frac{\mathcal{M}}{20})^{2}$ to the computation time. To ensure that the
average run time does not grow more than two fold, we set the maximum
${\mathcal{M}}$ allowed to be ten. The new set of normalized background
frequencies (with the most abundant PTMs included) may then be fed into
RAId_aPS to compute the corresponding APP score histogram. The histogram
obtained is then used to calculate the statistical significance of each
reported peptide.
Although rare PTMs in the peptide list might be omitted in constructing the
APP score histogram, the impact on the statistical significance accuracy is
minute. For if one were to include those PTMs, due to their small normalized
emission probabilities, peptides containing those PTMs would be weighted
substantially less than others and thus would not significantly affect the
shape of the score histogram. As for the emission probability $p_{0}(B)$
—needed in eqs. (5-6)— associated with amino acid $B$, one may use either
known amino acid background frequencies such as the Robinson-Robinson [29]
frequencies or can calculate the number of occurrences of all amino acids in a
parent-ion-mass-specific and database-specific manner. The former approach is
adopted by RAId_aPS when the number of peptides (provided by the user or
extracted from the database) is less than $2,000$; otherwise, the latter
approach is employed. There exists, of course, room for improvement in terms
of including PTMs in the APP statistics. Alternatives are currently under
investigations.
### Combining Search Results from Different Scoring Functions
When the user select multiple scoring functions in mode (iii) and mode (iv),
RAId_aPS is able to combine statistical significances reported by the
different scoring functions. For database search (mode (iv)), the protocol to
combine search results is identical to what was described before [9]. In this
section, we will briefly review this method.
For a given spectrum $\sigma$, to combine search results from $m$ scoring
functions (say scoring function $A_{1}$, $\ldots$, $A_{m}$), we first
construct a union peptide list
${\mathcal{L}}(\sigma)\equiv{\mathcal{L}}_{A_{1}}(\sigma)\cup\ldots\cup{\mathcal{L}}_{A_{m}}(\sigma)$,
where ${\mathcal{L}}_{A_{i}}(\sigma)$ is the reported list of peptide hits by
method $A_{i}$ for spectrum $\sigma$. A peptide in the union list has at least
one, and may have up to $m$ $E$-values derived from APP $P$-values, depending
on how many scoring functions reported that specific peptide in their
candidate lists. Each of the $E$-values associated with a peptide will be
first transformed into a database $P$-value [9], representing the probability
of seeing at least one hit in a given random database with quality score
larger than or equal to $S$. If one assumes that the occurrence of a high-
scoring random hit is a rare event and thus can be modeled by a Poisson
process with expected number of occurrence $E(S|\sigma)$, one may obtain the
database $P$-value mentioned earlier via
$P_{\rm db}(S|\sigma)=1-e^{-E(S|\sigma)}\;.$ (7)
The database $P$-value of peptide $\pi$ is set to one for methods that do not
report $\pi$ as a candidate. After this procedure, each peptide in the list
${\mathcal{L}}(\sigma)$ has $m$ database $P$-values
$(P_{1},P_{2},\ldots,P_{m})$. Assume that these $P$-values are independent,
the combined $P$-value (with $\tau\equiv\prod_{i=1}^{m}P_{i}$) for peptide
$\pi$ is given by [9]
$P_{\rm comb}(\pi)=\tau\sum_{k=0}^{m}{[\ln(1/\tau)]^{k}\over k!}\;$ (8)
Once $P_{\rm comb}(\pi)$ is obtained, we may invert the formula in Eq. (7) to
get a combined $E$-value $E_{\rm comb}$ via
$E_{\rm comb}(\pi)=\ln\left({1\over 1-P_{\rm comb}(\pi)}\right)\;.$ (9)
We then use $E_{\rm comb}(\pi)$ as the final $E$-value to determine the
statistical significance of peptide candidate $\pi$, similar to what is used
in reference [30]. From a theoretical stand point, one might ask whether or
not eq. (8) always gives rise to a smaller combined $P$-value than any of the
input $P$-values. The answer is no. For example, consider $P_{1}=p<1$ and
$P_{2}=1$. One then has combined $P$-value $p[1+\ln(1/p)]$ larger than
$P_{1}$. Readers interested in more details are referred to Appendix B of
reference [9].
The combining $P$-value strategy outlined by eqs. (7-9) is founded on the
assumption that $P$-values resulting from different search scores are
independent. That is, the resulting significance assignment is valid only when
scoring functions considered are uncorrelated, or at most weakly correlated.
In our earlier investigation [9], we found that although many scoring
functions are looking for similar scoring evidences, the pairwise correlations
among scoring functions investigated are weak, perhaps due to different
spectral filtering methods employed. The weak pairwise correlations among
different scoring functions implies that the outlined strategy above may still
provide decent significance assignment. How to properly take into account
method correlations while combining the search results is of course a very
important and open problem.
Suppose one has obtained a list of candidate peptides from some analysis tools
that provides only crude statistical significance assignment or no
significance assignment at all, it is possible to upload this list of peptides
along with the spectrum to RAId_aPS to get a reassignment of statistical
significance via mode (iii) of RAId_aPS. The fundamental idea here is to first
obtain the score histograms corresponding to the list of scoring functions
selected. With the histograms constructed, one can generate the $P$-values for
any score specified. Therefore, for a chosen scoring function and a given list
of peptides, RAId_aPS can provide for each peptide an APP $P$-value by scoring
each peptide and then inferring from the normalized score histogram.
In practical implementation, RAId_aPS sorts the list of peptides according to
their molecular masses and identifies their corresponding mass indices on the
mass grid. Using these indices as terminating points, but one at a time,
RAId_aPS constructs score histograms assuming that the parent ion weight is
given by the mass indices considered. Each peptide in the list is then
rescored using the user-selected scoring versions implemented in RAId_aPS and
the $P$-values corresponding to these scoring functions are obtained. If no
further information other than a flat list of peptides is given, RAId_aPS will
combine these $P$-values using eq. (8) and return a combined $P$-value for
each peptide in the list. When the number of qualified database peptides is
known –which is the case if one directly uploads to RAId_aPS any of the output
files of Mascot, SEQUEST, or X!Tandem– RAId_aPS will first transform the
$P$-values into $E$-values and then into database $P$-values (eq. (7)). For
each peptide in the list, RAId_aPS will then combine their database $P$-values
using eq. (8) and then obtain the final $E$-value via eq. (9).
## Results
### $\boldsymbol{E}$-value Accuracy
In the APP statistics subsection of Technical Background, it was demonstrated
that statistical significance assignment based on the APP score histogram is
spectrum-specific. However, one must verify $E$-value accuracy before claiming
that accurate spectrum-specific statistics are achieved via APP statistics. A
straightforward way to test $E$-value accuracy [11] is to compare the averaged
number of false positives (the textbook definition) versus reported $E$-value
using a spectral dataset resulting from a known mixture. To be specific, one
will first eliminate true positives from a database, and then use the spectra
from a known mixture as queries to look for peptide hits. Since the true
positives are removed from the database beforehand, all the peptide hits are
false positives. One then aggregates all the false positives together –there
might be many false positives from one spectrum– and then sorts them in
ascending order of $E$-value. Let $M$ be the total number of spectra used for
evaluation and let $N_{E\leq E_{c}}$ be the total number of false positives
with $E$-values smaller than or equal to $E_{c}$. If the $E$-values reported
are accurate, one expects to see that
$E_{c}=\frac{N_{E\leq E_{c}}}{M}\;,$
subject to fluctuations due to finite sampling.
Figures 5 and S4 assess $E$-value accuracy when $E$-values are obtained from
APP $P$-values. Figure 5 displays, based on searching a random database of
size 500MB, the measured average number of false positives as a function of
the reported $E$-value. The six-panel figure demonstrates statistical
stability against allowed mass error. For parent ion mass of $2,000$ Da, what
is displayed in Figure 5 covers the resolution range from $1,500$ ppm to $5$
ppm. Figure S5 displays the corresponding result for profile data. The
statistical stability shown is important since the use of high resolution mass
analyzers such as Orbitraps have gained popularity. Figure S4, using the
NCBI’s nr database, examines the $E$-value accuracy when used in biological
context. Since the biological database is not a collection of random peptides,
the validity of statistical theory founded on random databases should be
tested. As shown in Figure S5, the same statistical robustness holds for both
centroid and profile spectra while searching the biological protein database
tested.
Both the centroid data set and profile data set are tryptic and are identical
to the ones used in reference [4]. The $E$-value for a peptide hit is obtained
by multiplying that peptide hit’s APP $P$-value by a numerical factor $N_{d}$,
the number of qualified database peptides with similar masses. In terms of
enumerating qualified peptides, we employ the RAId_DbS strategy. Specifically,
we further divide the qualified peptides into ones with correct and incorrect
N-terminal cleavages [11] and have separate counters for them. If a candidate
peptide has correct N-terminal cleavage, its $N_{d}$ factor is the total
number of database peptides with both correct N-terminal cleavages and with
masses similar to that of the peptide considered; otherwise, it will have a
considerably larger $N_{d}$ factor that counts all database peptides with
masses similar to that of the peptide considered. The protein database used is
the NCBI’s nr (same version as in reference [11]) with identical cluster
removal procedure [11]. As shown in Figure 5 and Figures S4-S5, the $E$-values
reported by RAId_aPS using the various scoring functions implemented are
within a factor of five of the textbook definition. For any two scoring
functions, if they are independent, one may combine the statistics using eqs.
(7-9) and the combined $E$-value should also follow the theoretical curves.
How well the combined $E$-values reported trace the theoretical line can be
used as a measure of how independent these two scoring functions are, provided
that each scoring function already has $E$-value reported in agreement with
the textbook definition. As in reference [9], the combined $E$-value from any
two methods in general shows a larger deviation from the textbook definition.
This may be due to correlations between search methods. We are currently
investigating the possibility of taking into account the search method
correlation, which we suppose to be spectrum-specific too, while combining the
statistics. We will incorporate the corrected statistics into RAId_aPS if the
investigation along this direction turns out to be fruitful.
### Combine Database Search Results
The primary feature of RAId_aPS is the ability to combine, in a statistically
sound way, search results from different scoring functions. If the retrieval
performance of each scoring function implemented is poor, then even if one
combines the search results, the final outcome might still be poor. Below we
assess the retrieval performance of each scoring function implemented using
the Receiver Operating Characteristic (ROC) curves.
#### First assessment of scoring functions
Here we investigate the performance of the four implemented scoring functions
–RAId score, K-score, XCorr, and Hyperscore– each of which is a standard
scoring function, often employed with program-specific heuristics, for a known
search program. The retrieval efficiency is assessed using a centroid data set
(Figure 6, ISB data set). Since many search methods report only one or very
few candidate peptides per spectrum, we also include this type of ROC curve
(Figure 7) where only the best hit per spectrum is taken from the search
results. The performance of this ad hoc truncation apparently leads to better
retrieval at small number of false positives, indicating the existence of
false hits whose evidence peaks are homologous to that of the true positive(s)
associated with a spectrum. We are currently investigating the impact of the
existence of these types of false positives on the statistical significance
assignment. The results will be reported in a separate publication. The
corresponding plots when using a profile data set (NHLBI data set) are shown
respectively in Figure S6 (similar to Figure 6) and Figure S7 (similar to
Figure 7).
#### Different ROC analysis
When the true positive peptides are not known a priori, there exist various
strategies in classifying hits into true or false positives when making a ROC
plot. These strategies, unfortunately, will make a notable difference in
retrieval assessment. For example, in a cell lysate experiment of a certain
organism, it is customary to estimate the number of false positive hits by
introducing a decoy database during the data analysis. The main idea there is
to first sort the peptide hits according to their scores. Then for each decoy
hit, one assumes that there is just one corresponding false hit in the target
database. This strategy has been used extensively [24]. ROC analyses done this
way generally count false positives, which are highly homologous to the target
peptides, towards true positives. This has two effects: an overcount of true
positives and a undercount of false positives. As a consequence, the ROC
curves will appear more impressive. To mimick this situation, we used BLAST to
find in the NCBI’s nr database highly homologous proteins to the target
proteins used in the experiment and include those proteins in our true
positive set. This strategy produces ROC curves shown as the solid curves of
Figure S8. When compared to Figure 6 and Figure S6, the ROC curves produced by
this strategy seem much more impressive.
Not counting highly homologous proteins as false positives would probably be
agreeable. However, counting those peptides/proteins as true positives could
be exaggerating. Therefore one may use a slightly different strategy: removing
from consideration proteins homologous to the target proteins, which is called
the cluster removal strategy [11]. The dashed curves of Figure S9 are ROC
curves obtained this way. This strategy also produces slightly more impressive
ROC curves than in Figure 6 and Figure S6. Apparently, this indicates the
highly homologous false positive hits are the ones that degrade the retrieval
performance. Thus, it can be useful to remove those false positives from
consideration. Keeping only the best hit per spectrum turns out to be one way
to achieve this goal.
#### Combining Multiple Scoring Functions
Since different scoring functions have different spectral filtering
strategies, it is often advantageous to combine the search results from
several scoring functions. RAId_aPS provides a simple user interface, allowing
users to select several scoring functions at a time. A example output when
several scoring functions are selected is shown in Table 1.
Figure 8 illustrates the performance when RAId_aPS combines three different
scoring functions in its database search mode. Panels (A) and (B) of Figure 8
should be compared with Figure 6 and Figure S6 respectively. The ROC curves
obtained by combining three randomly chosen scoring functions indicate better
performance than individual scoring functions. Panels (C) and (D) should be
compared with Figures 7 and Figure S7 respectively. The results in those plots
are obtained from keeping only the best hit per spectrum prior to further
analysis. As shown in those plots, the ROC curves obtained by combining three
randomly chosen scoring functions indicate significantly better performance
than individual scoring functions, except for the case of RAId_DbS.
### Other modes
Examples of using mode (iv) were already shown above. We demonstrate here
other features of RAId_aPS to illustrate its versatility.
#### Compute TNPP: mode (i)
Given a parent ion mass, RAId_aPS is also able to compute efficiently the TNPP
associated with that molecular mass within a user-specified mass error. The
user interface for computing TNPP is self-explanatory. One simply types in the
molecular mass of interest, chooses a specific digesting enzyme or considers
no enzymatic restriction by choosing “no enzyme”, and then presses the “Submit
a job” button. If one wishes to change the default mass error tolerance, it
can be done under the “more parameter” toggle. One may also elect to include
PTMs or deselect certain amino acids from consideration, those choices are
available under the “Amino acids and PTMs” toggle. When using search methods
that do not have a theoretical model for the score distribution or when the
quality of the score model [11] is poor, one may wish to use a more
conservative statistical significance assignment. In this case, a user may set
$1/{\rm TNPP}$ as the lower bound for the best $P$-value for any given parent
ion mass. This may help in preventing exaggerated/inappropriate statistical
significance assignments.
#### Generate score histogram: mode (ii)
Extraction of the statistical significance from a score distribution often
requires a model, be it theoretically derived or empirically assumed, for the
score distribution. One may test the robustness of a score model by examining
how well the score model fits the database search score histograms. When using
search methods that have a score model, one may first test how well the same
score model applies when dealing with APP. If the score model loses stability,
this may indicate that the score model is not robust in general. Given a query
spectrum and a user-selected scoring function, RAId_aPS can be used to
generate a score histogram of APP under the selected scoring scheme. Using an
example spectrum, Figure 9 shows score histograms corresponding to the four
scoring functions implemented in RAId_aPS.
#### Reassign $\boldsymbol{E}$-value : mode (iii)
Statistical significance inference from RAId_aPS only depends on the total
number of qualified peptides inside the database searched but is not dependent
on the peptide content inside the database. This is because RAId_aPS bases its
statistics on the (weighted) score histogram obtained from scoring APP. As a
consequence, without going through the database search again, RAId_aPS can be
used to reassign statistical significance to a collection of candidate
peptides. The candidate peptides may come from a flat list provided by the
user, or they can also come from the output files of various search engines.
RAId_aPS allows users to upload the output files from SEQUEST, X!Tandem, and
Mascot for statistical significance reassignment.
Although scoring functions similar to XCorr, K-score and Hyperscore have been
implemented in RAId_aPS, other search engines’ scoring functions might not be
suitable for score histogram construction using dynamic programming. In this
case, the user may wish to compare the statistical significance reported by a
search engine with what is reported by RAId_aPS and even combine these
reported significances. As an example of this usage and to test RAId_aPS’s
performance, we use as queries $10,000$ profile spectra (the NHLBI data set)
as well as $12,628$ centroid spectra (A1-A4 of the ISB data set), each
produced from a known mixture of target proteins. Using Mascot as the search
engine, we searched in the NCBI’s nr database with proteins highly homologous
to the target proteins removed [11]. The output files were analyzed to produce
ROC curves, the black solid curves in Figure 10. We then reanalyzed the
candidate peptides’ statistical significance by combining the statistical
significance reported by Mascot with that reported by RAId_aPS using one
additional scoring function. For both profile and centroid spectra, when
combined with either the RAId score, K-score, or XCorr , one may obtain a
retrieval performance that is comparable with or slightly better than that
from Mascot alone (see Figure 10).
Since all the implemented scoring functions are accessible from RAId_aPS, one
can score any new PTM peptide using any of the scoring functions available to
RAId_aPS even when the original program does not yet include the PTMs of
interest. This way, annotated PTM found by RAId_DbS [23] may be confirmed with
other scoring functions in a natural manner and one may even combine the
statistical significance as described below to increase the sensitivity in
finding annotated PTMs and single amino acid polymorphisms (SAP).
## Discussion
In this section we will discuss another proposed use of the APP statistics in
confidence assignment, remark on the effectiveness of combining search results
using a different measure than ROC, propose avenues for improvement, and
describe future directions.
When combined with database searches, the score histogram obtained by RAId_aPS
also provides two useful quantities. First, it gives us the best peptide score
$S_{\rm APP}$ among APP. Although we did not pursue this way, it has been
advocated that the difference between $S_{\rm APP}$ and the best database hit
score per spectrum may serve as a statistical significance measure for the
highest-scoring peptide hits found in the database [24]. Second, the score
histogram provides us with $N_{s}$, the (weighted) number of APP with score
better than or equal to $S$. This number $N_{s}$ may also be used in
conjunction with the (relative) difference between $S_{\rm APP}$ and the best
database search score per spectrum while constructing statistical significance
measures other than $E$-value.
A natural question to ask is: how much retrieval gain can one anticipate if
one combines multiple scoring functions? Since FDR has been among the most
popular metrics for assessing the performance, we briefly investigate this
issue using FDR. Employing a frequently used procedure [31], we used the
reverse Homo sapiens protein database as the decoy database to estimate the
number of false positives and hence the FDR, by searching target database and
decoy database separately for each query spectrum. All $15$ possible
combinations of the four scoring functions available in RAId_aPS are tested
using the data set PRIDE_Exp_mzData_Ac_8421.xml (containing $15,916$ spectra),
downloaded from the PRoteomics IDEntifications (PRIDE) database
(http:www.ebi.ac.ukprideppp2_links.do). The results are summarized in Table 2
along with the average behavior associated with using one to four scoring
functions. Since it is known that performance of a search engine may vary when
the data to be analyzed changes[32], we like to focus more on the average
behavior rather than individual performance of a scoring function or any
specific combination of scoring functions. Based on the average retrieval
result of Table 2, we first observe that on average there is an overall
retrieval increase at $0\%-10\%$ FDR rates when one combine two scoring
functions versus using only one scoring function. We also note that there is
an increase in retrieval performance at medium FDR rate when more scoring
functions are combined. However, at very low FDR rates, it seems that
combining more than two scoring functions stop helping the retrieval.
Apparently, the performance boost does not continue indefinitely as more
scoring functions are included. This is evidenced by an observable performance
decline at low FDR rate when one combine all four scoring functions and
compared to combine only three. The saturation of performance gain is
reasonable if one takes into account the fact that most scoring functions seek
similar evidences, the scope covered by combining more scoring functions can’t
keep increasing indefinitely.
By integrating existing annotated information into organismal databases,
RAId_DbS is now able to incorporate during its data analysis annotated
information such as SAP, PTM, and their disease associations if they exist
[23]. This feature enables users to identify/include known
polymorphisms/modifications in their searches without needing to blindly allow
all possible SAPs and PTMs first and then post process to look up the
literature/databases for explanations. Since all the implemented scoring
functions of RAId_aPS are now within the same framework, we can let each plug-
in scoring function incorporate in its scoring the new SAP/PTM peptides. This
way, annotated SAP/PTM found by RAId_DbS may be confirmed by other implemented
scoring approaches in a natural manner and one may even combine the
statistical significances as described earlier to increase the sensitivity in
finding annotated SAPs/PTMs.
In the near future, we also plan to include more scoring functions in RAId_aPS
if their presence would enhance the retrieval performance without sacrifice
statistical accuracy. For example, we will investigate the effect of a new
scoring function, the compound Poisson. This is a natural way to incorporate
intensity information into Poisson count statistics. The other scoring
approach we will investigate is to deconvolute the peptide length information.
The reason to consider this alternative arises from the observation that many
scoring functions introduce different heuristics to correct for the scores
associated with candidate peptides of different lengths. The purpose of these
peptide length correction factors is to balance the fact that longer peptides
are likely to find more evidence peaks and thus the collected evidence scores
may require some length correction in order to make the comparison among
peptides of various lengths impartial. If we group peptides of the same
lengths and obtain statistical significance separately for peptide candidates
of each length, we no longer need to introduce any length correction factor.
This approach is not feasible for regular database searches since the sample
size of peptides of a fixed length may be too small. For our APP scheme,
however, we always have a large number of peptides participating in our score
histogram even if the peptide length is fixed. Therefore, the idea of
deconvoluting the peptide lengths becomes feasible for RAId_aPS.
## Acknowledgments
We thank Jimmy Eng for useful correspondence on the spectral filtering
strategy of SEQUEST’s XCorr. We also thank the administrative group of the NIH
Biowulf clusters, where all the computational tasks were carried out.
## References
* 1. Prakash A, Piening B, Whiteaker J, Zhang H, Shaffer SA, et al. (2007) Assessing bias in experiment design for large scale mass spectrometry-based quantitative proteomics. Mol Cell Proteomics 6: 1741–1748.
* 2. Taylor CF, Paton NW, Lilley KS, Binz PA, Julian RK, et al. (2007) The minimum information about a proteomics experiment (MIAPE). Nat Biotechnol 25: 887–893.
* 3. Oberg AL, Vitek O (2009) Statistical Design of Quantitative Mass spectrometry-Based Proteomics Experiments. J Proteome Res 8: 2144–2156.
* 4. Alves G, Ogurtsov AY, Wu WW, Wang G, Shen RF, et al. (2007) Calibrating E-values for MS2 library search methods. Biology Direct 2: 26.
* 5. Keller A, Nesvizhskii AI, Kolker E, R A (2002) Empirical statistical model to estimate the accuracy of peptide identifications made by ms/ms and database search. Anal Chem 74: 5383-5392.
* 6. Kall L, Canterbury JD, Weston J, Noble WS, MacCoss MJ (2007) Semi-supervised learning for peptide identification from shotgun proteomics datasets. Nat Methods 4: 923–925.
* 7. Searle BC, Turner M, Nesvizhskii AI (2008) Improving sensitivity by probabilistically combining results from multiple MS/MS search methodologies. J Proteome Res 7: 245–253.
* 8. Edwards N, Wu X, Tseng CW (2009) An unsupervised, model-free, machine-learning combiner for peptide identifications from tandem mass spectra. Clin Proteom 5: 23–36.
* 9. Alves G, Wu WW, Wang G, Shen RF, Yu YK (2008) Enhancing peptide identification confidence by combining search methods. J Proteome Res 7: 3102–3113.
* 10. Alves G, Yu YK (2008) Statistical characterization of a 1D random potential problem – with applications in score statistics of MS-based peptide sequencing . Physica A 387: 6538-6544.
* 11. Alves G, Ogurtsov AY, Yu YK (2007) RAId_DbS: Peptide identification using database searches with realistic statistics. Biology Direct 2: 25.
* 12. (2008) Developing and disseminating advances in computation and statistical proteomics. In: McIntosh M, editor, J. Proteome Res. pp. 18-456. Vol: 7, Special issue on Statistical Proteomics.
* 13. Fenyo D, Beavis RC (2003) A method for assessing the statistical significance of mass spectrometry-based protein identification using general scoring schemes. Anal Chem 75: 768-774.
* 14. Alves G, Yu YK (2005) Robust Accurate Identification of peptides (RAId): deciphering MS2 data using a structured library search with de novo based statistics. Bioinformatics 21: 3726-3732.
* 15. Doerr TP, Alves G, Yu YK (2005) Ranked solutions to a class of combinatorial optimizations with applications in mass spectrometry based peptide sequencing and a variant of directed paths in random media . Physica A 354: 558-570.
* 16. Zhang N, Aebersold R, Schwikowski B (2002) A probabilistic algorithm to identify peptides through sequence database searching using tandem mass spectral data. Proteomics 2: 1406-1412.
* 17. Eng JK, McCormack AL, Yates III JR (1994) An approach to correlate tandem mass spectral data of peptides with amino acid sequences in a protein database. J Amer Soc Mass Spectrom 5: 976-989.
* 18. Klammer AA, Park CY, Noble WS (2009) Statistical Calibration of the SEQUEST XCorr Function. J Proteome Res 8: 2106–2113.
* 19. Eng JK, Fischer B, Grossmann J, Maccoss MJ (2008) A fast SEQUEST cross correlation algorithm. J Proteome Res 7: 4598–4602.
* 20. Park CY, Klammer AA, Käll L, MacCoss MJ, Noble WS (2008) Rapid and accurate peptide identification from tandem mass spectra. J Proteome Res 7: 3022–3027.
* 21. Press WH, Teukolsky SA, Vetterling WT, Flannery BP (1999) Numerical Recipes in C. Cambridge: Cambridge University Press. 2nd ed.
* 22. Searle BC, Turner M, Nesvizhskii AI (2008) Improving sensitivity by probabilistically combining results from multiple MS/MS search methodologies. J Proteome Res 7: 245–253.
* 23. Alves G, Ogurtsov AY, Yu YK (2008) RAId_DbS: mass-spectrometry based peptide identification web server with knowledge integration. BMC Genomics 9: 505.
* 24. Kim S, Gupta N, Pevzner PA (2008) Spectral probabilities and generating functions of tandem mass spectra: a strike against decoy databases. J Proteome Res 7: 3354–3363.
* 25. Craig R, Beavis RC (2004) Tandem: matching proteins with tandem mass spectra. Bioinformatics 20: 1466-1467.
* 26. MacLean B, Eng JK, Beavis RC, McIntosh M (2006) General framework for developing and evaluating database scoring algorithms using the TANDEM search engine. Bioinformatics 22: 2830–2832.
* 27. Keller A, Eng J, Zhang N, Li XJ, Aebersold R (2005) A uniform proteomics MS/MS analysis platform utilizing open XML file formats. Mol Syst Biol 1: 2005.0017.
* 28. Keller A, Samuel P, Nesvizhskii AI, Stolyar S, Goodlett DR, et al. (2002) Experimental protein mixture for validating tandem mass spectral analysis. OMICS 6: 207-212.
* 29. Robinson AB, Robinson LR (1991) Distribution of glutamine and asparagine residues and their near neighbors in peptides and proteins. Proc Natl Acad Sci USA 88: 8880-8884.
* 30. Yu YK, Gertz E, Agarwala R, Schäffer A, Altschul S (2006) Retrieval accuracy, statistical significance and compositional similarity in protein sequence database searches. Nucl Acids Res 34: 5966-5973.
* 31. Elias JE, Gygi SP (2007) Target-decoy search strategy for increased confidence in large-scale protein identifications by mass spectrometry. Nat Methods 4: 207–214.
* 32. Jones AR, Siepen JA, Hubbard SJ, Paton NW (2009) Improving sensitivity in proteome studies by analysis of false discovery rates for multiple search engines. Proteomics 9: 1220–1229.
## Figure Legends
Figure 1: Illustration of APP mass grid with internal structure. In addition
to show the basic mass grid, this figure illustrates,using the peptide lengths
as an example, the possibility of including additional structures in the (raw)
score histogram associated with each mass index. The basic idea of obtaining
the score histogram via dynamic programming is explained in the Method
section. The key step to incorporate additional structure is to let the
(weighted) count associated with each (raw) score be further categorized by
the lengths of partial peptides reaching each mass index. In the end, one will
apply the length correction factor to the raw score to obtain the real score
histogram. Apparently, one may also keep track of the number of $b$ ($y$)
peaks accumulated within the raw score histogram. Again, the factorial
contribution can be added at the end prior to the construction of the final
score histogram.
Figure 2: Example processed spectra from different scoring functions versus
the original spectrum. The centroid spectrum used has a parent ion mass of
$1640.80$ Da. In panel (A), the original spectrum is displayed; (B) shows the
processed spectrum generated by the filtering protocol of RAId_DbS scoring
function; (C) exhibits the processed spectrum generated by the filtering
protocol of K-score; while (D) and (E) correspond respectively to the
processed spectra produced by XCorr and Hyperscore.
Figure 3: Histograms of correlations between filtering strategies. Used in
this plot are $38,424$ raw centroid spectra from the ISB data set [28]. Each
raw spectrum will have four different processed spectra come from each of the
four different filtering strategies. The mass fragments of every filtered
spectrum are then read to a mass grid. The spectrum is then viewed as a vector
with non-vanishing components only at the populated component/mass indices.
One then normalizes each filtered spectrum vector to unit length. An inner
product of any two filtered spectral vectors represents the correlation
between them. When the spectral quality does not pass a method-dependent
threshold, the corresponding filtering protocol may turn the raw spectrum into
a null spectrum without further searching the database. For a given pair of
filtering methods and a raw spectrum, if each of the two filtering methods
produces a nonempty filtered spectrum, one may turn those filtered spectra
into spectral vectors and compute their inner product, i.e., their
correlation. For each pair of filtering methods, these inner products are
accumulated and plotted as a correlation histogram. All six pairwise
combinations are shown.
Figure 4: Score correlations. A subset of the ISB centroid data set [28] was
used to perform this evaluation. For each scoring function, when the best hit
per spectrum (analyzed using the analysis program that the scoring function
was originally used for) is a true positive, that candidate peptide is scored
again using the corresponding scoring function implemented in RAId_aPS. Each
true positive best hit thus gives rise to two scores and plotted using the
following rule: the first score is used as the ordinate while the second score
(from RAId_aPS) is used as the abscissa. Including $500$ spectra, panel A is
for the RAId score. Panel B is for Hyperscore and contains $248$ spectra. The
result of K-score is shown in panel C with $220$ spectra. Shown with $500$
spectra, panel D documents the results for XCorr.
Figure 5: E-value accuracy assessment. The agreement between the reported
$E$-value and the textbook definition is examined using centroid data (A1-A4
subsets of ISB data set). The random database size used is 500 MB. The
molecular weight range considered while searching the database is
$[MW-\delta,MW+\delta]$. In each panel, the dashed lines, corresponding to
$x=5y$ and $x=y/5$, are used to provide a visual guide regarding how close/off
the experimental curves are from the theoretical curve.
Figure 6: ROC curves for the centroid data (A1-A4 of the ISB data set [28]).
For each of the four scoring functions considered, a set of ROC curves is
shown. These ROC curves include the results from running the designated
program associated with that scoring function, the results from running
RAId_aPS in the database search mode, and the results from combining with each
of the three other scoring functions. Panel (A) shows the results from RAId
score, whose designated program is RAId_DbS. Panel (B) displays the results
from K-score, whose designated program is X!Tandem. Panel (C) exhibits the
results from XCorr, which is mostly employed by SEQUEST. Panel (D) presents
the results from Hyperscore, whose designated program is also X!Tandem.
Instead of using only XCorr (like RAId_aPS), SEQUEST first selects the top
$500$ candidates using SP score. As shown in panel (C), for centroid data
there is an advantage to filtering candidates with the SP score. However, it
is also seen that by combining XCorr with either RAId score or Hyperscore,
equally good results can be attained without introducing the SP score
heuristics.
Figure 7: ROC curves for the centroid data (A1-A4 of the ISB data set [28])
when considering only the best hit per spectrum. For each of the four scoring
functions considered, a set of ROC curves is shown. These ROC curves include
in the consideration only the best hit per spectrum from running the
designated program associated with that scoring function, the best hit per
spectrum from running RAId_aPS in the database search mode, and the best hit
per spectrum from combining with each of the three other scoring functions.
Panel (A) shows the results from RAId score, whose designated program is
RAId_DbS. Panel (B) displays the results from K-score, whose designated
program is X!Tandem. Panel (C) exhibits the results from XCorr, which is
mostly employed by SEQUEST. Panel (D) presents the results from Hyperscore,
whose designated program is also X!Tandem. Instead of using only XCorr (like
RAId_aPS), SEQUEST first selects the top $500$ candidates using SP score. As
shown in panel (C), for centroid data there is advantage to filter candidates
with the SP score. However, it is also seen that by combining XCorr with
either RAId score or Hyperscore, equally good results can be attained without
introducing the SP score heuristics.
Figure 8: Illustration of RAId_aPS performance when combining three different
scoring functions. Panel (A) shows the results from the profile data (NHLBI
data set [4]), while panel (B) exhibits the results from the centroid data
(A1-A4 of the ISB data set [28]). Panel (C) shows the results from the profile
data but keeping only the best hit per spectrum, while panel (D) exhibits the
results from the centroid data but keeping only the best hit per spectrum.
Figure 9: Example score PDF (normalized histogram) output by RAId_aPS. An MS2
spectrum of parent ion mass $1640.80$ Da is queried with default parameters,
and the resulting score PDF for RAId, K-score, XCorr, and Hyperscore are shown
respectively in panels A, B, C, and D. The number of APP within $\pm$ 3Da of
parent ion mass is about $10^{19}$.
Figure 10: Example of reanalyzing output files from other search engine by
combining with statistical significance assignment from RAId_aPS. In this
example, we use the Mascot output files resulting from querying profile
spectra (panel (A), the NHLBI data set) and centroid spectra (panel (B), A1-A4
of the ISB data set [28]) to the NCBI’s nr database with proteins highly
homologous to those that were present in the mixture removed. Since each data
set is from a known mixture of proteins, it is possible to remove the proteins
homologous to the true positives from the nr database. We then combine the
calibrated $E$-value [4] of Mascot with the $E$-value obtained from RAId_aPS
when either RAId score, Hyperscore, K-score or XCorr is used.
## Tables
Table 1: An output example of the combined E-value from RAId_aPS. E_comb | RAId | Hyperscore | XCorr | K-score | Peptide peptidepeptide
---|---|---|---|---|---
$4.93{\rm e}{\small-24}$ | $1.69{\rm e}{\small-13}$ | $8.26{\rm e}{\small-11}$ | $5.87{\rm e}{\small-12}$ | $7.99{\rm e}{\small-13}$ | NYQEAKDAFLGSFLYEYSR
$1.43$ | $379.00$ | $0.08$ | $453.00$ | $101.00$ | APTSAGPWEKPTVEEALESGSR
$1.85$ | $28.50$ | $1.94$ | $9.01$ | $0.15$ | LERMTQALALQAGSLEDGGPSR
$3.38$ | $13.60$ | $0.30$ | $88.40$ | $4.32$ | TEDQRPQLDPYQILGPTSSR
$4.04$ | $15.80$ | $18.40$ | $0.38$ | $18.30$ | NYKAKQGGLRFAHLLDQVSR
$8.81$ | $257.00$ | $1.48$ | $1170.00$ | $1280.00$ | DTPMLLYLNTHTALEQMRR
$9.58$ | $8.76$ | $1.66$ | $353.00$ | $37.20$ | EKTESSGQETTAKCDRASKSR
$9.75$ | $1.71$ | $8.15$ | $82.80$ | $6.99$ | LLAQQSLNQQYLNHPPPVSR
$10.80$ | $358.00$ | $1.95$ | $311.00$ | $269.00$ | IQHGQCAYTFILPEHDGNCR
Table 2: Example retrieval tests based on FDR. All $15$ possible combinations of the four scoring functions available in RAId_aPS are shown along with the average behavior associated with using one to four scoring functions. The dataset PRIDE_Exp_mzData_Ac_8421.xml is used. The first column documents various combinations of scoring functions with the following abbreviations: R for RAId, K for K-score, H for hyperscore, and X for XCorr. The rest of the columns display the number of peptides identified at the false positive rate specified at the top of the column. The rows with bold characters indicate the average behavior of using a single (${\mathbf{S}}$) scoring function, combining two (${\mathbf{D}}$) scoring functions, combining three (${\mathbf{T}}$) scoring functions, and combining four (${\mathbf{Q}}$) scoring functions. Within these rows, except the last one where only one combination possible, the standard deviation associated with each average is shown inside the parentheses to the right of the average. Combination | FDR cutoff 0% | FDR cutoff 2.5% | FDR cutoff 5.0% | FDR cutoff 10%
---|---|---|---|---
R | 377 (3 86) | 822(7 8) | 8561 1 | 9481 1
K | 83 (3 86) | 709(7 8) | 7901 1 | 9771 1
H | 568 (3 86) | 775(7 8) | 8491 1 | 9081 1
X | 467 (3 86) | 821(7 8) | 8851 1 | 9961 1
$\overline{\mathbf{S}}(\sigma_{S})$ | 373 (182) | 781 (57)(1) | 845 (34)15 | 957 (39)15
RK | 485 (3 86) | 956(7 8) | 11271 1 | 16541 1
RH | 925 (3 86) | 1143(7 8) | 15991 1 | 23751 1
RX | 871 (3 86) | 1024(7 8) | 11401 1 | 15741 1
KH | 528 (3 86) | 1019(7 8) | 12101 1 | 16791 1
KX | 588 (3 86) | 860(7 8) | 9641 1 | 11461 1
HX | 895 (3 86) | 1064(7 8) | 12051 1 | 15321 1
$\overline{\mathbf{D}}(\sigma_{D})$ | 715 (186) | 1011 (87)(1) | 1207 (196) | 1660 (365)
RKH | 485 (3 86) | 849(7 8) | 26891 1 | 53281 1
RKX | 474 (3 86) | 792(7 8) | 10741 1 | 24251 1
RHX | 725 (3 86) | 867(7 8) | 19421 1 | 47951 1
KHX | 443 (3 86) | 658(7 8) | 9101 1 | 16911 1
$\overline{\mathbf{T}}(\sigma_{T})$ | 531 (116) | 791 (86)(1) | 1653 (716) | 3559 (1537)
RKHX (${\mathbf{Q}}$) | 332 (3 86) | 662(7 8) | 13361 1 | 41481 1
| arxiv-papers | 2008-06-16T23:02:54 | 2024-09-04T02:48:56.294010 | {
"license": "Public Domain",
"authors": "Gelio Alves, Aleksey Y. Ogurtsov, and Yi-Kuo Yu",
"submitter": "Yi-Kuo Yu",
"url": "https://arxiv.org/abs/0806.2685"
} |
0806.2740 | Transverse Wave Propagation in Relativistic
Two-fluid Plasmas in de Sitter Space
M. Atiqur Rahman111E-mail: atirubd@yahoo.com and M. Hossain Ali 222The Abdus
Salam International Centre for Theoretical Physics, Strada Costiera 11, 34014
Trieste, Italy. E-mail: mali@ictp.it, $m_{-}hossain_{-}ali_{-}bd@yahoo.com$
(Corresponding author).
Department of Applied Mathematics, University of Rajshahi , Rajshahi-6205,
Bangladesh.
Abstract
We investigate transverse electromagnetic waves propagating in a plasma in the
de Sitter space. Using the $3+1$ formalism we derive the relativistic two-
fluid equations to take account of the effects due to the horizon and describe
the set of simultaneous linear equations for the perturbations. We use a local
approximation to investigate the one-dimensional radial propagation of Alfvén
and high frequency electromagnetic waves and solve the dispersion relation for
these waves numerically.
Keywords: Two-fluid plasma, Alfvén and high frequency electromagnetic waves,
Cosmological event horizon.
## 1 Introduction
In recent years there have been renewed interests in investigating plasmas in
curved spacetimes of general relativity; because, a successful study of the
waves and emissions from plasmas falling into a compact body (e.g. black hole)
will be of great value in aiding the observational identification of black
hole candidates. The de Sitter (dS) space with positive cosmological constant
has properties similar to a black hole. We study two-fluid plasmas near the
horizon of the pure dS space.
Over the last few decades, physicists have a growing interest in dS space. In
the 1970s, the attention was due to the large symmetry group of dS space,
which made the field theory in dS space less ambiguous than, for example, in
the Schwarzschild spacetime. Researches in the 1980s focused the role it
played during inflation–accelerated expansion in the very early universe. The
universe is currently asymptotic dS and approach a pure dS space. Recent
cosmological observations [1, 2, 3, 4, 5, 6] suggest the possibility of
existing a positive cosmological constant ($\Lambda>0$) in our universe and
this possibility gives the picture, among many others, of some features
closely related to black holes: the existence of cosmological event horizons.
These causal horizons exist even in the absence of matter, namely in empty dS
space, and hide all the events which are not accessible for geodesic
observers. In addition, the success of the ADS/CFT correspondence [7, 8, 9,
10] has led to the intense study of dS space in the context of the quantum
gravity [11]. The attention has been to obtain an analogue of the ADS/CFT
correspondence in dS space, i.e. dS/CFT correspondence [12, 13, 14, 15, 16,
17, 18, 19, 20, 21, 22, 23, 24, 25, 26] in the light of which there has been
an extensive study of the semiclassical aspects of dS and asymptotic dS
spacetimes [27, 28, 29, 30, 31, 32, 33]. In view of these reasons, it may be
of special interest to investigate electromagnetic waves in a plasma in the dS
space.
Recently, Buzzi, et.al. [33, 34], using the 3+1 formulation [35, 36, 37, 38],
described a general relativistic version of two-fluid formulation of plasma
physics and investigated the nature of plasma waves (transverse waves in [33],
and longitudinal waves in [34]) near the horizon of the Schwarzschild black
hole. In this paper we apply the formalism of Buzzi et.al. [33] to investigate
the transverse electromagnetic waves propagating in a plasma close to the
(cosmological) event horizon of the pure dS space.
This paper is organized as follows. In section 2 we summarize the 3+1
formulation of general relativity. In section 3 we describe the horizon plasma
governing equations. In section 4 we consider one-dimensional wave propagation
in the radial $z$ (Rindler coordinate system) direction. We linearize the
equations in section 5 by considering a small perturbation to fields and fluid
parameters. In section 6 we discuss the local or mean-field approximation for
the lapse function $\alpha$ and obtain a dispersion relation for the
transverse wave. In section 7 we describe a procedure for solving the
dispersion relation numerically. In sections 8 we present our results.
Finally, in section 9 we give our remarks. We use units $G=c=k_{B}=1$.
## 2 3+1 Spacetime Formalism
In de sitter space, the simplest solution for the Einstein field equations
with $T_{\mu\nu}=0$ is written as
$\displaystyle ds^{2}$ $\displaystyle=$ $\displaystyle
g_{\mu\nu}dx^{\mu}dx^{\nu}$ (1) $\displaystyle=$
$\displaystyle-\left(1-\frac{r^{2}}{\ell^{2}}\right)dt^{2}+\left(1-\frac{r^{2}}{\ell^{2}}\right)^{-1}dr^{2}+r^{2}d\Omega_{2}^{2}.$
Here, $\ell$ is the curvature radius of the dS space [i.e.,
$\Lambda=\frac{3}{\ell^{2}}$ is the positive cosmological constant],
$d\Omega_{2}^{2}$ represents a unit 2-sphere, and the nonangular coordinates
range according to $0\leq r\leq\ell$ and $-\infty\leq t\leq\infty$. The
boundary at $r=\ell$ describes a cosmological horizon for an observer located
at $r=0$.
An absolute three-dimensional space defined by the hypersurfaces of constant
universal time $t$ is described by the metric
$ds^{2}=g_{ij}dx^{i}dx^{j}=\left(1-\frac{r^{2}}{\ell^{2}}\right)^{-1}dr^{2}+r^{2}d\Omega_{2}^{2}.$
(2)
The indices $i$, $j$ range over 1, 2, 3 and refer to coordinates in absolute
space. The Fiducial Observers (FIDOs), the observers remaining at rest with
respect to this absolute space, measure their proper time $\tau$ using clocks
that they carry with them and make local measurements of physical quantities.
Then all their measured quantities are defined as FIDO locally measured
quantities and all rates measured by them are measured using FIDO proper time.
The FIDOs use a local Cartesian coordinate system with unit basis vectors
tangent to the coordinate line
${\bf
e}_{\hat{r}}=\left(1-\frac{r^{2}}{\ell^{2}}\right)^{1/2}\frac{\partial}{\partial
r},\hskip 28.45274pt{\bf
e}_{\hat{\theta}}=\frac{1}{r}\frac{\partial}{\partial\theta},\hskip
28.45274pt{\bf e}_{\hat{\varphi}}=\frac{1}{r\,{\rm
sin}\theta}\frac{\partial}{\partial\varphi}.$ (3)
For a spacetime viewpoint rather than a 3 + 1 split of spacetime, the set of
orthonormal vectors also includes the basis vector for the time coordinate
given by
${\bf e}_{\hat{0}}=\frac{d}{d\tau}=\frac{1}{\alpha}\frac{\partial}{\partial
t},$ (4)
where $\alpha$ is the lapse function (or redshift factor) defined by
$\alpha(r)\equiv\frac{d\tau}{dt}=\left(1-\frac{r^{2}}{\ell^{2}}\right)^{1/2}.$
(5)
The gravitational acceleration felt by a FIDO is given by [35, 36, 37, 38]
${\bf a}=-\nabla{\rm ln}\alpha=\frac{1}{\alpha}\frac{r}{\ell^{2}}\,{\bf
e}_{\hat{r}},$ (6)
while the rate of change of any scalar physical quantity or any three-
dimensional vector or tensor, as measured by a FIDO, is defined by the
convective derivative
$\frac{D}{D\tau}\equiv\left(\frac{1}{\alpha}\frac{\partial}{\partial t}+{\bf
v}\cdot\nabla\right),$ (7)
$\bf v$ being the velocity of a fluid as measured locally by a FIDO.
## 3 Two-Fluid Plasma Equations in 3+1 Formalism
We consider a two-component plasma consisting of electrons and either
positrons or ions. In the 3+1 notation, the continuity equation for each of
the fluid species is
$\frac{\partial}{\partial
t}(\gamma_{s}n_{s})+\nabla\cdot(\alpha\gamma_{s}n_{s}{\bf v}_{s})=0.$ (8)
where $s$ is 1 for electrons and 2 for positrons (or ions). For a perfect
relativistic fluid of species $s$ in three-dimensions, the energy density
$\epsilon_{s}$, the momentum density ${\bf S}_{s}$, and stress-energy tensor
$W^{jk}_{s}$ are given by
$\epsilon_{s}=\gamma_{s}^{2}(\varepsilon_{s}+P_{s}{\bf v}_{s}^{2}),\hskip
17.07182pt{\bf S}_{s}=\gamma_{s}^{2}(\varepsilon_{s}+P_{s}){\bf v}_{s},\hskip
17.07182ptW_{s}^{jk}=\gamma_{s}^{2}(\varepsilon_{s}+P_{s})v_{s}^{j}v_{s}^{k}+P_{s}g^{jk}.$
(9)
where ${\bf v}_{s}$ is the fluid velocity, $n_{s}$ is the number density,
$P_{s}$ is the pressure, and $\varepsilon_{s}$ is the total energy density
defined by
$\varepsilon_{s}=m_{s}n_{s}+P_{s}/(\gamma_{g}-1).$ (10)
The gas constant $\gamma_{g}$ is $4/3$ for $T\rightarrow\infty$ and $5/3$ for
$T\rightarrow 0$.
Using the conservation of entropy, the equation of state can be expressed by
$\frac{D}{D\tau}\left(\frac{P_{s}}{n_{s}^{\gamma_{g}}}\right)=0,$ (11)
where $D/D\tau=(1/\alpha)\partial/\partial t+{\bf v}_{s}\cdot\nabla$. The full
equation of state for a relativistic fluid, as measured in the fluid’s rest
frame, is as follows [39, 40]:
$\varepsilon=m_{s}n_{s}+m_{s}n_{s}\left[\frac{P_{s}}{m_{s}n_{s}}-\frac{\textrm{i}H_{2}^{(1)^{\prime}}(\textrm{i}m_{s}n_{s}/P_{s})}{\textrm{i}H_{2}^{(1)}(\textrm{i}m_{s}n_{s}/P_{s})}\right],$
(12)
where the $H_{2}^{(1)}(x)$ are Hankel functions.
The quantities of (9) in the electromagnetic field are expressed by
$\displaystyle\epsilon_{s}$ $\displaystyle=$ $\displaystyle\frac{1}{8\pi}({\bf
E}^{2}+{\bf B}^{2}),\hskip 28.45274pt{\bf S}_{s}=\frac{1}{4\pi}{\bf
E}\times{\bf B},$ $\displaystyle W_{s}^{jk}$ $\displaystyle=$
$\displaystyle\frac{1}{8\pi}({\bf E}^{2}+{\bf
B}^{2})g^{jk}-\frac{1}{4\pi}(E^{j}E^{k}+B^{j}B^{k}).$ (13)
The equations for the conservation of energy and momentum are respectively
given by [35, 36, 37]
$\displaystyle\frac{1}{\alpha}\frac{\partial}{\partial t}\epsilon_{s}$
$\displaystyle=$ $\displaystyle-\nabla\cdot{\bf S}_{s}+2{\bf a}\cdot{\bf
S}_{s},$ (14) $\displaystyle\frac{1}{\alpha}\frac{\partial}{\partial t}{\bf
S}_{s}$ $\displaystyle=$ $\displaystyle\epsilon_{s}{\bf
a}-\frac{1}{\alpha}\nabla\cdot(\alpha{\stackrel{{\scriptstyle\leftrightarrow}}{{\bf
W}}}_{s}).$ (15)
When the two-fluid plasma couples to the electromagnetic fields, the Maxwell’s
equations take the following 3+1 form:
$\displaystyle\nabla\cdot{\bf B}$ $\displaystyle=$ $\displaystyle 0,$ (16)
$\displaystyle\nabla\cdot{\bf E}$ $\displaystyle=$ $\displaystyle 4\pi\sigma,$
(17) $\displaystyle\frac{\partial{\bf B}}{\partial t}$ $\displaystyle=$
$\displaystyle-\nabla\times(\alpha{\bf E}),$ (18)
$\displaystyle\frac{\partial{\bf E}}{\partial t}$ $\displaystyle=$
$\displaystyle\nabla\times(\alpha{\bf B})-4\pi\alpha{\bf J},$ (19)
where the charge and current densities are respectively defined by
$\sigma=\sum_{s}\gamma_{s}q_{s}n_{s},\hskip 34.14322pt{\bf
J}=\sum_{s}\gamma_{s}q_{s}n_{s}{\bf v}_{s}.$ (20)
Using (10) and (16–19), the energy and momentum conservation equations (14)
and (15) can be rewritten for each species $s$ in the form
$\displaystyle\frac{1}{\alpha}\frac{\partial}{\partial
t}P_{s}-\frac{1}{\alpha}\frac{\partial}{\partial
t}[\gamma_{s}^{2}(\varepsilon_{s}+P_{s})]-\nabla\cdot[\gamma_{s}^{2}(\varepsilon_{s}+P_{s}){\bf
v}_{s}]$ $\displaystyle+\gamma_{s}q_{s}n_{s}{\bf E}\cdot{\bf
v}_{s}+2\gamma_{s}^{2}(\varepsilon_{s}+P_{s}){\bf a}\cdot{\bf v}_{s}=0,$ (21)
$\displaystyle\gamma_{s}^{2}(\varepsilon_{s}+P_{s})\left(\frac{1}{\alpha}\frac{\partial}{\partial
t}+{\bf v}_{s}\cdot\nabla\right){\bf v}_{s}+\nabla
P_{s}-\gamma_{s}q_{s}n_{s}({\bf E}+{\bf v}_{s}\times{\bf B})$
$\displaystyle+{\bf v}_{s}\left(\gamma_{s}q_{s}n_{s}{\bf E}\cdot{\bf
v}_{s}+\frac{1}{\alpha}\frac{\partial}{\partial
t}P_{s}\right)+\gamma_{s}^{2}(\varepsilon_{s}+P_{s})[{\bf v}_{s}({\bf
v}_{s}\cdot{\bf a})-{\bf a}]=0.$ (22)
Although these equations are valid in a FIDO frame, they reduce for $\alpha=1$
to the corresponding special relativistic equations [41] which are valid in a
frame in which both fluids are at rest. The transformation from the FIDO frame
to the comoving (fluid) frame involves a boost velocity, which is simply the
freefall velocity, given by
$v_{\rm ff}=(1-\alpha^{2})^{\frac{1}{2}}.$ (23)
Then the relativistic Lorentz factor $\gamma_{\rm boost}\equiv(1-v_{\rm
ff}^{2})^{-1/2}=1/\alpha$.
For a good approximation near the horizon, we write the dS metric in the
Rindler coordinate system as follows:
$ds^{2}=-\left(1-\frac{r^{2}}{\ell^{2}}\right)dt^{2}+dx^{2}+dy^{2}+dz^{2},$
(24)
where
$x=\ell\left(\theta-\frac{\pi}{2}\right),\hskip 28.45274pty=\ell\varphi,\hskip
28.45274ptz=2\ell\left(1-\frac{r^{2}}{\ell^{2}}\right)^{1/2}.$ (25)
The standard lapse function in Rindler coordinates becomes $\alpha=z/2r_{h}$,
where $r_{h}=\ell$ is the location of the cosmological event horizon.
## 4 Radial Wave Propagation in One-Dimension
We consider one-dimensional wave propagation in the radial $z$ direction and
introduce the complex variables
$\displaystyle v_{sz}(z,t)=u_{s}(z,t),\hskip
8.5359ptv_{s}(z,t)=v_{sx}(z,t)+\textrm{i}v_{sy}(z,t),$ $\displaystyle
B(z,t)=B_{x}(z,t)+\textrm{i}B_{y}(z,t),\hskip
8.5359ptE(z,t)=E_{x}(z,t)+\textrm{i}E_{y}(z,t).$ (26)
Then
$\displaystyle v_{sx}B_{y}-v_{sy}B_{x}$ $\displaystyle=$
$\displaystyle\frac{\textrm{i}}{2}(v_{s}B^{\ast}-v_{s}^{\ast}B),$
$\displaystyle v_{sx}E_{y}-v_{sy}E_{x}$ $\displaystyle=$
$\displaystyle\frac{\textrm{i}}{2}(v_{s}E^{\ast}-v_{s}^{\ast}E),$ (27)
where the $\ast$ denotes the complex conjugate. The continuity equation (8)
takes the form
$\frac{\partial}{\partial t}(\gamma_{s}n_{s})+\frac{\partial}{\partial
z}(\alpha\gamma_{s}n_{s}u_{s})=0,$ (28)
while Poisson’s equation (17) becomes
$\frac{\partial E_{z}}{\partial
z}=4\pi(q_{1}n_{1}\gamma_{1}+q_{2}n_{2}\gamma_{2}).$ (29)
The ${\bf e}_{\hat{x}}$ and ${\bf e}_{\hat{y}}$ components of (18) and (19)
give
$\displaystyle\frac{1}{\alpha}\frac{\partial B}{\partial t}$ $\displaystyle=$
$\displaystyle-\textrm{i}\left(\frac{\partial}{\partial z}-a\right)E,$ (30)
$\displaystyle\textrm{i}\frac{\partial E}{\partial t}$ $\displaystyle=$
$\displaystyle-\alpha\left(\frac{\partial}{\partial
z}-a\right)B-\textrm{i}4\pi
e\alpha(\gamma_{2}n_{2}v_{2}-\gamma_{1}n_{1}v_{1}).$ (31)
Differentiating equation (31) with respect to $t$ and using (30), we obtain
$\left(\alpha^{2}\frac{\partial^{2}}{\partial
z^{2}}+\frac{3\alpha}{2r_{h}}\frac{\partial}{\partial
z}-\frac{\partial^{2}}{\partial t^{2}}+\frac{1}{4r_{h}^{2}}\right)E=4\pi
e\alpha\frac{\partial}{\partial
t}(n_{2}\gamma_{2}v_{2}-n_{1}\gamma_{1}v_{1}).$ (32)
The transverse component of the momentum conservation equation is obtained
from the ${\bf e}_{\hat{x}}$ and ${\bf e}_{\hat{y}}$ components of (22) as
follows:
$\rho_{s}\frac{Dv_{s}}{D\tau}=q_{s}n_{s}\gamma_{s}(E-\textrm{i}v_{s}B_{z}+\textrm{i}u_{s}B)-u_{s}v_{s}\rho_{s}a-v_{s}\left(q_{s}n_{s}\gamma_{s}{\bf
E}\cdot{\bf v}_{s}+\frac{1}{\alpha}\frac{\partial P_{s}}{\partial t}\right),$
(33)
where
${\bf E}\cdot{\bf v}_{s}=\frac{1}{2}(Ev_{s}^{\ast}+E^{\ast}v_{s})+E_{z}u_{s}$
and $\rho_{s}$ is the total energy density defined by
$\rho_{s}=\gamma_{s}^{2}(\varepsilon_{s}+P_{s})=\gamma_{s}^{2}(m_{s}n_{s}+\Gamma_{g}P_{s})$
(34)
with $\Gamma_{g}=\gamma_{g}/(\gamma_{g}-1)$.
## 5 Linearization
We use perturbation method to linearize the equations derived in the preceding
section by introducing the quantities
$\displaystyle u_{s}(z,t)$ $\displaystyle=$ $\displaystyle u_{os}(z)+\delta
u_{s}(z,t),\hskip 14.22636ptv_{s}(z,t)=\delta v_{s}(z,t),$ $\displaystyle
n_{s}(z,t)$ $\displaystyle=$ $\displaystyle n_{os}(z)+\delta n_{s}(z,t),\hskip
14.22636ptP_{s}(z,t)=P_{os}(z)+\delta P_{s}(z,t),$
$\displaystyle\rho_{s}(z,t)$ $\displaystyle=$
$\displaystyle\rho_{os}(z)+\delta\rho_{s}(z,t),\hskip 14.22636pt{\bf
E}(z,t)=\delta{\bf E}(z,t),$ $\displaystyle{\bf B}_{z}(z,t)$ $\displaystyle=$
$\displaystyle{\bf B}_{o}(z)+\delta{\bf B}_{z}(z,t),\hskip 14.22636pt{\bf
B}(z,t)=\delta{\bf B}(z,t),$ (35)
where magnetic field is chosen to lie along the radial ${\bf e}_{\hat{z}}$
direction. The relativistic Lorentz factor is also linearized such that
$\gamma_{s}=\gamma_{os}+\delta\gamma_{s},\qquad\mbox{where}\quad\gamma_{os}=\left(1-{\bf
u}_{os}^{2}\right)^{-\frac{1}{2}},\quad\delta\gamma_{s}=\gamma_{os}^{3}{\bf
u}_{os}\cdot\delta{\bf u}_{s}.$ (36)
Near the horizon the unperturbed radial velocity for each species as measured
by a FIDO along ${\bf e}_{\hat{z}}$ is assumed to be the freefall velocity so
that
$u_{os}(z)=v_{\textrm{ff}}(z)=[1-\alpha^{2}(z)]^{\frac{1}{2}}.$ (37)
It follows, from the continuity equation (28), that
$r^{2}\alpha\gamma_{os}n_{os}u_{os}=\mbox{const.}=r_{h}^{2}\alpha_{h}\gamma_{h}n_{h}u_{h},$
where the values with a subscript $h$ are the limiting values at the horizon.
The freefall velocity at the horizon becomes unity so that $u_{h}=1$. Since
$u_{os}=v_{\textrm{ff}}$, $\gamma_{os}=1/\alpha$; hence,
$\alpha\gamma_{os}=\alpha_{h}\gamma_{h}=1$. Also, because
$v_{\textrm{ff}}=r/r_{h}$, the number density for each species can be written
as follows:
$n_{os}(z)=n_{hs}v_{\textrm{ff}}^{-3}.$ (38)
The equation of state (11) and (38) lead to write the unperturbed pressure, in
terms of the freefall velocity, as follows:
$P_{os}(z)=P_{hs}v_{\textrm{ff}}^{-3\gamma_{g}}.$ (39)
Since $P_{os}=k_{B}n_{os}T_{os}$, then with $k_{B}=1$, the temperature profile
is
$T_{os}=T_{hs}v_{\rm ff}^{-3(\gamma_{g}-1)}(z).$ (40)
The unperturbed magnetic field is purely in the radial direction and it does
not experience effects of spatial curvature. From the flux conservation
$\nabla\cdot{\bf B}_{o}=0$ it follows that
$r^{2}B_{o}(r)=\mbox{const.}$
One can obtain from this the unperturbed magnetic field, in terms of the
freefall velocity, as follows:
$B_{o}(z)=B_{h}v_{\textrm{ff}}^{-2},$ (41)
where $v_{\textrm{ff}}=[1-\alpha^{2}(z)]^{1/2}$. Since
$\frac{dv_{\textrm{ff}}}{dz}=-\frac{\alpha}{2r_{h}}\frac{1}{v_{\rm ff}},$ (42)
we have
$\displaystyle\frac{du_{os}}{dz}$ $\displaystyle=$
$\displaystyle-\frac{\alpha}{2r_{h}}\frac{1}{v_{\textrm{ff}}},\qquad\frac{dB_{o}}{dz}=\frac{\alpha}{r_{h}}\frac{B_{o}}{v_{\textrm{ff}}^{2}},$
$\displaystyle\frac{dn_{os}}{dz}$ $\displaystyle=$
$\displaystyle\frac{3\alpha}{2r_{h}}n_{os},\qquad\frac{dP_{os}}{dz}=\frac{3\alpha}{2r_{h}}\frac{\gamma_{g}P_{os}}{v_{\rm
ff}^{2}}.$ (43)
When the linearized variables from (35) and (36) are substituted into the
continuity equation and products of perturbation terms are neglected, the
result gives
$\displaystyle\gamma_{os}\left(\frac{\partial}{\partial
t}+u_{os}\alpha\frac{\partial}{\partial
z}+\frac{u_{os}}{2r_{h}}+\gamma_{os}^{2}\alpha\frac{du_{os}}{dz}\right)\delta
n_{s}+\left(\alpha\frac{\partial}{\partial
z}+\frac{1}{2r_{h}}\right)(n_{os}\gamma_{os}u_{os})$
$\displaystyle+n_{os}\gamma_{os}^{3}\left[u_{os}\frac{\partial}{\partial
t}+\alpha\frac{\partial}{\partial
z}+\frac{1}{2r_{h}}+\alpha\left(\frac{1}{n_{os}}\frac{dn_{os}}{dz}+3\gamma_{os}^{2}u_{os}\frac{du_{os}}{dz}\right)\right]\delta
u_{s}=0.$ (44)
In the similar way, we obtain from the conservation of entropy, (11),
$\delta P_{s}=\frac{\gamma_{g}P_{os}}{n_{os}}\delta n_{s},$ (45)
and from the total energy density, (34),
$\delta\rho_{s}=\frac{\rho_{os}}{n_{os}}\left(1+\frac{\gamma_{os}^{2}\gamma_{g}P_{os}}{\rho_{os}}\right)\delta
n_{s}+2u_{os}\gamma_{os}^{2}\rho_{os}\delta u_{s},$ (46)
where $\rho_{os}=\gamma_{os}^{2}(m_{s}n_{os}+\Gamma_{g}P_{os})$. Linearizing
the transverse part of the momentum conservation equation, differentiating it
with respect to $t$ and then substituting from (30), we obtain
$\displaystyle\left(\alpha u_{os}\frac{\partial}{\partial
z}+\frac{\partial}{\partial t}-\frac{u_{os}}{2r_{h}}+\frac{\textrm{i}\alpha
q_{s}\gamma_{os}n_{os}B_{o}}{\rho_{os}}\right)\frac{\partial\delta
v_{s}}{\partial t}$ $\displaystyle-\frac{\alpha
q_{s}\gamma_{os}n_{os}}{\rho_{os}}\left(\alpha u_{os}\frac{\partial}{\partial
z}+\frac{\partial}{\partial t}-\frac{u_{os}}{2r_{h}}\right)\delta E=0.$ (47)
When linearized, Poisson’s equation (29) and (32) respectively give
$\displaystyle\frac{\partial\delta E_{z}}{\partial z}$ $\displaystyle=$
$\displaystyle 4\pi e(n_{o2}\gamma_{o2}-n_{o1}\gamma_{o1})+4\pi
e(\gamma_{o2}\delta n_{2}-\gamma_{o1}\delta n_{1})$ (48) $\displaystyle+4\pi
e(n_{o2}u_{o2}\gamma_{o2}^{3}\delta u_{2}-n_{o1}u_{o1}\gamma_{o1}^{3}\delta
u_{1}),$ $\left(\alpha^{2}\frac{\partial^{2}}{\partial
z^{2}}+\frac{3\alpha}{2r_{h}}\frac{\partial}{\partial
z}-\frac{\partial^{2}}{\partial t^{2}}+\frac{1}{4r_{h}^{2}}\right)\delta
E=4\pi e\alpha\left(n_{o2}\gamma_{o2}\frac{\partial\delta v_{2}}{\partial
t}-n_{o1}\gamma_{o1}\frac{\partial\delta v_{1}}{\partial t}\right).$ (49)
## 6 Dispersion Relation
Our consideration effects on a local scale for which the distance from the
horizon does not vary significantly. We use a local (or mean-field)
approximation for the lapse function and hence for the equilibrium fields and
fluid quantities. If the plasma is situated relatively close to the horizon,
$\alpha^{2}\ll 1$, then a relatively small change in distance $z$ will make a
significant difference to the magnitude of $\alpha$. Thus it is important to
choose a sufficiently small range in $z$ so that $\alpha$ does not vary much.
We consider thin layers in the ${\bf e}_{\hat{z}}$ direction, each layer with
its own $\alpha_{o}$, where $\alpha_{o}$ is some mean value of $\alpha$ within
a particular layer. Then a more complete picture can be built up by
considering a large number of layers within a chosen range of $\alpha_{o}$
values.
The local approximation imposes the restriction that the wavelength must be
smaller in magnitude than the scale of the gradient of the lapse function
$\alpha$, i.e., $\lambda<(\partial\alpha/\partial z)^{-1}=2\ell$, or
equivalently, $k>(\pi/\ell)$.
One of the disadvantages of the hydrodynamical approach is that it is
essentially a bulk, fluid approach and therefore the microscopic behavior of
the two-fluid plasma is treated in a somewhat approximate manner via the
equation of state. It means that the results are really only strictly valid in
the long wavelength limit. However, the restriction, imposed by the local
approximation, on the wavelength is not too severe and permits the
consideration of intermediate to long wavelengths so that the small $k$ limit
is still valid.
In the local approximation for $\alpha$, $\alpha\simeq\alpha_{o}$ is valid
within a particular layer. Hence, the unperturbed fields and fluid quantities
and their derivatives, which are functions of $\alpha$, take on their
corresponding “mean-field”values for a given $\alpha_{o}$. Then the
coefficients in (44), (47) and (48) are constants within each layer with
respect to $\alpha$ (and therefore $z$ as well). Hence, it is possible to
Fourier transform the equations with respect to $z$, assuming plane-wave-type
solutions for the perturbations of the form $\sim e^{(kz-\omega t)}$ for each
$\alpha_{o}$ layer.
When Fourier transformed, (47) and (49) turn out to be
$\delta E=\frac{\textrm{i}4\pi e\alpha_{o}\omega(n_{o2}\gamma_{o2}\delta
v_{2}-n_{o1}\gamma_{o1}\delta
v_{1})}{\alpha_{o}k(\alpha_{o}k-\textrm{i}3/2r_{h})-\omega^{2}-1/(2r_{h})^{2}},$
(50)
$\omega\left(\alpha_{o}ku_{os}-\omega+\frac{\textrm{i}u_{os}}{2r_{h}}+\frac{\alpha_{o}q_{s}\gamma_{os}n_{os}B_{o}}{\rho_{os}}\right)\delta
v_{s}-\textrm{i}\alpha_{o}\frac{q_{s}\gamma_{os}n_{os}}{\rho_{os}}\left(\alpha_{o}ku_{os}-\omega-\frac{\textrm{i}u_{os}}{2r_{h}}\right)\delta
E=0.$ (51)
The dispersion relation for the transverse electromagnetic wave modes may be
written as
$\displaystyle\left[K_{\pm}\left(K_{\pm}\pm\frac{\textrm{i}}{2r_{h}}\right)-\omega^{2}+\frac{1}{(2r_{h})^{2}}\right]=\alpha_{o}^{2}\left\\{\frac{\omega_{p1}^{2}(\omega-
u_{o1}K_{\pm})}{(u_{o1}K_{\mp}-\omega-\alpha_{o}\omega_{c1})}+\frac{\omega_{p2}^{2}(\omega-
u_{o2}K_{\pm})}{(u_{o2}K_{\mp}-\omega+\alpha_{o}\omega_{c2})}\right\\}$ (52)
for either the electron-positron or electron-ion plasma. Here,
$K_{\pm}=\alpha_{o}k\pm\textrm{i}/2r_{h}$,
$\omega_{cs}=e\gamma_{os}n_{os}B_{o}/\rho_{os}$, and $\omega_{ps}=\sqrt{4\pi
e^{2}\gamma_{os}^{2}n_{os}^{2}/\rho_{os}}$. The cyclotron frequency
$\omega_{cs}$, as well as the plasma frequency $\omega_{ps}$, is frame
independent. Although the fluid quantities are measured in the fluid frame,
the field $B_{o}$ is measured in the FIDO frame. Hence, the factors of
$\gamma_{os}$ do not cancel out explicitly. The transformation
$B_{o}\rightarrow\gamma_{os}B_{o}$ boosts the fluid frame for either fluid and
thereby cancels the $\gamma_{os}$ factors. The $+$ and $-$ denote the left $L$
and right $R$ modes, respectively. The complex conjugate of the dispersion
relation for the $R$ mode gives the dispersion relation for the $L$ mode. In
the special relativistic case, the two modes have the same dispersion
relation.
## 7 Numerical Solution Modes
The dispersion relations (52) are complicated enough even in the simplest
cases for the electron-positron plasma where both species are assumed to have
the same equilibrium parameters, and an analytical solution is cumbersome and
unprofitable. We therefore solve numerically the dispersion relation in order
to determine all the physically meaningful modes for the transverse waves. We
put the equations in the form of a matrix equation as follows:
$(A-kI)X=0,$ (53)
where the eigenvalue is chosen to be the wave number $k$, the eigenvector $X$
is given by the relevant set of perturbations, and $I$ is the identity matrix.
In order to write the perturbation equations in an appropriate form, we
introduce the following set of dimensionless variables:
$\displaystyle\tilde{\omega}=\frac{\omega}{\alpha_{o}\omega_{\ast}},\quad\tilde{k}=\frac{kc}{\omega_{\ast}},\quad
k_{h}=\frac{1}{2r_{h}\omega_{\ast}},$
$\displaystyle\delta\tilde{u}_{s}=\frac{\delta
u_{s}}{u_{os}},\quad\tilde{v}_{s}=\frac{\delta
v_{s}}{u_{os}},\quad\delta\tilde{n}_{s}=\frac{\delta n_{s}}{n_{os}},$
$\displaystyle\delta\tilde{B}=\frac{\delta
B}{B_{o}},\quad\tilde{E}=\frac{\delta
E}{B_{o}},\quad\delta\tilde{E}_{z}=\frac{\delta E_{z}}{B_{o}}.$ (54)
For an electron-positron plasma, $\omega_{p1}=\omega_{p2}$ and
$\omega_{c1}=\omega_{c2}$; so, $\omega_{\ast}$ is defined as
$\omega_{\ast}=\left\\{\begin{array}[]{rl}&\omega_{c},\quad\mbox{Alfv\'{e}n
modes},\\\ &\\\
&(2\omega_{p}^{2}+\omega_{c}^{2})^{\frac{1}{2}},\quad\mbox{high frequency
modes},\end{array}\right.$ (55)
where $\omega_{p}=\sqrt{\omega_{p1}\omega_{p2}}$ and
$\omega_{c}=\sqrt{\omega_{c1}\omega_{c2}}$. However, for the case of an
electron-ion plasma, the plasma frequency and the cyclotron frequency are
different for each fluid; so, the choice of $\omega_{\ast}$ is a more
complicated matter. We assume, for simplicity, that
$\omega_{\ast}=\left\\{\begin{array}[]{rl}&\frac{1}{\sqrt{2}}(\omega_{c1}^{2}+\omega_{c2}^{2})^{\frac{1}{2}},\quad\mbox{Alfv\'{e}n
modes},\\\ &\\\ &(\omega_{\ast 1}^{2}+\omega_{\ast
2}^{2})^{\frac{1}{2}},\qquad\mbox{high frequency modes},\end{array}\right.$
(56)
where $\omega_{\ast s}^{2}=(2\omega_{ps}^{2}+\omega_{cs}^{2})$.
The dimensionless eigenvector for the transverse set of equations is
$\tilde{X}_{\rm transverse}=\left[\begin{array}[]{c}\delta\tilde{v}_{1}\\\
\delta\tilde{v}_{2}\\\ \delta\tilde{B}\\\ \delta\tilde{E}\end{array}\right].$
(57)
When linearized and Fourier transformed, equations (30) and (31) turn out to
be
$\left(k-\frac{\textrm{i}}{2r_{h}\alpha_{o}}\right)\delta
E+\frac{\textrm{i}\omega}{\alpha_{o}}\delta B=0,$ (58)
$\frac{\textrm{i}\omega}{\alpha_{o}}\delta
E=\left(k-\frac{\textrm{i}}{2r_{h}\alpha_{o}}\right)\delta B+4\pi
e(\gamma_{o2}n_{o2}\delta v_{2}-\gamma_{o1}n_{o1}\delta v_{1}).$ (59)
Using (54), we write (51), (58), and (59) in the dimensionless form:
$\displaystyle\tilde{k}\delta\tilde{v}_{s}$ $\displaystyle=$
$\displaystyle\left(\frac{\tilde{\omega}}{u_{os}}-\left(\frac{q_{s}}{e}\right)\frac{\omega_{cs}}{u_{os}\omega_{\ast}}-\frac{\textrm{i}k_{h}}{\alpha_{o}}\right)\delta\tilde{v}_{s}+\left(\frac{q_{s}}{e}\right)\frac{\omega_{cs}}{u_{os}\omega_{\ast}}\delta\tilde{B}-\textrm{i}\left(\frac{q_{s}}{e}\right)\frac{\omega_{cs}}{u_{os}\omega_{\ast}}\delta\tilde{E},$
(60) $\displaystyle\tilde{k}\delta\tilde{E}$ $\displaystyle=$
$\displaystyle-\textrm{i}\tilde{\omega}\delta\tilde{B}+\frac{\textrm{i}k_{h}}{\alpha_{o}}\delta\tilde{E},$
(61) $\displaystyle\tilde{k}\delta\tilde{B}$ $\displaystyle=$ $\displaystyle
u_{o1}\frac{\omega_{p1}^{2}}{\omega_{c1}\omega_{\ast}}\delta\tilde{v}_{1}-u_{o2}\frac{\omega_{p2}^{2}}{\omega_{c2}\omega_{\ast}}\delta\tilde{v}_{2}+\frac{\textrm{i}k_{h}}{\alpha_{o}}\delta\tilde{B}+\textrm{
i}\tilde{\omega}\delta\tilde{E}.$ (62)
These are the equations in the required form to be used as input to (53).
## 8 Results
We carried out the numerical analysis using the well known MATLAB. We have
considered both the electron-positron plasma and the electron-ion plasma. The
limiting horizon values for the electron-positron plasma are taken to be
$n_{hs}=10^{18}\textrm{cm}^{-3},\quad T_{hs}=10^{10}\textrm{K},\quad
B_{h}=3\times 10^{6}\textrm{G},\quad\mbox{and}\quad\gamma_{g}=\frac{4}{3}.$
(63)
For the electron-ion plasma, the ions are essentially non relativistic, and
the limiting horizon values are chosen to be
$n_{h1}=10^{18}\textrm{cm}^{-3},\quad T_{h1}=10^{10}\textrm{K};\quad
n_{h2}=10^{15}\textrm{cm}^{-3},\quad T_{h2}=10^{12}\textrm{K}.$ (64)
The equilibrium magnetic field has the same value as it has for the electron-
positron case. The gas constant is $\gamma_{g}=4/3$.
### 8.1 Alfvén Modes
#### 8.1.1 Electron-Positron Plasma
For the ultrarelativistic electron-positron plasma in the special relativistic
case, only one purely real Alfvén mode exists [41], while for the
Schwarzschild case there are two Alfvén modes [33]. In our analysis we find
three Alfvén modes, as shown in Fig. 1 and Fig. 2., for the electron-positron
plasma. The Alfvén modes in de Sitter space are interesting in that, there
exists three Alfvén modes for the electron-positron plasma compared with four
modes for the electron-ion plasma. These three modes for the electron-positron
plasma coalesce into a single mode on taking the special relativistic limit,
giving the result of ref. [41]. Since we have used the convention
$e^{\textrm{i}kz}=e^{\textrm{i}[\textrm{Re}(k)+\textrm{iIm}(k)]}$, the damping
corresponds to $\textrm{Im}(\tilde{k})>0$ and growth to
$\textrm{Im}(\tilde{k})<0$.
#### 8.1.2 Electron-Ion Plasma
In this case four modes are found, two of which are growth and the other two
are damped. The modes shown in Fig. 3 and Fig. 4 are damped and the remaining
two modes shown in Fig. 5 and Fig. 6 are growth. The first two modes are
equivalent to the modes shown in Fig. 1 and the other two are equivalent to
the modes shown in Fig. 2 for electron-positron plasma.
The differences in the magnitudes of the $\omega_{c1}$ and $\omega_{c2}$ for
the first two modes apparently lead to take the frequencies from their
negative (and therefore unphysical) values for the electron-positron case to
positive physical values for the electron-ion case. These changes are thus
because of the difference in mass and density factors as between the positrons
and ions.
These four modes for electron-ion plasma are equivalent to those of the
Schwarzschild case [33]. It is evident that the growth and damping rates are
independent of the frequency, but depended only on the value of $\alpha_{o}$.
### 8.2 High Frequency Modes
#### 8.2.1 Electron-Positron Plasma
In this case three high frequency electromagnetic modes are found for the
electron-positron plasma, as shown in Figs. 7–9. High frequency modes in the
horizon of dS space for each fluid are interesting in that all the modes are
both damping and growth modes.
The two modes, shown in Figs. 7 and 8, are similar and both modes are damping
and growth very near to the horizon. These two modes are growth for most of
the frequency domain but shows damped for lower frequencies as
$\omega\rightarrow.04$ and $\alpha_{o}\rightarrow 0$. Thus at a distance from
the horizon corresponding to $\alpha_{o}\rightarrow 0.2$ it appears that
energy is no longer fed into wave mode by the gravitational field but begins
to be drained from the waves. The third mode shown in Fig. 9 is also growth
and damping mode. This mode is damped for most of the frequency domain but
shows growth for lower frequencies as $\omega\rightarrow.04$ and
$\alpha_{o}\rightarrow 0$. These three modes are equivalent to the three modes
of ref. [33] for this case. Also these three modes coalesce with a single
modes in the special relativistic case [41] as $\alpha_{o}\rightarrow 1$.
#### 8.2.2 Electron-Ion Plasma
Similar as for the electron-positron plasma, the electron-ion plasma has three
high frequency modes. Two of these, shown in Fig. 10 and Fig. 11, shows
damping and growth very near to the horizon for lower frequencies, and for
upper frequencies they are growth modes. These two modes are growth for higher
frequency but shows damped for lower frequencies as $\omega\rightarrow.02$ and
$\alpha_{o}\rightarrow 0$. The third mode, shown in Fig. 12, is growth for
lower frequencies as $\omega\rightarrow.02$ and $\alpha_{o}\rightarrow 0$ and
shows damped for higher frequencies. These three modes are similar to the
modes for high frequency electron-ion plasma in the Schwarzschild case [33].
## 9 Concluding Remarks
The main concern of this study has been exclusively the investigation, within
the local approximation, of Alfvén and high frequency transverse
electromagnetic waves in a two-plasma in the purely de Sitter space. We derive
the dispersion relations for the Alfv en and high frequency electromagnetic
waves by using a local approximation and give their numerical solutions. In
the limit $\ell\rightarrow 0$ our results reduce to that in special relativity
as obtained by Sakai and Kawata [41] (i.e., only one purely real mode for
Alfvén and high frequency electromagnetic waves). In contrast to the work of
Sakai and Kawata [41], new modes (damped or growth) arise for the Alfv en and
high frequency electromagnetic waves in the pure dS space. In our work all the
modes for Alfvén waves are either damped or growing, but for high frequency
electromagnetic waves all the modes are both damped and growing. This is
because of the singularity of the de Sitter space. For the electron-positron
plasma, the damping and growth rates are similar with the electron-ion plasma
but different by several orders of magnitude, compared with the real
components of the wave number. For both the fluid components the damping and
growth rates are obviously frequency independent, but are dependent on the
radial distance from the horizon as denoted by the mean value of the lapse
function $\alpha_{o}$. This is of course not for the case of the high
frequency waves. In that case the rate of damping or growth is dependent on
both frequency and radial distance from the horizon. Damped modes demonstrate,
at least in this approximation, that energy is being drained from some of the
waves by the gravitational field. The majority of the modes are growth rates
and that indicate that the gravitational field is feeding energy into the
waves.
In the light of recent astronomical observations, it has been suggested that
our universe will asymptotically approach a de Sitter space [1]. Hence,
aspects of the de Sitter space might be of interest in a broader context. Our
study of plasmas in the de Sitter space is thus well motivated.
Acknowledgement
One of the authors (MHA) thanks the SIDA as well as the Abdus Salam
International Centre for Theoretical Physics (ICTP), Trieste, Italy, for
supporting with an Associate position of the Centre.
Figure 1: Top: Real part of Alfvén mode for the electron-positron plasma.
Middle: Imaginary part of Alfvén damped mode. Bottom: Imaginary part of Alfvén
growth mode.
Figure 2: Left: Real part of Alfvén mode for the electron-positron plasma.
Right: Imaginary part of Alfvén growth mode.
Figure 3: Left: Real part of Alfvén mode for the electron-ion plasma. Right:
Imaginary part of Alfvén damped mode.
Figure 4: Left: Real part of Alfvén mode for the electron-ion plasma. Right:
Imaginary part of Alfvén damped mode.
Figure 5: Left: Real part of Alfvén mode for the electron-ion plasma. Right:
Imaginary part of Alfvén growth mode.
Figure 6: Left: Real part of Alfvén mode for the electron-ion plasma. Right:
Imaginary part of Alfvén growth mode.
Figure 7: Left: Real part of high frequency mode for the electron-positron
plasma. Right: Imaginary part of high frequency damping and growth mode.
Figure 8: Left: Real part of high frequency mode for the electron-positron
plasma. Right: Imaginary part of high frequency damping and growth mode.
Figure 9: Left: Real part of high frequency mode for the electron-positron
plasma. Right: Imaginary part of high frequency growth and damping mode.
Figure 10: Left: Real part of high frequency mode for the electron-ion plasma.
Right: Imaginary part of high frequency damping and growth mode.
Figure 11: Left: Real part of high frequency mode for the electron-ion plasma.
Right: Imaginary part of high frequency damping and growth mode.
Figure 12: Left: Real part of high frequency mode for the electron-ion plasma.
Right: Imaginary part of high frequency growth and damping mode.
## References
* [1] N. Bahcall, J. P. Ostriker, S. Perlmutter, and P. J. Steinhardt, Science 284, (1999)1481.
* [2] A.G. Reiss, et al., Astron. J. 116, (1998)1009.
* [3] S. Perlmutter, et al., Astron. J. 517, (1999)565.
* [4] J.P. Ostriker, P.J. Steinhardt, Nature 377, (1995)600.
* [5] S. Perlmutter et al., Astrophys. J. 483, (1997)565.
* [6] B. Schmidt et al., Astrophys. J. 507, (1998)46.
* [7] J. Maldacena, Adv. Theor. Math. Phys. 2, (1998)231.
* [8] E. Witten, Adv. Theor. Math. Phys. 2, (1998)253.
* [9] S. Gubser, I. Klebanov, and A. Polyakov, Phys. Lett. B 428, (1998)105.
* [10] O. Aharony, S. Gubser, J. Maldacena, H. Ooguri, and Y. Oz, Phys. Rep. 323, (2000)183.
* [11] E. Witten, “Quantum gravity in de Sitter space,”hep-th/0106109.
* [12] A. Strominger, JHEP 0110, (2001)034.
* [13] A. Strominger, JHEP 0111, (2001)049.
* [14] D. Klemm, Nucl. Phys. B, 625, (2002)295.
* [15] C. M. Hull, J. High Energy Phys. 07, (1998)021.
* [16] Mu-In Park, Phys. Lett. B 440, (1998)275.
* [17] Mu-In Park, Nucl. Phys. B 544, (1999)377.
* [18] I. Antoniadis, P. Mazur, and E. Mottola, astro-ph/9705200.
* [19] A. Volovich, hep-th/0101176.
* [20] V. Balasubramanian, P. Horava, and D. Minic, J. High Energy Phys. 05, (2001)043.
* [21] C.M. Hull, and R.R. Khuri, Nucl. Phys. B 575, (2000)231.
* [22] P.O. Mazur and E. Mottola, Phys. Rev. D 64, (2001)104022.
* [23] M. Spradlin, A. Strominger, and A. Volovich, “Les Houches Lectures on De Sitter Space,”hep-th/0110007.
* [24] B. McInnes, Nucl. Phys. B 627, (2002)311.
* [25] R. Bousso, A. Maloney, and A. Strominger, Phys. Rev. D 65, (2002)104039.
* [26] M. Spradlin and A. Volovich, Phys. Rev. D 65, (2002)104037.
* [27] A. J. M. Medved, Phys. Rev. D 66, (2002)124009.
* [28] M. Parikh, Phys. Lett. B 546, (2002)189.
* [29] R. Bousso, A. Maloney, and A. Strominger, Phys. Rev. D 65, (2002)104039.
* [30] Kyung-Seok Cha, Bum-Hoon Lee, and Chanyong Park, J. Korean Phys. Soc. 42, (2003)735 [hep-th/0207194].
* [31] S. Nojiri and S. Odintsov, J. High Energy Phys. 12, (2001)033.
* [32] S. Nojiri and S. Odintsov, Phys. Lett. B 523, (2001)165.
* [33] V. Buzzi, K. C. Hines, R. A. Treumann: Phys. Rev. D 51, (1995)6663.
* [34] V. Buzzi, K. C. Hines, R. A. Treumann: Phys. Rev. D 51, (1995)6677.
* [35] K.S. Thorne, D.A. Macdonald: Mon. Not. R. Astron. Soc. 198,(1982)339.
* [36] D.A. Macdonald, K.S. Thorne: Mon. Not. R. Astron. Soc. 198,(1982)345.
* [37] R.H. Price, K.S. Thorne: Phys. Rev. D 33,(1986)915.
* [38] K.S. Thorne, R.H. Price, D.A. Macdonald: Black Holes: The Membrane Paradigm. Yale University Press, New Haven (1986).
* [39] E. Harris: Phys. Rev. 108,(1957)1357.
* [40] F. Jüttner: Ann. Phys. 34,(1911)856(Leipzig).
* [41] J. Sakai, T. Kawata: J. Phys. Soc. Jpn. 49,(1980)747.
| arxiv-papers | 2008-06-17T10:53:03 | 2024-09-04T02:48:56.302769 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "M. Atiqur Rahman and M. Hossain Ali",
"submitter": "M Hossain Ali",
"url": "https://arxiv.org/abs/0806.2740"
} |
0806.2804 | # Statistical Ensembles with Volume Fluctuations
Mark I. Gorenstein Bogolyubov Institute for Theoretical Physics, Kiev,
Ukraine
###### Abstract
The volume fluctuations in statistical mechanics are discussed. First, the
volume fluctuations in ensembles with a fixed external pressure, the so called
pressure ensembles, are considered. Second, a generalization of the pressure
ensembles is suggested. Namely, the statistical ensembles with the volume
fluctuating according to externally given distributions are considered.
Several examples and possible applications in statistical models of hadron
production are discussed.
statistical ensembles, pressure ensembles, volume fluctuations, particle
number fluctuations
###### pacs:
24.10.Lx, 24.60.Ky, 25.75.-q
## I Introduction
Successful application of the statistical model to description of mean hadron
multiplicities in high energy collisions (see, e.g., recent papers stat1 and
references therein) has stimulated investigations of properties of statistical
ensembles. Whenever possible, one prefers to use the grand canonical ensemble
(GCE) due to its mathematical convenience. The canonical ensemble (CE) CE
should be applied when the number of carriers of a conserved charges is small
(of the order of 1), such as strange hadrons strange , anti-baryons
antibaryons , or charmed hadrons charm . The micro-canonical ensemble (MCE)
MCE has been used to describe small systems with fixed energy, e.g. mean
hadron multiplicities in proton-antiproton annihilation at rest. In all these
cases, calculations performed in different statistical ensembles yield
different results. This happens because the systems are ‘small’ and they are
‘far away’ from the thermodynamic limit (TL). The mean multiplicity of hadrons
in relativistic heavy ion collisions ranges from $10^{2}$ to $10^{4}$, and
mean multiplicities (of light hadrons) obtained within GCE, CE, and MCE
approach each other. One refers here to the thermodynamical equivalence of
statistical ensembles and uses the GCE for calculating the hadron yields.
Measurements of a hadron multiplicity distribution $P(N)$ in interactions,
including nucleus-nucleus collisions, open a new field of applications of the
statistical models. The particle multiplicity fluctuations are usually
quantified by the ratio of variance to mean value of a multiplicity
distribution $P(N)$, the scaled variance, and are a subject of current
experimental activities. In statistical models there is a qualitative
difference in the properties of mean multiplicity and scaled variance of
multiplicity distributions. It was recently found fluc ; mce1 ; mce2 ; res ;
exp ; clt ; acc that even in the TL corresponding results for the scaled
variance are different in different ensembles. Hence the equivalence of
ensembles holds for mean values in the TL, but does not extend to
fluctuations.
A statistical system is characterized by the extensive quantities: volume $V$,
energy $E$, and conserved charge(s)111In statistical description of the hadron
or quark-gluon systems, these conserved charges are usually the net baryon
number, strangeness, and electric charge. In non-relativistic statistical
mechanics, the number of particles plays the role of a conserved ‘charge’.
$Q$. The MCE is defined by the postulate that all micro-states with given $V$,
$E$, and $Q$ have equal probabilities of being realized. This is the basic
postulate of the statistical mechanics. The MCE partition function just
calculates the number of microscopic states with given fixed $(V,E,Q)$ values.
In the CE the energy exchange between the considered system and ‘infinite
thermal bath’ is assumed. Consequently, a new parameter, temperature $T$ is
introduced. To define the GCE, one makes a similar construction for conserved
charge $Q$. An ‘infinite chemical bath’ and the chemical potential $\mu$ are
introduced. The CE introduces the energy fluctuations. In the GCE, there are
additionally the charge fluctuations. The MCE, CE, and GCE are the most
familiar statistical ensembles. In several textbooks (see, e.g., Ref. RR ;
Tolpygo ), the pressure (or isobaric) canonical ensemble has been also
discussed. The ‘infinite bath of the fixed external pressure’ $p_{0}$ is then
introduced. This leads to the volume fluctuations around the average value.
A more general concept of the statistical ensembles was suggested in Ref.
alpha . The statistical ensemble is defined by an externally given
distribution of extensive quantities, $P_{\alpha}(\vec{A})$. The construction
of distribution of any property $O$ in such an ensemble proceeds in two steps.
Firstly, the MCE $O$-distribution, $P_{mce}(O;\vec{A})$, is calculated at
fixed values of the extensive quantities $\vec{A}=(V,E,Q)$. Secondly, this
result is averaged over the external distribution $P_{\alpha}(\vec{A})$ alpha
,
$\displaystyle P_{\alpha}(O)~{}=~{}\int
d\vec{A}~{}P_{\alpha}(\vec{A})~{}P_{mce}(O;\vec{A})~{}.$ (1)
The ensemble defined by Eq. (1), the $\alpha$-ensemble, includes the standard
statistical ensembles as particular cases.
Recently, the micro-canonical ensemble with the volume fluctuations was
introduced MCEsVF for modelling the hadron production in high energy
interactions. An introduction of the volume fluctuations was necessary in
order to reproduce the KNO-scaling KNO of the hadron multiplicity
distribution observed in proton-proton and proton-antiproton collisions at
high energies. The volume fluctuations lead also to the power law shape of the
single particle spectra at high transverse momenta.
In the present paper the statistical ensembles with volume fluctuations are
studied. First, the pressure ensembles are considered. In general, there are 3
pairs222 In the present study we do not discuss the role of the total
3-momentum. As shown in Ref. acc the total momentum conservation is not
important in the TL for thermodynamical functions and fluctuations in the full
phase space. It may however influence the particle number fluctuations in the
limited segments of the phase space. of variables –
$(V,p_{0}),~{}(E,T),~{}(Q,\mu)$ – and, thus, the 8 statistical ensembles333For
several conserved charges $\\{Q_{i}\\}$ the number of possible ensembles is
larger, as each charge can be treated either canonically or grand canonically.
can be constructed. Among these 8 ensembles there are 4 pressure ensembles:
$(p_{0},E,Q)$, $(p_{0},T,Q)$, $(p_{0},E,\mu)$, and $(p_{0},T,\mu)$. In
addition to the pressure canonical ensemble known from the literature, three
other possibilities – pressure micro-canonical, pressure grand micro-
canonical, and pressure grand canonical ensembles – are constructed and
studied. In Section II, non-relativistic statistical systems are discussed,
whereas Section III presents the results for an ultra-relativistic ideal gas.
In Section IV the concept of pressure ensembles is extended to the case of
more general volume fluctuations. Namely, the statistical ensembles with
volume fluctuating according to externally given distributions are introduced.
The Summary presented in Section V closes the paper.
## II Non-Relativistic Boltzmann Gas
### II.1 Canonical and Micro-Canonical Ensembles
In this Section the system of non-relativistic Boltzmann particles is
discussed. The $(V,T,N)$ Canonical Ensemble (CE) partition function of $N$
particles reads landau :
$\displaystyle Z_{ce}(V,T,N)~{}=~{}\frac{1}{N!}\,\int\frac{d{\bf x}_{1}d{\bf
p}_{1}}{(2\pi)^{3}}~{}\ldots~{}\frac{d{\bf x}_{N}d{\bf
p}_{N}}{(2\pi)^{3}}~{}\exp\left[-~{}\frac{E({\bf x}_{1},\ldots,{\bf
x}_{N};{\bf p}_{1},\ldots,{\bf p}_{N})}{T}\right]~{},$ (2)
where $T$ is the system temperature, the particle degeneracy factor is assumed
to be equal to 1, and $E$ is the microscopic $N$-particle energy usually
presented as the sum of potential and kinetic terms,
$\displaystyle E({\bf x}_{1},\ldots,{\bf x}_{N};{\bf p}_{1},\ldots,{\bf
p}_{N})~{}=~{}U({\bf x}_{1},\ldots,{\bf x}_{N})~{}+~{}\sum_{i=1}^{N}\frac{{\bf
p}_{i}^{2}}{2m}~{},$ (3)
with $m$ being the particle mass. For the $N$-particle energy given by (3) the
integration over momentum in Eq. (2) can be done explicitly,
$\displaystyle
Z_{ce}(V,T,N)~{}=~{}\frac{1}{N!}~{}\left(\frac{mT}{2\pi}\right)^{3N/2}~{}\int_{V}d{\bf
x}_{1}~{}\ldots~{}d{\bf x}_{N}~{}\exp\left[-~{}\frac{U({\bf x}_{1},\ldots,{\bf
x}_{N})}{T}\right]~{}.$ (4)
The particle coordinate ${\bf x}_{1},\ldots,{\bf x}_{N}$ are integrated over
the system volume $V$. The CE thermodynamical functions can be expressed in
terms of the free energy,
$\displaystyle F(V,T,N)~{}=-~{}T~{}\ln~{}Z_{ce}(V,T,N)~{}.$ (5)
The CE pressure, entropy, chemical potential, and average energy are:
$\displaystyle p~{}=~{}-~{}\left(\frac{\partial F}{\partial
V}\right)_{T,N}~{},~{}~{}~{}~{}S~{}=~{}-\left(\frac{\partial F}{\partial
T}\right)_{V,N}~{},~{}~{}~{}~{}\mu~{}=~{}\left(\frac{\partial F}{\partial
N}\right)_{T,V}~{},~{}~{}~{}~{}\overline{E}~{}=~{}F~{}+~{}TS~{}.$ (6)
For the non-interacting particles, the potential energy vanishes, $U({\bf
x}_{1},\ldots,{\bf x}_{N})=0$, thus, the ideal gas free energy is:
$\displaystyle
F_{id}(V,T,N)~{}\cong~{}-NT~{}-~{}NT~{}\ln\left[\frac{V}{N}~{}\left(\frac{mT}{2\pi}\right)^{3/2}\right]~{},$
(7)
where we have assumed $N\gg 1$ and, thus, $\ln N!\cong N\ln N-N$. The
thermodynamical functions of the ideal gas from Eqs. (6,7) read:
$\displaystyle
p=\frac{NT}{V},~{}~{}S=\frac{5}{2}N+N\ln\left[\frac{V}{N}\left(\frac{mT}{2\pi}\right)^{3/2}\right],~{}~{}\mu=-T\ln\left[\frac{V}{N}\left(\frac{mT}{2\pi}\right)^{3/2}\right],~{}~{}\overline{E}=\frac{3}{2}NT~{}.$
(8)
The $(V,E,N)$ Micro-Canonical Ensemble (MCE) partition function of $N$ non-
relativistic Boltzmann particles reads:
$\displaystyle Z_{mce}(V,E,N)~{}=~{}\frac{1}{N!}\,\int\frac{d{\bf x}_{1}d{\bf
p}_{1}}{(2\pi)^{3}}~{}\ldots~{}\frac{d{\bf x}_{N}d{\bf
p}_{N}}{(2\pi)^{3}}~{}\delta\left[E-~{}U({\bf x}_{1},\ldots,{\bf
x}_{N})-\sum_{j=1}^{N}~{}\frac{{\bf p}_{j}^{2}}{2m}\right]$ (9)
The MCE entropy is defined as,
$\displaystyle S(V,E,N)~{}=~{}\ln\left[E_{0}~{}Z_{mce}(V,E,N)\right]~{},$ (10)
where $E_{0}$ is an arbitrary constant with a dimension of energy. For non-
interacting particles, $U({\bf x}_{1},\ldots,{\bf x}_{N})=0$, one finds,
$\displaystyle
Z_{mce}(V,E,N)=\frac{V^{N}(Em/2\pi)^{3N/2}}{E~{}N!~{}\Gamma(3N/2)}~{},~{}~{}~{}S(V,E,N)\cong\frac{5}{2}N+N\ln\left[\frac{V}{N}\left(\frac{mE}{3N\pi}\right)^{3/2}\right]~{}.$
(11)
It has been assumed that $N\gg 1$ and $E\gg E_{0}$, thus, $\Gamma(3N/2)\cong
3N/2~{}[\ln(3N/2)-1]$, $N!\cong N(\ln N-1)$, and $\ln(E_{0}\cdot
E^{3N/2-1})\cong\ln(E^{3N/2})$.
The MCE temperature, pressure, and chemical potential are the following:
$\displaystyle\frac{1}{T}~{}=~{}\left(\frac{\partial S}{\partial
E}\right)_{V,N}~{},~{}~{}~{}~{}~{}~{}\frac{p}{T}~{}=~{}\left(\frac{\partial
S}{\partial
V}\right)_{E,N}~{},~{}~{}~{}~{}~{}\frac{\mu}{T}~{}=~{}-~{}\left(\frac{\partial
S}{\partial N}\right)_{V,E}~{}.$ (12)
For non-interacting particles they read,
$\displaystyle\frac{1}{T}~{}=~{}\frac{3}{2}~{}\frac{N}{E}~{},~{}~{}~{}~{}~{}~{}\frac{p}{T}~{}=~{}\frac{N}{V}~{},~{}~{}~{}~{}~{}\frac{\mu}{T}~{}=~{}-~{}\ln\left[\frac{V}{N}\left(\frac{mE}{3N\pi}\right)^{3/2}\right]~{}.$
(13)
If $\overline{E}=E$, the ideal gas system of $N$ particles in the volume $V$
has the same temperature, pressure, and entropy in the MCE (11-13) and in the
CE (8). This means the thermodynamical equivalence of CE and MCE at $N\gg 1$.
### II.2 Pressure Canonical and Pressure Micro-Canonical Ensembles
The particle coordinates and momenta for the considered system are denoted as
${\bf x}_{1},~{}\ldots,~{}{\bf x}_{N}$ and ${\bf p}_{1},~{}\ldots,~{}{\bf
p}_{N}$, respectively. The $E$ given by Eq. (3) denotes the system energy, and
$V$ is the system volume. The ‘thermostat’ is now introduced with
corresponding particle coordinates, ${\bf X}_{1},~{}\ldots,~{}{\bf
X}_{N_{T}}$, and momenta, ${\bf P}_{1},~{}\ldots,~{}{\bf P}_{N_{T}}$, and with
$E_{T}$ and $V_{T}$ being the thermostat energy and volume, respectively. The
system plus thermostat is described by the MCE, i.e. the total energy and the
total volume are assumed to be fixed: $E+E_{T}=E^{*}=$ const ,
$V+V_{T}=V^{*}=$ const . The probability distribution of system particle
coordinates and momenta is proportional to,
$\displaystyle f_{V}({\bf x}_{1},~{}\ldots,~{}{\bf x}_{N};~{}{\bf
p}_{1},~{}\ldots,~{}{\bf p}_{N})~{}\propto~{}\int\prod_{i=1}^{N_{T}}d^{3}{\bf
X}_{i}d^{3}{\bf P}_{i}~{}\delta\left(E~{}+~{}\sum_{j=1}^{N_{T}}\frac{{\bf
P}_{j}^{2}}{2M}~{}-~{}E^{*}\right)~{},$ (14)
where the thermostat particle coordinates ${\bf X}_{1},\ldots,{\bf X}_{N_{T}}$
are integrated over the thermostat volume $V_{T}$, and the system particle
coordinates ${\bf x}_{1},\ldots,{\bf x}_{N}$ are contained in the volume $V$.
The particles in the thermostat are assumed to be non-interacting. Thus,
$E_{T}=\sum_{j=1}^{N_{T}}{\bf P}_{j}^{2}/(2M)$, with $M$ being the particle
mass. The momentum integration in Eq. (14) gives,
$\displaystyle\int d^{3}{\bf P}_{1}\ldots d^{3}{\bf
P}_{N_{T}}~{}\delta\left(E+\sum_{j=1}^{N_{T}}\frac{{\bf
P}_{j}^{2}}{2M}~{}-~{}E^{*}\right)\propto\left(E^{*}~{}-~{}E\right)^{3N_{T}/2~{}-1}~{}\propto\left(1-\frac{E}{E^{*}}\right)^{3N_{T}/2~{}-1}~{}$
$\displaystyle\cong~{}\left(1~{}-~{}\frac{E}{E_{T}}\right)^{3N_{T}/2~{}-1}\cong~{}\left(1~{}-~{}\frac{E}{3N_{T}T/2}\right)^{3N_{T}/2}~{}\cong~{}\exp(-~{}E/T)~{},$
(15)
where it has been assumed, $N_{T}\rightarrow\infty$, $E/E_{T}\rightarrow 0$,
and the ideal gas equation for the thermostat energy, $E_{T}=3TN_{T}/2$, has
been used. The integration over thermostat particle coordinates ${\bf X}_{i}$
in Eq. (14) gives the factor depending on the system volume $V$,
$\displaystyle\int_{V_{T}}~{}d^{3}{\bf X}_{1}\ldots d^{3}{\bf X}_{N_{T}}~{}$
$\displaystyle=~{}V_{T}^{N_{T}}~{}=~{}(V^{*}~{}-~{}V)^{N_{T}}~{}\propto~{}\left(1~{}-~{}\frac{V}{V^{*}}\right)^{N_{T}}~{}\cong~{}\left(1~{}-~{}\frac{V}{V_{T}}\right)^{N_{T}}~{}$
$\displaystyle=~{}\left(1~{}-~{}\frac{V}{(TN_{T}/p_{0})}\right)^{N_{T}}~{}\cong~{}\exp\left(~{}-~{}\frac{p_{0}~{}V}{T}\right)~{},$
(16)
where it has been assumed, $V_{T}\rightarrow\infty$, $N_{T}\rightarrow\infty$,
$V/V_{T}\rightarrow 0$, and the ideal gas equation for the thermostat
pressure, $p_{0}=TN_{T}/V_{T}$, has been used.
One obtains from Eqs. (14-II.2),
$\displaystyle f_{V}({\bf x}_{1},\ldots,{\bf x}_{N},~{}{\bf p}_{1},\ldots,{\bf
p}_{N})~{}=~{}\frac{1}{Z_{pce}(p_{0},T,N)}~{}\frac{1}{N!}~{}\exp\left(~{}-~{}\frac{p_{0}V~{}+~{}E}{T}\right)~{},$
(17)
where the system energy $E$ in (17) depends on particle coordinates and
momenta according to Eq. (3). The function $Z_{pce}(p_{0},T,N)$ in Eq. (17) is
the $(p_{0},T,N)$ Pressure Canonical Ensemble (PCE) partition function. It is
defined by the normalization condition,
$\displaystyle\int_{0}^{\infty}dV~{}\int\prod_{i=1}^{N}\left[\frac{d^{3}{\bf
x}_{i}d^{3}{\bf p}_{i}}{(2\pi)^{3}}\right]~{}f_{V}({\bf x}_{1},\ldots,{\bf
x}_{N};~{}{\bf p}_{1},\ldots,{\bf p}_{N})~{}=~{}1~{},$ (18)
and it equals to,
$\displaystyle
Z_{pce}(p_{0},T,N)~{}=~{}\int_{0}^{\infty}dV~{}\exp\left(-~{}\frac{p_{0}~{}V}{T}\right)~{}Z_{ce}(V,T,N)~{},$
(19)
where $Z_{ce}(V,T,N)$ is the CE partition function given by Eq. (2).
Using $Z_{ce}=\exp(-F/T)$, the probability volume distribution in the PCE can
be presented as:
$\displaystyle
W_{pce}(V)~{}=~{}\frac{1}{Z_{pce}(p_{0},T,N)}~{}\exp\left[-~{}\frac{F(V,T,N)~{}+~{}p_{0}V}{T}\right]~{}.$
(20)
First, one finds the maximum of $W_{pce}(V)$, which defines the most probable
value of the volume, $V=V_{0}$,
$\displaystyle\left[\frac{\partial F(V,T,N)}{\partial
V}\right]_{V=V_{0}}~{}+~{}p_{0}~{}=~{}0~{}.$ (21)
By definition, $-~{}\partial F/\partial V$ is the CE pressure (8). Thus, the
PCE equation of state (21) has clear physical meaning: the internal pressure
$p$ for the most probable volume $V_{0}$ equals to the fixed external pressure
$p_{0}$,
$\displaystyle p(V_{0},N,T)~{}=~{}p_{0}~{}.$ (22)
In the TL the volume distribution (20) can be approximated as,
$\displaystyle
W_{pce}(V)~{}\cong~{}\frac{1}{Z_{pce}(p_{0},T,N)}~{}\exp\left[-~{}\frac{F(V_{0},T,N)~{}+~{}p_{0}V_{0}}{T}\right]$
(23)
$\displaystyle\times~{}\exp\left[-~{}\frac{1}{2T}\left(\frac{\partial^{2}F(V,T,N)}{\partial
V^{2}}\right)_{V=V_{0}}\left(V~{}-~{}V_{0}\right)^{2}\right]\equiv\left(2\pi\omega_{V}~{}V_{0}\right)^{-1/2}~{}\exp\left[~{}-~{}\frac{(V-V_{0})^{2}}{2\omega_{V}~{}V_{0}}\right]~{}.$
Thus, the most probable volume, $V_{0}$, and the average volume $\overline{V}$
are equal to each other in the TL,
$\displaystyle\overline{V}~{}\equiv~{}\int_{0}^{\infty}dV~{}V~{}W_{pce}(V)~{}\cong~{}V_{0}~{},$
(24)
and $\omega_{V}$ introduced in Eq. (23) defines the scaled variance of the
volume fluctuations,
$\displaystyle\omega_{V}~{}\equiv~{}\frac{\overline{V^{2}}~{}-~{}\overline{V}^{2}}{\overline{V}}~{}=~{}\frac{T}{V_{0}}~{}\left[\frac{\partial^{2}F(V,T,N)}{\partial
V^{2}}\right]_{V=V_{0}}^{-1}~{}=~{}-~{}\frac{T}{V_{0}}~{}\left[\frac{\partial
p(V,T,N)}{\partial V}\right]_{V=V_{0}}^{-1}~{}.$ (25)
For the ideal gas pressure (8) one gets:
$\displaystyle\overline{V}~{}\cong~{}V_{0}~{}=~{}\frac{NT}{p_{0}}~{},~{}~{}~{}~{}\omega_{V}~{}\cong~{}\frac{T}{p_{0}}~{}.$
(26)
The Eqs. (23-25) are valid if $(\partial p/\partial V)$ is negative. This is
the case for the ‘normal’ equation of state, but is not valid for either a 1st
order phase transition or critical point. Note that the TL
$N,V_{0}\rightarrow\infty$, with $N/V_{0}$ being finite, is assumed to make
clear notions of the phase transition or critical point.
1st Order Phase Transition. In the casei of the 1st order phase transition
$\partial p/\partial V=0$ landau for $V_{1}<V<V_{2}$ in the mixed phase.
According to Eq. (20) this leads to $W_{pce}(V)=const\cong(V_{2}-V_{1})^{-1}$
for $V_{1}<V<V_{2}$. Introducing the notation, $V_{2}=\gamma V_{1}$, with
$\gamma=const>1$, one finds at $V_{1}\rightarrow\infty$,
$\displaystyle\overline{V}~{}$
$\displaystyle\cong~{}\int_{V_{1}}^{V_{2}}V~{}W_{pce}(V)dV~{}=~{}\frac{\gamma+1}{2}~{}V_{1}~{},$
(27) $\displaystyle\omega_{V}$
$\displaystyle~{}\cong~{}\frac{1}{\overline{V}}~{}\left[\int_{V_{1}}^{V_{2}}V^{2}~{}W_{pce}(V)dV~{}-~{}\overline{V}^{2}\right]~{}~{}=~{}\frac{1}{3}~{}\left(\frac{\gamma-1}{\gamma+1}\right)^{2}~{}\overline{V}~{}.$
(28)
Thus, the volume fluctuations are anomalously large,
$\omega_{V}\propto\overline{V}\rightarrow\infty$.
Critical Point. At the critical point it follows landau ,
$\displaystyle\frac{\partial p}{\partial
V}~{}=~{}\frac{\partial^{2}p}{\partial V^{2}}~{}=~{}0~{}.$ (29)
In this case, Eq. (20) takes the form,
$\displaystyle
W_{pce}(V)~{}\cong~{}\frac{1}{Z_{pce}(p_{0},T,N)}~{}\exp\left[-~{}\frac{F(V_{0},T,N)~{}+~{}p_{0}V_{0}}{T}\right]~{}$
(30)
$\displaystyle\times\exp\left[-~{}\frac{1}{T}~{}\frac{1}{4!}\left(\frac{\partial^{4}F(V,T,N)}{\partial
V^{4}}\right)_{V=V_{0}}~{}(V~{}-~{}V_{0})^{4}\right]\equiv~{}\frac{2~{}A_{0}^{1/4}}{\Gamma(1/4)}~{}\exp\left[~{}-~{}A_{0}~{}(V~{}-~{}V_{0})^{4}\right]~{},$
where $A_{0}\equiv-(24T)^{-1}(\partial^{3}p/\partial V^{3})_{V=V_{0}}$. The
volume distribution (30) leads to the scaled variance for the volume
fluctuations,
$\displaystyle\omega_{V}~{}=~{}\frac{\overline{V^{2}}-\overline{V}^{2}}{\overline{V}}~{}\cong~{}\frac{1}{V_{0}}~{}\int_{0}^{\infty}dV~{}(V-V_{0})^{2}~{}W_{pce}(V)~{}=~{}\frac{1}{V_{0}}~{}\frac{\Gamma(3/4)}{\Gamma(1/4)}~{}A_{0}^{-1/2}~{},$
(31)
where the Gamma functions in Eqs. (30-31) are $\Gamma(1/4)\cong 3.626$ and
$\Gamma(3/4)\cong 1.225$. Finally, one finds for (31),
$\displaystyle\omega_{V}~{}\cong~{}\frac{1.656~{}T^{1/2}}{V_{0}}~{}\left[-~{}\left(\frac{\partial^{3}p}{\partial
V^{3}}\right)_{V=V_{0}}\right]^{-1/2}~{}.$ (32)
In order to estimate the scaled variance (32) of the volume fluctuations at
the critical point we use the van der Waals (VdW) equation of state landau ,
$\displaystyle\left(p~{}+~{}a~{}\frac{N^{2}}{V^{2}}\right)~{}\left(V~{}-~{}b~{}N\right)~{}=~{}NT~{}.$
(33)
The critical point is defined by the following equations
$\displaystyle\frac{\partial p}{\partial V}~{}$
$\displaystyle=~{}-~{}\frac{NT}{(V-Nb)^{2}}~{}+~{}\frac{2N^{2}a}{V^{3}}~{}=~{}0~{},$
(34) $\displaystyle\frac{\partial^{2}p}{\partial V^{2}}~{}$
$\displaystyle=~{}\frac{2NT}{(V-Nb)^{3}}~{}-~{}\frac{6N^{2}a}{V^{4}}~{}=~{}0~{}.$
(35)
They give,
$\displaystyle
T_{cr}~{}=~{}\frac{8}{27}~{}\frac{a}{b}~{},~{}~{}~{}~{}V_{cr}~{}=~{}3Nb~{},~{}~{}~{}~{}p_{cr}~{}=~{}\frac{1}{27}\frac{a}{b^{2}}~{}.$
(36)
At the critical point one finds,
$\displaystyle\left(\frac{\partial^{3}p}{\partial
V^{3}}\right)_{cr}~{}=~{}\left[-~{}\frac{6NT}{(V-Nb)^{4}}~{}+~{}\frac{24N^{2}a}{V^{5}}\right]_{cr}~{}=~{}-~{}\frac{a}{9^{2}b^{5}}~{}\frac{1}{N^{3}}~{}.$
(37)
Thus, for $T=T_{cr}$, $p_{0}=p_{cr}$, and $\overline{V}\cong V_{0}=V_{cr}$,
and Eqs. (32,37) give,
$\displaystyle\omega_{V}~{}\cong~{}\frac{1.656~{}T^{1/2}}{V_{cr}}~{}\left[-~{}\left(\frac{\partial^{3}p}{\partial
V^{3}}\right)_{V=V_{cr}}\right]^{-1/2}~{}\cong~{}2.703~{}b~{}\sqrt{N}~{}.$
(38)
The volume fluctuations at the critical point are anomalously large,
$\omega_{V}\propto N^{1/2}\propto\overline{V}^{1/2}\rightarrow\infty$.
Similarly to the PCE we introduce now the $(s_{0},E,N)$ Pressure Micro-
Canonical ensemble (PMC),
$\displaystyle
Z_{pmc}(s_{0},E,N)~{}=~{}\int_{0}^{\infty}dV\exp\left(-s_{0}V\right)Z_{mce}(V,E,N)~{},$
(39)
where the parameter $s_{0}=p_{0}/T_{0}$ is defined by the external conditions
of the thermostat pressure $p_{0}$ and the thermostat temperature $T_{0}$. The
micro-canonical energy $E$ is assumed to be fixed. Thus, there is no energy
exchange between the ‘thermostat’ and considered system, and the internal MCE
temperature $T$ (12) may differ from $T_{0}$. Using Eq. (10), the volume
distribution in the PMC can be presented in the form,
$\displaystyle W_{pmc}(V)~{}$
$\displaystyle=~{}\frac{1}{Z_{pmc}(s_{0},E,N)}~{}\exp\left[-~{}s_{0}~{}V~{}+~{}S(V,E,N)\right]~{}$
(40)
$\displaystyle\cong~{}\frac{1}{Z_{pmc}(s_{0},E,N)}~{}\exp\left[-~{}s_{0}~{}V_{0}~{}+~{}S(V_{0},E,N)~{}+~{}\frac{1}{2}~{}\left(\frac{\partial^{2}S}{\partial
V^{2}}\right)_{V=V_{0}}\left(V-V_{0}\right)^{2}\right]~{}.$
The most probable (and average) volume $V_{0}$ is defined by the condition,
$\displaystyle-~{}s_{0}~{}+~{}\left(\frac{\partial S(V,E,N)}{\partial
V}\right)_{V=V_{0}}~{}=~{}0~{}.$ (41)
The Eq. (41) corresponds to,
$\displaystyle\frac{p}{T}~{}=~{}\frac{p_{0}}{T_{0}}~{},$ (42)
which is similar to Eq. (22) in the PCE. The scaled variance of the volume
fluctuations $\omega_{V}$ in the PMC equals to:
$\displaystyle\omega_{V}~{}\cong~{}-~{}\left[V_{0}~{}(\partial^{2}S/\partial
V^{2})_{V=V_{0}}\right]^{-1}~{}=~{}-~{}\frac{T}{V_{0}}~{}\left[\frac{\partial
p(V,E,N)}{\partial V}\right]_{V=V_{0}}^{-1}~{},$ (43)
which is again similar to Eq. (25) in the PCE. Thus, the average volume
$V_{0}$ and scaled variance of the volume fluctuations $\omega_{V}$ are
identical in the PCE and PMC in the TL. For the ideal gas one finds,
$\displaystyle\langle
V\rangle_{pmc}~{}\cong~{}V_{0}~{}=~{}\frac{N}{s_{0}}~{},~{}~{}~{}~{}~{}~{}\omega_{V}~{}=~{}\frac{1}{s_{0}}~{},$
(44)
which coincide with (26) in the PCE.
### II.3 Grand Canonical Ensemble
The $(V,T,\mu)$ Grand Canonical Ensemble (GCE) partition function of non-
relativistic non-interacting Boltzmann particles reads:
$\displaystyle
Z_{gce}(V,T,\mu)~{}=~{}\sum_{N=0}^{\infty}~{}\exp\left(\frac{\mu~{}N}{T}\right)~{}Z_{ce}(V,T,N)~{},$
(45)
where $Z_{ce}(V,T,N)$ is given by Eq. (2) and $\mu$ is the chemical potential.
The GCE pressure is calculated as,
$\displaystyle
p(T,\mu)~{}=~{}\frac{T}{V}~{}\ln\left[Z_{gce}(V,T,\mu)\right]~{},$ (46)
and it does not depend on the volume $V$ in the TL. The CE, MCE, and GCE are
thermodynamically equivalent at $V\rightarrow\infty$. Note that the GCE is the
most convenient one from the technical point of view.
There is, however, an evident problem in the formulation of the
$(p_{0},T,\mu)$ Pressure Grand Canonical ensemble (PGC). The PGC partition
function is obtained by extending Eq. (19),
$\displaystyle
Z_{pgc}(p_{0},T,\mu)~{}=~{}\int_{0}^{\infty}dV~{}\exp\left(-~{}\frac{p_{0}~{}V}{T}\right)~{}Z_{gce}(V,T,\mu)~{}=~{}\frac{T}{p_{0}~{}-~{}p(T,\mu)}~{},$
(47)
where $p(T,\mu)$ is the GCE pressure (46). The $(p_{0},T,\mu)$-ensemble has a
unique property. Among 8 possible ensembles this is the only one where the
system description includes only intensive quantites, $p_{0},T$ and $\mu$. For
$p(T,\mu)=p_{0}$, the system volume is undefined. In the domain $p(T,\mu)\geq
p_{0}$, the PGC partition function does not exist as the integral over the
volume in Eq. (47) diverges. For $p(T,\mu)<p_{0}$, the volume distribution has
the form:
$\displaystyle
W_{pgc}(V)~{}=~{}\frac{1}{Z_{pgc}(p_{0},T,\mu)}~{}\exp\left[-~{}V~{}\left(\frac{p_{0}~{}-~{}p}{T}\right)\right]~{}.$
(48)
The most probable volume $V_{0}$ equals to zero, but the average volume is:
$\displaystyle\overline{V}~{}=~{}\int_{0}^{\infty}dV~{}V~{}W_{pgc}(V)~{}=~{}\frac{T}{p_{0}~{}-~{}p(T,\mu)}~{}.$
(49)
This special ensemble will be discussed further in details for the ultra-
relativistic gas in the next section.
## III Ultra-Relativistic Gas
In this section several examples of the pressure ensembles for the ultra-
relativistic ($m=0$) ideal gas of Boltzmann particles are considered. For
simplicity only statistical systems without conserved charges are discussed.
Thus, the number of particles is not restricted and chemical potential equals
to zero. Furthermore, the Boltzmann statistics is used and the degeneracy
factor is assumed to be one.
### III.1 Grand Canonical and Grand Micro-Canonical Ensembles
The $(V,T)$ GCE444The chemical potential connected to the number of particles
equals to zero. partition function of massless non-interacting neutral
Boltzmann particles reads:
$\displaystyle
Z_{gce}(V,T)=\sum_{N=0}^{\infty}\frac{1}{N!}\left(\frac{V}{2\pi^{2}}\right)^{N}\int_{0}^{\infty}\prod_{i=1}^{N}p^{2}_{i}dp_{i}\exp\left(-\frac{p_{i}}{T}\right)=\sum_{N=0}^{\infty}\frac{1}{N!}~{}\left(\frac{VT^{3}}{\pi^{2}}\right)^{N}=\exp(\overline{N})~{},$
(50)
where $\overline{N}\equiv\langle N\rangle_{gce}=VT^{3}/\pi^{2}$ is the GCE
average number of particles. The GCE system pressure and average energy are:
$\displaystyle p~{}$ $\displaystyle=~{}\frac{T}{V}~{}\ln
Z_{gce}~{}=~{}\frac{T^{4}}{\pi^{2}}~{}=~{}T~{}n(T)~{},$ (51)
$\displaystyle\langle E\rangle_{gce}~{}$
$\displaystyle\equiv~{}\overline{E}~{}=~{}T^{2}~{}\frac{\partial\ln
Z_{gce}}{\partial
T}~{}=~{}\frac{3}{\pi^{2}}~{}VT^{4}~{}=~{}\varepsilon(T)~{}V~{},$ (52)
where $n(T)=\overline{N}/V=T^{3}/\pi^{2}$ and $\varepsilon(T)=3T^{4}/\pi^{2}$
are the particle number density and energy density, respectively. The GCE
multiplicity distribution has the Poisson form,
$\displaystyle
P_{gce}(N;V,T)~{}=~{}\frac{\overline{N}^{N}}{N!}~{}\exp\left(-~{}\overline{N}\right)~{},$
(53)
and the scaled variance of particle number distribution (53) equals to:
$\displaystyle\omega_{gce}~{}\equiv~{}\frac{\langle
N^{2}\rangle_{gce}~{}-~{}\langle N\rangle^{2}_{gce}}{\langle
N\rangle_{gce}}~{}=~{}1~{}.$ (54)
The $(V,E)$ Grand Micro-Canonical ensemble555The energy is fixed, but the
number of particles is not. The chemical potential connected to the number of
particles equals to zero. Thus, this ensemble is named the Grand Micro-
Canonical Ensemble mce2 . (GMC) partition function is mce1 :
$\displaystyle Z_{gmc}(V,E)\;$
$\displaystyle=~{}\sum_{N=1}^{\infty}\frac{1}{N!}\,\left(\frac{V}{2\pi^{2}}\right)^{N}~{}\int_{0}^{\infty}\prod_{i=1}^{N}p_{i}^{2}dp_{i}~{}\delta\left(E~{}-~{}\sum_{j=1}^{N}p_{j}\right)\;$
$\displaystyle\equiv\;\sum_{N=1}^{\infty}W_{N}(V,E)\;=\;\frac{1}{E}\sum_{N=1}^{\infty}\frac{A^{N}}{N!~{}(3N-1)!}~{}=~{}\frac{A}{2E}\;\;_{0}F_{3}\left(;\,\frac{4}{3},\frac{5}{3},2;\,\frac{A}{27}\right)~{},$
(55)
where ${}_{0}F_{3}$ is the generalized hyper-geometric function I , and
$\displaystyle A~{}\equiv~{}\frac{VE^{3}}{\pi^{2}}~{}.$ (56)
The GMC particle number distribution function equals to:
$\displaystyle
P_{gmc}(N;V,E)~{}\equiv~{}\frac{W_{N}(V,E)}{Z_{gmc}(V,E)}\;=\;\frac{1}{Z_{gmc}(V,E)}~{}\frac{A^{N}}{E~{}N!~{}(3N-1)!}\;.$
(57)
It is defined for $N\geq 1$. The average number of particles in the GMC equals
to mce1 :
$\displaystyle\langle
N\rangle_{gmc}~{}\cong~{}\left(\frac{A}{27}\right)^{1/4}\;.$ (58)
In the large volume limit the mean multiplicities in the GME and the GCE are
equal, $\langle N\rangle_{gmc}=\overline{N}$ providing $E=\overline{E}$ and
the GMC and the GCE volumes are equal. The temperature and the pressure in the
GME are equal to:
$\displaystyle
T~{}=~{}\frac{E}{3\overline{N}}~{}=~{}\left(\frac{\pi^{2}E}{3V}\right)^{1/4},~{}~{}~{}~{}~{}~{}p~{}=~{}\frac{\overline{N}T}{V}~{}=~{}\frac{T^{4}}{\pi^{2}}~{},$
(59)
and they coincide with the corresponding quantities (51-52) in the GCE.
For $\overline{N}\gg 1$ the particle number distribution in the GME (57) can
be approximated by the Gaussian one:
$\displaystyle
P_{gmc}(N;V,E)\;\cong~{}\left(2\pi~{}\omega_{gce}~{}\overline{N}\right)^{-1/2}~{}\exp\left[~{}-~{}\frac{\left(N-\overline{N}\right)^{2}}{2~{}\omega_{gmc}~{}\overline{N}}\right]~{},$
(60)
with the GME scaled variance
$\displaystyle\omega_{gmc}~{}=~{}\frac{\langle
N^{2}\rangle_{gmc}~{}-~{}\langle N\rangle_{gmc}^{2}}{\langle
N\rangle_{gmc}}~{}=~{}\frac{1}{4}~{}.$ (61)
The Poisson distribution $P_{gce}(N)$ (53) for $\overline{N}\gg 1$ can be also
approximated by the Gauss distribution (60), but with $\omega_{gce}=1$. In the
TL the particle number distributions in the GME and GCE have both the Gauss
form clt . The average number of particles is the same, $\langle
N\rangle_{gmc}\cong\overline{N}$, but the scaled variance is different,
$\omega_{gce}=1$ and $\omega_{gmc}=1/4$.
### III.2 Pressure Grand Canonical Ensemble
The $(p_{0},T)$ Pressure Grand Canonical ensemble (PGC) is defined as the
ensemble with the fixed external pressure $p_{0}$, temperature $T$, and the
chemical potential connected to the number of particles equals to zero. The
PGC partition function equal to pressure :
$\displaystyle
Z_{pgc}(p_{0},T)~{}\equiv~{}\int_{0}^{\infty}dV~{}\exp\left(-~{}\frac{p_{0}V}{T}\right)~{}Z_{gce}(V,T)~{}=~{}\frac{T}{p_{0}~{}-~{}p(T)}~{},$
(62)
where the relation (51) between $p(T)$ and $Z_{gce}$ has been used. The value
of $p_{0}$ has a physical meaning of the external pressure. A convergence of
the integral over the volume in Eq. (62) requires the inequality,
$\displaystyle p(T)~{}=~{}\frac{1}{\pi^{2}}~{}T^{4}~{}<~{}p_{0}~{}.$ (63)
Thus, at each fixed value of $p_{0}$, there is a ‘limiting temperature’
$T^{*}$ in the PGC:
$\displaystyle T~{}<~{}T^{*}~{}=~{}\left(\pi^{2}~{}p_{0}\right)^{1/4}~{}.$
(64)
The probability volume distribution in the PGC is:
$\displaystyle
W_{pgc}(V)~{}=~{}\frac{1}{Z_{pgc}(p_{0},T)}~{}\exp\left[-~{}V\left(\frac{p_{0}~{}-~{}p(T)}{T}\right)\right]~{}.$
(65)
It gives the most probable volume equal to zero, $V_{0}=0$, and the average
value,
$\displaystyle\langle
V\rangle_{pgc}~{}\equiv~{}\overline{V}~{}=~{}\int_{0}^{\infty}dV~{}V~{}W_{pgc}(V)~{}=~{}-~{}T~{}\frac{\partial\ln
Z_{pgc}(p_{0},T)}{\partial p_{0}}~{}=~{}\frac{T}{p_{0}~{}-~{}p(T)}~{},$ (66)
which is not equal to zero and may even go to infinity at $T\rightarrow
T^{*}$. Using Eq. (66), the volume distribution (65) can be written as,
$\displaystyle
W_{pgc}(V)~{}=~{}\overline{V}^{~{}-1}~{}\exp\left(-~{}V/\overline{V}\right)~{}.$
(67)
For the mean values of energy and particle number in the PGC one finds:
$\displaystyle\langle E\rangle_{pgc}~{}$
$\displaystyle=~{}\int_{0}^{\infty}dV~{}W_{pgc}(V)~{}\langle
E\rangle_{gce}~{}=~{}\frac{3}{\pi^{2}}~{}T^{4}~{}\overline{V}~{}=~{}\varepsilon(T)~{}\overline{V},$
(68) $\displaystyle\langle N\rangle_{pgc}~{}$
$\displaystyle=~{}\int_{0}^{\infty}dV~{}W_{pgc}(V)~{}\langle
N\rangle_{gce}~{}=~{}\frac{1}{\pi^{2}}~{}T^{3}~{}\overline{V}~{}=~{}n(T)~{}\overline{V}~{},$
(69)
where $n(T)$ and $\varepsilon(T)$ are the particle number density and energy
density of the GCE given by Eq. (51) and Eq. (52), respectively.
The independent variables are $(V,T)$ in the GCE and $(p_{0},T)$ in the PGC.
If the GCE volume is chosen to be equal to the average volume of the PGC, then
Eqs. (68) and (69) give: $\langle E\rangle_{pgc}=\overline{E}$ and $\langle
N\rangle_{pgc}=\overline{N}$, i.e. the GCE and PGC are thermodynamically
equivalent. The fluctuations of $E$ and $N$ are however different in these two
ensembles. Calculating,
$\displaystyle\langle
V^{2}\rangle_{pgc}~{}=~{}\int_{0}^{\infty}dV~{}V^{2}~{}W_{pgc}(V)~{}=~{}\frac{T^{2}}{Z_{pgc}}~{}\frac{\partial^{2}Z_{pgc}}{\partial
p_{0}^{2}}~{}=~{}2~{}\left[\frac{T}{p_{0}~{}-~{}p(T)}\right]^{2}~{}=~{}2\overline{V}^{2}~{},$
(70)
one finds,
$\displaystyle\omega_{V}~{}=~{}\frac{\langle V^{2}\rangle_{pgc}~{}-~{}\langle
V\rangle_{pgc}^{2}}{\langle V\rangle_{pgc}}~{}=~{}\overline{V}~{}.$ (71)
Thus, in the TL limit $\overline{V}\rightarrow\infty$ the volume fluctuations
become anomalously large. This, in turn, leads to anomalous energy and
particle number fluctuations. The particle number distribution in the PGC has
the form of the geometrical distribution:
$\displaystyle
P_{pgc}(N;p_{0},T)~{}=~{}\int_{0}^{\infty}dV~{}W_{pgc}(V)~{}P_{gce}(N,V,T)~{}=~{}(1-\eta)~{}\eta^{N}~{},$
(72)
where $P_{gce}(N,V,T)$ has been taken from Eq. (53), and
$\eta\equiv(T/T^{*})^{4}<1$ . The most probable number of particles is $N=0$,
whereas the average value (69) is larger than zero. It equals to $\langle
N\rangle_{pgc}=\eta(1-\eta)^{-1}$ and may even go to infinity at
$\eta\rightarrow 1$. This happens if $T\rightarrow T^{*}$. From Eq. (72) it
follows:
$\displaystyle\omega_{pgc}~{}=~{}\frac{\langle
N^{2}\rangle_{pgc}~{}-~{}\langle N\rangle_{pgc}^{2}}{\langle
N\rangle_{pgc}}~{}=~{}1~{}+~{}\frac{\eta}{1-\eta}~{}=~{}1~{}+n(T)~{}\overline{V}~{}.$
(73)
The first term in the r.h.s. of Eq. (73) corresponds to the Poisson
fluctuations (54) of the GCE at fixed volume, whereas the second term comes
from the volume fluctuations at fixed particle number density. The
multiplicity distribution (72) can be rewritten as,
$\displaystyle
P_{pgc}(N;p_{0},T)~{}\equiv~{}P_{pgc}(N;\overline{N})~{}=~{}\frac{1}{\overline{N}+1}~{}\exp\left[-~{}N~{}\ln\left(1+\frac{1}{\overline{N}}\right)\right]~{}.$
(74)
For $\overline{N}\gg 1$ the distribution $P_{pgc}$ approches:
$\displaystyle
P_{pgc}(N;\overline{N})~{}\cong~{}\frac{1}{\overline{N}}~{}\exp\left(-~{}\frac{N}{\overline{N}}\right)~{}.$
(75)
The particle number distribution $P_{pgc}(N)$ (75) satisfies the so called
KNO-scaling KNO .
### III.3 Pressure Grand Micro-Canonical Ensemble
The $(s_{0},E)$ Pressure Grand Micro-canonical ensemble (PGM) is defined as
the statistical ensemble with fixed energy $E$ and fixed external parameter
$s_{0}=p_{0}/T_{0}$. As before, the chemical potential connected to the number
of particles equals to zero. The PGM partition function is equal to:
$\displaystyle
Z_{pgm}(s_{0},E)~{}\equiv~{}\int_{0}^{\infty}dV~{}\exp(-~{}s_{0}V)~{}Z_{gmc}(V,E)~{}$
(76)
$\displaystyle=~{}\frac{1}{E}~{}\sum_{N=1}^{\infty}\left(\frac{E^{3}}{\pi^{2}}\right)^{N}~{}\frac{1}{N!~{}(3N-1)!}\int_{0}^{\infty}dV~{}\exp(-~{}s_{0}V)~{}V^{N}~{}=~{}\frac{1}{s_{0}~{}E}~{}\sum_{N=1}^{\infty}\frac{B^{N}}{(3N-1)!}~{},$
where
$\displaystyle B~{}\equiv~{}\frac{E^{3}}{s_{0}~{}\pi^{2}}~{}.$ (77)
Using Eq. Prud ,
$\displaystyle\sum_{k=0}^{\infty}~{}\frac{x^{3k}}{(3k)!}~{}=~{}\frac{1}{3}\left[\exp(x)~{}+~{}2\exp(-x/2)\cos\left(\frac{\sqrt{3}}{2}~{}x\right)\right]~{},$
(78)
one finds at $B\gg 1$,
$\displaystyle\ln[Z_{pgm}(s_{0},E)]~{}\cong~{}B^{1/3}~{}.$ (79)
In the TL, the volume distribution in the PGM can be found from (76) using the
asymptotic behavior of ${}_{0}F_{3}$ function I ,
$\displaystyle\ln[Z_{gmc}(V,E)]~{}=~{}\ln\left[\frac{A}{2E}\;\;_{0}F_{3}\left(;\,\frac{4}{3},\frac{5}{3},2;\,\frac{A}{27}\right)\right]~{}\cong~{}4~{}\left(\frac{VE^{3}}{27\pi^{2}}\right)^{1/4}~{}.$
(80)
The volume distribution in the PGM is then proportional to,
$\displaystyle W_{pgm}(V)~{}$
$\displaystyle\propto~{}\exp\left[4\left(\frac{VE^{3}}{27\pi^{2}}\right)^{1/4}~{}-~{}s_{0}V\right]~{}\equiv~{}\exp[\phi(V)]~{}$
$\displaystyle\cong~{}\exp\left[\phi(V_{0})~{}+~{}\frac{1}{2}\left(\frac{\partial^{2}\phi}{\partial
V^{2}}\right)_{V=V_{0}}~{}\left(V~{}-~{}V_{0}\right)^{2}\right]~{}.$ (81)
The most probable volume $V_{0}$ in the PGM is defined by the condition,
$\displaystyle\left(\frac{\partial\phi}{\partial
V}\right)_{V=V_{0}}~{}=~{}\left(\frac{E^{3}}{27\pi^{2}V_{0}^{3}}\right)^{1/4}~{}-~{}s_{0}~{}=~{}0~{}.$
(82)
This gives,
$\displaystyle V_{0}~{}=~{}\frac{E}{3\pi^{2/3}s_{0}^{4/3}}~{}.$ (83)
One also finds,
$\displaystyle\left(\frac{\partial^{2}\phi}{\partial
V^{2}}\right)_{V=V_{0}}~{}=~{}-~{}\frac{3s_{0}}{4V_{0}}~{}.$ (84)
Thus, the volume distribution in the PGM can be approximated as:
$\displaystyle
W_{pgm}(V)~{}\cong~{}\left(2\pi~{}\omega_{V}~{}V_{0}\right)^{-1/2}~{}\exp\left[-~{}\frac{\left(V~{}-~{}V_{0}\right)^{2}}{2~{}\omega_{V}~{}V_{0}}\right]~{},$
(85)
where $\omega_{V}=4/(3s_{0})$ is the scaled variance of the volume
fluctuations in the PGM. The average volume and its fluctuations in the PGM
can be also calculated in terms of the partition function $Z_{pgm}(s_{0},E)$
using Eqs. (76-79),
$\displaystyle\langle V\rangle_{pgm}~{}$
$\displaystyle=~{}\int_{0}^{\infty}dVW_{pgm}(V)~{}V~{}=~{}-~{}\frac{\partial\ln~{}Z_{pgm}}{\partial
s_{0}}~{}\cong~{}\frac{E}{3\pi^{2/3}s_{0}^{4/3}}~{},$ (86)
$\displaystyle\langle V^{2}\rangle_{pgm}~{}$
$\displaystyle=~{}\int_{0}^{\infty}dVW_{pgm}(V)~{}V^{2}~{}=~{}\frac{1}{Z_{pgm}}~{}\frac{\partial^{2}Z_{pgm}}{\partial
s_{0}^{2}}~{}\cong~{}\frac{E^{2}}{9\pi^{4/3}s_{0}^{8/3}}~{},$ (87)
$\displaystyle\omega_{V}~{}$ $\displaystyle=~{}\frac{\langle
V^{2}\rangle_{pgm}~{}-~{}\langle V\rangle^{2}_{pgm}}{\langle
V\rangle_{pgm}}~{}\cong~{}\frac{4}{3s_{0}}~{}.$ (88)
The condition (83) can be written as,
$\displaystyle s_{0}~{}=~{}\frac{1}{\pi^{2}}~{}T^{3}~{}=~{}\frac{p}{T}~{},$
(89)
and it means that the internal pressure $p$ equals to $s_{0}T$. The most
probable volume $V_{0}$ (83) and average volume $\langle V\rangle_{pgm}$ (86)
are then equal to each other, and both are equal to $E\pi^{2}/(3T^{4})$. This
corresponds to the fixed volume in the MCE with fixed energy $E$ and the MCE
temperature $T$. The volume distribution in the PGM is therefore different
from that in the PGC (67). In contrast to the PGC, there is an extensive
variable, the energy $E$, in the PMG. This leads to the system average volume
proportional to the energy and given by Eq. (83). The internal pressure equals
to $p=s_{0}T$ (89). As a result, the volume fluctuations in the PGM are
Gaussian (85) with the finite scaled variance (88). Using Eq. (89) the scaled
variance (88) of the volume fluctuations can be expressed in terms of the MCE
temperature, $\omega_{V}=4\pi^{2}/(3T^{3})$. One also finds:
$\displaystyle\langle N\rangle_{pgm}~{}$
$\displaystyle=~{}\frac{1}{Z_{pgm}}~{}B~{}\frac{\partial Z_{pgm}}{\partial
B}~{}\cong~{}\frac{1}{3}~{}B^{1/3}~{},$ (90) $\displaystyle\langle
N^{2}\rangle_{pgm}~{}$
$\displaystyle=~{}\frac{1}{Z_{pgm}}~{}B\frac{\partial}{\partial
B}B~{}\frac{\partial Z_{pgm}}{\partial
B}~{}\cong~{}\frac{1}{9}~{}B^{1/3}~{}+~{}\frac{1}{9}~{}B^{2/3},$ (91)
$\displaystyle\omega_{pgm}~{}$ $\displaystyle=~{}\frac{\langle
N^{2}\rangle_{pgm}~{}-~{}\langle N\rangle^{2}_{pgm}}{\langle
N\rangle_{pgm}}~{}\cong~{}\frac{1}{3}~{}.$ (92)
In the TL, the multiplicity distribution in the PGM can be approximated as:
$\displaystyle
P_{pgm}(N;s_{0},E)\;\cong~{}\left(2\pi~{}\omega_{pgm}~{}\overline{N}\right)^{-1/2}~{}\exp\left[~{}-~{}\frac{\left(N-\overline{N}\right)^{2}}{2~{}\omega_{pgm}~{}\overline{N}}\right]~{},$
(93)
where $\overline{N}=B^{1/3}/3\cong\langle N\rangle_{pgm}$ and
$\omega_{pgm}=1/3$ . The scaled variance in the PGM can be presented in the TL
as the following,
$\displaystyle\omega_{pgm}~{}=~{}\omega_{mce}~{}+~{}\frac{1}{16}~{}\omega_{V}~{}n~{}=~{}\frac{1}{4}~{}+~{}\frac{1}{12}~{}=~{}\frac{1}{3},$
(94)
where $n=T^{3}/\pi^{2}$ is the MCE particle number density. The first term in
the r.h.s. of Eq. (94), $\omega_{mce}=1/4$ , is due to the particle number
fluctuations in the MCE with fixed volume, and the second term is the
contribution due to the volume fluctuations.
## IV Ensembles with External Volume Fluctuations
The multiplicity distributions $P(N)$ in relativistic gases fluc ; mce1 ; mce2
; res ; exp ; clt ; acc are sensitive to conservation laws obeyed by the
system, and therefore to fluctuations of extensive quantities $E$ and $Q$. The
examples considered in the previous section demonstrate that the volume
fluctuations also influence the particle number fluctuations. Thus, for the
calculation of multiplicity distributions, the choice of the statistical
ensemble is then not a matter of convenience, but a physical question. On the
other hand, the fluctuations of extensive quantities $\vec{A}\equiv(V,E,Q)$
depend not on the system’s physical properties, but rather on external
conditions. One can imagine a huge variety of these conditions, thus, 8
statistical ensembles discussed in the previous sections are only some special
examples.
A more general concept of the statistical ensembles based on Eq. (1) was
recently suggested in Ref. alpha . The system volume may exhibit fluctuations
described by the externally given distribution. When $V$ is the only
fluctuating variable, Eq. (1) is reduced to
$\displaystyle P_{\alpha}(O)=\int
dV~{}P_{\alpha}(E,V,Q)~{}P_{mce}(O;E,V,Q)~{},$ (95)
where $P_{\alpha}(V)$ is externally given volume distribution.
The effect of volume fluctuations on the particle number fluctuations is
calculated for the system of non-interacting massless Boltzmann particles with
zero chemical potential. At fixed volume the system is treated within the GMC,
and the particle number distribution is $P_{gmc}(N;V,E)$ (57).
In the first example, the volume distribution is assumed to be:
$\displaystyle
P_{\alpha}(V)~{}=~{}\left(2\pi\omega_{V}~{}a_{V}^{2}\overline{V}\right)^{-1/2}~{}\exp\left[-~{}\frac{\left(V~{}-~{}\overline{V}\right)^{2}}{2\omega_{V}~{}a_{V}^{2}\overline{V}}\right]~{},$
(96)
where $\omega_{V}=4\pi^{2}/(3T^{3})$ coincides with (88) in the PGM. The
choice of $P_{\alpha}(V)$ results in a simple correspondence to the GMC and
PGM in the TL. In Eq. (96), $a_{V}$ is a dimensionless tuneable parameter
which determines the width of the distribution. In the limit $a_{V}\rightarrow
0$, Eq. (96) becomes a Dirac $\delta$-function, $\delta(V-\overline{V})$. This
corresponds to the GMC. For $a_{V}=1$, Eq. (96) results in the PGM volume
fluctuations (85) in the TL. The particle number distribution reads,
$\displaystyle
P_{\alpha}(N)~{}=~{}\int_{0}^{\infty}dV~{}P_{\alpha}(V)~{}P_{gmc}(N;V,E)~{}.$
(97)
A substitution of $P_{\alpha}(V)$ in Eq. (97) by the distribution (96) results
in $P_{\alpha}(N)$ in the TL given by the Gaussian,
$\displaystyle
P_{\alpha}(N)~{}\cong~{}\left(2\pi\omega_{\alpha}\overline{N}\right)^{-1/2}~{}\exp\left[-~{}\frac{\left(N~{}-~{}\overline{N}\right)^{2}}{2\omega_{\alpha}~{}\overline{N}}\right]~{},$
(98)
where the average number of particles $\overline{N}$ is defined by the energy
and average volume,
$\overline{N}=~{}[\overline{V}~{}E^{3}/(27\pi^{2})]^{1/4}$, and the scaled
variance of the particle number distribution (98) equals,
$\displaystyle\omega_{\alpha}~{}=~{}\frac{1}{4}~{}+~{}\frac{1}{12}~{}a_{V}^{2}~{}.$
(99)
The first term in the r.h.s. of Eq. (99) equals to the GMC scaled variance at
fixed $E$ and $V$, the second term is due to the volume fluctuations. As it
can be expected, $\omega_{\alpha}=\omega_{gmc}$ for $a_{V}=0$, and
$\omega_{\alpha}=\omega_{pgm}$ for $a_{V}=1$. It also follows,
$\omega_{gmc}<\omega_{\alpha}<\omega_{pgm}$ for $0<a_{V}<1$, and
$\omega_{\alpha}>\omega_{pgm}$ for $a_{V}>1$. The $\alpha$-ensemble defined by
Eqs. (97, 96) presents an extension of the GMC ($a_{V}=0$) and PGM ($a_{V}=1$)
to a more general volume distribution.
In the second example both $E$ and $V$ are assumed to fluctuate. Equation (97)
should be then extended as:
$\displaystyle P_{\alpha}(N)~{}=~{}\int
dEdV~{}P_{\alpha}(E,V)~{}P_{gmc}(N;E,V)~{}.$ (100)
First, uncorrelated volume and energy distributions are considered:
$\displaystyle P_{\alpha}(E,V)~{}=~{}P_{1}(V)\times P_{2}(E)~{},$ (101)
where $P_{1}(V)$ is given by Eq. (96) and $P_{2}(E)$ is taken in the following
form,
$\displaystyle
P_{2}(E)~{}=~{}\left(2\pi\omega_{E}~{}a_{E}^{2}\overline{E}\right)^{-1/2}~{}\exp\left[-~{}\frac{\left(E~{}-~{}\overline{E}\right)^{2}}{2\omega_{E}~{}a_{E}^{2}~{}\overline{E}}\right]~{},$
(102)
where $\overline{E}=3T^{4}\overline{V}/\pi^{2}$ and $\omega_{E}=4T$. A
substitution of $P_{\alpha}(E,V)$ in Eq. (100) by the above distributions
results in $P_{\alpha}(N)$ in the TL given again by the Gaussian (98) with the
average number of particles $\overline{N}$ defined by the average energy and
average volume,
$\overline{N}=~{}[\overline{V}~{}\overline{E}^{3}/(27\pi^{2})]^{1/4}$, and the
scaled variance equal to:
$\displaystyle\omega_{\alpha}~{}=~{}\frac{1}{4}~{}+~{}\frac{3}{4}~{}a_{E}^{2}~{}+~{}\frac{1}{12}a_{V}^{2}~{}.$
(103)
The first term in the r.h.s. of Eq. (103) is the MCE scaled variance at fixed
$E$ and $V$, the second term is due to energy fluctuations, and the third one
comes from volume fluctuations. The results of the GMC, PGM, and GCE for the
scaled variance of the particle number distribution can be obtained from Eq.
(103) at $a_{V}=a_{E}=0$; $a_{V}=1,~{}a_{E}=0$; and $a_{V}=0,~{}a_{E}=1$,
respectively
Examples of correlated $V$ and $E$ distributions can be constructed as
follows. One assumes the factorized form (101) with Eq. (102) for the energy
distribution, but with $\overline{E}=3T^{4}V/\pi^{2}$ and $\omega_{E}=4T$,
i.e. the average energy $\overline{E}$ depends on the volume $V$. Assuming,
$\displaystyle
P_{1}(V)~{}=~{}\frac{1}{\overline{V}}~{}\exp(-~{}V/\overline{V})~{},$ (104)
with $\overline{V}$ given by Eq. (66), from Eqs. (101) and (102) with
$a_{E}=1$, the results for the PGC are obtained.
The relation, $\overline{E}=3T^{4}V/\pi^{2}$, can be used together with any
form of the volume distribution $P_{1}(V)$. This yields a generalization of
the GCE to the systems with externally given volume fluctuations. The energy
fluctuations at fixed $V$ can be also selected in accordance with the physics
requirements.
As the third example of the statistical ensembles with volume fluctuations one
refers to the recent paper MCEsVF where the micro-canonical ensemble with the
scaling volume fluctuations for the ultra-relativistic ideal gas has been
considered. The volume fluctuations were assumed to have the specific scaling
properties. They were chosen to describe the KNO scaling KNO of the particle
multiplicity distributions measured in proton-proton collisions at high
energies. A striking feature of the model is power law form of the single
momentum spectrum at high momenta, instead of the exponential Boltzmann
distribution in the systems with fixed volume.
## V Summary
In this paper the volume fluctuations in the statistical mechanics have been
studied. The statistical ensembles with fixed external pressure were
considered. Statistical systems of classical non-relativistic particles and
non-interacting massless particles were discussed. The volume fluctuations in
the pressure canonical ensemble become anomalously large in the thermodynamic
limit for the first order phase transition, $\omega_{V}\propto N$, or at the
critical point, $\omega\propto N^{1/2}$. Another type of anomalous volume
fluctuations takes place for the pressure grand canonical ensemble. In this
special ensemble, all thermodynamical variables are the intensive quantities,
$(p_{0},T,\mu)$. In the thermodynamical limit the mean particle multiplicity
obtained within the considered ensembles with volume fluctuations is equal to
the mean calculated within the micro-canonical, canonical, and grand canonical
ensembles. This is not valid, however, for the scaled variance. The influence
of the volume fluctuations in the pressure ensembles on particle number
fluctuations have been discussed for ultra-relativistic ideal gas.
Following Ref. alpha the ensembles with volume fluctuations determined by the
externally given distribution function were introduced. Multiplicity
fluctuations are sensitive to the volume fluctuations. The volume fluctuations
may also influence the behavior of the particle momentum spectra MCEsVF .
Thus, we believe that the concept of statistical ensembles with fluctuating
extensive quantities, in particular, with fluctuating volume of the
statistical system, may be appropriate for the statistical description of
hadron production in relativistic collisions. It may be also useful for other
physical systems. In fact, in all cases when the equilibrium statistical
mechanics is used to calculate the fluctuations of the system properties.
###### Acknowledgements.
We would like to thank D.V. Anchishkin, V.V. Begun, M. Gaździcki, W. Greiner,
M. Hauer, B.I. Lev, I.N. Mishustin, O.N. Moroz, and Yu.M. Sinyukov for
numerous discussions. This work was in part supported by the Program of
Fundamental Researches of the Department of Physics and Astronomy of NAS,
Ukraine.
## References
* (1) J. Cleymans, H. Oeschler, K. Redlich, and S. Wheaton, Phys. Rev. C 73, 034905 (2006); F. Becattini, J. Manninen, and M. Gaździcki, ibid. 73, 044905 (2006); A. Andronic, P. Braun-Munzinger, and J. Stachel, Nucl. Phys. A 772, 167 (2006).
* (2) F. Becattini, Z. Phys. C 69, 485 (1996); F. Becattini and U. Heinz, Z. Phys. C 76, 269 (1997).
* (3) J. Cleymans, K. Redlich, and E. Suhonen, Z. Phys. C 51, 137 (1991).
* (4) M.I. Gorenstein, M. Gaździcki, and W. Greiner, Phys. Lett. B 483, 60 (2000).
* (5) M.I. Gorenstein, A.P. Kostyuk, H. Stöcker, and W. Greiner, Phys. Lett. B 509, 277 (2001).
* (6) F. Becattini and L. Ferroni, Eur. Phys. J. C 35, 243 (2004); 38, 225 (2004); V.V. Begun, L. Ferroni, M.I. Gorenstein, M. Gaździcki, F. Becattini, J. Phys. G 32, 1003 (2006); F. Becattini and L. Ferroni, Eur. Phys. J. C 51, 899 (2007); 52, 597 (2007).
* (7) V.V. Begun, M. Gaździcki, M.I. Gorenstein, and O.S. Zozulya, Phys. Rev. C 70, 034901 (2004); V.V. Begun, M.I. Gorenstein, and O.S. Zozulya, Phys. Rev. C 72, 014902 (2005); A. Keränen, F. Becattini, V.V. Begun, M.I. Gorenstein, and O.S. Zozulya, J. Phys. G 31, S1095 (2005); F. Becattini, A. Keränen, L. Ferroni, and T. Gabbriellini, Phys. Rev. C 72, 064904 (2005); J. Cleymans, K. Redlich, and L. Turko, Phys. Rev. C 71, 047902 (2005); J. Phys. G 31, 1421 (2005); V.V. Begun and M.I. Gorenstein, Phys. Rev. C 73, 054904 (2006).
* (8) V.V. Begun, M.I. Gorenstein, A.P. Kostyuk, and O.S. Zozulya, Phys. Rev. C 71, 054904 (2005).
* (9) V.V. Begun, M.I. Gorenstein, A.P. Kostyuk, and O.S. Zozulya, J. Phys. G 32, 935 (2006).
* (10) V.V. Begun, M.I. Gorenstein, M. Hauer, V.P. Konchakovski, and O.S. Zozulya, Phys. Rev. C 74, 044903 (2006);
* (11) V.V. Begun, M. Gaździcki, M.I. Gorenstein, M. Hauer, B. Lungwitz, and V.P. Konchakovski, Phys. Rev. C 76, 024902 (2007).
* (12) M. Hauer, V.V. Begun, and M.I. Gorenstein, arXiv:0706.3290 [nucl-th].
* (13) M. Hauer, arXiv:0710.3938 [nucl-th].
* (14) Yu.B. Rumer and M. Sh. Rivkin, Thermodynamics, Statistical Physics, and Kinetics (Nauka, Moscow, 1972) [in Russian].
* (15) K.B. Tolpygo, Thermodynamics and Statistical Physics (Kiev University, 1966) [in Russian].
* (16) M.I Gorenstein and M. Hauer, arXiv:08014219 [nucl-th].
* (17) V.V. Begun, M. Gaździcki, and M.I. Gorenstein, arXiv:0804.0075 [hep-ph].
* (18) Z. Koba, H. B. Nielsen, P. Olesen, Nucl. Phys. B 40, 317 (1972).
* (19) L.D. Landau and E.M. Lifshitz. Statistical Physics (Course of Theoretical Physics, Volume 5), Pergamon Press, Oxford, 1980.
* (20) M.I. Gorenstein, Yad. Fiz. 31, 1630 (1980) (Russ.); Sov. J. Nucl. Phys. 31, 845 (1980).
* (21) Weisstein, Eric W. ”Generalized Hypergeometric Function.” From MathWorld – A Wolfram Web Resource. http://mathworld.wolfram.com/GeneralizedHypergeometricFunction.html
* (22) A.P. Prudnikov, Yu.A. Brychkov, and O.I. Marichev, Integrals and Series, (Moscow, Nauka, 1986).
| arxiv-papers | 2008-06-17T14:43:05 | 2024-09-04T02:48:56.309038 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Mark I. Gorenstein",
"submitter": "Mark Gorenstein I.",
"url": "https://arxiv.org/abs/0806.2804"
} |
0806.2808 | # A tight closure approach
to a result of G. Faltings
Tirdad Sharif Tirdad Sharif, School of Mathematics, Institute for Studies in
Theoretical Physics and Mathematics, P. O. Box 19395-5746, Tehran, Iran
sharif@ipm.ir
http://www.ipm.ac.ir/IPM/people/personalinfo.jsp?PeopleCode=IP0400060
###### Abstract.
Using a result of M. Hochster and C. Huneke on $F$-rational rings a criterion
for complete intersection rings of characteristic $p>0$ is presented. As an
application, we give a completely different proof for an algebraic result of
G. Faltings that was used by Taylor and Wiles in [9] for a simplification of
the proof of the minimal deformation problem.
###### Key words and phrases:
Complete Intersection Algebras, Deformation Algebras, Hecke Algebras, Tight
Closure
###### 2000 Mathematics Subject Classification:
13A35, 13H10
The author was supported by a grant from IPM, (No. 83130311).
## 1\. Introduction
The theory of tight closure was created by Melvin Hochster and Craig Huneke.
An important notion in the theory of tight closure is the notion of _$F$
-rationality_. There is a close connection between this notion and _rational
singularity_ , see [8].
The main purpose of this note is to use a result of M. Hochster and C. Huneke
in [3] on $F$-rational rings, to give a criterion for complete intersection
rings of characteristic $p>0$, see Theorem 2.5 for the precise statement.
In [9] G. Faltings proved an algebraic result that was used by Taylor and
Wiles to reprove the _minimal deformation problem_ by a simpler method. As an
application of Theorem 2.5 we prove Proposition 2.8, which is an extension of
Faltings result. Our approach to prove this proposition is fundamentally
different from Faltings method where he simplifies Taylor and Wiles argument,
see Remark 2.6.
In order to give a view for the reader we recall the minimal deformation
problem very briefly.
The modularity conjecture for semistable elliptic curves depends on a critical
conjecture of Wiles that was proved in [9, 10], see Conjecture (2.16) of [10].
Let $\mathcal{D}$ be a deformation theory and let $R_{\mathcal{D}}$ and
$T_{\mathcal{D}}$ be the universal deformation and Hecke algebras associated
to $\mathcal{D}$, respectively. The universal property of $R_{\mathcal{D}}$
implies that there is a homomorphism
$\varphi_{\mathcal{D}}:R_{\mathcal{D}}\rightarrow T_{\mathcal{D}}$. Wiles
conjecture asserts that $\varphi_{\mathcal{D}}$ is an isomorphism. In the
special case, if $\mathcal{D}$ is a minimal deformation theory, then the above
conjecture is called the minimal deformation problem. For different types of
deformation theories and their relations see Chapter 1 and 2 of [10].
Wiles method in [10] to prove the above conjecture, was closely related to
show that certain Hecke algebras are complete intersection that was shown in
[9].
There is no need for the reader to know the number theoretic materials of
Wiles proof in detail. We refer the interested readers to [6] for an elegant
exposition of the above matters.
## 2\. Definitions, Notations and The Main Theorem
Throughout this note all rings are commutative and Noetherian of
characteristic $p>0$ and all of modules are finite (that is, finitely
generated).
In this note we use the notations $\nu_{R}(M)$ and $\ell_{R}(M)$ respectively
for the minimal number of generators of $M$ and the length of $M$ over $R$.
###### Definition 2.1.
Let $(R,\mathfrak{m},k)$ be a local ring and let $R[[X]]=R[[X_{1},...,X_{n}]]$
be the formal power series of $n$ variables over $R$. Write $A=R[[X]]/J$ for
an ideal $J$ of $R[[X]]$. If the local ring $A\otimes_{R}k$ is complete
intersection, then $A$ is called a complete intersection $R$-algebra.
###### Definition 2.2.
Let $R$ be a ring and let $J=(a_{1},...,a_{t})$ be an ideal of $R$. An element
$x\in R$ is said to be in the tight closure of $J$ and write $x\in J^{*}$ if
there is an element $c\in R^{0}$ such that for all large $q=p^{n}$ we have
$cx^{q}\in J^{[q]}$, wherein $R^{0}$ is the complement of the union of all
minimal primes of $R$ and $J^{[q]}=(a_{1}^{q},...,a_{t}^{q})$.
An ideal $J$ is called tightly closed if $J=J^{*}$. The ring $R$ is called
$F$-rational if every parameter ideal of $R$ is tightly closed and in
particular, if every ideal of $R$ is tightly closed, then it is called
$F$-regular.
###### Fact 2.3.
Let $Q$ be a regular local ring, then it is $F$-regular, see [2, (4.4)].
To state the main theorem of this note we will make use of the following
simple lemma the proof of which is omitted.
###### Lemma 2.4.
Let I be an ideal of R, then $(I^{*})^{n}\subseteq(I^{n})^{*}$ for $n\geqslant
1$.
###### Theorem 2.5.
Let $(A,\mathfrak{n},k)$ be a local ring of dimension $t$ and let
$\mathfrak{q}$ be an ideal of $A$, generated by a system of parameters. Let
$\eta\colon A\to Q$ and $\theta\colon Q\to B$ be local ring homomorphisms such
that $Q$ is regular of dimension $t$, $\theta$ is surjective, and $\eta$
induces an isomorphism between the residue fields of $A$ and $Q$. Set
$\delta=\theta\eta\colon A\to B$ and $\ell(B/\mathfrak{q}^{*}B)=d$. Assume for
an integer $q=p^{n}>d^{t}t^{t-1}$ we have
$\ell(B/{\mathfrak{q}}^{[q]}B)\geqslant\ell(A/{\mathfrak{q}}^{[q]})$. If the
ideal ${\mathfrak{q}}^{[q]}$ is tightly closed, then the canonical
homomorphism $\pi:Q/q^{*}Q\rightarrow B/\mathfrak{q}^{*}B$ is an isomorphism.
In particular, if $A$ is equidimensional homomorphic image of a Cohen-Macaulay
ring, then $\pi$ is an isomorphism between complete intersection rings.
###### Proof.
Write $\mathfrak{m}$ for the maximal ideal of $Q$ and
$J=\operatorname{Ker}\theta.$ We may assume that $B=Q/J$. From the assumptions
it follows that $\mathfrak{m}^{d}\subseteq J+{\mathfrak{q}}^{*}Q.$ Therefore
we have the next inclusions
$\mathfrak{m}^{tqd}\subseteq(J+{\mathfrak{q}}^{*}Q)^{tq}\subseteq(J+(\mathfrak{q}^{*})^{tq}Q)$.
Lemma 2.4 implies that $\mathfrak{m}^{tqd}\subseteq
J+(\mathfrak{q}^{tq})^{*}Q$. Since $\mathfrak{q}$ is generated by $t$ elements
it is easy to see that ${\mathfrak{q}}^{tq}\subseteq{\mathfrak{q}}^{[q]}$.
This yields that $\mathfrak{m}^{tqd}\subseteq J+(\mathfrak{q}^{[q]})^{*}Q$ and
from our assumptions we have $\mathfrak{m}^{tqd}\subseteq
J+\mathfrak{q}^{[q]}Q$.
We claim that $J\subseteq\mathfrak{m}^{d+1}$, otherwise choose $u$ in $J$ not
in $\mathfrak{m}^{d+1}$ and consider the following exact sequence
$0\longrightarrow\operatorname{Ker}(\alpha)\longrightarrow
Q/\mathfrak{m}^{c}\stackrel{{\scriptstyle\alpha}}{{\longrightarrow}}Q/\mathfrak{m}^{c}\longrightarrow\operatorname{Coker}(\alpha)\longrightarrow
0\hskip 72.26999pt(2.5.1)$
wherein the map $\alpha$ is the homothety by element $u$ and $c=tdq$.
Obviously, there is a surjective map as the following
$\operatorname{Coker}(\alpha)=Q/(uQ+\mathfrak{m}^{c})\longrightarrow
Q/(J+{\mathfrak{q}}^{[q]}Q)=B/\mathfrak{q}^{[q]}B.$
Therefore
$\ell_{Q}(\operatorname{Coker}(\alpha))\geqslant\ell_{Q}(B/{\mathfrak{q}}^{[q]}B)$.
It is clear that $B/{\mathfrak{q}}^{[q]}B$ has finite length over $Q$. The
ideal $\mathfrak{q}$ is an $\mathfrak{n}$-primary ideal of $A$, therefore
$\ell_{A}(B/{\mathfrak{q}}^{[q]}B)$ is finite and since
$k=A/\mathfrak{n}\simeq Q/\mathfrak{m}$, clearly the length of
$B/{\mathfrak{q}}^{[q]}B$ over both of $Q$ and $A$ are equal. By assumptions
$\ell(B/{\mathfrak{q}}^{[q]}B)\geqslant\ell(A/{\mathfrak{q}}^{[q]})$,
therefore we have
$\ell_{Q}(\operatorname{Coker}(\alpha))\geqslant\ell(B/{\mathfrak{q}}^{[q]}B)\geqslant\ell(A/{\mathfrak{q}}^{[q]})$
and from [5, (14.10)] we find that $\ell(A/\mathfrak{q}^{[q]})\geqslant
q^{t}e(A)\geqslant q^{t}$ wherein the symbol $e(A)$ is the multiplicity of
$A$. Thus $\ell_{Q}(\operatorname{Coker}(\alpha))\geqslant q^{t}$.
Let $x_{1},x_{2},...,x_{t}$ be a minimal generator for $\mathfrak{m}.$ The
local ring $Q$ is regular, so the associated graded ring
$gr_{\mathfrak{m}}(Q)$ is a polynomial ring in $t$ variables over $k$, see [7,
Page 76,Theorem 9(d)]. This fact implies that $x_{j}$ for $1\leqslant
j\leqslant t$ are analytically independent in $\mathfrak{m}$ and so that
$\nu_{Q}(\mathfrak{m}^{c-i})={c-i+t-1\choose t-1}$ for $i\geqslant 1$. Since
$u$ is not an element of $\mathfrak{m}^{d+1}$ therefore
$\operatorname{Ker}(\alpha)\subseteq\mathfrak{m}^{c-d}/\mathfrak{m}^{c}$ and
hence
$\ell_{Q}(\operatorname{Ker}(\alpha))\leqslant\ell_{Q}(\mathfrak{m}^{c-d}/\mathfrak{m}^{c})$.
The following equality is clear
$\ell_{Q}(\mathfrak{m}^{c-d}/\mathfrak{m}^{c})=\displaystyle{\sum_{i=1}^{d}}\nu_{Q}(\mathfrak{m}^{c-i}).\hskip
72.26999pt(2.5.2)$
Now from (2.5.1) we have
$\ell_{Q}(\operatorname{Ker}\alpha)=\ell_{Q}(\operatorname{Coker}\alpha)$ and
from (2.5.2) we find that
$\ell_{Q}(\mathfrak{m}^{c-d}/\mathfrak{m}^{c})\leqslant d{c+t-2\choose
t-1}\leqslant dc^{t-1}=d(tdq)^{t-1}.$ Therefore $q^{t}\leqslant
d^{t}t^{t-1}q^{t-1}$ and this contradicts the choice of $q$. Hence we must
have $J\subseteq\mathfrak{m}^{d+1}$ and consequently
$\mathfrak{m}^{d}\subseteq
J+{\mathfrak{q}}^{*}Q\subseteq\mathfrak{m}^{d+1}+{\mathfrak{q}}^{*}Q.$ Thus
$\mathfrak{m}^{d}+{\mathfrak{q}}^{*}Q=\mathfrak{m}^{d+1}+{\mathfrak{q}}^{*}Q$
and by using Nakayama’s Lemma
$\mathfrak{m}^{d+1}\subseteq{\mathfrak{q}}^{*}Q$. This yields that
$J\subseteq{\mathfrak{q}}^{*}Q$ and this shows that the canonical homomorphism
$\pi:Q/{\mathfrak{q}}^{*}Q\rightarrow B/{\mathfrak{q}}^{*}B$ is an
isomorphism.
Now let $A$ be an equidimensional ring which is homomorphic image of a Cohen-
Macaulay ring. In this case, since ${\mathfrak{q}}^{[q]}$ is generated by a
system of parameters and it is tightly closed it follows from [4, (4.2.e)]
that $A$ is $F$-rational and hence $\mathfrak{q}={\mathfrak{q}}^{*}$. Now it
is easy to see that $\pi$ is an isomorphism between complete intersection
rings. ∎
###### Remark 2.6.
Let $(\mathcal{O},\mathfrak{m},k)$ be a complete discrete valuation ring with
finite residue field $k$ and let $\mathcal{D}$ be a minimal deformation
theory. Let $R$ and $T$ be the universal deformation and Hecke
$\mathcal{O}$-algebras associated to $\mathcal{D}$, respectively and let
$\varphi=\varphi_{\mathcal{D}}$ be the homomorphism described in Section 1.
Taylor and Wiles in the appendix of [9] for a simplification of some arguments
in Section 3 of [9] and Chapter 3 of [10] use a commutative algebraic result
of G. Faltings to reprove that $\varphi$ is an isomorphism between complete
intersection algebras. In Proposition 2.8 and Corollary 2.11 we give an
extension of corresponding results due to Taylor-Wiles and Faltings without
the assumption of finiteness of $k$. Our method to prove the above results is
totally different from their methods.
In the following we refine the definition of a _(level) n-structure_ due to
Wiles and Faltings in [9].
###### Definition 2.7.
Let $(\mathcal{O},\mathfrak{m},k)$ be a local ring and let
$\mathcal{O}[[Y]]=\mathcal{O}[[Y_{1},...,Y_{t}]]$. Set
$\mathfrak{q}=(Y_{1},...,Y_{t})$ and assume that for an integer $n\geqslant 1$
we have a commutative diagram of local $\mathcal{O}$-algebras as the following
in which $\varphi$ is surjective and $T$ is a finite free
$\mathcal{O}$-module.
$\begin{array}[]{llll}\mathcal{O}[[Y]]\stackrel{{\scriptstyle\psi_{n}}}{{\longrightarrow}}&R_{n}&\stackrel{{\scriptstyle\varphi_{n}}}{{\longrightarrow}}&T_{n}\\\
&\downarrow&&\downarrow\\\
&R&\stackrel{{\scriptstyle\varphi}}{{\longrightarrow}}&T\\\ \end{array}$
By a level $n$-structure we mean the above diagram with the following
properties:
$(1)$ There is a surjective homomorphism
$\lambda_{n}:\mathcal{O}[[X_{1},X_{2},...,X_{t}]]\longrightarrow R_{n}$.
$(2)$ The ring homomorphism $R_{n}\longrightarrow T_{n}$ is surjective.
$(3)$ From the vertical homomorphisms we get the isomorphisms
$R_{n}/{\mathfrak{q}}R_{n}\rightarrow R$ and
$T_{n}/{\mathfrak{q}}T_{n}\rightarrow T$.
$(4)$ The ring $T_{n}/{\mathfrak{q}^{[p^{n}]}}T_{n}$ is finite and free as a
module over $\mathcal{O}[[Y]]/{\mathfrak{q}^{[p^{n}]}}$.
For convenience we represent a level $n$-structure by the notation
$L_{n}(\mathcal{O},\varphi,\varphi_{n},\psi_{n},\lambda_{n})$.
###### Proposition 2.8.
Let $m\geqslant 1$ be an integer such that for every $n\geqslant m$, the level
$n$-structure $L_{n}(\mathcal{O},\varphi,\varphi_{n},\psi_{n},\lambda_{n})$
has the following additional property:
$(5)$ For each $x\in\mathfrak{m}$ we have $\psi_{n}(x)\in\mathfrak{m}R_{n}$.
Then $\varphi$ is an isomorphism between complete intersection
$\mathcal{O}$-algebras.
###### Proof.
Set $\widehat{\square}=\square\otimes_{\mathcal{O}}k$,
$\mathfrak{q}_{n}=\mathfrak{q}^{[p^{n}]}$ and
$\mathcal{O}[[X]]=\mathcal{O}[[X_{1},X_{2},...,X_{t}]]$. From (1) there is a
surjective homomorphism $k[[X]]\rightarrow\widehat{R_{n}}$. The fifth property
implies that $\psi_{n}(\mathfrak{m}[[Y]])\subseteq\mathfrak{m}R_{n}$. Thus
$\psi_{n}$ induces a local homomorphism
$\widehat{\psi_{n}}:k[[Y]]\longrightarrow\widehat{R_{n}}$. Since $k[[X]]$ is a
complete local ring, using [1, (7.16)] we can lift the homomorphism
$k[[Y]]\rightarrow\widehat{R_{n}}$ to a unique homomorphism $k[[Y]]\rightarrow
k[[X]]$. Thus from (1), (2) and (3) we find that there is a surjective
homomorphism as the following
$\alpha:k[[X]]/{\mathfrak{q}}k[[X]]\longrightarrow\widehat{T_{n}}/{\mathfrak{q}}\widehat{T_{n}}\simeq\widehat{T}.$
From (3) it follows that $\widehat{T_{n}}/{\mathfrak{q}}\widehat{T_{n}}$ is a
finite vector space over $k$. We may assume that the length of
$\widehat{T_{n}}/{\mathfrak{q}}\widehat{T_{n}}$ over $k[[X]]$ is equal $d$.
Now choose $n\geqslant m$ such that $p^{n}>t^{t-1}d^{t}$. From the fourth
property it is clear that the length of
$\widehat{T_{n}}/{\mathfrak{q}_{n}}{\widehat{T_{n}}}$ over $k[[Y]]$ is equal
or greater than the length of $k[[Y]]/{\mathfrak{q}_{n}}k[[Y]]$.
Now by writing $A=k[[Y]]$, $Q=k[[X]]$ and $B=\widehat{T_{n}}$ from Theorem 2.5
and Fact 2.3 we get that $\alpha$ is an isomorphism between complete
intersection rings.
On the other hand, $\varphi_{n}$ is surjective, from this and the third
property we have the following surjective homomorphisms
$\beta:k[[X]]/{\mathfrak{q}}k[[X]]\longrightarrow\widehat{R_{n}}/{\mathfrak{q}}\widehat{R_{n}}$
$\gamma:\widehat{R_{n}}/{\mathfrak{q}}\widehat{R_{n}}\longrightarrow\widehat{T_{n}}/{\mathfrak{q}}\widehat{T_{n}}$
such that $\alpha=\gamma\beta$. It was shown that $\alpha$ is an isomorphism
thus from (3) it is clear that $\widehat{\varphi}$ is an isomorphism between
$\widehat{R}$ and $\widehat{T}$ as complete intersection rings. Since $T$ is a
finite free module and $\varphi$ is surjective, Nakayama’s Lemma implies that
$\varphi$ is an isomorphism. It is obvious that the local ring $T$ is a
homomorphic image of $\mathcal{O}[[X]]$. Since $\widehat{T}$ is a complete
intersection ring from Definition 2.1 it follows that $T$ and so $R$ are
complete intersection $\mathcal{O}$-algebras. ∎
###### Remark 2.9.
In the proof of the above proposition the fifth property enables us to reduce
the level $n$-structure
$L_{n}(\mathcal{O},\varphi,\varphi_{n},\psi_{n},\lambda_{n})$ to the level
$n$-structure
$L_{n}(k,\widehat{\varphi},\widehat{\varphi_{n}},\widehat{\psi_{n}},\widehat{\lambda_{n}})$
and then by lifting property which is based on [1, (7.16)] we can prove the
proposition as a simple corollary of our main theorem. If we assume that
$\mathcal{O}$ is $\mathfrak{m}$-adic complete, then $\mathcal{O}[[X]]$ is
$(\mathfrak{m}[[X]]+(X))$-adic complete. From (1) we have a surjective
homomorphism $\mathcal{O}[[X]]\longrightarrow R_{n}$. Now again by using [1,
(7.16)] we can lift the local homomorphism $\psi_{n}$ to a unique local
homomorphism $\mathcal{O}[[Y]]\longrightarrow\mathcal{O}[[X]]$. Now by a
similar method to our proof we can show that $\varphi$ is an isomorphism
between complete intersection $\mathcal{O}$-algebras. In this case, we don’t
need to use the fifth property directly. However, in the following corollary
as a simple application of the level $n$-structures, we make clear the
relation between the completeness of $\mathcal{O}$ and the fifth property. In
a sense, we can say that the assumption of completeness of our base ring
implies the fifth property. We use the next simple lemma the proof of which is
omitted.
###### Lemma 2.10.
Let $L_{n}(\mathcal{O},\varphi,\varphi_{n},\psi_{n},\lambda_{n})$ be a level
$n$-structure, then there is a level $n$-structure
$L^{\prime}_{n}(\mathcal{O},\varphi,\varphi^{\prime}_{n},\psi^{\prime}_{n},\lambda^{\prime}_{n})$
such that $\lambda^{\prime}_{n}$ is the canonical epimorphism.
###### Corollary 2.11.
Let $(\mathcal{O},\mathfrak{m},k)$ be an $\mathfrak{m}$-adic complete local
ring, then $\varphi$ is an isomorphism between complete intersection
$\mathcal{O}$-algebras.
###### Proof.
From Lemma 2.10 it follows that there is a level $n$-structure such that the
homomorphism
$\lambda_{n}:\mathcal{O}[[X]]\longrightarrow\mathcal{O}[[X]]/J_{n}=R_{n}$ is a
canonical epimorphism. Therefore for each $g\in\mathcal{O}[[X]]$ we have
$\lambda_{n}(g)=g+J_{n}$, where $J_{n}=\operatorname{Ker}\lambda_{n}$. Since
$\mathcal{O}$ is $\mathfrak{m}$-adic complete hence $\mathcal{O}[[X]]$ is
$(\mathfrak{m}[[X]]+(X))$-adic complete. Thus from [1, (7.16)] we can lift
$\psi_{n}$ to a unique homomorphism
$\xi:\mathcal{O}[[Y]]\longrightarrow\mathcal{O}[[X]]$ such that $\xi(x)=x$ for
each $x\in\mathfrak{m}$. On the other hand, we have $\lambda_{n}\xi=\psi_{n}$.
Hence $\psi_{n}(x)=x+J_{n}$ and this shows that
$\psi_{n}(x)\in\mathfrak{m}R_{n}$. Now from Proposition 2.8 assertion holds. ∎
## Acknowledgments
The author is grateful to Sean Sather-Wagstaff and Irena Swanson for their
useful comments on this note.
## References
* [1] D. Eisenbud, Commutative algebra with a view toward algebraic geometry, Springer, (1995).
* [2] M. Hochster, C. Huneke, Tight closure, invariant theory and the Briancon-Skoda theorem, J. Amer. Math. Soc. 3 (1990), 31–116.
* [3] M. Hochster, C. Huneke, Tight closure of parameter ideals and splitting in module-finite extensions, J. Algebraic Geom. 3 (1994), no. 4, 599–670.
* [4] Huneke. C, Tight closure and it’s applications, With an appendix by Melvin Hochster. C.B.M.S. Regional Conference Series in Mathematics, 88, Amer. Math. Soc. (1996).
* [5] H. Matsumura, Commutative ring theory, Cambridge Studies in Advanced Mathematics, 8. Cambridge University Press, Cambridge, (1989).
* [6] K. A. Ribet, Galois representations and modular forms, Bult. AMS. 32 (1995), no. 4, 375–402.
* [7] J. P. Serre, Local Algebra, Springer Monographs in Matematics, Springer-Verlag, Berlin, (2000).
* [8] K. Smith, F-rational rings have rational singularities, Amer. J. Math. 119 (1997), 159–180.
* [9] R. Taylor, A. Wiles, Ring theoretic properties of certain Hecke Algebras, Annals of Math. 141 (1995), 553–572.
* [10] A. Wiles, Modular elliptic curves and Fermat’s last theorem. Ann. of Math, (2) 141 (1995), no. 3, 443–551.
| arxiv-papers | 2008-06-17T14:51:57 | 2024-09-04T02:48:56.314801 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Tirdad Sharif",
"submitter": "Tirdad Sharif",
"url": "https://arxiv.org/abs/0806.2808"
} |
0806.2882 | # Stability of Distant Satellites of the Giant Planets in the Solar System
Yue Shen11affiliationmark: and Scott Tremaine22affiliationmark: 1Princeton
University Observatory, Princeton, NJ 08544; 2Institute for Advanced Study,
Princeton, NJ 08540
###### Abstract
We conduct a systematic survey of the regions in which distant satellites can
orbit stably around the four giant planets in the solar system, using orbital
integrations of up to $10^{9}$ yr. In contrast to previous investigations, we
use a grid of initial conditions on a surface of section to explore phase
space uniformly inside and outside the planet’s Hill sphere (radius $r_{\rm
H}$; satellites outside the Hill sphere sometimes are also known as quasi-
satellites). Our confirmations and extensions of old results and new findings
include the following: (i) many prograde and retrograde satellites can survive
out to radii $\sim 0.5r_{\rm H}$ and $\sim 0.7r_{\rm H}$, respectively, while
some coplanar retrograde satellites of Jupiter and Neptune can survive out to
$\sim r_{\rm H}$; (ii) stable orbits do not exist within the Hill sphere at
high ecliptic inclinations when the semi-major axis is large enough that the
solar tide is the dominant non-Keplerian perturbation; (iii) there is a gap
between $\sim r_{\rm H}$ and $2r_{\rm H}$ in which no stable orbits exist;
(iv) at distances $\gtrsim 2r_{\rm H}$ stable satellite orbits exist around
Jupiter, Uranus and Neptune (but not Saturn). For Uranus and Neptune, in
particular, stable orbits are found at distances as large as $\sim 10r_{\rm
H}$; (v) the differences in the stable zones beyond the Hill sphere arise
mainly from differences in the planet/Sun mass ratio and perturbations from
other planets; in particular, the absence of stable satellites around Saturn
is mainly due to perturbations from Jupiter. It is therefore likely that
satellites at distances $\gtrsim 2r_{\rm H}$ could survive for the lifetime of
the solar system around Uranus, Neptune, and perhaps Jupiter.
###### Subject headings:
celestial mechanics – planets and satellites: formation – planets and
satellites: general – minor planets, asteroids
††slugcomment: AJ in press
## 1\. Introduction
Most of the satellites of the four giant planets in the solar system can be
divided into two groups, usually called the regular and irregular satellites.
Regular satellites orbit close to the planet (within $\sim 0.05r_{\rm H}$,
where $r_{\rm H}$ is the Hill radius111The Hill radius is defined as $r_{\rm
H}=a_{p}(\mu/3)^{1/3}$, where $a_{p}$ is the semi-major axis of the planet
orbit and $\mu\equiv m_{p}/(m_{p}+M_{\odot})$ with $m_{p}$ the planet mass.),
and move on nearly circular, prograde orbits that lie close to the planetary
equator. Irregular satellites are found at distances $\sim 0.05r_{\rm
H}-0.6r_{\rm H}$, with large orbital eccentricities and inclinations, on both
prograde and retrograde orbits. An alternative division between regular and
irregular satellites is given by the critical semi-major axis (e.g., Goldreich
1966; Burns 1986), $a_{\rm crit}=(2\mu J_{2}R^{2}a_{p}^{3})^{1/5}$; those with
$a>a_{\rm crit}$ are classified as irregular satellites. Here $J_{2}$ is the
planet’s second zonal harmonic coefficient (augmented by any contribution from
the inner regular satellites) and $R$ is the planet’s radius. This critical
radius marks the location where the precession of the satellite’s orbital
plane is dominated by the Sun rather than by the planet’s oblateness. The
current number ratios of irregular to regular satellites are $55/8$ for
Jupiter, $35/21$ for Saturn, $9/18$ for Uranus, and $7/6$ for Neptune (e.g.,
Jewitt & Haghighipour 2007). The regular satellites are likely to have formed
within a circumplanetary disk of gas and solid bodies. The kinematic
differences between regular and irregular satellites suggest that the latter
must have formed through a quite different mechanism, most likely capture from
the circumstellar disk (for a recent review, see Jewitt & Haghighipour 2007).
The search for irregular satellites of the giant planets has been fruitful in
recent years, owing mainly to modern high-sensitivity, large-scale CCDs (e.g.,
Gladman et al. 1998, 2000, 2001; Holman et al. 2004; Kavelaars et al. 2004;
Sheppard & Jewitt 2003; Sheppard et al. 2005, 2006). Up to 2007, 106 irregular
satellites of the giant planets had been discovered, compared to 53 regular
satellites. Two features stand out in the distributions of orbital parameters
of these irregular satellites. First, retrograde irregular satellites extend
to larger semi-major axes than prograde ones ($\sim 0.6r_{\rm H}$ compared to
$\sim 0.4r_{\rm H}$); second, satellites with orbital inclination in the range
$\sim 60^{\circ}-130^{\circ}$ relative to the ecliptic are absent.
A number of authors have shown that these features can be explained reasonably
well by the requirement that the satellite orbits be stable. Hénon (1969,
1970) studied the planar circular restricted three-body problem in Hill’s
(1886) approximation, where the mass ratio $\mu\rightarrow 0$ while the radii
of interest shrink to zero as $\mu^{1/3}$. He showed that prograde satellite
orbits are stable up to a mean distance from the planet $\sim 0.4r_{\rm H}$,
while retrograde satellite orbits can be stable at much larger distances from
the planet. Thus it is not surprising that retrograde satellites are found at
larger distances than prograde ones. Hamilton & Krivov (1997) studied the
dynamics of distant satellites of asteroids in heliocentric orbits using a
“generalized Tisserand constant” and, among other conclusions, confirmed that
retrograde orbits are more stable than prograde ones. Carruba et al. (2002)
used a combination of analytic arguments and numerical integrations to show
that high-inclination orbits inside the Hill sphere exhibit large eccentricity
oscillations (Kozai oscillations; Kozai 1962) due to secular solar
perturbations. They found that orbits with inclinations (relative to the
planetary orbital plane) between $55^{\circ}$ and $130^{\circ}$ are generally
unstable, thus explaining the absence of irregular satellites on high-
inclination orbits. Nesvorný et al. (2003) performed detailed orbital
integrations of the four giant planets plus a grid of test-particle satellites
for intervals of $10^{6}$–$10^{8}$ yr. They confirmed that retrograde
satellites can be stable at larger radii than prograde ones, and that highly
inclined orbits are unstable. They argued that the largest semi-major axes at
which satellites of the four giant planets could survive for times comparable
to the lifetime of the solar system were $\sim 0.7r_{\rm H}$ for retrograde
satellites and $\sim 0.4r_{\rm H}$ for prograde ones, and that these upper
limits were achieved only for nearly circular orbits close to the plane of the
ecliptic.
Other authors have examined the possibility that stable satellite orbits exist
with mean distance from the planet $\gtrsim r_{\rm H}$. In the planetocentric
frame, the dominant force on such satellites is due to the Sun, rather than
the planet222Hence these are sometimes called “quasi-satellites” (Lidov &
Vashkov’yak 1994a,b; Mikkola & Innanen 1997).. Nevertheless, the satellite
remains close to the planet because it is in a 1:1 resonance in the sense that
its heliocentric mean longitude librates around that of the planet; the
resulting orbit relative to the planet is a retrograde ellipse with axis ratio
2:1, the short axis pointing towards the Sun, and synodic period equal to the
planet’s orbital period. The analytical theory of such orbits is described by
Jackson (1913), Lidov & Vashkov’yak (1994a,b), Mikkola & Innanen (1997),
Namouni (1999), Mikkola et al. (2006), and others. Hénon’s (1970) numerical
analysis of Hill’s approximation to the planar circular restricted three-body
problem suggests that stable retrograde satellites can exist at arbitrarily
large distance from the planet. Benest (1971) confirmed that stable retrograde
orbits at large distances persist in the elliptic restricted three-body
problem, where the mass ratio and eccentricity were chosen to match those of
Jupiter. Wiegert et al. (2000) demonstrated that retrograde satellites of
Uranus and Neptune could be stable for up to $10^{9}$ yr at distances up to
$\sim 10r_{\rm H}$, suggesting that primordial objects of this type could
still exist in the solar system although none are currently known.
Despite the number and quality of these investigations, there are several
unanswered questions that lead us to revisit the problem of orbital stability
of satellites at large distances from the host planet. (i) Wiegert et al.
(2000) found stable satellite orbits beyond the Hill radius only for Uranus
and Neptune, not Jupiter or Saturn. What is the reason for this difference?
The possibilities include differences in the planetary masses and orbital
eccentricities, or different perturbations from neighboring planets. (ii)
Wiegert et al. (2000) explored orbits outside the Hill radius, while Nesvorný
et al. (2003) explored orbits inside the Hill radius (indeed, in the former
paper the integrations were terminated when the particles entered the Hill
sphere of radius $r_{\rm H}$ around the planet, while in the latter paper the
integrations were terminated when the particles exited the Hill sphere). Are
there stable satellite orbits that cross the Hill sphere? (iii) As we shall
describe further in §2, the grids of initial conditions used by Nesvorný et
al. and Wiegert et al. do not provide a complete exploration of the phase
space in which stable satellite orbits exist.
The primary goal of this paper is to map out the entire stability region in
phase space—both inside and outside the Hill sphere—in which satellite orbits
that can survive around the four giant planets for times comparable to the age
of the solar system (our main integrations last for up to 100 Myr). We
describe our setup in §2, and present the results in §3. We conclude and
discuss our results in §4.
Following Fabrycky (2008), we shall define a “satellite” of a planet to be a
small body whose distance from the planet never exceeds the semi-major axis of
the planet, $a_{p}$. This definition excludes bodies on Trojan orbits around
the triangular Lagrange points, bodies on horseshoe orbits, and objects such
as asteroid 2003 YN107 (Connors et al. 2004), which oscillates between a
horseshoe orbit and an orbit centered on Earth. This definition seems simple
and reasonable to us, but other definitions are common in the literature. Many
authors define “satellite” to be an object that always remains within the Hill
sphere of the planet or whose Jacobi constant constrains it to remain within
the last closed zero-velocity surface around the planet. Benest (1971) defines
a satellite to be a body whose heliocentric orbital frequency is the same as
the planet’s, but whose synodic frequency around the planet is non-zero.
Wiegert et al. use the term “quasi-satellite” for an object that remains
outside the Hill sphere but whose heliocentric longitude difference from the
planet never exceeds $120^{\circ}$ and regularly passes through zero. However,
the term “quasi-satellite” is confusing because it is also used for objects
such as 2003 YN107 that spend part of their time on horseshoe orbits and thus
are only temporarily satellites in our sense.
## 2\. Methods
Although all of our results are based on direct numerical integrations of the
N-body problem (Sun, one or four giant planets, plus a test particle orbiting
one planet), we shall find it useful to interpret our results in terms of the
coordinates and notation used by Hénon (1970) in the exploration of satellite
orbits in Hill’s approximation.
### 2.1. Hill’s approximation
When studying satellite motions near a planet ($r\ll r_{\rm H}$) it is
conventional to employ a non-rotating planetocentric coordinate system, which
we denote as $(xyz)$. However, in Hill’s approximation to the circular
restricted three-body problem, it is more convenient to use a rotating
planetocentric coordinate system $(\xi\eta\zeta)$, where $\xi$, $\eta$ and
$\zeta$ are scaled coordinates in the rotating frame in which the planet is at
the origin, the $\xi$ axis is along the direction opposite the Sun and the
$\zeta$ axis is perpendicular to the Sun-planet orbital plane. In Hill’s
formulation the unit of length is $\mu^{1/3}a_{p}$, and the unit of time is
$n^{-1}$ where $n\equiv[G(M_{\odot}+m_{p})/a_{p}^{3}]^{1/2}$ is the mean
motion of the planet. As usual, the orbit of the planet in the inertial frame
is counter-clockwise as viewed from the positive $z$ or $\zeta$ axis. In
Hill’s coordinate system the collinear Lagrangian points $L_{1}$ and $L_{2}$
are located at $\eta=\zeta=0$, $\xi=\pm 3^{-1/3}\simeq 0.6934$, and the Hill
radius is $r_{\rm H}=3^{-1/3}$. Similar definitions are used in this paper
when the planet orbit is eccentric and/or perturbed by other planets; in this
case the $\xi$ axis points away from the instantaneous position of the Sun,
the $\zeta$ axis is perpendicular to the instantaneous orbital plane of the
planet around the Sun, and $a_{p}$ is the initial semi-major axis of the
planet.
In the circular restricted three-body problem, Hill’s approximation is
achieved by taking the limit $\mu\rightarrow 0$, where the equations of motion
reduce to (e.g., Hénon 1974; Murray & Dermott 1999):
$\displaystyle\ddot{\xi}=2\dot{\eta}+3\xi-\frac{\xi}{(\xi^{2}+\eta^{2}+\zeta^{2})^{3/2}}\
,\ \ \ddot{\eta}=-2\dot{\xi}-\frac{\eta}{(\xi^{2}+\eta^{2}+\zeta^{2})^{3/2}}\
,$
$\displaystyle\qquad\qquad\qquad\ddot{\zeta}=-\zeta-\frac{\zeta}{(\xi^{2}+\eta^{2}+\zeta^{2})^{3/2}}\
.$ (1)
There exists an integral of motion for these equations,
$\Gamma=3\xi^{2}+\frac{2}{(\xi^{2}+\eta^{2}+\zeta^{2})^{1/2}}-\zeta^{2}-(\dot{\xi}^{2}+\dot{\eta}^{2}+\dot{\zeta}^{2})\
,$ (2)
which corresponds to the Jacobi constant in the circular restricted three-body
problem.
Figure 1.— Sampling of initial conditions in terms of the Hénon diagram. The
gray-shaded regions are forbidden. The Lagrange points $L_{1}$ and $L_{2}$ are
marked by $*$. The horizontal dashed line at $\xi=0$ separates retrograde
orbits from prograde ones (as defined in the rotating planetocentric frame).
Left: The region shaded by vertical lines is an approximate reproduction of
the stable region as estimated in Hénon (1970). The dotted region is where the
osculating Kepler elements correspond to bound elliptical orbits ($a>0$,
$e<1$). Right: Curved lines represent the initial conditions derived using
osculating elements in the high-resolution survey of Nesvorný et al. (2003),
color-coded according to eccentricity. Solid and long-dashed lines represent
orbits which are prograde and retrograde in the non-rotating planetocentric
frame respectively. There are two sets of lines for each eccentricity
corresponding to argument of pericenter $\omega=0^{\circ}$ and $180^{\circ}$
respectively. The short segments at the lower right below $L_{1}$ are
extensions of the $e=0.50$ and $0.75$ branches which are prograde in the non-
rotating frame. Thus orbits that are retrograde in the rotating frame can be
prograde in the non-rotating frame. The initial conditions for zero-
inclination orbits sampled by Wiegert et al. (2000) are shown as filled
circles for Uranus and open circles for Jupiter.
For the moment let us restrict ourselves to motion in the Sun-planet orbital
plane, so $\zeta=\dot{\zeta}=0$ at all times. Then to study orbital motions we
may use a surface of section defined by $\eta=0$, $\dot{\eta}>0$. The
trajectory in the four-dimensional $(\xi,\eta,\dot{\xi},\dot{\eta})$ phase
space is then represented by a set of points in the $(\xi,\dot{\xi})$ plane,
and for a given value of the Jacobi constant $\Gamma$ the other two phase-
space coordinates can be derived from
$\eta=0,\qquad\dot{\eta}=\left(3\xi^{2}-\dot{\xi}^{2}+{2\over|\xi|}-\Gamma\right)^{1/2}.$
(3)
We define “prograde” and “retrograde” in the rotating frame unless otherwise
noted. Thus retrograde orbits have $\xi<0$ in this surface of section and
prograde orbits have $\xi>0$.
A drawback of this surface of section is that a different plot is needed for
each value of the Jacobi constant $\Gamma$. To obtain a global view of the
dynamics, we use a different surface of section defined by $\eta=0$,
$\dot{\xi}=0$, $\dot{\eta}=(3\xi^{2}+2/|\xi|-\Gamma)^{1/2}$. A trajectory is
represented by a point in the $(\Gamma,\xi)$ plane. This surface of section
was introduced by Hénon (1969; 1970), and we shall call it the Hénon surface
of section or Hénon diagram. The Hénon diagram, like any surface of section,
will not show orbits that do not cross it; the usefulness of the Hénon diagram
derives from the observation that most stable orbits periodically pass close
to the point $\eta=\dot{\xi}=0$—for example, this occurs for nearly Keplerian
orbits close to the planet when their line of apsides precesses past the Sun-
planet line. The orbits not shown on the Hénon diagram include those confined
to some resonant islands, which should occupy a small fraction of phase space,
and escape orbits, which we are not interested in anyway.
Fig. 1 (left panel) is a Hénon diagram modeled on Figure 12 of Hénon (1970).
The Lagrange points are at $(\Gamma,\xi)=(3^{4/3},\pm 3^{-1/3})=(4.32675,\pm
0.69336)$. Forbidden regions, in which
$\dot{\eta}^{2}=3\xi^{2}+2/|\xi|-\Gamma$ would be negative, are shaded in
gray. The stable regions of phase space, as estimated by Hénon, are denoted by
vertical stripes. The diagram shows that retrograde satellites ($\xi<0$) have
a larger stable region than prograde satellites ($\xi>0$), a conclusion
consistent with the numerical studies described in §1. Moreover, the stable
band in this diagram that begins at $(\Gamma,\xi)=(-1.4,-1.2)$ and stretches
downward to the left shows that retrograde satellites can be stable at
distances much larger than the Hill radius; in fact, this band continues to
arbitrarily large negative values of $\Gamma$ and $\xi$ (see Figure 13 of
Hénon 1970), so retrograde satellites can be stable at arbitrarily large
distances from the planet, at least in Hill’s approximation to the planar
circular restricted three-body problem.
In future discussions we divide the stable regions in Fig. 1 into three
branches: the inner prograde branch ($\xi>0$), the inner retrograde branch
($\xi<0$ and $\Gamma>0$), and the outer retrograde branch ($\xi<0$ and
$\Gamma<0$).
A simple and rather complete way to sample initial conditions in the planar
three-body problem is to use the Hénon diagram, i.e., to sample uniformly in
the $(\Gamma,\xi)$ plane. As described above, this approach is based on the
assumption that most stable orbits periodically have their apocenter or
pericenter on the Sun-planet line. Note that even without invoking Hill’s
approximation the question of which initial conditions on the Hénon diagram
correspond to stable orbits is well-posed. Accordingly, we may present our
stability results in terms of the Hénon diagrams, even though our orbit
integrations do not use Hill’s approximation.
We may compare this approach to the grids of initial conditions used in other
investigations of the stability of satellite orbits. The initial conditions
for Nesvorný et al.’s “high-resolution survey” were chosen from a grid of
planet-centered osculating Keplerian orbital elements, with semi-major axis
$a$ given typically by $a/r_{\rm H}=0.1$–$1$, eccentricity $e=0$–$0.75$,
inclination $i=0^{\circ}$–$180^{\circ}$, argument of pericenter
$\omega=0^{\circ},90^{\circ}$, and the other elements distributed uniformly
between $0^{\circ}$ and $360^{\circ}$. The right panel of Fig. 1 shows similar
initial conditions on the Hénon diagram ($i=0^{\circ}$ or $180^{\circ}$ and
$\omega=0^{\circ}$ or $180^{\circ}$). The conversions from osculating elements
to $(\Gamma,\xi)$ were done using equations (8) and (10) in Hénon (1970). It
is clear that the initial conditions sampled in Nesvorný et al. do not provide
a complete exploration of the phase space in which stable satellite orbits
could exist; in particular, they completely missed the stable region that
extends beyond the Hill sphere (of course, such orbits are also excluded from
their study by their artificially imposed escape criterion $r>r_{\rm H}$). In
fact, most of the stable orbits beyond the Hill sphere have hyperbolic
osculating elements. In Fig. 1 we plot the boundaries that separate regions of
hyperbolic osculating elements from those with elliptical osculating elements,
where the latter are shaded by a dotted pattern. The functional forms of these
boundaries are:
$\displaystyle\Gamma$ $\displaystyle=$ $\displaystyle
2\xi^{2}-2^{3/2}|\xi|^{1/2}\ ,\qquad(\xi<0)\ ,$ (4) $\displaystyle\Gamma$
$\displaystyle=$ $\displaystyle 2\xi^{2}+2^{3/2}|\xi|^{1/2}\
,\qquad(0\leq\xi\leq 2^{1/3})\ ,$ (5) $\displaystyle\Gamma$ $\displaystyle=$
$\displaystyle 2\xi^{2}+2^{3/2}|\xi|^{1/2}\ ,\qquad(\xi<-2^{1/3})\ .$ (6)
The initial conditions explored by Wiegert et al. (2000) were chosen from a
grid of heliocentric osculating Keplerian elements, these being the same as
the elements of the host planets except for the eccentricity and inclination.
The eccentricity was typically chosen in the range $e=0$–$0.5$ and inclination
in the range $0^{\circ}$–$30^{\circ}$. With this procedure, zero-inclination
orbits appear in the Hénon diagram along the locus
$\Gamma=2/|\xi|-\xi^{2}\qquad\hbox{with}\qquad\xi=-e\mu^{-1/3}\ ,$ (7)
where the expression for $\Gamma$ is evaluated using Hill’s approximation. The
grid sampled by Wiegert et al. for $i=0$ is also shown in the right panel of
Fig. 1, converted from the heliocentric frame using Hill’s units (but without
Hill’s approximation); for clarity, only Jupiter and Uranus are shown.
Although Wiegert et al.’s initial conditions do probe the stability region
found by Hénon beyond the Hill sphere, the coverage is far from complete.
Figure 2.— Hénon diagrams for Jupiter under various circumstances (see §3.1
for details). The satellite orbital plane initially coincides with Jupiter’s.
In each panel, the dotted region is the grid of initial conditions in the
$(\Gamma,\xi)$ plane. The two blank regions in the upper-right corner are
forbidden. The filled circles in different colors represent the initial
conditions of orbits that survive for various times.
Figure 3.— Two-dimensional Hénon diagrams for the four planets. Each orbit is
integrated up to $10^{8}$ yr under the gravitational influence of its host
planet, the Sun, and the other three giant planets. Notations are the same as
Fig. 2.
To extend our study to three-dimensional orbital motions we use a surface of
section at $\eta=\zeta=\dot{\xi}=0$, $\dot{\eta}>0$. In the rotating frame, we
define the initial inclination angle $I$ by
$\tan I=\frac{\dot{\zeta}}{\dot{\eta}}\bigg{|}_{t=0}\ ,$ (8)
such that the initial $\eta$ and $\zeta$ component velocities are
$\dot{\eta}=\cos
I\bigg{(}3\xi^{2}+\frac{2}{|\xi|}-\Gamma\bigg{)}^{1/2}\\!\\!,\quad\dot{\zeta}=\sin
I\bigg{(}3\xi^{2}+\frac{2}{|\xi|}-\Gamma\bigg{)}^{1/2}\\!\\!.$ (9)
Since $\dot{\eta}>0$, the inclination is restricted to the range
$-90^{\circ}<I<90^{\circ}$.
Therefore each point in the $(\Gamma,\xi)$ plane represents a unique set of
initial conditions for a given inclination. The usefulness of the Hénon
diagram in this case is based on the assumption that most stable orbits
periodically have their line of apsides and their line of nodes simultaneously
on the Sun-planet line. This assumption is not always valid: it requires that
the argument of pericenter $\omega$ is periodically 0 or $\pi$, while a
satellite trapped in the Kozai resonance has an argument of pericenter that
librates around $\frac{1}{2}\pi$ or $\frac{3}{2}\pi$ (Kozai 1962; Carruba et
al. 2002). We estimate the incompleteness in our survey due to such orbits in
§3.2.
Because the equations of motion are symmetric around the $\zeta=0$ plane, we
may further restrict the inclination to the range $0^{\circ}\leq
I<90^{\circ}$. As in the two-dimensional case, we define “prograde” and
“retrograde” in the rotating frame unless otherwise noted. Thus retrograde
orbits have $\xi<0$ and prograde orbits have $\xi>0$ at this surface of
section $\eta=\zeta=\dot{\xi}=0$, $\dot{\eta}>0$.
### 2.2. Numerical orbit integrations
Even in the two-dimensional case, we expect that the stable regions for
distant satellites of the giant planets will be somewhat different from those
derived by Hénon (1970) and shown in Fig. 1, since (i) Hénon’s results are
based on Hill’s approximation $\mu\to 0$, while the giant planets have $\mu$
in the range 0.00096 (Jupiter) to 0.000044 (Uranus); (ii) Hénon’s results
assume that the planet orbit is circular, while the giant planets have
eccentricities between 0.0086 and 0.056; (iii) both the satellites and their
host planets are subject to perturbations from the other planets. We must
carry out long-term numerical integrations of the satellite orbits to assess
the influence of these effects on the stability region shown in Fig. 1.
We sample the initial conditions using a fine grid on the Hénon diagram, with
$d\Gamma=0.1$ and $d\xi=0.06$. This is shown as the dotted grid in Fig. 2. We
then convert them to the non-rotating $(xyz)$ planetocentric coordinate system
where we do the integrations of satellite orbits. We require that in the
rotating frame the Sun is always located at the $-\xi$ axis, and the angular
velocity of the rotating frame equals the instantaneous angular velocity of
the Sun relative to the planet in the non-rotating planetocentric frame; thus
the angular speed of the rotating frame is time-varying if the planet’s orbit
is eccentric, and the direction of the $\zeta$ axis may vary if the planet’s
orbit is perturbed by other planets. We use a unit of length $\mu^{1/3}a_{p}$
and unit of time $n^{-1}$ to scale the coordinates/velocities between the two
frames, where $a_{p}$ is taken to be the initial semi-major axis of the
planet.
The system to be numerically integrated is composed of the four outer giant
planets (or sometimes just one of them), the Sun, and a satellite around one
of the planets; the satellite is treated as a massless test particle. We use a
second-order Wisdom-Holman symplectic scheme (Wisdom & Holman 1991), as
implemented in the Swift package (Levison & Duncan 1994). Following Nesvorný
et al. (2003), we have modified the Swift code such that the integration of
the planets is done in the Jacobi coordinate system while that of the
satellites is done in the non-rotating planetocentric coordinate system. We
tried different timesteps to optimize between speed and accuracy, and found
$dt=20$ days is short enough to produce the correct results with reasonable
computational cost, for all four planets.
One potential concern is that the Wisdom-Holman symplectic scheme, as we have
implemented it, is designed for nearly Keplerian orbits relative to the planet
and might break down at large distances from the planet, where the orbits are
nearly Keplerian relative to the Sun. However, the characteristic orbital
period at large distances is equal to the planetary orbital period, and this
is much longer than the orbital periods of satellites inside the Hill radius
that the integrator is designed to follow, so even a crude integrator should
work well. Moreover, our ability to reproduce the Hénon diagram (compare Fig.
1 and the lower right panel of Fig. 2), the long-term stability of many of our
orbits, and the similarity of the characteristic orbit shapes to those found
by Hénon (see §3.1), all indicate that even at the largest distances probed
here, the symplectic integrator seems to work pretty well. As a further check,
we have used the Bulirsch-Stoer integrator to follow satellite orbits around
Uranus for $10^{6}$ yr and found almost identical results to the Wisdom-Holman
integrator.
We terminate the integration if the distance of the satellite from the planet
exceeds the semi-major axis of the planet since at this point the satellite
has escaped according to our definition at the end of §1, or if the distance
is less than the semi-major axis of the outermost regular satellite of each
planet (being Callisto, Iapetus, Oberon and Triton respectively), since at
this point the satellite lifetime against ejection or collision with the
regular satellite or the planet is likely to be short. Any test particles that
cross either of these two radii are considered lost. We have experimented with
including the quadrupole moment $J_{2}$ of the planet (including the
contribution from the inner regular satellites) but this has no detectable
effect on our results.
## 3\. Results
Figure 4.— Two-dimensional Hénon diagrams for Saturn only (left) and Uranus
only (right); the effects of the other three planets are not included in the
integrations. In contrast to the results in Fig. 3, Saturn can host stable
outer retrograde orbits, and most of the satellites that survive for $10^{6}$
yr also survive for $10^{8}$ yr around both planets.
### 3.1. Two-dimensional Hénon diagrams
We first study cases in which the initial velocity vectors of satellites lie
in the planet orbital plane, i.e., $\zeta=\dot{\zeta}=0$. As we have
described, this is different from Hénon’s problem because: (i) we do not use
Hill’s approximation; (ii) planets such as Jupiter have non-zero eccentricity;
and (iii) there are gravitational perturbations from other planets.
As an illustration we show how the stable region changes under various
conditions in Fig. 2, for satellites around Jupiter and an integration time
$10^{6}$ yr. We consider four situations: (a) Jupiter moves on its actual
(slightly eccentric) orbit, including perturbations from the other three giant
planets (upper left); (b) the planar restricted three-body problem, in which
Jupiter travels on an orbit with its current eccentricity of 0.048 and the
other planets are absent (upper right); (c) the planar circular restricted
three-body problem, in which Jupiter travels on a circular orbit with its
current semi-major axis (bottom left); (d) same as (c) except that the planet
mass is $1/100$ of the Jupiter mass (bottom right).
By comparing Figs. 1 and 2 it is clear that case (d) in the bottom right panel
best reproduces the original Hénon diagram; this is not surprising since
$\mu\simeq 10^{-5}$ is smallest so Hill’s approximation is satisfied best, and
the other conditions assumed in Hénon’s problem (circular planet orbit, no
other planets) are also satisfied. When using the actual Jupiter mass in (c),
the outer retrograde stable region (i.e., the lower-left branch) shifts and
shrinks. The overall stability region shrinks further—but does not vanish—when
Jupiter’s orbit is eccentric as in case (b), and for the most realistic case
(a).
Note that $10^{6}$ yr is only a small fraction of the lifetime of the solar
system, and possible erosion of the stable region over longer times is
somewhat indicated by the presence of a few red and blue dots in the upper
panels of Fig. 2, indicating orbits that are unstable on timescales of
$10^{4}$ and $10^{5}$ yr.
These illustrative calculations show that some Jovian satellites orbiting well
outside the Hill radius can survive for at least $10^{6}$ yr, although the
stable region is substantially smaller than in Hill’s approximation to the
circular restricted three-body problem and appears to erode slowly with time.
They also show that the stable region is larger (relative to the Hill radius)
if the planet mass $\mu$ is smaller, suggesting that the stable regions of the
other giant planets may be larger than Jupiter’s.
We now extend these calculations in the following ways: (i) we examine
satellite orbits around all four giant planets, using the actual planetary
orbits including perturbations from the three other planets; (ii) since stable
orbits are found at the most negative Jacobi constant ($\Gamma=-6$) examined
in Fig. 2, we extend the grid of initial conditions to $\Gamma=-16$; (iii) we
extend the integration time from $10^{6}$ yr to $10^{8}$ yr.
The results are shown in Fig. 3. There are large regions of inner
retrograde/prograde orbits that are stable for $10^{8}$ yr. For Jupiter, there
are a few outer retrograde orbits that survive for $10^{8}$ yr; Wiegert et al.
(2000) found no orbits that survived for $\gtrsim 10^{7}$ yr, but this may
reflect their less complete coverage of phase space. For Saturn, the outer
retrograde stable region completely disappears in less than $10^{6}$ yr, a
conclusion already reached by Wiegert et al. For Uranus and Neptune, in
contrast, there is a large stable region of outer retrograde orbits remaining
after $10^{8}$ yr. We expect that the stable regions around Jupiter, Uranus,
and Neptune will shrink somewhat further between $10^{8}$ yr and $5\times
10^{9}$ yr, the approximate age of the solar system, so we integrated some
outer retrograde satellite orbits around these three planets for $10^{9}$ yr.
We found that about a third of the Jovian orbits and over half of the outer
retrograde orbits for Uranus and Neptune shown in Fig. 3 still survive. Thus
it is very likely that Uranus and Neptune could still host primordial
satellites on such orbits to the present time. It is likely, but not certain,
that similar satellites could survive around Jupiter, at least in small
volumes of phase space.
The shrinkage of the stable region of the outer retrograde branch between
$10^{6}$ and $10^{8}$ yr, as well as the lack of stable outer retrograde
orbits around Saturn, appear to be mainly due to perturbations from the other
planets. To demonstrate this, we ran two $10^{8}$ yr integrations for Saturn
only and Uranus only. The results are shown in Fig. 4; in this case, Saturn
can host stable outer retrograde satellites for at least $10^{8}$ yr, and
there is almost no difference in the size of the stable region between
$10^{6}$ and $10^{8}$ yr for either planet.
The stable region around Uranus is larger than the one around Saturn in Fig.
4, and the stable regions around Uranus and Neptune are larger than the one
around Jupiter in Fig. 3. These differences are probably caused mostly by
their different planet-to-Sun mass ratios $\mu$. As $\mu$ increases, the outer
retrograde stability branch in the lower left of the Hénon diagram shrinks,
and shifts upward (see Hénon 1965; 1970, or compare the two lower panels of
Fig. 2).
We also notice in Figs. 2d and in the right panel of 4 that there is a little
tail or branch to the stable region around $(\Gamma,\xi)=(-5,-2.5)$. We
suspect this comes from Hénon’s periodic family $g_{3}$, which bifurcates from
the periodic retrograde orbits at $(\Gamma,\xi)=(-2,-1.2)$ and passes close to
the point $(\Gamma,\xi)=(-5,-2.5)$ (see Hénon 1970, Fig. 13).
Figure 5.— Examples of stable orbits around Uranus. All orbits initially lie
in the orbital plane of Uranus. Dots are instantaneous locations for the first
$10^{6}$ yr, plotted at intervals of 100 yr, with the green and red stars
marking the starting and ending locations. We also plot a few revolutions as
red curves. The left column is in the non-rotating planetocentric frame and
the right column is in the rotating planetocentric frame. In the left column,
the blue circles indicate the Hill sphere. Upper: an inner prograde orbit;
middle: an inner retrograde orbit; bottom: an outer retrograde orbit.
Figure 6.— Examples of three-dimensional Hénon diagrams for Uranus. The
initial inclinations $I=15,30,45,60,75$ degrees in the rotating frame. The
notation is the same as in Figs. 2 and 3. The stable regions shrink as the
inclination increases. In particular, no stable outer retrograde orbit exists
for $I\geq 30^{\circ}$.
Figure 7.— Examples of three-dimensional Hénon diagrams for Uranus. In these
diagrams the surface of section is taken when the satellite is at maximum
height above the planet’s orbital plane (in contrast to Figure 6 where the
surface of section is taken when the satellite crosses the plane). The initial
inclinations $I^{*}=15,30,45,60,75$ degrees in the rotating frame.
Figure 8.— Spatially accessible regions of stable satellite orbits for
Jupiter. The left column is in the non-rotating planetocentric frame and the
right column is in the rotating planetocentric frame. The bottom two panels
show expanded views of the inner portions of the upper two panels. In each
panel, retrograde orbits and prograde orbits are plotted separately in the
left and right halves, with “prograde/retrograde” defined in each frame used.
The blue circles show the Hill sphere and the smaller central red circles show
the inner boundary in the numerical integrations (the orbital radius of the
outermost regular satellite, in this case Callisto). The extreme thinness of
the zone of stable retrograde orbits outside the Hill sphere is an artifact of
our simulations, which sampled the initial inclinations only at
$0^{\circ},15^{\circ},\ldots$.
Figure 9.— Spatially accessible regions of stable satellite orbits for Saturn.
The notation is the same as in Fig. 8.
Figure 10.— Spatially accessible regions of stable satellite orbits for
Uranus. The notation is the same as in Fig. 8.
Figure 11.— Spatially accessible regions of stable satellite orbits for
Neptune. The notation is the same as in Fig. 8.
What do these stable orbits look like? In Hill’s approximation, the stable
(outer and inner) retrograde and inner prograde orbits are generated from the
periodic $f$ and $g$ families respectively using the terminology of Hénon
(1969). We show examples of stable orbits (i.e., those that survived for
$10^{8}$ yr) around Uranus in Fig. 5. For each example orbit, we plot the
instantaneous locations for the first one million years as dots with the two
stars marking the starting and ending locations. We also plot the trajectory
for several revolutions. Each orbit is plotted in both the non-rotating
planetocentric $x$-$y$ plane (left column) and the rotating $\xi$-$\eta$ plane
(right column). In the rotating frame, the inner prograde orbit (top panel) is
elongated along the Sun-planet axis while the inner retrograde orbit (middle
panel) is elongated perpendicular to the Sun-planet axis. The outer retrograde
orbit (bottom panel) is also elongated perpendicular to the Sun-planet axis
and oscillates about the planet as an ellipse with an axis ratio of
approximately 2:1 (compare Fig. 11 of Hénon 1970), as one would expect from
epicycle theory.
Note that the stable retrograde orbits with $\xi$ close to $-0.6934$ in Fig. 3
regularly cross the Hill radius. Hence such orbits are missed by the surveys
of both Wiegert et al. and Nesvorný et al., who terminate their integrations
if $r<r_{\rm H}$ or $r>r_{\rm H}$, respectively.
Figure 12.— Left: Heliocentric angular velocities of the stable outer
irregular satellites as a function of angular distance from the planet, as
viewed from the Sun in the non-rotating frame. Middle: Histograms of the
difference between the heliocentric satellite semi-major axis and the planet
semi-major axis for the stable outer irregular satellites, where the peaks of
these distributions are arbitrarily scaled. Right: Histograms of the
heliocentric eccentricity for the stable outer irregular satellites. The
sampling of points is the same as we used to produce the spatial stability
regions in §3.3.
### 3.2. Three-dimensional Hénon diagrams
We now extend the initial conditions in §3.1 to three dimensions by including
the scaled vertical coordinate $\zeta$. As discussed in §2.1, we consider a
surface of section $\eta=\zeta=\dot{\xi}=0$, $\dot{\eta}>0$ at $t=0$. Similar
to the two-dimensional case, we sample the initial conditions using a fine
grid in the $(\Gamma,\xi)$ plane, and use equation (9) to generate initial
velocities. We choose a sequence of inclinations in the rotating frame,
$I=15^{\circ},30^{\circ},45^{\circ},60^{\circ},75^{\circ}$. Each satellite
orbit is then integrated for $10^{8}$ yr along with the four giant planets and
the Sun.
The general behavior when incorporating inclination is the erosion of stable
regions in the Hénon diagram. As an example, we show the results for Uranus in
Fig. 6. The outer retrograde orbits quickly become unstable when the initial
inclination exceeds $\approx 20^{\circ}$. The inner retrograde stable region
erodes with increasing inclination and disappears at $I\gtrsim 75^{\circ}$.
The inner prograde stable region can survive even at $I\approx 75^{\circ}$.
This asymmetry between inner retrograde and prograde orbits is due in part to
the definition of inclination in the rotating frame333When translated into the
non-rotating planetocentric frame, the inclinations of “prograde”
(“retrograde”) orbits are actually smaller (larger) than in the rotating
frame. As we already noted in Fig. 1, under certain circumstances retrograde
orbits in the rotating frame can even be prograde in the non-rotating frame..
However, even when inclination is defined in the non-rotating frame such an
asymmetry may still be present (see Ćuk & Burns 2004 for a discussion of the
dynamical reasons for the asymmetry).
To separate the destabilizing effects of inclination from the effects of
perturbations from other planets, we also ran these three-dimensional
simulations for Uranus without the other planets. We found that for $I\lesssim
20^{\circ}$, perturbations from other planets do play a major role in eroding
the region of stable outer retrograde orbits, as illustrated by comparing the
upper left and lower right panels of Fig. 6, which show the Hénon diagram for
$I=15^{\circ}$ with and without the other planets. However, for
$I=30^{\circ},45^{\circ},60^{\circ},75^{\circ}$ the results for Uranus alone
are almost identical to the realistic case which includes perturbations from
the three other planets. This result suggests that bound retrograde orbits
outside the Hill sphere may not exist at all for $I\gtrsim 20^{\circ}$. This
is expected because the Coriolis force, which stabilizes outer retrograde
orbits by bending their trajectory towards the planet in the rotating frame,
is reduced when the inclination angle $I$ increases. This shrinkage of the
stable outer retrograde branch with inclination is already noticed by
comparing the right panel of Fig. 4 ($I=0$) and the bottom-right panel of Fig.
6 ($I=15^{\circ}$).
Note that the inclinations $I$ are planetocentric and measured in the rotating
frame. For low-inclination orbits along the outer retrograde branch they can
be converted to heliocentric inclinations $i$ using the approximate formula
for small $e$
$i\approx 2eI\ ,$ (10)
where $e$ is the heliocentric eccentricity. Thus our stability region
$I\lesssim 20^{\circ}$ corresponds roughly to $i\lesssim 4^{\circ}$ for
$e\approx 0.1$, the typical heliocentric eccentricity of surviving satellites
for Uranus and Neptune (see Fig. 12); this is in reasonable agreement with
Wiegert et al.’s estimate that most of their long-lived orbits had $i\lesssim
2^{\circ}$, especially considering that our sampling of initial conditions is
more complete than theirs. Mikkola & Innanen (1997) and Mikkola et al. (2006)
estimate analytically that the outer retrograde orbits are unstable if $i>e$
(for a circular planet orbit), which implies instability if $I\gtrsim
30^{\circ}$. Our own results show that all the test particles with initial
positions outside the Hill sphere cross the escape radius $a_{p}$ well before
$10^{4}$ yr for $I\geq 30^{\circ}$.
As described at the end of §2.1, a limitation of these results is that the
Hénon diagram we have used will not display orbits trapped in a Kozai
resonance, or other stable orbits whose argument of pericenter does not
periodically pass through 0 or $\pi$. To estimate the contribution of such
orbits, we have constructed a different set of three-dimensional Hénon
diagrams in which the initial conditions are changed from our usual choice
$\eta=\dot{\xi}=\zeta=0,\dot{\eta}>0$ to
$\eta=\dot{\xi}=\dot{\zeta}=0,\dot{\eta}>0$ (i.e., when the orbit is at its
maximum height above the planet’s orbital plane, rather than crossing the
orbital plane). In this case, we define the initial inclination $I^{*}$ in the
rotating frame by
$\tan I^{*}=\frac{\zeta}{\xi}\bigg{|}_{t=0}.$ (11)
The results are shown in Figure 7, which should be compared to Figure 6. Each
point in either set of Hénon diagrams corresponds to a unique orbit, but
orbits appearing in the Hénon diagrams of one figure at a given value of
$(\Gamma,\xi,I)$ may or may not appear in the other figure, where they will
have the same value of $\Gamma$ but possibly different values of $\xi$ and
inclination. The stable regions are somewhat larger in Figure 7 at a given
inclination—for example, a few outer retrograde satellites survive for
$10^{8}$ yr at $I^{*}=30^{\circ}$—but the conclusions described above are not
significantly altered. In the following discussion, we neglect stable orbits
that do not appear in our fiducial Hénon diagrams (i.e., using the surface of
section $\eta=\dot{\xi}=\zeta=0,\dot{\eta}>0$); thus we may slightly
underestimate the size of the stable regions. More discussion on orbits
trapped in the Kozai resonance inside the Hill sphere can be found in Carruba
et al. (2002).
### 3.3. Spatial stability regions
We now project the phase-space volume that hosts stable orbits onto coordinate
space, to explore where stable satellites might be found.
We plot the positions of stable orbits in the two-dimensional plane with
coordinates $\left[(x^{2}+y^{2})^{1/2},z\right]$ (non-rotating frame) or
$\left[(\xi^{2}+\eta^{2})^{1/2},\zeta\right]$ (rotating frame). Prograde and
retrograde orbits are plotted separately on the left and right sides of a
given figure panel, where “prograde” and “retrograde” are defined in the frame
used. We plot the position of each stable (up to $10^{8}$ yr) point in the
Hénon diagram at uniformly spaced times (every Myr) between $5\times 10^{7}$
and $10^{8}$ yr in Figs. 8 (Jupiter) to 11 (Neptune).
The stability regions within the Hill sphere are very similar to those shown
in Figs. 9-12 of Nesvorný et al. (2003), though slightly larger because we
show instantaneous position rather than semi-major axis. For Jupiter and
Saturn, the stable prograde orbits generally extend to $\sim 0.5r_{\rm H}$;
the stable retrograde orbits can extend further to $\sim 0.7r_{\rm H}$, and
nearly coplanar retrograde orbits even extend to $\sim r_{\rm H}$ for Jupiter.
For Uranus and Neptune, both the prograde and retrograde stable orbits can
extend a little bit further relative to the Hill sphere. No stable orbits
exist at high latitudes, presumably because of Kozai oscillations (Kozai 1962;
Carruba et al. 2002; Nesvorný et al. 2003).
It is also notable that there are stable regions beyond the Hill sphere for
Jupiter, Uranus and Neptune, as we discussed in previous sections. This is
particularly the case for Uranus and Neptune. Most of these distant stable
satellites are concentrated close to the orbital plane of the planet as for
Jupiter, although the appearance of a very thin layer in Figs. 8 (Jupiter)- 11
(Neptune) is somewhat an artifact of the coarse sampling of inclinations in
our initial conditions
($I=15^{\circ},30^{\circ},45^{\circ},60^{\circ},75^{\circ}$). Stable
satellites can be found as far as $\sim 5r_{\rm H}$ from Jupiter and even
$\sim 10r_{\rm H}$ for Uranus and Neptune, and as high as $2.5r_{\rm H}$ above
the orbital plane for the latter two planets. In the following section we
discuss briefly the strategy of searches for these distant satellites.
## 4\. Discussion and conclusions
We have conducted a systematic survey of the stable regions of satellites
around giant planets in the solar system, using numerical orbital integrations
that include gravitational perturbations from the other planets. We confirm
previous results for satellites within the Hill sphere: stable retrograde
satellites can exist further out than prograde satellites (e.g., Hénon 1970;
Hamilton & Krivov 1997; Nesvorný et al. 2003); and stable orbits cannot exist
at high inclinations (e.g., Carruba et al. 2002; Nesvorný et al. 2003).
We also confirm and extend the conclusions of Wiegert et al. (2000) that
distant retrograde satellites (“retrograde” as defined in the rotating frame)
can survive well beyond the Hill sphere for at least $10^{8}$–$10^{9}$ yr, and
probably for the lifetime of the solar system. Uranus and Neptune are the most
promising host planets for such distant satellites, since their stability
regions have the largest extent (e.g., Fig. 10-11). Jupiter has a smaller
stable region (Fig. 8), and Saturn appears to have no stable regions beyond
the Hill radius (Fig. 9).
Remarkably, there is a gap between the inner and outer stability zones for
retrograde satellites, extending from about $r_{\rm H}$ to $2r_{\rm H}$, in
which almost no stable orbits exist.
To check whether any of the proposed distant satellites have already been
discovered as Centaurs, we take the positions and velocities of the known
Centaurs from the IAU Minor Planet Center444http://cfa-
www.harvard.edu/iau/mpc.html that have planetocentric distances smaller than
the semi-major axes of each of the four giant planets at the last observed
epoch. There are 31 Centaurs (Jupiter 1; Uranus 16; Neptune 14) that satisfy
the criterion. We numerically integrated these objects along with the Sun and
giant planets for $10^{8}$ yr, but none of them survived as satellites
according to the definition in §1.
Searches for satellites far beyond the Hill radius can be carried out either
with dedicated deep, wide-angle surveys around the giant planets, or through
all-sky surveys such as Pan-STARRS and LSST. The most promising search areas
are close to the orbital plane of the planet, since only low-inclination
orbits survive (e.g., compare Fig. 3 and Fig. 6). In Fig. 12 (left panel) we
show the heliocentric angular velocity of the stable outer retrograde
satellites as a function of angular distance from the planet as viewed from
the Sun, for Jupiter (black), Uranus (blue) and Neptune (red) respectively,
sampled every Myr between $5\times 10^{7}$ and $10^{8}$ yr.
In the middle and right panels of Fig. 12 we show histograms of the difference
in heliocentric semi-major axis from their host planet and heliocentric
eccentricities for the stable outer retrograde satellites. These distributions
can be used to cull a large sample for potential satellites. Once a candidate
is identified with reliable orbit elements, a long-term orbital integration
should be run to confirm its satellite nature.
The discovery and characterization of satellites beyond the Hill sphere would
provide rich information about the early formation of the solar system.
Fabrycky (2008; also see Kortenkamp 2005) recently performed simulations of
capture of neighboring planetesimals from the circumstellar disk during slow
planet growth, and found that such distant satellites are a natural outcome
for Uranus and Neptune. Thus an inventory of this potential population of
bodies would enhance our understanding of the formation of planets and their
satellites in the early solar system, and the properties of the primordial
planetesimal disk.
This research was supported in part by NASA grant NNX08AH83G. We thank Dan
Fabrycky and the anonymous referee for comments that greatly improved the
paper.
## References
* (1) Benest, D. 1971, A&A, 13, 157
* (2) Burns, J. A. 1986, in Satellites, ed. J. A. Burns & M. S. Matthews (Tucson: University of Arizona Press), 1
* (3) Carruba, V., Burns, J. A., Nicholson, P. D., & Gladman, B. J. 2002, Icarus, 158, 434
* (4) Chambers, J. E. 1999, MNRAS, 304, 793
* (5) Connors, M., et al. 2004, Meteoritics & Planetary Science, 39, 1251
* Ćuk & Burns (2004) Ćuk, M., & Burns, J. A. 2004, AJ, 128, 2518
* (7) Fabrycky, D. 2008, submitted to Icarus
* (8) Gladman, B. J., Nicholson, P. D., Burns, J. A., Kavelaars, JJ, Marsden, B. G., Williams, G. V., & Offutt, W. B. 1998, Nature, 392, 897
* (9) Gladman, B., Kavelaars, JJ, Holman, M., Petit, J.-M., Scholl, H., Nicholson, P., & Burns, J. A. 2000, Icarus, 147, 320 (erratum in Icarus, 148, 320)
* (10) Gladman, B., et al. 2001, Nature, 412, 163
* (11) Goldreich, P. 1966, Rev. Geophys. Sp. Phys., 4, 411
* (12) Hamilton, D. P. & Krivov, A. V. 1997, Icarus, 128, 241
* (13) Hénon, M. 1965, Ann. Astr., 28, 992
* (14) Hénon, M. 1969, A&A, 1, 223
* (15) Hénon, M. 1970, A&A, 9, 24
* (16) Hénon, M. 1974, A&A, 30, 317
* (17) Hill, G. W. 1886, Acta Math., 8, 1
* (18) Holman, M. J., et al. 2004, Nature, 430, 865
* (19) Jackson, J. 1913, MNRAS, 74, 62
* (20) Jewitt, D., & Haghighipour, N. 2007, ARA&A, 45, 261
* (21) Kavelaars, J. J., et al. 2004, Icarus, 169, 474
* (22) Kortenkamp, S. J. 2005, Icarus, 175, 409
* (23) Kozai, Y. 1962, AJ, 67, 591
* (24) Levison, H. F., & Duncan, M. J. 1994, Icarus, 108, 18
* (25) Lidov, M. L., & Vashkov’yak, M. A. 1994a, Astro. Lett., 20, 188
* (26) Lidov, M. L., & Vashkov’yak, M. A. 1994b, Astro. Lett., 20, 676
* (27) Mikkola, S., & Innanen, K. 1997, in The Dynamical Behaviour of our Planetary System, ed. R. Dvorak and J. Henrard (Dordrecht: Kluwer), 345
* (28) Mikkola, S., Innanen, K., Wiegert, P., Connors, M., & Brasser, R. 2006, MNRAS, 369, 15
* (29) Murray, C. D., & Dermott, S. F. 1999, Solar Systam Dynamics (Cambridge: Cambridge University Press)
* (30) Namouni, F. 1999, Icarus, 137, 293
* (31) Nesvorný, D., Alvarellos, J. L. A., Dones, L., & Levison, H. F. 2003, AJ, 126, 398
* (32) Sheppard, S. S., & Jewitt, D. 2003, Nature, 423, 261
* (33) Sheppard, S. S., Jewitt, D., & Kleyna, J. 2005, AJ, 129, 518
* (34) Sheppard, S. S., Jewitt, D., & Kleyna, J. 2006, AJ, 132, 171
* (35) Wiegert, P., Innanen, K., & Mikkola, S. 2000, AJ, 119, 1978
* (36) Wisdom, J., & Holman, M. 1991, AJ, 102, 1528
| arxiv-papers | 2008-06-17T22:17:53 | 2024-09-04T02:48:56.320183 | {
"license": "Public Domain",
"authors": "Yue Shen (1), Scott Tremaine (2) ((1) Princeton, (2) IAS)",
"submitter": "Yue Shen",
"url": "https://arxiv.org/abs/0806.2882"
} |
0806.2938 | # Surface operators and magnetic degrees of freedom in Yang-Mills theories
A. Di Giacomo
University of Pisa and INFN, Sezione di Pisa,
Largo Pontecorvo 3, 56127, Pisa, Italy
V.I. Zakharov111Electronic address: vzakharov@itep.ru
INFN, Sezione di Pisa, Largo Pontecorvo 3, 56127, Pisa, Italy;
ITEP, B. Cheremushkinskaya 25, Moscow, 117218, Russia
###### Abstract
Magnetic degrees of freedom are manifested through violations of the Bianchi
identities and associated with singular fields. Moreover, these singularities
should not induce color non-conservation. We argue that the resolution of the
constraint is that the singular fields, or defects are Abelian in nature.
Recently proposed surface operators seem to represent a general solution to
this constraint and can serve as a prototype of magnetic degrees of freedom.
Some basic lattice observations, such as the Abelian dominance of the
confining fields, are explained then as consequences of the original non-
Abelian invariance. Generically, the properties of the two-dimensional defects
associated with the surface operators are close to the predictions of the dual
models for the magnetic D2 branes. 222This paper is prepared for the volume
dedicated to the 80th birthday of Lev Borisovich Okun. The authors are
acknowledging deep influence which Lev’s Okun style and works have had on
them. One of the main themes of papers of L.B. Okun is that it is only Nature,
or experiment which can decide whether a phenomenon is exotic or not, reality
or an artefact. We would include nowadays lattice data, or better to say
lattice experimentation with field theory into the means to discover new
phenomena. Magnetic monopoles were introduced first, as a pure theoretic
construct more than 70 years ago by P.M.A Dirac [1]. Reality of descendants of
the Dirac monopoles, magnetic degrees of freedom in Yang-Mills theories was
established, we believe, via lattice experiment. In these notes we are trying
to make a step towards theoretical interpretation of these magnetic degrees of
freedom invoking for this purpose the notion of surface operators [2] which
proved useful in such mathematical discipline as number theory. We are amused
by the idea that a kind of Abelian dominance exists both in Yang-Mills theory
and in the number theory and the same mathematical means can be adapted to
describe the two phenomena. We hope that the simplicity of the basic means
used will match Lev’s Okun quest for clarity of the basic concepts of physics.
PACS: 12.38.Aw; 11.15.Ha
## 1 Introduction
It is commonly accepted that confinement is due to condensation of magnetic
degrees of freedom. Such a mechanism is established, both theoretically and
via lattice measurements in the Abelian cases of a pure gauge field in the
compact version of the theory and $Z_{2}$ gauge theory, for an early review
see [3]. In this case the magnetic degrees of freedom are in fact the Dirac
monopoles.
Moreover on the lattice one observes monopole trajectories and has to use the
polymer approach to field theory, see, e.g. [4], to describe the monopole
properties in the language of the quantum geometry. In the geometrical
language the confining field configurations are identified as infinite
clusters of 1d defects, or trajectories. Another Abelian example is the
$Z_{2}$ gauge theory. In this case the confining configuration is a
percolating cluster of surfaces, or vortices.
In the non-Abelian case 333For simplicity we consider pure SU(2) gauge theory,
without matter. there is no consensus yet on the particular choice of the
magnetic degrees of freedom. The general idea is to reduce the non-Abelian
degrees of freedom to the pattern of Abelian theories where the mechanism of
confinement is well understood. For this purpose one introduces the so called
projected fields:
$A_{\mu}^{a}(x)~{}\to~{}A_{\mu}^{3}(x)~{},or~{}~{}A_{\mu}^{a}(x)~{}\to~{}(Z_{2})_{\mu}(x)~{}~{},$
(1)
where $A_{\mu}^{3}(x)$ is an Abelian gauge field, and $(Z_{2})_{\mu}(x)$ is a
$Z_{2}$ gauge field ($(Z_{2})_{\mu}(x)=\pm 1$ for links). The U(1) projections
emphasize the role of the monopole trajectories or 1d defects while $Z_{2}$
projections allow to find center vortices, or 2d defects 444The literature on
the confinement is vast. For the purpose of orientation and further references
we can mention reviews in Ref. [5] and in Ref. [6] exposing in detail the
$U(1)$ and $Z_{2}$ projections, respectively..
The projected fields (1) are related to the original Yang-Mills fields in a
nonlocal way and theoretical interpretation of the lattice data is obscured by
their use. It turns impossible to reconstruct configurations of the original
non-Abelian fields $A_{\mu}^{a}(x)$ which correspond to the monopoles or
vortices. However, it seems indisputable that effectively Abelian degrees of
freedom are responsible for the confinement.
In this note we will reverse the problem and instead of trying to directly
interpret the lattice data consider classification of non-Abelian singular
fields in the continuum which could be responsible for violations of the
Bianchi identities while not inducing violations of the color conservation. We
argue that a natural candidate for such defects are singular fields living on
two-dimensional surfaces. The central point is that such fields are Abelian in
nature and, therefore, violations of the Bianchi identities and of the
equations of motion are disentangled.
As far as the mathematical tools are concerned, we utilize the so called
surface operators [2] which describe four-dimensional (4d) non-Abelian fields
living on 2d surfaces. It is interesting that from the phenomenological point
of view the use of the surface operators suggests a unification of the two
alternative Abelian pictures mentioned above. Namely, the defects are 2d, as
emphasized by the center projections, while the fields living on them are
Abelian, as commonly emphasized by the $U(1)$ projections.
## 2 Abelian case
To set up the framework, let us review first the compact U(1) theory [7]. The
Lagrangian is the same as for a free electromagnetic field:
$L_{U(1)}~{}=~{}{1\over 4e^{2}}(F_{\mu\nu})^{2}~{}~{},$ (2)
supplemented, however, by the condition that the Dirac string carries no
action. The condition is automatically satisfied in the lattice version of the
theory.
Admitting singular fields, or monopoles into the theory violates Bianchi
identities and the modified Maxwell equations now read:
$\partial_{\mu}F_{\mu\nu}~{}=~{}0~{}~{},~{}~{}\partial_{\mu}\tilde{F}_{\mu\nu}~{}\equiv~{}j_{\nu}^{mon}~{},~{}~{}\partial_{\nu}j_{\nu}^{mon}~{}=~{}0~{}.$
(3)
where $j_{\nu}^{mon}$ is the monopole current.
The non-vanishing, conserved current $j_{\nu}^{mon}$ can be traded for a
magnetically charged scalar field $\phi_{M}$. This is a generic field
theoretic phenomenon (in the Euclidean space-time). The derivation can be
found in quantum-geometry courses, see, e.g., [4] while specific applications
to lattice monopoles are discussed, in particular, in Refs [8]. Here we will
remind, for a later use, a few basic steps in relating the monopole current to
a magnetically charged field.
Observing $j_{\mu}^{mon}$, say, in lattice simulations is equivalent to
observing particle trajectories. Therefore, it is reasonable to start from the
so called polymer formulation of a free scalar field theory with the action
$S_{cl}~{}=~{}M_{bare}\cdot L~{}~{},$ (4)
where $M_{bare}$ is the (bare) mass and $L$ is the length of trajectory.
The mapping of the polymer representation (4) into the standard field
theoretic representation is achieved through evaluating the path integral for
the particle propagator:
$D(x,y,M)~{}=~{}\Sigma_{paths}\exp\big{(}~{}-S_{cl}(x,y,M)\big{)}~{}~{}.$ (5)
The sum (5) can be evaluated exactly and one establishes [4] a relation
between the physical mass $m^{2}_{\phi}$ and the polymer-representation mass
$M$:
$m^{2}_{\phi}~{}\approx~{}{const\over a}\big{(}M_{bare}(a)~{}-~{}{{\ln 7}\over
a}\big{)}~{}~{},$ (6)
where $\ln 7$ is a geometric factor specific for a cubic lattice in $d=4$ and
we introduced explicitly dependence of the bare mass (4) on the lattice
spacing $a$ which is an ultraviolet cut off.
To relate the polymer approach to the physics of the lattice monopoles one
identifies the bare mass $M_{bare}$ in (4) with the radiative mass of the
monopole:
$M_{bare}~{}\to~{}M_{mon}~{}=~{}{const\over ae^{2}}~{}~{},$ (7)
where $e^{2}$ is the electric charge squared. The Higgs, or confining phase
corresponds to $m^{2}_{\phi}<0$. Once $m^{2}_{\phi}=0$ there emerges an
infinite, or percolating cluster of the monopole trajectories. The relation
between the monopole density $\rho_{mon}$ and the field-theoretic vacuum
expectation value reads as:
$\langle 0||\phi|^{2}|0\rangle~{}=~{}const\cdot
a\rho_{mon}~{}~{},~{}~{}(\langle 0|\phi|0\rangle)^{2}~{}=~{}const\cdot
a\rho_{mon}^{perc}~{}~{},$ (8)
where $\rho_{mon}$ is the total monopole density and $\rho_{mon}^{perc}$ is
the density of the percolating monopoles.
## 3 Non-Abelian singular fields
The Abelian construction just described does not generalize to the non-Abelian
case. Indeed, if
$(D_{\mu}\tilde{G}_{\mu\nu})^{a}~{}=~{}j_{\nu}^{a}~{}~{},$
where $j_{\nu}^{a}$ is the monopole current, then the colored current
$j_{\nu}^{a}$ would modify also the equations of motion, not only the Bianchi
identities since any colored current is a source of gluons. Moreover, if we
trade the current $j_{\nu}^{a}\neq 0$ for a scalar field then this field is
colored, $\phi^{a}$ and its vacuum expectation value would violate color
conservation.
Now, we are coming to a crucial point. We do not take these difficulties as a
proof that singular fields have no role to play in the non-Abelian case.
Instead, we merely conclude that trajectories, or 1d defects are not adequate
to the non-Abelian case and will be looking for defects of other dimensions.
Note that the fact that monopoles are intrinsically U(1) (not $SU(2)$) objects
has been emphasized since long, see, in particular, [9].
Turn now to two-dimensional defects, or surfaces. Classification of singular
field living on a surface, or 2d defects is actually contained in [2] 555The
observation on Abelian nature of the non-Abelian fields living on a surface,
crucial for our considerations, was intensely exploited also earlier [10].
However, the surface operators also allow to ascribe to the surfaces density
of topological charge, see below, and phenomenological consequences from this
observation have not been considered, to our knowledge.. The central point is
that non-Abelian fields living on a surface can in fact be rotated to an
Abelian direction and as a result violations of the Bianchi identities
($D_{\mu}\tilde{G}_{\mu\nu}\neq 0$) can be consistent with the validity of the
equations of motion ($D_{\mu}G_{\mu\nu}=0$).
Introduce first the action associated with the surfaces in the form:
$S_{surface}~{}=const\int d\sigma_{\mu\nu}G_{\mu\nu}^{a}~{}~{}~{}.$
This action is not gauge invariant. However, it can be redefined in such a way
as to respect the non-Abelian invariance. The reason is that the surface
interaction at each point $x$ involves only a single component of the field
strength tensor $G_{\mu\nu}^{a}$. Therefore, one can use gauge invariance to
rotate this particular component to the Cartan subgroup:
$d\sigma_{\mu\nu}G_{\mu\nu}^{a}(x)~{}\to~{}d\sigma_{\mu\nu}G_{\mu\nu}^{3}(x)~{},~{}~{}$
(9)
where for simplicity we considered the gauge group $SU(2)$.
Note that the projection (9) is determined up to a sign. One can fix the sign
by imposing the condition
$\big{(}d\sigma_{\mu\nu}G_{\mu\nu}^{3}\big{)}(x)\cdot\big{(}d\sigma_{\mu\nu}G_{\mu\nu}^{3}\big{)}(y)~{}>~{}0~{}~{}.$
(10)
As is argued in [2] the surface can be endowed also with a dual field
$\tilde{G}_{\mu\nu}^{a}$ which can also be rotated to the Cartan subgroup:
$d\sigma_{\mu\nu}\tilde{G}_{\mu\nu}^{a}(x)~{}\to~{}d\sigma_{\mu\nu}\tilde{G}_{\mu\nu}^{3}(x)~{},~{}~{}$
(11)
where the point $x$ belongs to the surface. Note that only one of the two
rotations (9), (11) is uncertain in sign.
From the point of view of the lattice formulation, the possibility of two
independent (apart from a sign) rotations assumes a particular regularization.
Indeed, the dual field is defined on the dual-lattice plaquettes and in this
sense can be rotated independently. However, such a procedure would assume
simultaneous use of both direct and dual lattices which is not necessarily
legitimate. We will not go into details of this issue here and just follow
Ref. [2] in postulating, in the continuum-theory language, existence of closed
surfaces, with surface element $d\sigma_{\mu\nu}$, and with the fields
$G_{\mu\nu}^{3},\tilde{G}_{\mu\nu}^{3}$ associated with the surfaces. One
ascribes then to the surface the following action
$S_{surface}~{}=~{}\alpha\int d\sigma_{\mu\nu}G_{\mu\nu}^{3}~{}+~{}\beta\int
d\sigma_{\mu\nu}\tilde{G}_{\mu\nu}^{3}~{}~{},$ (12)
where $\alpha,\beta$ are constants.
Note that in terms of invariants the two-dimensional defects considered have
both non-vanishing action and topological-charge densities,
$G^{2}(x)~{}>~{}0,~{}~{}G\tilde{G}(x)~{}\neq~{}0~{}~{},$ (13)
as far as the point $x$ belongs to the surface. The fields
$G_{\mu\nu}^{3},\tilde{G}_{\mu\nu}^{3}$ can be directly defined in terms of
the invariants as
$G^{3}_{\mu\nu}(x)~{}\equiv~{}+\sqrt{G^{2}(x)}{d\sigma_{\mu\nu}\over|d\sigma_{\mu\nu}|}~{},~{}~{}\tilde{G}^{3}_{\mu\nu}(x)~{}\equiv~{}{G\tilde{G}(x)\over\sqrt{G^{2}(x)}}{d\sigma_{\mu\nu}\over|d\sigma_{\mu\nu}|}~{},$
(14)
where the point $x$ belongs to the surface.
## 4 Various facets of defects
### 4.1 Wilson lines
Two-dimensional surfaces introduced above can be considered on classical or
quantum level, as external objects or as dynamical degrees of freedom. Indeed,
turn first to the example of the Wilson line. Classically (and in Abelian
case)
$\int_{C}A_{\mu}dx_{\mu}~{}=~{}\int_{surface}d\sigma_{\mu\nu}H_{\mu\nu}$
where the surface is spanned on the contour $C$. In the classical case the
contour integral merely ’measures’ the external magnetic flux. If the
potential $A_{\mu}$ is given locally by a pure gauge, the contour integral is
quantized.
If we introduce the Wilson line as an external object and consider it quantum
mechanically, then there emerge divergences due to the self energy:
$\langle
TrP\exp\\{-\int_{C}\hat{A}_{\mu}dx_{\mu}\\}\rangle~{}\sim~{}\exp(-const~{}g^{2}L/a)~{},$
(15)
where $L$ is the perimeter of the Wilson line $C$, $a$ is the lattice spacing,
$g^{2}$ is a coupling and we keep only the most divergent piece.
Finally, one can consider the loops as dynamical objects, generated within the
theory itself. In particular, it is quite common nowadays to consider a
condensate of Polyakov’s lines [11] which are defined at finite temperature:
$Tr~{}L~{}\equiv~{}TrP\exp\\{-\int_{0}^{1/T}dx_{0}\hat{A}_{0}({\bf
x})dx_{0}\\}~{}~{}$
and are the Wilson lines stretched in the time direction.
Phenomenologically, it is appealing to assume that there exists effective
potential for the vacuum expectation value of the Polyakov’s lines:
$V(<trL>)~{}=~{}-c_{2}<trL>^{2}~{}+~{}c_{4}<TrL>^{4}~{}~{},$ (16)
where in the deconfining phase, or at $T>T_{c}$ the constants $c_{2,4}$ are
positive, see, in particular [12].
Let us emphasize that for the vacuum expectation value $<TrL>$ to be non-
vanishing,
$<TrL>~{}\neq~{}0~{}~{},$
the ultraviolet divergence exhibited by (15) is to be canceled by the entropy.
We discussed above such a cancelation on the example of theory of a free
particle within the polymer approach.
### 4.2 Surface integrals classically
Surface integrals similar to (12) appear in the GNO classification [15] of the
non-Abelian monopoles 666The surfaces are closed in this case. Hereafter we
will consider closed surfaces in all the cases. . One assumes that there exist
classical solutions such that at large distances $r$ from the position of the
monopole the (space-space) components of the gauge field tensor take the form
$G_{ij}~{}=~{}{\epsilon_{ijk}r_{k}\over r^{3}}G(r)~{}~{},$ (17)
where $G(r)$ are magnetic charges. The solution can possess an electric charge
as well and then the dual field $\tilde{G}_{ij}$ is also non-vanishing. The
observation [15] is that in gauge theories, though electric charge takes
values in the weight lattice of the gauge group, the (quantized) magnetic
charge $G$ takes values in the weight lattice of a dual group.
In case of the gauge group $SU(N)$ the dual group is $SU(N)/Z_{N}$ where
$Z_{N}$ is the center group. In the simplest case of SU(2) group which we
concentrate on the magnetic charges $Q_{M}=\pm 1$ are to be identified since
the corresponding magnetic fields (17) can actually be gauged transformed into
each other.
### 4.3 Surface operators as external objects
As is mentioned above, Ref. [2] introduces the surface operators as external
objects on the quantum level. One can measure vacuum expectations value of the
surface operators by substituting the the vacuum fields (14) defined for each
configuration on the lattice and then averaging over the configurations.
Clearly, the vacuum expectation value is suppressed by the ultraviolet
divergent self energy proportional to the area of the closed surface:
$\langle
S\rangle~{}\sim~{}\exp\big{(}-\alpha\cdot(const)(Area)/a^{2}\big{)}~{}~{},$
(18)
where $\alpha$ enters definition of the action (12) and $(const)$ is related
to the average value of the plaquette action. In other words,
$G^{3}_{\mu\nu}(x)$ living on the surface appears singular in the limit of the
vanishing lattice spacing $a\to 0$. The divergence corresponds to self energy
of colored dipoles living on the surface and oriented in the third direction
of the color space. It is an analog of the divergence (15) in case of the
Wilson lines.
One can measure also density of the topological charge $G\tilde{G}(x)$ for the
points $x$ belonging to the surface, $x\epsilon\Sigma$. The results of the
measurements would depend on the properties of the Yang-Mills vacuum and
cannot immediately be predicted.
As in case of the Wilson line one can consider quantities not sensitive to the
ultraviolet divergence (18). For this purpose one can consider, for example,
closed surfaces unifying two planes separated by distances $r$ and measure
potential energy as function of $r$. Studying expectation values of the
surface operators could probe the confinement, similar to the case of the
Wilson lines [2]. Indeed, the vacuum expectation value of the surface operator
(apart from the just mentioned ultraviolet divergent piece) might exhibit
either volume or area laws:
$\langle S\rangle\sim\exp\big{(}-(const)(Volume)\big{)}~{},~{}or~{}\langle
S\rangle\sim\exp\big{(}-(const)(Area)\big{)}~{}.$ (19)
Note that by measuring the vacuum expectation values of the surface operators
one studies the interaction of particles having color magnetic and dipole
moments, but no color charge. Therefore, it is our guess that there is no
volume law even in the confining phase.
### 4.4 Surfaces as dynamical defects
In this paper we are going to make the next step and assume that surfaces
become dynamical and condense in the vacuum in the confining phase so that
there is a net of surfaces endowed with fields living on them. The picture is
similar to condensation of the Polyakov’s lines in the deconfinement phase.
It is a strong assumption concerning the dynamics of non-Abelian theories
which in no way can be derived from first principles. Our assumption is
motivated by the lattice data that do demonstrate existence of 2d surfaces
percolating through the vacuum [6, 13]. For a more theoretically oriented
reader let us mention that holographic models predict both percolation of the
Polyakov’s lines in the deconfining phase and percolation of magnetic strings
in the confining phase [14].
In more detail, we will rely on the Sakai-Sugimoto holographic model [16]
which is most successful in describing low-energy QCD. The metrics of this
model is formulated in terms of the ordinary 4d space $(t,x^{i},i=1,2,3)$,
extra fifth dimension $u$ with horizon at $u=u_{\Lambda}$, one more extra
compact dimension $x^{4}$ and compact sphere $\Omega_{4}$:
$\displaystyle ds^{2}=\big{(}{u\over
R_{0}}\big{)}^{3/2}(-dt^{2}+\delta_{ij}dx^{i}dx^{j}+f(u)dx_{4}^{2})+\big{(}{u\over
R_{0}}\big{)}^{-3/2}\big{(}{du^{2}\over f(u)}+u^{2}d\Omega_{4}^{2}\big{)}$
$\displaystyle f(u)~{}=~{}1~{}-~{}\big{(}{u_{\Lambda}\over u}\big{)}^{3}.$
(20)
It is also crucial that the defects wrapped on the compact dimension $x^{4}$
have non-trivial $\theta$-angle dependence. Note that the metric in dimensions
$(x^{4},u)$ has a cigar form so that the radius of the $x^{4}$ compact
dimension tends to zero at the horizon $u=u_{\Lambda}$. Which means that the
defects wrapped around this dimension have tension vanishing at the horizon.
In particular, the magnetic strings are defined as D2 branes wrapped around
the $x^{4}$ dimension. They look as strings, or 2d surfaces in the ordinary 4d
space and, in case of large enough area, have vanishing (classically) tension.
Because the D2 branes are wrapped on the compact $x^{4}$ dimension one expects
that they carry density of the topological charge.
The metrics (4.4) refers to the case of zero temperature $T=0$ which we are
mostly interested in. At finite T it is convenient to use Euclidean space,
with a compact time direction $\tau$. Thus, at $T\neq 0$ there are two compact
coordinates, $x^{4}$ and $\tau$. The deconfining phase transition is then
understood as change of geometries in two-dimensional $(x^{4},u)$ and
$(\tau,u)$ subspaces. Namely, at $T>T_{c}$ the cigar-shape geometry sets in
the $\tau,u$ coordinates while in the $(x^{4},u)$ coordinates the geometry
becomes cylinder-like. As a result, defects wrapped in (Euclidean) time
direction become tensionless and could condense. As is argued in [14] this is
a general phenomenon within the dual model: below $T_{c}$ defects condensed in
the vacuum have nontrivial $\theta$-angle dependencies while at $T>T_{c}$ it
is the defects wrapped on the compact time direction which are tensionless.
The Polyakov line is wrapped around the time direction and becomes tensionless
at $T>T_{c}$ according to the dual model. This fits well the phenomenological
models which postulate condensate of the Polyakov’s lines. The magnetic D2
branes are wrapped around the compact $x^{4}$ direction and are tensionless at
small temperatures. They are expected to condense at small T.
## 5 Surface operators and magnetic branes
The hint which we get from the dual models is that in the Yang-Mills vacuum
there could exist magnetic strings, tensionless. We conclude that their
tensionlessness is to be manifested as percolation of 2d surfaces, or
existence of an infinite cluster of such vortices. As usual in quantum
geometry, the surfaces are to be endowed with action. Moreover, the vanishing
string tension implies action-entropy balance. Another important prediction of
the dual models is that the surfaces are carrying topological charge density
(since the D2 branes are wrapped around the compact $x^{4}$ coordinate).
A crucial observation is that the surface operators are an adequate
mathematical tool to consider the magnetic strings because they also describe
2d surfaces endowed both with action and topological charge. In this section
we will consider further phenomenological consequences of such identification.
### 5.1 Non-Abelian ’monopoles’
Assumption on the dynamical nature of the 2d defects brings about a conclusion
that there should exist also 1d defects living on surfaces. Indeed, consider
first two external surfaces with
$(G^{2})_{1}~{}=~{}(G^{2})_{2}~{},~{}(\tilde{G}G)_{1}~{}=~{}-(\tilde{G}G)_{2}~{}.$
As far as the surfaces are treated as external objects they are different
although degenerate in energy. However, once we allow for dynamical surfaces
to exist the two degenerate states mix up and the actual ground state would
consist of their mixture. Lines on which the $\tilde{G}G$ changes its sign are
one-dimensional defects. Which we will call non-Abelian monopoles (since they
are associated with lines, or trajectories).
There is an important difference in the structure of gauge fields associated
with the Abelian and non-Abelian monopoles. Consider the Abelian case first
and introduce a time slice of the 4d space. Then at some point a singular
magnetic field may not have a particular direction:
${\bf H}_{Abelian}~{}\sim~{}{{\bf r}\over r^{3}}~{}~{},$ (21)
where ${\bf r}=0$ is the position of the singularity. Such a singularity
corresponds to the Dirac monopole.
In the non-Abelian case consider a time slice of the surface which is a line,
with coordinate $\tau$. At some point $\tau_{0}$ the magnetic field living on
the line may change direction:
${\bf H}_{non-Abelian}~{}=~{}|{\bf
H}|{(\tau-\tau_{0})\over|\tau-\tau_{0}|}~{}~{}.$ (22)
It is crucial that the magnetic field in the non-Abelian case is not
spherically symmetric but is line-like. Such field configuration is allowed by
the non-Abelian invariance as a defect while the spherically symmetric field
(21), familiar from the Abelian case, is not allowed. If we now include
development in time both the Abelian and non-Abelian monopoles become lines,
or trajectories. However, in the Abelian case the trajectories live on 4d
space while in the non-Ableian case they live on 2d surfaces.
Let us emphasize that the very existence of the 1d defects depends crucially
on the assumption that the 2d defects (surfaces) carry both fields
$G^{3}_{\mu\nu}$ and $\tilde{G}^{3}_{\mu\nu}$. If there were only magnetic
fields living on the surface the change of the direction of the magnetic
field, see Eq. (22) would be obscured by the non-Abelian gauge invariance
which allows to change the sign of the magnetic field.
### 5.2 Open strings
So far we were discussing closed surfaces, or strings. One might ask whether
it is possible to have an open string as a defect. Thus, we wish to introduce
a 1d boundary. For the whole construction to be non-Abelian invariant, this
boundary, or line is to have also non-Abelian invariant meaning. Physicswise,
the magnetic field which exists along the strings ’flows’ to its end in case
of an open string. Thus, it is natural to have a monopole at the end points.
Moreover, this ’monopole’ is to be defined in an $SU(2)$ invariant way. There
is only a single monopole of this type, that is the $Z_{2}$ monopole [18], or
the ’t Hooft line [19] (in case of $SU(N)$ group there are $(N-1)$ species of
$Z_{N}$ monopoles).
The $Z_{2}$ monopole can be viewed also as an ordinary Abelian monopole with
magnetic flux $2\pi/g$:
$\int_{Z_{2}~{}monopole}d\sigma_{\alpha\beta}G_{\alpha\beta}~{}=~{}{2\pi\over
g}$ (23)
where $g$ is the gauge coupling. Note that the corresponding Dirac string is
having quadratically divergent action and in this sense is visible (for
further discussion see [20]). Finally, we note that the magnetic field living
on the 2d surface is not quantized and the magnetic flux flowing to the end
points of the string is generically not equal to the flux (23).
## 6 Lattice data
As we have already mentioned, the lattice data demonstrate existence of 2d
surfaces with properties close to the predicted above. Mini-reviews of the
data can be found, e.g., in Ref. [13]. For the sake of completeness, we
reiterate main points here.
Definition of the surfaces on the lattice are algorithmic and is difficult to
translate into the continuum-theory language. This is the main difficulty in
interpretation of the data. As a partial compensation for this difficulty the
lattice itself allows for specific and powerful tests of the hypothesis that
some particular lattice defects have meaning in the continuum limit. First of
all, one can check scaling laws which are obeyed by physical objects, for
reviews see [6, 5].
Let us give an example. Imagine that there is a technical definition of closed
2d surfaces in Yang-Mills vacuum. One can measure global characteristics of
it. In particular, one measures the total area of the surface. Imagine,
furthermore, that one finds:
$(Area)_{tot}~{}\approx~{}\Lambda^{2}_{QCD}V_{tot}~{}~{},$ (24)
where $V_{tot}$ is the total volume of the lattice. The proportionality to
$V_{tot}$ is triviality for large enough volumes. However, the proportionality
to the physical scale, $\Lambda_{QCD}^{2}$ is highly nontrivial.
On the lattice, one changes the lattice spacing $a$ and the bare coupling
$g^{2}(a)$, in accordance with the renormgroup. Absence of explicit dependence
of the area (24) on the lattice spacing implies strong ’conspiracy’ of the
data. Indeed the statement (24) can be rewritten as observation that
probability of a given plaquette to belong to the surface is proportional to
$\theta_{plaq}~{}\approx~{}const(a\cdot\Lambda_{QCD})^{2}~{}~{},$ (25)
and changing the lattice spacing by, say, factor of 2 changes the probability
(25) by a factor of 4 which is a well defined prediction and a large effect
numerically.
Once relation (24) is found empirically, in terms of a certain $Z_{2}$
projection of the original non-Abelian fields it becomes a strong evidence
that using this projection allows to detect a gauge invariant physical object.
Indeed, the property (24) is perfectly gauge invariant. Moreover, it implies
that the tension of the strings is in physical units:
$T_{string}~{}\sim~{}\Lambda_{QCD}^{2}~{}~{},$ (26)
since it is the tension which controls the area. Of course it is not a
complete proof yet of the physical nature of the surfaces. Moreover, one could
reverse the procedure and just try to define such a projection that Eq. (24)
holds with good accuracy. Thus, we trade the property (24) for the definition
of the surface. The central problem then, are there further gauge invariant
properties possessed by the surface. In case of the center vortices the answer
is in positive (for review and details see [6, 13]).
In particular, there is an extra non-Abelain action associated with the
surfaces [20]:
$\delta S_{surface}~{}\approx~{}(const){(Area)\over a^{2}}~{}~{},$ (27)
where the use of the symbol ‘approximate’ always means that the evidence is
numerical, i.e., true within some error bars. We do not discuss the error bars
here and only sketch the overall picture.
The physical meaning of (27) is that the construction is consistent at short
distances. Indeed, in the language of the quantum geometry having (26) is
consistent only if there is action-entropy balance:
$(Tension)(Area)~{}=~{}(Action)_{surface}~{}-~{}(Entropy)_{surface}$ (28)
where both the action and entropy associated with the surface are ultraviolet
divergent while their difference is finite.
Another confirmation of this amusing picture is provided by measurement of the
extra action associated with plaquettes next to those belonging to our
surfaces. It vanishes identically:
$\delta S_{next~{}to~{}surface}~{}=~{}0~{}~{}.$ (29)
In physical terms, it is a confirmation that the action is indeed associated
with the field living on the surface as we have been conjecturing.
A crucial property of the surfaces we are discussing is that they are expected
to possess not only action but density of the topological charge as well.
Amusingly enough, this prediction has already been tested and, within error
bars, confirmed. Namely, one measures on the lattice for each particular
configuration space distribution of wave functions of topological modes of
quarks. It turns out that the intensity of the topological charge (as measured
by intensity of the topological fermionic modes) is correlated positively with
the position of the surfaces and the correlation grows as an inverse power of
the lattice spacing [21]:
$\\{(Intensity~{}(topological~{}modes),~{}~{}Intensity~{}(surfaces)\\}~{}\sim~{}a^{-\gamma}~{}~{},$
(30)
where the critical exponent $\gamma\approx 2$ 777The precise definition of the
correlator (30) is rather technical in nature. It can be found in the original
papers [21]..
Apart from the surfaces, or 2d defects one observes on the lattice 1d defects,
or monopole trajectories as well. The monopoles observed on the lattice 888So
called monopoles of the Maximal Abelian Projection. do seem to have properties
close to those predicted theoretically. Namely, they lie on the surfaces just
discussed and their non-Abelian field is rather line-like, aligned with the
surfaces.
### 6.1 Localization of classical solutions
Thus, the center vortices on the lattice do have properties similar to those
inherent to the surfaces discussed in the preceding sections and based on the
surface operators. However, the reader might feel uncomfortable to discuss
confinement in terms of singular fields (18). Indeed, in the continuum theory
one usually discusses non-perturbative physics in Yang-Mills theory in terms
of soft fields. On the other hand, the lattice data reveals that hard fields
are associated with the confinement and this cannot be disregarded. Thus, we
have a paradox.
It is worth emphasizing therefore that it would be wrong to consider the
infrared and ultraviolet sensitive descriptions as mutually excluding each
other. It is much more adequate to think in terms of a kind of duality, that
the nonperturbative physics can be described either in infrared or ultraviolet
terms. Moreover, one description probably smoothly goes into the other one as
a functions of resolution of measurements, or of the value of the lattice
spacing [22]. The term ’resolution’ is used here because generating each filed
configuration on the lattice, $\\{A_{\mu}^{a}(x)\\}$ can be considered as a
measurement of the gauge fields on the whole lattice with resolution of the
lattice spacing $a$.
Unfortunately, there is no detailed understanding of the role of the
resolution. Probably, one deals with quite a general phenomenon which can be
called stabilization of classical solutions with high-frequency oscillating
fields. At this time the guess can be substantiated, however, only by some
analogies and intuitive considerations. It would be extremely interesting to
clarify further these issues 999Analytically, interplay between the infrared
and ultraviolet sensitive descriptions was discovered first in 2d CP(N) models
[24]. Namely, it was proven that upon summation over all the instanton
solutions (including multi-instantons) the instanton ensemble is described by
the Coulomb-gas model. The constituents are point-like and possess topological
charge equal to $1/N_{c}$ fraction of that of the instanton. .
Let us start with a far fetched analogy from classical mechanics. Consider a
pendulum with frequency $\Omega_{slow}$. Apply now a fast oscillating force to
the fixed point of the pendulum, with amplitude $a_{fast}$ and frequency
$\omega_{fast}$. Then the effective potential which determines the resulting
motion of the pendulum becomes:
$U_{eff}~{}=~{}mgl(1-\cos\phi)+a^{2}_{fast}\omega^{2}_{fast}\sin^{2}\phi$ (31)
where $\phi$ is the angle of deviation of the pendulum from the vertical line.
If the ratio $\omega_{fast}/\Omega_{slow}$ is large the effective potential
(31) is much steeper than the original one. The resulting amplitude of
oscillation of the pendulum around $\phi=0$ is much smaller than without the
oscillating force. Moreover, the point $\phi=\pi$ becomes a minimum of the
effective potential as well 101010This is a famous example due to S. Kapitza,
see L.D. Landau and E.M. Lifshitz, Course of Theoretical Physics, vol. 1..
In the Yang-Mills case, the analogy to slow motion is provided by a classical
solution, say instanton, of size $\rho_{inst}\sim\Lambda_{QCD}^{-1}$. Imagine
that we would like to next include effect of zero-point fluctuations which are
fast oscillating fields. The zero-point fluctuations are dominated by the
ultraviolet scale:
$[a_{\mu}^{a}]_{quant}~{}\sim~{}a^{-1}_{latt}$
where $a_{\mu}^{a}$ is the quantum gauge field and $a_{latt}$ is the lattice
spacing.
The classical instanton solution $(A^{a}_{\mu})_{class}$ is no longer a
solution. Indeed $(A^{a}_{\mu})_{class}\sim\rho_{inst}/g$ and is much smaller,
by the factor of order $(a_{latt}/\rho_{inst})$ than the quantum field
$a_{\mu}^{a}$. In analogy with the mechanical example mentioned above one
should rather first average over fast quantum oscillations and derive an
effective potential.
Of course, there are no means to solve this problem analytically. However, it
seems rather obvious that the ’instanton’ shrinks, the same as the motion of
the pendulum. Indeed, instantons are in fact generic non-perturbative
fluctuations and the probability to find such a fluctuation is of order
$\theta_{non-
pert}~{}\sim~{}\exp\big{(}~{}-~{}(const)/\alpha_{s}(a^{2}_{latt})\big{)}~{}~{},$
(32)
where $\alpha_{s}(a^{2}_{latt})$ is the running coupling. The crucial point is
that $\alpha_{s}$ is normalized on the scale on which measurements are
performed, i.e. on the scale $a_{latt}$.
Estimate (32) implies that the instanton volume shrinks to zero as a power of
the lattice spacing $a_{latt}$. Moreover, relying bluntly on the analogy with
the pendulum, we come to a bizarre conclusion that
$V_{instanton}~{}\sim~{}(const)\cdot a^{4}_{lattice}~{}~{}.$ (33)
Strange to say, but this guess might well be true, according to the lattice
data 111111 The phenomenon of the shrinking of classical solutions was
discovered by the ITEP lattice group [20]. However, the classical solutions
relevant to this paper belong to a dual formulation of the Yang-Mills theories
and the status of these solutions is rather vague. The idea to check this
phenomenon on the case of ’instantons’, or chiral defects belongs to the
authors of the first paper in Ref. [23]..
Let us specify the notion of the ’volume occupied by an instanton’. The
definition is provided actually by considering zero- or near-zero fermionic
modes. In the classical instanton case the topological modes are located in
the same region of the 4d space as the instanton itself:
$|q_{\lambda=0}(x)|^{2}~{}\sim~{}{\rho^{4}\over(\rho^{2}+x^{2})^{3}}~{}~{},$
(34)
where $q_{\lambda=0}$ is a zero mode of the Dirac equation for a quark in
external gluon field:
$\hat{D_{\mu}}\gamma_{\mu}q_{\lambda=0}~{}=~{}0~{},$
and to derive (34) one substitutes the instanton field for the external gluon
field.
On the lattice, one can solve the Dirac equation numerically substituting into
the covariant derivative $D_{\mu}$ the gluon fields generated on the lattice
for a particular configuration $\\{A_{\mu}^{a}(x)\\}$. Then the volume
occupied by the topological fermionic modes can be understood as the volume
occupied by ‘instanton modified by high-frequency perturbative fields’.
Lattice data rather support estimate (33).
Let us emphasize that the shrinking volume of the instanton implies also
singular fields inside the small region occupied by the modified instanton.
Indeed, the average value of the topological charge squared, or topological
susceptibility,
$\int
d^{4}x<0|~{}G\tilde{G}(x),~{}G\tilde{G}~{}|0>~{}\sim~{}\Lambda_{QCD}^{4}~{}~{},$
(35)
represents a matrix element and its value cannot depend on details of the
measurements, in particular on the value of $a_{latt}$. On the other hand, the
value of $Q_{top}^{2}$ for each configuration is equal to the number of exact
zero modes of the quarks. Thus, the total topological charge of configurations
cannot depend on the shrinking of each particular lump of the topological
charge. Hence, the topological fields become singular in the limit of the
vanishing volume of the lumps.
Assuming all this to be correct, what are implications for interpretation of
2d surfaces? To answer the question note that in terms of the dual model
instantons are D0 branes. These branes have one extended coordinate which is
wrapped around the $x^{4}$ circle. In the geometrical language of the dual
models (4.4) this wrapping is responsible for the non-trivial topological
charge of instantons. To summarize: the instantons are points in ordinary 4d
space, within the defects classification of the dual models.
On the other hand, in the ordinary field theoretic language the instantons are
rather extended objects of the size of order $\rho\sim\Lambda_{QCD}^{-1}$.
Remarkably enough, the shrinking caused by $a_{latt}\to 0$ just brings an
instanton to a point:
$\Big{(}G^{2}\sim{\rho^{4}\over(\rho^{2}+x^{2})^{4}}\Big{)}~{}\to~{}\Big{(}G^{2}\sim\delta^{4}(x)\Big{)}~{}~{},a_{latt}\to
0.$ (36)
Thus, one can speculate that in the limit of vanishing lattice spacing the
geometry of the classical solutions becomes similar to the geometry of the
classical solutions in the dual formulation.
Coming back to our surfaces populated with singular non-Abelian fields they
could well be a high-frequency image of classical solution of Yang-Mills
theory in the dual representation. This guess cannot be supported however by
any straightforward derivation.
## 7 Open questions. Conclusions
As we have demonstrated above, different approaches to magnetic degrees of
freedom converge to a common and highly non-trivial picture. In particular,
the dual models [14] relate percolation of the magnetic strings to the
prediction that they should carry density of the topological charge. The
surface operators [2] provide us with an adequate construct to describe such
surfaces endowed with both gauge fields and their duals. On the lattice, one
does observe percolating surfaces, with tension in physical units and endowed
with topological charge.
In this note we emphasized that the monopole mechanism of confinement,
whatever it means in detail, assumes significant role to be played by singular
non-Abelian fields. Indeed, ’electric’ degrees of freedom are introduced
directly in the Lagrangian, while ’magnetic degrees of freedom’ are emerging
through violations of Bianchi identities. Hence magnetic degrees of freedom
are associated with singular fields.
The construction is well understood and known since long [7] in the Abelian
case. One can either start with a gauge field plus a scalar, Higgs field, or
one can start with pure gauge field but admit singular field configurations,
monopoles. Then these singular fields, or defects can be traded for a scalar
field and one comes back to an effective Higgs-like formulation.
In the non-Abelian case monopoles, or 1d defects do not match the group
structure [9, 18]. Instead, one should look for classification of singular
fields consistent with the non-Abelian symmetry. As is argued in fact recently
[2] such defects are 2d dimensional surfaces with non-Abelian fields living on
them. In other words, the surfaces are endowed with densities of action and of
topological charge. We argued that such surfaces can be profiled by 1d
defects, or non-Abelian monopoles, whose field is not spherically symmetric
but rather line-like.
For the defects to become dynamical degrees of freedom there should be a fine
tuning which allows infrared and ultraviolet scales to coexist. Such fine
tuning is well understood in the Abelian case, (see Eq. (6)). A central point:
for the dual-superconductor mechanism of confinement to be relevant, similar
fine tuning should be realized in the non-Abelian case as well. This time,
however, it is self-tuning between the action and entropy of the 2d surfaces.
No explicit form of the fine tuning is known theoretically but coexistence of
the two scale is seen in the lattice simulations.
Also, there should be consistency between dynamics in four dimensions, or of
the original Yang-Mills fields and dynamics on the defects world-sheet. Let us
conclude with an example of such a consistency check. For dynamics of a
particle (or 1d defect) to be in fact independent on the ultraviolet scale the
action associated with the ’monopoles’ should be approximately
$L\cdot M_{non-Abelian}~{}\approx~{}L\cdot{ln7\over a}~{}~{},$ (37)
see Eq. (6). And this has been checked on the lattice [25]. On the other hand,
the radiative mass (37) is associated now with the non-Abelian fields which
live not in the bulk but on the world-sheet (see discussion above). Which
seems to be also true on the lattice, for references and review see [13]. The
condition (37) replaces the quantization condition for the magnetic field in
case of the non-Abelian monopoles living on the surfaces.
## Acknowledgments
We are thankful to M.N. Chernodub, A.S. Gorsky, F.V. Gubarev, Ph. de Forcrand
for useful discussions. This note was worked out mostly during the time when
both authors were participants to the workshop “Non-Perturbative Methods in
Strongly Coupled Gauge Theories” at the Galileo Galilei Institute (Florence).
We are thankful to the GGI for the invitation and hospitality.
## References
* [1] Paul Dirac, ”Quantised Singularities in the Electromagnetic Field”, Proc. Roy. Soc. A 133, (1931) 60.
* [2] S. Gukov, E. Witten, “Gauge Theory, Ramification, And The Geometric Langlands Program”, [arXiv:hep-th/0612073];
E. Witten, Fortsch. Phys., 55 (2007) 545.
* [3] R. Savit, Rev. Mod. Phys. 52 (1980) 453.
* [4] A.M. Polyakov, ”Gauge Fields and Strings”, Harvard Academic Publishers, (1987);
J. Ambjorn, ”Quantization of geometry”, [arXiv:hep-th/9411179].
* [5] M.N. Chernodub, F.V. Gubarev, M.I. Polikarpov, A. I. Veselov , Progr. Theor. Phys. Suppl., 131. 309 (1998), [arXiv:hep-lat/9802036];
A. Di Giacomo, Progr. Theor. Phys. Suppl., 131 161 (1998, [arXiv:hep-
lat/9802008];
T. Suzuki, Progr. Theor. Phys. Suppl., 131 633 (1998).
* [6] J. Greensite, Prog. Part. Nucl. Phys., 51 (2003) 1, [arXiv:hep-lat/0301023].
* [7] A.M. Polyakov, Phys. Lett., B59 (1975) 82 ;
T. Banks, R. Myerson, J. B. Kogut, Nucl. Phys., B129 (1977) 493;
H. Shiba, T. Suzuki, Phys. Lett., B333 (1994) 461, [arXiv:hep-lat/9404015];
A. Di Giacomo, G. Paffuti, Phys. Rev. D56 (1997) 6816, [arXiv:hep-
lat/9707003].
* [8] M. Stone, P. R. Thomas, Phys. Rev. Lett., 41 (1978) 351;
M.N. Chernodub, V.I. Zakharov, Nucl. Phys., B669 (2003) 233, [arXiv:hep-
th/0211267].
* [9] S. R. Coleman, “The Magnetic Monopole Fifty Years Later”, published in Erice Subnuclear 1981:21 (QCD161:I65:1981).
* [10] M.N. Chernodub, F.V. Gubarev, M.I. Polikarpov, V. I. Zakharov, Nucl. Phys., B600 (2001) 165, [arXiv:hep-th/0010265].
* [11] A.M. Polyakov, Phys. Lett. 72B (1978) 477; B. Svetitsky, L.G. Yaffe, Nucl. Phys. B210 (1982) 423.
* [12] R. D. Pisarski, Phys. Rev. D62, 111501 (2000); [arXiv:hep-ph/0101168].
* [13] V.I. Zakharov, Braz. J. Phys., 37 (2007) 165 [arXiv:hep-ph/0612342]; AIP Conf. Proc., 756 (2005) 182, [arXiv:hep-ph/0501011].
* [14] A. Gorsky, V. Zakharov, Phys. Rev. D77 (2008) 045017, arXiv:0707.1284 [hep-th];
A.S. Gorsky, V.I. Zakharov, A. R. Zhitnitsky, “On Classification of QCD
defects via holography”’ [arXiv:0902.1842 [hep-ph]].
* [15] P. Goddard, J. Nuyts, D. I. Olive, Nucl. Phys. B125 (1977) 1.
* [16] T. Sakai, Sh. Sugimoto, Prog. Theor. Phys. 113 (2005) 843, [arXiv:hep-th/0412141].
* [17] G. ’t Hooft, Nucl. Phys., B190 (1981) 455.
* [18] E. Lubkin, Ann. Phys., 23 (1963) 233.
* [19] G. ’t Hooft, Nucl. Phys., B153 (1979) 141.
* [20] F.V. Gubarev, A.V. Kovalenko, M.I. Polikarpov, S.N. Syritsyn, V.I. Zakharov, Phys. Lett. B574 (2003)136,[arXiv:hep-lat/0212003].
* [21] A. V. Kovalenko, S. M. Morozov, M. I. Polikarpov, V. I. Zakharov, Phys. Lett. B 648 (2007) 383 [arXiv:hep-lat/0512036];
R. Hollwieser, M. Faber, J. Greensite, U. M. Heller, S. Olejnik, Phys. Rev.
D78 (2008) 054508, [arXiv:0805.1846 [hep-lat]].
* [22] V.I. Zakharov, “Matter of resolution: From quasiclassics to fine tuning.”, in Sense of Beauty in Physics: Miniconference in Honor of Adriano Di Giacomo on his 70th Birthday, Pisa, Italy, Jan 2006. [arXiv:hep-ph/0602141].
* [23] MILC Collaboration (C. Aubin et al.), Nucl. Phys. Proc. Suppl. 140 (2005) 626, [arXiv:hep-lat/0410024];
E.M. Ilgenfritz et al., Phys. Rev. D76 (2007) 034506, [arXiv:0705.0018].
* [24] A. A. Belavin, V. A. Fateev, A. S. Schwarz and Yu. S. Tyupkin, Phys. Lett. B83 (1979);
V.A. Fateev, I.V. Frolov, A.S. Shvarts, Nucl. Phys. B154 (1979)1.
* [25] V.G. Bornyakov et al., Phys. Lett., B537 (2002) 291, [arXiv:hep-lat/0103032].
| arxiv-papers | 2008-06-18T09:24:46 | 2024-09-04T02:48:56.327124 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "A. Di Giacomo, V.I. Zakharov",
"submitter": "Adriano Di Giacomo",
"url": "https://arxiv.org/abs/0806.2938"
} |
0806.2940 | # Construction of A Lattice on the completion space of an algebra and an
isomorphism to its Caratheodory Extension
Jun Tanaka University of California, Riverside, USA juntanaka@math.ucr.edu,
yonigeninnin@gmail.com, junextension@hotmail.com and Peter F. Mcloughlin
Crafton Hills College, USA pmcloughlin@aol.com
(Date: August, 10, 2008)
###### Abstract.
In this paper, we will show how one is able to construct a lattice on the
completion of an algebra and to obtain an isomorphism to its Caratheodory
Extension. In addition, it will be shown that the lattice form a
$\sigma$-algebra and a complete Heyting algebra of countable type.
###### Key words and phrases:
Caratheodory Extension Theorem, Lattice Theory, Measure Theory
###### 2000 Mathematics Subject Classification:
Primary: 28A12, 28B20
## 1\. Introduction
The Caratheodory Extension Theorem is a crucial part of any advanced Real
Analysis course; the process extends an algebra to a $\sigma$-algebra and a
measure on an algebra to a measure on a $\sigma$-algebra. This is a powerful
tool in both measure theory and statistics (see for example, [5],[2]). A
Boolean Algebra structure on the Caratheodory Extension was discussed in
several papers (e.g. Kolmogorov, [8], Coquand and Palmgren [3]). Furthermore,
the Caratheodory Extension is complete with a pseudometric
$d(.,.)=\mu^{\ast}(.\triangle.)$ where $\mu^{\ast}(.)$ is the outer measure on
the power set P(X) (Please refer to [4]). The Caratheodory Extension has
various rich structures which are open to further investigation.
In this paper, we will show how the Caratheodory Extension process is
intimately related to the metric completion process. In particular, it will be
shown how one is able to construct a lattice on the completion of an algebra
and to obtain an isomorphism to its Caratheodory Extension. The author
believes this method to be new, intuitive, and constructive. Especially
noteworthy is that, in Definition 4.14 of this paper, we give a precise
definition of $\vee^{\infty}$ on the completion space
$(\bar{d},\overline{\Omega}$) of an algebra $\Omega$ and show that the lattice
is a complete Heyting algebra of countable type.
Let X be an arbitrary nonempty set, and let $\mu$ be a finite measure on an
algebra $\Omega\subset\mathbf{P}(X)$. Note that
$d(.,.)=\mu^{\ast}(.\triangle.)$ defines a pseudometric on $\Omega$. Denote by
$(\bar{d},\overline{\Omega})$ the completion of $(d,\Omega)$. Also, let $S$ be
the set of all Cauchy sequences in $(d,{\Omega})$. By the metric completion
procedures, we know
$\bar{d}(\\{\overline{B^{\alpha_{1}}_{n}}\\},\\{\overline{B^{\alpha_{2}}_{n}}\\})=\lim
d({B^{\alpha_{1}}_{n}},{B^{\alpha_{2}}_{n}})$ where
$\\{{B^{\alpha_{i}}_{n}}\\}$ is in $S$ with
$\\{\overline{B^{\alpha_{i}}_{n}}\\}=\\{\\{B_{n}\\}\in S:\lim
d(B_{n},B_{n}^{\alpha_{i}})=0\\}$ for $i=1$or 2.
###### Definition 1.1.
Let
$\bar{\mu}(\\{\overline{B^{\alpha}_{n}}\\})=\bar{d}(\\{\overline{B^{\alpha}_{n}}\\},\\{\bar{\phi}\\})=\lim\mu({B^{\alpha}_{n}})$.
###### Definition 1.2.
If $B_{n}\in\Omega$, then $\\{B_{n}\\}$ is called a $\mu$-Cauchy sequence if
$\mu(B_{n}\triangle B_{m})\rightarrow 0$ as n,m $\rightarrow\infty$.
###### Definition 1.3.
For A, B in $\mathbf{P}(X)$, A = B a.e. if $\mu^{\ast}(A\triangle B)=0.$
In Section 4, we will show that a lattice structure can be naturally defined
on $\overline{\Omega}$ which makes it a $\sigma$-algebra. Let
$\widetilde{\mathbf{S}}$ = $\\{S\in\mathbf{P}(X)\ |$ $\exists\ \mu$-Cauchy
sequence $\\{B_{n}\\}$ s.t. $\lim\mu^{\ast}(B_{n}\triangle S)=0\\}$. In [9],
we proved that $\widetilde{\mathbf{S}}$ is a $\sigma$-algebra where, for any
$\mu$-Cauchy sequence $\\{B_{n}\\}$ such that $\lim\mu^{\ast}(B_{n}\triangle
S)=0$, the measure $\widetilde{\mu}(S)$ on $\widetilde{\mathbf{S}}$ is defined
as $\widetilde{\mu}(S)$ = $\lim\mu(B_{n})$. In addition, we proved that
$\widetilde{\mu}$ is a countably additive measure on $\widetilde{\mathbf{S}}$.
Thus, ($\widetilde{\mu}$, $\widetilde{\mathbf{S}}$) is a measure space. We
showed that the Caratheodory Extension of $\Omega$ can be expressed as the set
of limit points of $\mu$-Cauchy sequences under the pseudometric $d(A,B)$ =
$\mu^{\ast}(A\triangle B)$. Moreover, when the measure is a $\sigma$-finite
measure, we obtained an equivalent expression of the Caratheodory Extension,
$\widetilde{\mathbf{S}}$ = $\\{S\in\mathbf{P}(X)$ $|$ $\exists\ \mu$-Cauchy
sequence $\\{B_{n}\\}$ s.t. $\lim\mu^{\ast}(B_{n}\triangle S)=0\\}$. Theorem 2
in [9] shows that $E$ is a measurable set iff $E$ is in
$\widetilde{\mathbf{S}}$. Thus, the measure space ($\widetilde{\mu}$,
$\widetilde{\mathbf{S}}$) agrees with the Caratheodory Extension when $\mu$ is
a finite measure. Moreover, it shows that measurable sets are exactly limit
points of $\mu$-Cauchy sequences. The $\sigma$-finite case follows from the
finite case. In Section 5, we will define a $\sigma$-algebra lattice
isomorphism between two $\sigma$-algebras, and define a map $F:$
$\overline{\Omega}\rightarrow\mathbf{P}(X)$ given by
$F(\overline{\\{B_{n}\\}})=B$ where $\lim\mu^{\ast}(B_{n}\triangle B)=0$. We
will show that $F$ is an isometry and a $\sigma$-algebra lattice isomorphism
between the completion $\overline{\Omega}$ and the Caratheodory Extension of
$\Omega$ under the equivalence relation $\sim$ defined as A $\sim$ B iff
$\mu^{\ast}(A\triangle B)$ =0.
## 2\. Main Definitions and Notations
Our notation agrees with that in [9]. For emphasis, let X be a nonempty set.
Throughout the paper, unless otherwise stated, $\mu$ will be a finite measure
on an algebra $\Omega\subset\mathbf{P}(X)$. Unless otherwise stated,
$\\{B^{\alpha}_{n}\\}$ will be a $\mu$-Cauchy sequence.
In addition, for each $\mu$-Cauchy sequence $\\{B_{n}\\}$, a capital letter
corresponds to the limit point of the $\mu$-Cauchy sequence. For instance,
$\lim\mu^{\ast}(A_{n}\triangle A)=0$, $\lim\mu^{\ast}(B_{n}\triangle B)=0$,
$\lim\mu^{\ast}(Y_{L}\triangle Y)=0$. There exist such limit points in
$\mathbf{P}(X)$ since the Caratheodory Extension is complete under the
pseudometric d $($[4]$)$.
We recall that by a $\sigma$-algebra of subsets of X, we mean a family
$\Omega$ of subsets such that (1) every countable union of sets in $\Omega$ is
in $\Omega$, and (2) $A^{C}$ is in $\Omega$ whenever A is.
###### Definition 2.1.
Define a map $F:$ $\overline{\Omega}\rightarrow\mathbf{P}(X)$ given by
$F(\overline{\\{B_{n}\\}})=B$
where $\lim\mu^{\ast}(B_{n}\triangle B)=0$. Note that such a B always exists
in $\mathbf{P}(X)$ since the Caratheodory Extension is complete $($[4]$)$.
###### Definition 2.2.
$\sigma$-algebra lattice isomorphism
Suppose lattices $($X, $\vee_{X}$,$\wedge_{X}$ $)$ and $($Y,
$\vee_{Y}$,$\wedge_{Y}$ $)$ are $\sigma$-algebras, and H: X $\rightarrow$ Y is
a one-to-one, onto well-defined map. Then H is called a $\sigma$-algebra
lattice isomorphism if
$\displaystyle H(\cdot\vee_{X}\cdot)=$ $\displaystyle
H(\cdot)\vee_{Y}H(\cdot),\ \ \ \ \ \ \ H(\vee^{\infty}_{X}\ \cdot)$
$\displaystyle=$ $\displaystyle\vee^{\infty}_{Y}H(\cdot),$ $\displaystyle
H(\cdot\wedge_{X}\cdot)=$ $\displaystyle H(\cdot)\wedge_{Y}H(\cdot),\ \ \ \ \
\ \ H(\cdot^{\textbf{C}})$ $\displaystyle=$ $\displaystyle
H(\cdot)^{\textbf{C}}.$
We will show that F is a $\sigma$-algebra lattice isomorphism between
$\overline{\Omega}$ and $\widetilde{\mathbf{S}}_{\diagup_{\sim}}$ where $\sim$
is an equivalence relation in the sense of Definition 1.3.
## 3\. Preliminaries
In this section, we shall briefly review the well-known facts about lattice
theory (e.g. Birkhoff [1], Iwamura [7]), propose an extension lattice, and
investigate its properties, as well as some lemmas from [9]. A nonempty set L
is called a lattice if L is closed under the operations of meet ($\wedge$) and
join ($\vee$). It is denoted by (L,$\wedge$,$\vee$) or simply L. If it
satisfies, in addition, the distributive law, then it is called a distributive
lattice. For two lattices L and $L^{\prime}$, a bijection from L to
$L^{\prime}$, which preserves lattice operations is called a lattice
isomorphism, or simply an isomorphism. If there is an isomorphism from L to
$L^{\prime}$, then L is called lattice-isomorphic with $L^{\prime}$, and we
write L $\cong$ $L^{\prime}$. We write x $\leq$ y if x $\wedge$ y = x or,
equivalently, if x $\vee$ y = y. L is called complete if, for any subset A of
L, L contains the supremum $\vee$A and the infimum $\wedge$A, with respect to
the above order. A complete lattice L includes the maximum and minimum
elements, which are denoted by I and O, or 1 and 0, respectively. A
distributive lattice is called a Boolean algebra or a Boolean lattice, if, for
any element x in L, there exists a unique complement $x^{C}\in L$ such that x
$\vee$ $x^{C}$ = 1 and $x\wedge x^{C}$ = 0. Let L be a lattice and
$\cdot^{c}$: L $\rightarrow$ L be an operator. Then $\cdot^{c}$ is called a
lattice complement in L if the following conditions are satisfied:
(1) $\forall x$ $\in$ L, $x^{C}\wedge x$ = 0, $x^{C}\vee x$ = I,
(2) $\forall$ x, y $\in$ L, x $\leq$ y $\Rightarrow$ $x^{C}\geq y^{C}$,
(3) $\forall$ x $\in$ L, $(x^{C})^{C}$ = x.
###### Definition 3.1.
A complete lattice L is called a lattice $\sigma$-algebra or simply a
$\sigma$-algebra if the following conditions are satisfied:
(1) $\forall$ A $\in$ L, $A^{C}\in$L where $\cdot^{C}$ is a lattice
complement,
(2) $\forall$ $A_{i}\in$ L, $\vee^{\infty}A_{i}\in$ L.
###### Definition 3.2.
A complete lattice is called a complete Heyting algebra $($cHa$)$, if
$\displaystyle\vee_{i\in I}\ (x_{i}\wedge y)=(\vee_{i\in I}\ x_{i})\wedge y$
holds for $\forall x_{i},y\in L$ $(i\in I)$; where I is an index set of
arbitrary cardinal number. In the case that the cardinality of I is countable,
it is called a complete Heyting Algebra of countable type.
It is well-known that for a set E, $|P(E)|$= $2^{|E|}$. The set of all subsets
of E is a Boolean algebra.
###### Lemma 3.3.
If $\\{B_{n}\\}$ is a $\mu$-Cauchy sequence, then $\\{\mu(B_{n})\\}$ is a
Cauchy sequence of real numbers.
###### Proof.
Note $B_{n}\subseteq(B_{n}\triangle B_{m})\cup B_{m}$ and
$B_{m}\subseteq(B_{n}\triangle B_{m})\cup B_{n}$ implies
$\mu(B_{n})\leq\mu(B_{n}\triangle B_{m})+\mu(B_{m})$ and
$\mu(B_{m})\leq\mu(B_{n}\triangle B_{m})+\mu(B_{n})$. Hence,
$|\mu(B_{n})-\mu(B_{m})|\leq\mu(B_{n}\triangle B_{m})$. ∎
###### Lemma 3.4.
$d(A_{1}\cup A_{2},A_{3}\cup A_{4})\leq d(A_{1},A_{3})+d(A_{2},A_{4})$ for any
$A_{i}\in\Omega$.
###### Proof.
The proof is in [9] ∎
###### Lemma 3.5.
$\\{B^{\alpha}_{n}\cup B^{\gamma}_{n}\\}$ and $\\{(B^{\alpha}_{n})^{C}\\}$ are
$\mu$-Cauchy sequences.
###### Proof.
By Lemma 3.4,
$d(B^{\alpha}_{n}\cup B^{\gamma}_{n},B^{\alpha}_{m}\cup B^{\gamma}_{m})\leq
d(B^{\alpha}_{n},B^{\alpha}_{m})+d(B^{\gamma}_{n},B^{\gamma}_{m}).$
$\displaystyle\mbox{Moreover},\,\,d((B^{\alpha}_{n})^{C},(B^{\alpha}_{m})^{C})=\mu((B^{\alpha}_{n})^{C}\triangle(B^{\alpha}_{m})^{C})=\mu(B^{\alpha}_{n}\Delta
B^{\alpha}_{m})=d(B^{\alpha}_{n},B^{\alpha}_{m}).$ $\displaystyle\mbox{Hence,
the claim follows}.$
∎
## 4\. The Completion Space $(\bar{d},\overline{\Omega}$) and the Lattice
(L,$\wedge$,$\vee$)
In this section, we introduce the completion of ($d$,$\Omega$) where $\Omega$
is an algebra of subsets of X.
###### Definition 4.1.
Let $E_{\alpha}=\\{\overline{B^{\alpha}_{n}}\\}$ and
$E_{A}=\overline{\\{A\\}}$ when $A\in\Omega$ and $\\{B^{\alpha}_{n}\\}$ is a
sequence in $\Omega$.
###### Definition 4.2.
Define $E_{\alpha}\vee E_{\gamma}=\\{\overline{B^{\alpha}_{n}\cup
B^{\gamma}_{n}}\\}$, $E_{\alpha}\wedge
E_{\gamma}=\\{\overline{B^{\alpha}_{n}\cap B^{\gamma}_{n}}\\}$, and
$E^{C}_{\alpha}=\\{\overline{(B^{\alpha}_{n})^{C}}\\}$.
###### Remark 4.3.
$E_{\alpha}\wedge E_{\gamma}=E_{\phi}$ iff $\bar{\mu}(E_{\alpha}\wedge
E_{\gamma})=0$ iff $\lim\mu(B^{\alpha}_{n}\cap B^{\gamma}_{n})=0$. Moreover,
$E_{\alpha}\subset E_{\gamma}$ iff $\bar{\mu}(E_{\alpha}\wedge
E^{C}_{\gamma})=0$. The containment may be written in the usual lattice sense;
namely, $E_{\alpha}\wedge E_{\gamma}=E_{\alpha}$.
###### Lemma 4.4.
The map
$\vee:\overline{\Omega}\times\overline{\Omega}\rightarrow\overline{\Omega}$
defined by $\vee(E_{\alpha}\times E_{\gamma})=E_{\alpha}\vee
E_{\gamma}=\\{\overline{B^{\alpha}_{n}\cup B^{\gamma}_{n}}\\}$ is well-
defined.
###### Proof.
Suppose $\\{\overline{B^{\alpha}_{n}}\\}=\\{\overline{A^{\alpha}_{n}}\\}$ and
$\\{\overline{A^{\gamma}_{n}}\\}=\\{\overline{B^{\gamma}_{n}}\\}$.
Then $\lim d(B^{\alpha}_{n},A^{\alpha}_{n})=\lim
d(B^{\gamma}_{n},A^{\gamma}_{n})=0$. Also, by Lemma 3, $\\{B^{\alpha}_{n}\cup
B^{\gamma}_{n}\\}$ and $\\{A^{\alpha}_{n}\cup A^{\gamma}_{n}\\}$ are
$\mu$-Cauchy sequences.
Hence, by Lemma 3.4, $d(B^{\alpha}_{n}\cup B^{\gamma}_{n},A^{\alpha}_{n}\cup
A^{\gamma}_{n})\leq
d(B^{\alpha}_{n},A^{\alpha}_{n})+d(B^{\gamma}_{n},A^{\gamma}_{n})$ and the
claim follows. ∎
###### Lemma 4.5.
The map $\cdot{}^{C}:\overline{\Omega}\rightarrow\overline{\Omega}$ defined by
$(\\{\overline{B^{\alpha}_{n}}\\})^{C}=\overline{\\{(B^{\alpha}_{n})^{C}\\}}$
is well-defined.
###### Proof.
Suppose $\\{\overline{B^{\alpha}_{n}}\\}=\\{\overline{A^{\alpha}_{n}}\\}$.
Then $\lim d(B^{\alpha}_{n},A^{\alpha}_{n})=0$. Now,
$d((B^{\alpha}_{n})^{C},(A^{\alpha}_{n})^{C})=\mu((B^{\alpha}_{n})^{C}\triangle(A^{\alpha}_{n})^{C})=\mu(B^{\alpha}_{n}\triangle
A^{\alpha}_{n})=d(B^{\alpha}_{n},A^{\alpha}_{n})$ which implies
$\\{(B^{\alpha}_{n})^{C}\\}\sim\\{(A^{\alpha}_{n})^{C}\\}$. ∎
###### Lemma 4.6.
The map
$\wedge:\overline{\Omega}\times\overline{\Omega}\rightarrow\overline{\Omega}$
defined by $\wedge(E_{\alpha}\times E_{\gamma})=E_{\alpha}\wedge
E_{\gamma}=\\{\overline{B^{\alpha}_{n}\cap B^{\gamma}_{n}}\\}$ is well-
defined.
###### Proof.
The proof follows immediately from the two previous lemmas. ∎
###### Lemma 4.7.
$\cdot{}^{C}$ in Lemma 4.5 is a lattice complement.
###### Proof.
The proof is obvious. ∎
###### Lemma 4.8.
$\bar{\mu}(E_{\alpha}\vee
E_{\gamma})\leq\bar{\mu}(E_{\alpha})+\bar{\mu}(E_{\gamma})$ for
$E_{\alpha},E_{\gamma}\in\overline{\Omega}$.
###### Proof.
$E_{\alpha}=\\{\overline{B^{\alpha}_{n}}\\}$ and
$E_{\gamma}=\\{\overline{B^{\gamma}_{n}}\\}$ implies $E_{\alpha}\vee
E_{\gamma}=\\{\overline{B^{\alpha}_{n}\cup B^{\gamma}_{n}}\\}$. Hence
$\bar{\mu}(E_{\alpha}\vee E_{\gamma})=\lim\mu(B^{\alpha}_{n}\cup
B^{\gamma}_{n})\leq\lim[\mu(B^{\alpha}_{n})+\mu(B^{\gamma}_{n})]=\lim\mu(B^{\alpha}_{n})+\lim\mu(B^{\gamma}_{n})\leq\bar{\mu}(E_{\alpha})+\bar{\mu}(E_{\gamma})$.
∎
###### Lemma 4.9.
If $E_{\alpha_{1}}$ and $E_{\alpha_{2}}$ are disjoint, then
$\bar{\mu}(E_{\alpha_{1}}\vee
E_{\alpha_{2}})=\bar{\mu}(E_{\alpha_{1}})+\bar{\mu}(E_{\alpha_{2}})$.
###### Proof.
By Remark 4.3, we must have $\lim\mu(B^{\alpha_{1}}_{n}\cap
B^{\alpha_{2}}_{n})=0$. Note, $B^{\alpha_{i}}_{n}=(B^{\alpha_{i}}_{n}\cap
B^{\alpha_{j}}_{n})\ \cup\ (B^{\alpha_{i}}_{n}\cap(B^{\alpha_{j}}_{n})^{C})$
implies $\mu(B^{\alpha_{i}}_{n})=\mu(B^{\alpha_{i}}_{n}\cap
B^{\alpha_{j}}_{n})+\mu(B^{\alpha_{i}}_{n}\cap(B^{\alpha_{j}}_{n})^{C})$ and
$\lim\mu(B^{\alpha_{i}}_{n})=\lim\mu(B^{\alpha_{i}}_{n}\cap(B^{\alpha_{j}}_{n})^{C})$
for $\\{i,j\\}=\\{1,2\\}$.
Moreover, $\mu(B^{\alpha_{1}}_{n}\cup B^{\alpha_{2}}_{n})=\mu(\
(B^{\alpha_{1}}_{n}\cap B^{\alpha_{2}}_{n})\ \cup\
(B^{\alpha_{1}}_{n}\cap(B^{\alpha_{2}}_{n})^{C})\ \cup\
(B^{\alpha_{2}}_{n}\cap(B^{\alpha_{1}}_{n})^{C})\ )=\mu(B^{\alpha_{1}}_{n}\cap
B^{\alpha_{2}}_{n})+\mu(B^{\alpha_{1}}_{n}\cap(B^{\alpha_{2}}_{n})^{C})+\mu(B^{\alpha_{2}}_{n}\cap(B^{\alpha_{1}}_{n})^{C})$
implies $\bar{\mu}(E_{\alpha_{1}}\vee
E_{\alpha_{2}})=\lim\mu(B^{\alpha_{1}}_{n}\cup
B^{\alpha_{2}}_{n})=\lim\mu(B^{\alpha_{1}}_{n})+\lim\mu(B^{\alpha_{2}}_{n})=\bar{\mu}(E_{\alpha_{1}})+\bar{\mu}(E_{\alpha_{2}})$.
∎
###### Theorem 4.10.
$\overline{\Omega}$ is an algebra and $\bar{\mu}$ is a measure on
$\overline{\Omega}$.
###### Proof.
By Lemmas 3.5, 4.6 and 4.5, $\overline{\Omega}$ is an algebra. Moreover, the
fact that $\bar{\mu}$ takes on values in $\mathbb{R}^{+}$ along with Lemmas 6
and 7 imply $\bar{\mu}$ is a measure on $\overline{\Omega}$. ∎
###### Remark 4.11.
$\bar{d}(E_{\alpha},E_{\gamma})=\lim
d(B^{\alpha}_{n},B^{\gamma}_{n})=\lim[\mu(B^{\alpha}_{n}\cap(B^{\gamma}_{n})^{C})+\mu(B^{\gamma}_{n}\cap(B^{\alpha}_{n})^{C})]=\lim\mu(B^{\alpha}_{n}\cap(B^{\gamma}_{n})^{C})+\lim\mu(B^{\gamma}_{n}\cap(B^{\alpha}_{n})^{C})=\bar{\mu}(E_{\alpha}\wedge(E_{\gamma})^{C})+\bar{\mu}(E_{\gamma}\wedge(E_{\alpha})^{C})=\bar{\mu}(E_{\alpha}\triangle
E_{\gamma})$.
###### Lemma 4.12.
Let $E_{i}$ = $\overline{\\{B^{i}_{n}\\}}$ $\in\overline{\Omega}$ for i $\geq$
1 and by following the proof of Lemma 8 in [9], construct $Y_{L}$ =
$\cup_{i=1}^{N_{L}}B^{i}_{K_{L}}$ for each $L$ such that
$\mu^{\ast}(\cup_{i=1}^{\infty}S_{i}\triangle\cup_{i=1}^{N_{L}}B^{i}_{K_{L}})\leq\mu^{\ast}(\cup_{i=N_{L}+1}^{\infty}S_{i})+\mu^{\ast}(\cup_{i=1}^{N_{L}}S_{i}\triangle\cup_{i=1}^{N_{L}}B^{i}_{K_{L}})<\frac{1}{L}.$
Then E := $\overline{\\{Y_{L}\\}}$ satisfies the following conditions;
$\displaystyle 1.$ $\displaystyle\vee^{n}_{i=1}E_{i}\leq E\ \ \mbox{for all}\
n,$ $\displaystyle 2.$
$\displaystyle\lim\bar{\mu}(E\wedge(\vee^{n}_{i=1}E_{i})^{C})=0.$
In particular, E is uniquely determined in $\overline{\Omega}$.
###### Proof.
Note that $E_{i}$ = $\overline{\\{B^{i}_{n}\\}}$ =
$\overline{\\{B^{i}_{K_{L}}\\}}$.
$(\vee_{i=1}^{n}E_{i})\wedge\overline{\\{Y_{L}\\}}$ =
$\overline{\\{\cup_{i=1}^{n}B^{i}_{K_{L}}\cap Y_{L}\\}}$ =
$\overline{\\{\cup_{i=1}^{n}B^{i}_{K_{L}}\\}}$ = $\vee_{i=1}^{n}E_{i}$ for any
n.
Let $N_{L}>n$.
$\displaystyle\mu(\cup_{i=1}^{N_{L}}B^{i}_{K_{L}}\cap(\cup_{i=1}^{n}B^{i}_{K_{L}})^{\textbf{C}})=$
$\displaystyle\mu(\cup_{i=1}^{N_{L}}B^{i}_{K_{L}}\triangle\cup_{i=1}^{n}B^{i}_{K_{L}})=\mu^{\ast}(\cup_{i=1}^{N_{L}}B^{i}_{K_{L}}\triangle\cup_{i=1}^{n}B^{i}_{K_{L}})$
$\displaystyle\leq$
$\displaystyle\mu^{\ast}(\cup_{i=1}^{\infty}S_{i}\triangle\cup_{i=1}^{N_{L}}B^{i}_{K_{L}})+\mu^{\ast}(\cup_{i=1}^{\infty}S_{i}\triangle\cup_{i=1}^{n}B^{i}_{K_{L}}).$
This implies that
$\lim\overline{\mu}(\overline{\\{Y_{L}\\}}\wedge(\vee_{i=1}^{n}E_{i})^{\textbf{C}})$
= 0.
Suppose $E,E^{\prime}\in\overline{\Omega}$ both satisfy condition 1 and 2
above. Then $\vee^{n}_{i=1}E_{i}\leq E\wedge E^{\prime}\leq E$ for all n.
Hence, we must have $E=E\wedge E^{\prime}$. Similarly, we must have
$E^{\prime}=E\wedge E^{\prime}$. Therefore E is unique.
∎
###### Lemma 4.13.
Let $E_{i}$ = $\overline{\\{B^{i}_{n}\\}}$ $\in\overline{\Omega}$ for i $\geq$
1 and $E_{i}\wedge E_{j}=E_{\phi}$ for $i\neq j$. Then
$\bar{\mu}(E)=\sum^{\infty}_{i=1}\bar{\mu}(E_{i})$ where E is defined as in
Lemma 4.12.
###### Proof.
Suppose that $E_{i}$ = $\overline{\\{B^{i}_{n}\\}}$ $\in\overline{\Omega}$ for
i $\geq$ 1 and $E_{i}\wedge E_{j}=E_{\phi}$ for $i\neq j$. Then it follows
that
$\bar{\mu}(\vee^{n}_{i=1}E_{i})=\sum^{k}_{i=1}\bar{\mu}(E_{i})\leq\bar{\mu}(E)$
for all $k$.
Hence, we must have
(1) $\sum^{\infty}_{i=1}\bar{\mu}(E_{i})\leq\bar{\mu}(E).$
Now since
$\bar{\mu}(E)=\bar{\mu}(\vee^{n}_{i=1}E_{i})+\bar{\mu}(E\wedge(\vee^{n}_{i=1}E_{i})^{C})$
for all n and $\lim\bar{\mu}(E\wedge(\vee^{n}_{i=1}E_{i})^{C})=0$ , we must
have $\bar{\mu}(E)=\sum^{\infty}_{i=1}\bar{\mu}(E_{i})$ by (1).
∎
###### Definition 4.14.
For ${E}_{i}$ in $\overline{\Omega}$, we define $\vee^{\infty}_{i=1}{E}_{i}=E$
where $E$ is as in Lemma 4.12.
###### Theorem 4.15.
$(\bar{\mu},\overline{\Omega})$ is a measure space.
###### Proof.
This follows by Theorem 4.10, 4.12, and 4.13, and Definition 4.14. ∎
## 5\. The Isomorphism
In this section, we introduce an isomorphism $F:$
$\overline{\Omega}\rightarrow\mathbf{P}(X)$ given by
$F(\overline{\\{B_{n}\\}})=B$.
###### Lemma 5.1.
$F$ : $\overline{\Omega}\rightarrow\widetilde{\mathbf{S}}_{\diagup_{\sim}}$ is
a well-defined map.
###### Proof.
Suppose that $\overline{\\{A_{n}\\}}=\overline{\\{B_{n}\\}}$.
There exist A and B in $\mathbf{P}(X)$ such that
$\lim\mu^{\ast}(A_{n}\triangle A)=0$ and $\lim\mu^{\ast}(B_{n}\triangle B)=0$.
$\mu^{\ast}(A\triangle B)\leq$ $\mu^{\ast}(A\triangle
A_{n})+\mu^{\ast}(A_{n}\triangle B_{n})+\mu^{\ast}(B_{n}\triangle B)$ by the
triangle inequality.
By taking limits on both sides, $\mu^{\ast}(A\triangle B)=0$.
Thus, A = B a.e.. Therefore, F is well-defined.
∎
###### Theorem 5.2.
F is an isometry between $\overline{\Omega}$ and
$\widetilde{\mathbf{S}}_{\diagup_{\sim}}$.
###### Proof.
First, we show F is onto $\widetilde{\mathbf{S}}$. Let
$X\in\widetilde{\mathbf{S}}$. Then there exists a $\mu$-Cauchy sequence
$\\{B_{n}\\}$ such that $\lim\mu^{\ast}(B_{n}\triangle X)=0$. Thus
$F(\overline{\\{B_{n}\\}})=X$ a.e.. Therefore, F is onto.
Second, we will show F preserves the metric. Let $\overline{\\{A_{n}\\}}$,
$\overline{\\{B_{n}\\}}$ $\in$ $\overline{\Omega}$. Letting A and B be as
before, we have
$\lim\mu(A_{n}\triangle B_{n})=\mu^{\ast}(A\triangle B).$
Therefore, $\overline{d}(\overline{\\{A_{n}\\}},\overline{\\{B_{n}\\}})$ =
$\lim\mu(A_{n}\triangle B_{n})$ = $\mu^{\ast}(A\triangle B)$
= $\mu^{\ast}(F(\overline{\\{A_{n}\\}})\triangle F(\overline{\\{B_{n}\\}}))$ =
$d(F(\overline{\\{A_{n}\\}}),F(\overline{\\{B_{n}\\}}))$. Thus, F preserves
the metric.
Lastly, we will show that F is one-to-one. Let
$F(\overline{\\{A_{n}\\}}),F(\overline{\\{B_{n}\\}})\in\widetilde{\mathbf{S}}$
such that $F(\overline{\\{A_{n}\\}})=F(\overline{\\{B_{n}\\}})$ a.e..
Then $A=B$ a.e. implies $\mu^{\ast}(A\triangle B)=0$. Then, as in the proof of
F being onto, $\lim\mu(A_{n}\triangle B_{n})$ = $\mu^{\ast}(A\triangle B)$.
Thus $\overline{\\{A_{n}\\}}$ = $\overline{\\{B_{n}\\}}$ and F is one-to-one.
Therefore, F is an isometry between $\overline{\Omega}$ and
$\widetilde{\mathbf{S}}_{\diagup_{\sim}}$.
∎
###### Theorem 5.3.
F is a $\sigma$-algebra lattice isomorphism between $\overline{\Omega}$ and
$\widetilde{\mathbf{S}}_{\diagup_{\sim}}$.
###### Proof.
We already showed that F is a one-to-one, onto map in Theorem 5.2.
$F(\overline{\\{A_{n}\\}}\vee\overline{\\{B_{n}\\}})$ =
$F(\overline{\\{A_{n}\cup B_{n}\\}})$ = $F(\overline{\\{A_{n}\\}})\cup
F(\overline{\\{B_{n}\\}})$ a.e. since $\lim\mu(A_{n}\cup B_{n}\ \ \triangle\ \
A\cup B)$ = 0.
Let $\overline{\\{B_{n}\\}}\in\overline{\Omega}$.
Then,
$F(\overline{\\{B_{n}\\}}^{\textbf{C}})=F(\overline{\\{(B_{n})^{\textbf{C}}\\}})=B^{\textbf{C}}\
\ a.e..$
Thus, $F(\cdot^{\textbf{C}})$ = $F(\cdot)^{\textbf{C}}$ in
$\widetilde{\mathbf{S}}_{\diagup_{\sim}}$.
Similarly, $F(\cdot\wedge\cdot)$ =
$F(\cdot^{\textbf{C}}\vee\cdot^{\textbf{C}})^{\textbf{C}}$ =
$[F(\cdot^{\textbf{C}})\cup F(\cdot^{\textbf{C}})]^{\textbf{C}}$=
$F(\cdot)\cap F(\cdot)$.
Let $E_{\alpha_{i}}$ $\in\overline{\Omega}$ for i $\geq$ 1 and
$E_{\alpha_{i}}$ = $\overline{\\{B^{\alpha_{i}}_{n}\\}}$.
Then for each i, there exists an $S_{i}\in\widetilde{\mathbf{S}}$ such that
$\lim\mu^{\ast}(B^{\alpha_{i}}_{n}\triangle S_{i})=0$.
Now suppose we have $\\{Y_{L}\\}$ in the same manner as in Lemma 4.12. By
design, $\\{Y_{L}\\}$ converges to $\cup_{i=1}^{\infty}S_{i}$. Then
$\overline{\\{Y_{L}\\}}$ = $\vee_{i=1}^{\infty}E_{\alpha_{i}}$ by Lemma 4.12.
Now we have
$F(\vee_{i=1}^{\infty}E_{\alpha_{i}})=F(\vee_{i=1}^{\infty}\overline{\\{B^{\alpha_{i}}_{n}\\}})=F(\overline{\\{Y_{L}\\}})=Y.$
Since $\lim\mu^{\ast}(Y_{L}\triangle Y)=0$ and
$\lim\mu^{\ast}(Y_{L}\triangle\cup^{\infty}_{i=1}S_{i})=0$, we have $Y$ =
$\cup^{\infty}_{i=1}S_{i}$ a.e.. In addition, $\cup^{\infty}_{i=1}S_{i}$ =
$\cup^{\infty}_{i=1}F(\overline{\\{B^{\alpha_{i}}_{n}\\}})$.
Thus,
$F(\vee_{i=1}^{\infty}E_{\alpha_{i}})=\cup^{\infty}_{i=1}F(E_{\alpha_{i}}).$
Therefore, the claim follows.
∎
###### Theorem 5.4.
$(\overline{\Omega}$,$\wedge$,$\vee)$ is a complete Heyting algebra of
countable type. More precisely, it is the case that the cardinality of I in
Definition 3.2 is countable.
###### Proof.
Since $\widetilde{\mathbf{S}}$ is a complete Heyting algebra of countable
type, the proof follows from Theorem 5.3. ∎
## 6\. Conclusion
Theorems 5.2 and 5.3 show that the completion of $\Omega$ is isometric and
$\sigma$-algebra lattice isomorphic to
$\widetilde{\mathbf{S}}_{\diagup_{\sim}}$. Thus, by the conclusion in [9] the
completion of $\Omega$ is isometric to, as well as $\sigma$-algebra isomorphic
to, the Caratheodory Extension under the equivalence relation $\sim$ . An
important note is that the isomorphism shows that Caratheodory Extension
Theorem and the Metric Completion Process are essentially the same, differing
only on measure zero sets. In addition, the completion $\overline{\Omega}$ of
$\Omega$ is a complete Heyting algebra of countable type by Theorem 5.4. The
$\sigma$-finite case follows from the finite case.
## 7\. Acknowledgement
The first author would like to thank Professors who gave him very professional
advice and suggestions, which he truly believes improved the presentation of
this paper.
## References
* [1] G. Birkhoff, Lattice Theory, 3rd ed. AMS colloquim Publication, Providence, RI, 1967.
* [2] R.N. Bhattacharya and E.C.Waymire, Stochastic Processes With Applications, Wiley, New York, 1990.
* [3] T. Conquand and E. Palmgren, Metric boolean algebras and constructive measure theory, Arch Mathematical Logic 41 , (2002), 687-704
* [4] N. Dunford and J. T. Schwartz, Linear Operators Part I General Theory, Willy Interscience Publication, New York, 1988
* [5] T. R. Fleming and D. P. Harrington, Counting Processes And Survival Analysis, Wiley, New York, 1991.
* [6] N. Nakajima, A generalized fuzzy set and its representation, Fuzzy Computing, Elsevier Science Publishers B.V. , (1988), 139-153.
* [7] T. Iwamura, Sokuron, Kyoritsu Shuppan, Tokyo, 1966. (In Japanese)
* [8] A. N. Kolmogorov, Complete metric boolean algebra, Philosophical studies 77, (1995) 57-66
* [9] J. Tanaka and P. F. Mcloughlin, A realization of measurable sets as limit points, submitted
| arxiv-papers | 2008-06-18T09:34:10 | 2024-09-04T02:48:56.333350 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Jun Tanaka, Peter McLoughlin",
"submitter": "Jun Tanaka",
"url": "https://arxiv.org/abs/0806.2940"
} |
0806.2943 | # Modern Set
Jun Tanaka University of California, Riverside, USA juntanaka@math.ucr.edu,
yonigeninnin@gmail.com, junextension@hotmail.com
(Date: May, 25, 2008)
###### Key words and phrases:
Fuzzy set, L-fuzzy set, generalized fuzzy set
###### 2000 Mathematics Subject Classification:
Primary: 03E72
## 1\. Introduction
In this paper, we intend to generalize the classical set theory as much as
possible. we will do this by freeing sets from the regular properties of
classical sets; e.g., the law of excluded middle, the law of non-
contradiction, the distributive law, the commutative law,etc…. The fuzzy set
theory succeeded in freeing sets from the law of excluded middle and the law
of contradiction. However, in order to extend our language, it is more or less
unreasonable to keep the commutative law and the distributive law. This modern
idea of sets keeps the concept of membership functions but their value are not
necessarily in [0,1]; nor do these modern sets form a lattice necessarily.
Especially noteworthy is that modern sets are more general than generalized
fuzzy set, (Please refer to Nakajima [6]). Here is the hierarchy of
generality: Modern sets $\geq$ L-fuzzy set $\geq$ Generalized Fuzzy set $\geq$
fuzzy set $\geq$ classical set. These modern sets become the classical sets
under the restriction to $O_{x}$, $I_{x}$ for all x in X, and no further
conditions are required
The world of natural languages can not be ruled by a single classical logic
because the real physical world consists of many aspects, each of which obeys
a different logic. As is well-known, the distributive law does not always hold
in every logic system. For example, in classical and intuitionistic logic it
holds, but in quantum logic it fails.
The Commutative law seems invalid when we talk about anything with temporal
order in our daily conversation since every event in time is not reversible.
”He is a student and a male”. I can also say, ”He is a male and a student” but
mean the same thing. The two significations commute of course. ”I have a
health insurance and a car insurance”. In this case, the two significations
commute. But how about this case ”He went to a supermarket and then a drug
store”. In this case, the two significations do not commute anymore since
there is temporal order. Furthermore, we have no expression for ”He went to a
supermarket and then a drug store” in classical logic or fuzzy logic. In other
words, the properties from classical logic, which we take for granted, do not
express our thoughts that well.
## 2\. Preliminaries : Extension of Lattice
In this section, we shall briefly review the well-known facts about lattice
theory (e.g. Birkhoff [1], Iwamura [4] ), propose an extension lattice, and
investigate its properties. (L,$\wedge$,$\vee$) is called a lattice, if it is
closed under operations $\wedge$ and $\vee$, and satisfies, for any elements
x,y,z in L:
(L1) the commutative law: x $\wedge$ y = y $\wedge$ x and x $\vee$ y = y
$\vee$ x
(L2) the associative law:
$\displaystyle x\wedge(y\wedge z)=(x\wedge y)\wedge z\ \ \text{and}\ \
x\vee(y\vee z)=(x\vee y)\vee z$
(L3) the absorption law: x $\vee$ ( y $\wedge$ x ) =x and x $\wedge$ ( y
$\vee$ x ) = x.
Hereinafter, the lattice (L,$\wedge$,$\vee$) will often be written L for
simplicity.
A mapping h from a lattice L to another $L^{\prime}$ is called a lattice-
homomorphism, if it satisfies
$\displaystyle h(x\wedge y)=h(x)\wedge h(y)\ \ \text{and}\ \ h(x\vee
y)=h(x)\vee h(y),\forall x,y\in L.$
If h is a bijection, that is, if h is one-to-one and onto, it is called a
lattice-isomorphism; and in this case, $L^{\prime}$ is said to be lattice-
isomorphic to L.
A lattice (L,$\wedge$,$\vee$) is called distributive if, for any x,y,z in L,
(L4) the distributive law holds:
$\displaystyle x\vee(y\wedge z)=(x\vee y)\wedge(y\vee z)\ \ \text{and}\ \
x\wedge(y\vee z)=(x\wedge y)\vee(y\wedge z)$
A lattice L is called complete if, any subset A of L, L contain the supremum
$\vee$ A and the infimum $\wedge$ A. If L is complete, then L itself includes
the maximum and minimum elements, which are often denoted by 1 and 0, or I and
O, respectively.
###### Definition 1.
Complete Heyting algebra (cHa)
A complete lattice is called a complete Heyting algebra (cHa), if
$\displaystyle\vee_{i\in I}\ (x_{i}\wedge y)=(\vee_{i\in I}\ x_{i})\wedge y$
holds for $\forall x_{i},y\in L$ ($i\in I$); where I is an index set of
arbitrary cardinal number.
A distributive lattice is called a Boolean algebra or a Boolean lattice, if
for any element x in L, there exists a unique complement $x^{C}$ such that x
$\vee$ $x^{C}$ = 1 and $x\wedge x^{C}$ = 0.
It is well-known that for a set E, the power set P(E) = $2^{E}$. The set of
all subsets of E, is a Boolean algebra.
## 3\. Preliminaries : Generalized fuzzy sets
In this section we will consider an algebraic structure of a family of fuzzy
sets. We will show that the following three families are mutually equivalent:
a ring of generalized fuzzy subsets, an extension of (Boolean) Lattice P(X),
and a set of L-fuzzy sets (introduced by Gouen [3]).
###### Definition 2.
A family GF(X), which is closed under operations $\vee$ and $\wedge$, is
called a ring of generalized fuzzy subsets of X, if it satisfies:
(1) GF(X) is a complete Heyting algebra with respect to $\vee$ and $\wedge$,
(2) GF(X) contains P(X) = $2^{X}$ as a sublattice of GF(X),
(3) the operations $\vee$ and $\wedge$ coincide with set operations $\bigcup$
and $\bigcap$, respectively, in P(X), and
(4) for any element A in P(X), A$\vee$X= X and A$\wedge\emptyset=\emptyset$.
###### Definition 3.
Let $L_{x}$ be a lattice which is assigned to each x in X, and let L denote
$\\{L_{x}\mid x\in X\\}$. An L-fuzzy set A is characterized by an L-valued
membership function $\mu_{A}$ which associates to each point x in X an element
$\mu_{A}(x)$ in $L_{x}$.
###### Theorem 1.
If each $L_{x}$ is a cHa, then LF(x), the family of all L-fuzzy sets, is a
ring of generalized fuzzy subsets of X.
We will show that LF(X) satisfies all the conditions in Definition 2. First,
by the definition, LF(X) is closed under $\wedge$ and $\vee$. (1) LF(X) is a
cHa, because each $L_{x}$ is a cHa. Condition (2) LF(X) $\supseteqq$ P(X) and
(3) $\vee$ = $\cup$ and $\wedge$ = $\cap$ in P(X) follow from the fact that
any element of P(X) is defined with the element of LF(X) whose membership
function takes just two values, 1 and 0. (4) It follows from $\mu_{X}(x)$=1
and $\mu_{\emptyset}(x)$ = 0 that A $\cup$ X = X and A $\cap$ $\emptyset$ =
$\emptyset$, for any A in LF(X).
## 4\. Modern Sets
###### Definition 4.
Boolean Algebra
In conformity with Birkhoff’s book [1], the fundamental operations of
intersection and union of elements will be defined by
$\displaystyle x\ \cap\ y$ intersection $\displaystyle x\ \cup\ y$ union
As is well known, it follows from the definition of Boolean algebra that there
exists a unit element I and a null element O for which we have the following:
$\displaystyle x\cap I=x\ \text{and}$ $\displaystyle\ x\cap O=O$
$\displaystyle x\cup I=I\ \text{and}$ $\displaystyle\ x\cup O=x\ \ \ \forall
x\in X$
Note that $\cap$ and $\cup$ commute at this point.
###### Definition 5.
Weak Boolean algebra
Let H be an algebraic space with two distinct operators $\ast_{\wedge}$ ,
$\ast_{\vee}$ from H to itself. H is called a Weak Boolean Algebra if
$\exists$ distinct O, I $\in$ H such that
$\displaystyle O\ast_{\wedge}I=O\ \text{and}$ $\displaystyle\
I\ast_{\wedge}O=O\ \ \ \ \ \ \text{O and I commute}$ $\displaystyle
O\ast_{\wedge}O=O\ \text{and}$ $\displaystyle\ I\ast_{\wedge}I=I$
$\displaystyle O\ast_{\vee}I=I\ \text{and}$ $\displaystyle\ I\ast_{\vee}O=I\ \
\ \ \ \ \text{O and I commute}$ $\displaystyle O\ast_{\vee}O=O\ \text{and}$
$\displaystyle\ I\ast_{\vee}I=I$
Please note that $\ast_{\wedge}$ $\ast_{\vee}$ are associated with $\wedge$
and $\vee$, respectively. O and I are associated with the minimum element and
the maximum element, respectively.
###### Definition 6.
Modern Sets
Suppose a weak Boolean Algebra $H_{x}$ is assigned to each x in X and let H
denote $\\{H_{x}|x\in X\\}$. Each modern set A is characterized by a
membership function $\mu_{A}$ such that for each x $\in$ X, $\mu_{A}$ assigns
an element $\mu_{A}(x)\in H_{x}$. We define $H(X)$ as the family of all modern
sets. When A is a set in the ordinary sense of the term (in P(X)), its
membership function can take on only two values $O_{x}$ and $I_{x}$ with
$\mu_{A}(x)$ = $O_{x}$ or $I_{x}$ according to whether x does or does not
belong to A. The operations $\ast_{\vee_{x}}$ and $\ast_{\wedge_{x}}$ coincide
with the set operations $\bigcup$ and $\bigcap$, respectively, in P(X).
A modern set is empty if and only if its membership function is identically
$O_{x}$ for all x in X.
Two modern sets A and B are equal if and only if $\mu_{A}(x)$ = $\mu_{B}(x)$
for all x in X.
If $H_{x}$ is partially ordered by $\leq$ for each x, then we can define
containment as follows: A is contained in B if and only if
$\mu_{A}(x)\leq\mu_{B}(x)$ for all x in X.
Union. The union of two modern sets A and B with respective membership
functions $\mu_{A}(x)$ and $\mu_{B}(x)$ is a modern set, written as C = A
$\vee$ B, whose membership function is related to those of A and B by
$\mu_{C}(x)=\mu_{A}(x)\ast_{\vee_{x}}\mu_{B}(x),\ x\in X$
Note that the order does matter if $H_{x}$ is not commutative.
Intersection The intersection of two modern sets A and B with respective
membership functions $\mu_{A}(x)$ and $\mu_{B}(x)$ is a modern set, written as
C = A $\wedge$ B, whose membership function is related to those of A and B by
$\mu_{C}(x)=\mu_{A}(x)\ast_{\wedge_{x}}\mu_{B}(x),\ x\in X$
The notion of complement was not given in Definition 5 since we can trivially
define the complement on a Modern Set for each x in X such as
$\cdot{}^{C}:H_{x}\rightarrow H_{x}$ where
$(\\{O_{x}\\})^{C}=I_{x}\ \ \text{and}\ \ (\\{I_{x}\\})^{C}=O_{x}$
even while $(\\{A_{x}\\})^{C}$ can be anything for all $A_{x}\in
H_{x},A_{x}\neq O_{x},I_{x}$ such that $((\\{A_{x}\\})^{C})^{C}$ =
$\\{A_{x}\\}$
###### Theorem 2.
For any modern sets A, B , C whose membership function take on values $O_{x}$
or $I_{x}$, (L1)-(L4) hold.
###### Proof.
Since the proof is more or less clear, herein we will briefly indicate the
proof for the commutative and distributive cases. Let’s check the commutative
law. Call C = A $\vee$ B. We only have to check the four possible cases:
$\displaystyle O_{x}\ast_{\wedge}I_{x}=O_{x}\ \text{and}$ $\displaystyle\
I_{x}\ast_{\wedge}O_{x}=O_{x}$ $\displaystyle O_{x}\ast_{\wedge}O_{x}=O_{x}\
\text{and}$ $\displaystyle\ I_{x}\ast_{\wedge}I_{x}=I_{x}\ \ \text{for all x}\
\in X$
Thus, $\mu_{C}(x)=\mu_{A}(x)\ast_{\vee_{x}}\mu_{B}(x)$ =
$\mu_{B}(x)\ast_{\vee_{x}}\mu_{A}(x)$. Therefore, C = A $\vee$ B = B $\vee$ A.
We can show the commutative law for intersection similarly.
We will show the two most important cases of the distributive law briefly.
Call C = A $\vee$ ( B $\wedge$ C ).
$\displaystyle
O_{x}\ast_{\vee}(I_{x}\ast_{\wedge}O_{x})=O_{x}\ast_{\vee}O_{x}=O_{x}=I_{x}\ast_{\wedge}O_{x}=(I_{x}\ast_{\vee}O_{x})\ast_{\wedge}(O_{x}\ast_{\vee}O_{x})$
$\displaystyle
O_{x}\ast_{\vee}(I_{x}\ast_{\wedge}I_{x})=O_{x}\ast_{\vee}I_{x}=I_{x}=I_{x}\ast_{\wedge}I_{x}=(O_{x}\ast_{\vee}I_{x})\ast_{\wedge}(O_{x}\ast_{\vee}I_{x})$
$\displaystyle\text{for all x}\ \in X$
Please check the rest of cases. That should not be too difficult.
∎
###### Theorem 3.
$H_{x}$ is commutative for all x in X iff H = $\\{H_{x}\ |x\in X\\}$ is
commutative.
###### Proof.
It is obvious. ∎
###### Theorem 4.
$H_{x}$ is distributive for all x in X iff H = $\\{H_{x}\ |x\in X\\}$ is
distributive.
###### Proof.
It is obvious. ∎
###### Theorem 5.
Similar equivalent statement (as above) hold for the absorption law, the law
of excluded middle, the law of non-contradiction, the associative law, etc…
###### Remark 1.
As you see from the previous theorems, if ($H_{x}$, $\ast_{\vee_{x}}$,
$\ast_{\wedge_{x}}$, $\cdot{}^{C}$ ) is defined to be in the sense of fuzzy
sets such as $\ast_{\vee_{x}}$ = max, $\ast_{\wedge_{x}}$ = min, $(\cdot)^{C}$
= 1- ($\cdot$) in the interval [0,1], then $H(X)$ becomes the family of fuzzy
sets.
If $H_{x}$ satisfy the definition of Lattice for each x in X , then $H(X)$
becomes a family of L-fuzzy set. If $H_{x}$ additionally satisfy the
definition of a complete Heyting algebra, it is a ring of generalized fuzzy
subsets.
###### Proof.
The proof is clear. ∎
###### Example 6.
We will present an example of a non-commutative modern set. Let $H_{x}$ be a
space of linear bounded operators on a Hilbert space $\mathcal{H}$ for each x
in X. Then we take the zero $O_{x}$ and the identity $I_{x}$ in $H_{x}$. We
define the composition $\circ$ = $\ast_{\wedge_{x}}$ and the addition + =
$\ast_{\vee_{x}}$. In order to create a weak Boolean Algebra, we must define
an equivalence class $\sim$ as $A\sim B$ iff A = B or there exist n, m
$\in\mathbb{N}-\\{0\\}$ such that A = $nI_{x}$ and B = $mI_{x}$.
$\displaystyle O_{x}(I_{x})=O_{x}\ \text{and}$ $\displaystyle\
I_{x}(O_{x})=O_{x}\ \ \ \ \ \ O_{x}\ \text{and}\ I_{x}\ \text{commute}$
$\displaystyle O_{x}(O_{x})=O_{x}\ \text{and}$ $\displaystyle\
I_{x}(I_{x})=I_{x}$ $\displaystyle O_{x}+I_{x}=I_{x}\ \text{and}$
$\displaystyle\ I_{x}+O_{x}=I_{x}\ \ \ \ \ \ O_{x}\ \text{and}\ I_{x}\
\text{commute}$ $\displaystyle O_{x}+O_{x}=O_{x}\ \text{and}$ $\displaystyle\
I_{x}+I_{x}=2I_{x}=I_{x}$
Now we have a weak Boolean Algebra $H_{x}$ where the composition operation is
typically not commutative. Thus H(X) is a family of modern sets of non-
commutative type if one of $H_{x}$ is not commutative under the composition.
Needless to say, if all of $H_{x}$ is commutative under the composition, then
H(X) satisfies the commutative law.
###### Example 7.
Let (M, n$\times$n , + , $\cdot$) be an n$\times$n matrix space closed under
addition + and matrix multiplication $\cdot$. Then call the matrix
multiplication identity I and the addition identity matrix O. Now we take
$\cdot$ = $\ast_{\wedge_{x}}$ and + = $\ast_{\vee_{x}}$. By considering the
same equivalence class as in Example 6, we can create a modern set from the
n$\times$n matrix space.
###### Theorem 8.
Gelfand Theorem
If $\mathfrak{U}$ is a commutative a $C^{\ast}$-algebra, then $\mathfrak{U}$
is $\ast$-isomorphism to $\mathbf{C}(X)$, some compact Hausdorff space X.
###### Remark 2.
For a given commutative $C^{\ast}$-algebra, we have a representation of it as
a space of continuous functions on some compact Hausdorff by theorem 8. Thus
we can always create a commutative Modern set from any commutative
$C^{\ast}$-algebra under the same construction as in Example 6.
We give the following definition and theorem as more examples of Modern Sets.
###### Definition 7.
By a representation of a $C^{\ast}$-algebra $\mathfrak{U}$ on a Hilbert space
$\mathcal{H}$, we mean a ∗ homomorphism $\varphi$ from $\mathfrak{U}$ into
$\mathfrak{B}(\mathcal{H})$. If in addition, $\varphi$ is one-to-one, it is
called a faithful representation.
###### Theorem 9.
The Gelfand-Neumark Theorem
Each $C^{\ast}$-algebra has a faithful representation on some Hilbert space.
###### Example 10.
For a given $C^{\ast}$-algebra, we have a representation of it as bounded
linear operators on a Hilbert space by theorem 9. Thus we can always create a
Modern set from any $C^{\ast}$-algebra under the same construction as in
Example 6.
## 5\. Conclusion and observation
As we mentioned in the Introduction, systematic expression of our thought
requires room for at least the non-commutative property. I strongly believe
that this new logic system will open up a new blanch of Artificial
Intelligence. Property-like verbs such as ”be”, ”have”, and ”own” seem valid
in classical logic. However, most of the other verbs are required to be non-
commutative with respect to objects and time. This modern set does not need to
be commutative, in some sense, this is closer to the system of our thought. We
need further investigation to improve the systematic expression of our thought
in order to create a real Artificial Intelligence. I dream the day will come,
when we make a real AI.
## References
* [1] G. Birkhoff, Lattice Theory, 3rd ed. AMS colloquim Publication, Providence, RI, 1967.
* [2] J.A. Goguen, L-fuzzy sets, J. Math. Anal. Appl., 18, 145-174, 1967
* [3] M. Heidegger, being and time, HarperOne; Revised edition, 1962 (English)
* [4] T. Iwamura, Sokuron (Kyoritsu Shuppan, Tokyo, 1966).
* [5] R. V. Kadison, J.R. Ringrose, Fundamentals of the Theory of Opreator Algebra, AMS, 1997.
* [6] N. Nakamura, generalized fuzzy sets, Fuzzy sets and Systems 32 (1989), 307-314.
* [7] G. Takeuti, Senkei-Daisu to ryoushi-rikigaku, 131-162, Shokabo, Tokyo, 1981 (Japanese)
* [8] L.A. Zadeh, Fuzzy sets, Information and Control 8 (1965), 338-353.
| arxiv-papers | 2008-06-18T09:45:18 | 2024-09-04T02:48:56.338398 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Jun Tanaka",
"submitter": "Jun Tanaka",
"url": "https://arxiv.org/abs/0806.2943"
} |
0806.2993 | # Reply to ” Comment on ’Once more about the $K\bar{K}$ molecule approach to
the light scalars’ ”
N.N. Achasov achasov@math.nsc.ru A.V. Kiselev kiselev@math.nsc.ru Laboratory
of Theoretical Physics, Sobolev Institute for Mathematics, Novosibirsk,
630090, Russia
###### Abstract
The need to regularize loop integrals in a manner that preserves gauge
invariance, for example, using the Pauli-Villars method, requires a
subtraction that in the large mass limit hides its high momentum origin. This
gives rise to the illusion that only nonrelativistic kaon loop momenta are
relevant, when in fact this is not the case, as we show.
###### pacs:
12.39.-x 13.40.Hq 13.66.Bc
The authors of Ref. Comment adduced the following argument against our
criticism achasov2007 . Since the $\phi\to K^{+}K^{-}\to\gamma(a_{0}/f_{0})$
amplitude vanishes for gauge invariance when the photon momentum vanishes,
only those terms of the integrand have the physical sense which vanish with
vanishing the photon momentum. What actually happens is that cancellation of
contributions from different places of momentum (or co-ordinate) space is
realized. It is commonplace in electrodynamics. In particular, low energy
theorems are based on this. Discarding the integrand in the third term of Eq.
(3), the second term in Eq. (4), and the contribution of Eq. (5) the authors
of Comment distort the physical significance of the $K^{+}K^{-}$ loop model
because these contributions represent the high momentum and charge flow
distributions of kaons. Below we show that the $K^{+}K^{-}$ loop model
describes the relativistic physics.
When basing the experimental investigations of the light scalar mesons
production in the $\phi$ radiative decays
$\phi\to\gamma\bigl{[}a_{0}(980)/f_{0}(980)\bigr{]}\to\gamma\bigl{[}(\pi^{0}\eta)/(\pi^{0}\pi^{0})\bigr{]}$,
there was suggested achasov1989 the kaon loop model $\phi\to
K^{+}K^{-}\to\gamma\bigl{[}a_{0}(980)/f_{0}(980)\bigr{]}$ with the pointlike
interaction, see Fig. 1. This model is used in the data treatment and ratified
by experiment. In Refs. achasov2001 ; achasov2003 an analysis of mechanisms
of decays under consideration was carried out, which gave the clear arguments
for this kaon loop model.
| |
---|---|---
(a) | (b) | (c)
Figure 1: Diagrams contributing to the radiative decay amplitude.
Every diagram contribution in
$T\left\\{\phi(p)\to\gamma[a_{0}(q)/f_{0}(q)]\right\\}=(a)+(b)+(c)$ (1)
is divergent hence should be regularized in a gauge invariant manner, for
example, in the Pauli-Wilars one.
$\displaystyle\overline{T}\left\\{\phi(p)\to\gamma[a_{0}(q)/f_{0}(q)],M\right\\}=\overline{(a)}+\overline{(b)}+\overline{(c)}\,,$
$\displaystyle\overline{T}\left\\{\phi(p)\to\gamma[a_{0}(q)/f_{0}(q)],M\right\\}=\epsilon^{\nu}(\phi)\epsilon^{\mu}(\gamma)\overline{T}_{\nu\mu}(p,q)$
$\displaystyle=\epsilon^{\nu}(\phi)\epsilon^{\mu}(\gamma)\left[\overline{a}_{\nu\mu}(p,q)+\overline{b}_{\nu\mu}(p,q)+\overline{c}_{\nu\mu}(p,q)\right],$
(2)
$\displaystyle\overline{a}_{\nu\mu}(p,q)=-\frac{i}{\pi^{2}}\int\Biggl{\\{}\frac{(p-2r)_{\nu}(p+q-2r)_{\mu}}{(m_{K}^{2}-r^{2})[m_{K}^{2}-(p-r)^{2}][m_{K}^{2}-(q-r)^{2}]}$
$\displaystyle-\frac{(p-2r)_{\nu}(p+q-2r)_{\mu}}{(M^{2}-r^{2})[M^{2}-(p-r)^{2}][M^{2}-(q-r)^{2}]}\Biggr{\\}}dr\,,$
(3)
$\displaystyle\overline{b}_{\nu\mu}(p,q)=-\frac{i}{\pi^{2}}\int\Biggl{\\{}\frac{(p-2r)_{\nu}(p-q-2r)_{\mu}}{(m_{K}^{2}-r^{2})[m_{K}^{2}-(p-r)^{2}][m_{K}^{2}-(p-q-r)^{2}]}$
$\displaystyle-\frac{(p-2r)_{\nu}(p-q-2r)_{\mu}}{(M^{2}-r^{2})[M^{2}-(p-r)^{2}][M^{2}-(p-q-r)^{2}]}\Biggr{\\}}dr\,,$
(4) $\overline{c}_{\nu\mu}(p,q)=-\frac{i}{\pi^{2}}\,2g_{\nu\mu}\int
dr\Biggl{\\{}\frac{1}{(m_{K}^{2}-r^{2})[m_{K}^{2}-(q-r)^{2}]}-\frac{1}{(M^{2}-r^{2})[M^{2}-(q-r)^{2}]}\Biggr{\\}}\,,$
(5)
where $M$ is the regulator field mass. $M\to\infty$ in the end
$\overline{T}\bigl{[}\phi\to\gamma(a_{0}/f_{0}),M\to\infty\bigr{]}\to
T^{Phys}\bigl{[}\phi\to\gamma(a_{0}/f_{0})\bigr{]}\,.$ (6)
We can shift the integration variables in the regularized amplitudes and
easily check the gauge invariance condition
$\epsilon^{\nu}(\phi)k^{\mu}\overline{T}_{\nu\mu}(p,q)=\epsilon^{\nu}(\phi)(p-q)^{\mu}\overline{T}_{\nu\mu}(p,q)=0\,.$
(7)
It is instructive to consider how the gauge invariance condition
$\epsilon^{\nu}(\phi)\epsilon^{\mu}(\gamma)\overline{T}_{\nu\mu}(p,p)=0$ (8)
holds true,
$\displaystyle\epsilon^{\nu}(\phi)\epsilon^{\mu}(\gamma)\overline{T}_{\nu\mu}(p,p)=$
$\displaystyle\epsilon^{\nu}(\phi)\epsilon^{\mu}(\gamma)T^{\,m_{K}}_{\nu\mu}(p,p)-\,\epsilon^{\nu}(\phi)\epsilon^{\mu}(\gamma)T^{M}_{\nu\mu}(p,p)=(\epsilon(\phi)\epsilon(\gamma))(1-1)=0\,.$
(9)
The superscript $m_{K}$ refers to the nonregularized amplitude and the
superscript $M$ refers to the regulator field amplitude. So, the contribution
of the (a), (b), and (c) diagrams does not depend on a particle mass in the
loops ($m_{K}$ or $M$) at $p=q$ typical . But, the physical meaning of these
contributions is radically different. The
$\epsilon^{\nu}(\phi)\epsilon^{\mu}(\gamma)T^{\,m_{K}}_{\nu\mu}(p,p)$
contribution is caused by intermediate momenta (a few GeV) in the loops ,
whereas the regulator field contribution is caused fully by high momenta
($M\to\infty$) and teaches us how to allow for high $K$ virtualities in a
gauge invariant way.
Needless to say the integrand of
$\epsilon^{\nu}(\phi)\epsilon^{\mu}(\gamma)\overline{T}_{\nu\mu}(p,p)$ is not
equal to 0.
It is clear that
$\displaystyle\epsilon^{\nu}(\phi)\epsilon^{\mu}(\gamma)T^{M\to\infty}_{\nu\mu}(p,q)\to\epsilon^{\nu}(\phi)\epsilon^{\mu}(\gamma)T^{M\to\infty}_{\nu\mu}(p,p)$
$\displaystyle\equiv\epsilon^{\nu}(\phi)\epsilon^{\mu}(\gamma)T^{M}_{\nu\mu}(p,p)\equiv(\epsilon(\phi)\epsilon(\gamma))\,.$
(10)
So, the regulator field contribution tends to the subtraction constant when
$M\to\infty$.
The finiteness of the subtraction constant hides its high momentum origin and
gives rise to an illusion of a nonrelativistic physics in the $K^{+}K^{-}$
model with the pointlike interaction. See, for example, Ref. kalash ; see
Section 2 in this paper.
This work was supported in part by the Presidential Grant No. NSh-1027.2008.2
for Leading Scientific Schools and by RFFI Grant No. 07-02-00093 from the
Russian Foundation for Basic Research. A.V.K. thanks very much the Dynasty
Foundation and ICFPM for support, too.
## References
* (1) Yu. S. Kalashnikova, A. E. Kudryavtsev, A. V. Nefediev, J. Haidenbauer, and C. Hanhart, Phys. Rev. D 78, 058501 (2008).
* (2) N.N. Achasov and A.V. Kiselev, Phys. Rev. D 76, 077501 (2007)
* (3) N.N. Achasov and V.N. Ivanchenko, Nucl. Phys. B 315, 465 (1989).
* (4) N.N. Achasov and V.V. Gubin, Phys. Rev. D 63, 094007 (2001);
Yad. Fiz. 65, 1566 (2002) [Phys. At. Nucl. 65, 1528 (2002)].
* (5) N.N. Achasov, Nucl. Phys. A 728, 425 (2003);
Yad. Fiz. 67, 1552 (2004) [Phys. At. Nucl. 67, 1529 (2004)].
* (6) A typical example of such integrals is $2\int_{0}^{\infty}\frac{m^{2}x}{(x+m^{2})^{3}}dx=1$.
* (7) Yu. S. Kalashnikova, A. E. Kudryavtsev, A. V. Nefediev, C. Hanhart, and J. Haidenbauer, Eur.Phys.J. A 24, 437 (2005).
| arxiv-papers | 2008-06-18T13:20:37 | 2024-09-04T02:48:56.342457 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "N.N. Achasov and A.V. Kiselev",
"submitter": "Nikolay Achasov",
"url": "https://arxiv.org/abs/0806.2993"
} |
0806.3047 | # Deeply Virtual Compton Scattering at HERA and perspectives at CERN
Laurent Schoeffel
###### Abstract
Standard parton distribution functions contain neither information on the
correlations between partons nor on their transverse motion, then a vital
knowledge about the three dimensional structure of the nucleon is lost. Hard
exclusive processes, in particular DVCS, are essential reactions to go beyond
this standard picture. In the following, we examine the most recent data in
view of the dipole model predictions and their implication on the
quarks/gluons imaging (tomography) of the nucleon.
###### Keywords:
deeply virtual scattering, dipole model, parton distributions, proton
tomography
###### :
11.25.Db, 12.38.Bx, 13.60.Fz
## 1 Introduction
Measurements of the deep-inelastic scattering (DIS) of leptons and nucleons,
$e+p\to e+X$, allow the extraction of Parton Distribution Functions (PDFs)
which describe the longitudinal momentum carried by the quarks, anti-quarks
and gluons that make up the fast-moving nucleons. These functions have been
measured over a wide kinematic range in the Bjorken scaling variable $x_{Bj}$
and the photon virtuality $Q^{2}$. While PDFs provide crucial input to
perturbative Quantum Chromodynamic (QCD) calculations of processes involving
hadrons, they do not provide a complete picture of the partonic structure of
nucleons. In particular, PDFs contain neither information on the correlations
between partons nor on their transverse motion, then a vital knowledge about
the three dimensional structure of the nucleon is lost. Hard exclusive
processes, in which the nucleon remains intact, have emerged in recent years
as prime candidates to complement this essentially one dimentional picture
lolopic .
Indeed, within the investigation of hard exclusive reactions in the Bjorken
limit, the probe provided by the photon works as a clean tool reactions in
order to extract reliable knowledge on the substructure of strongly
interacting particles complementar to exclusive process.
The simplest exclusive process is the deeply virtual Compton scattering (DVCS)
or exclusive production of real photon, $e+p\rightarrow e+\gamma+p$. This
process is of particular interest as it has both a clear experimental
signature and is calculable in perturbative QCD. The DVCS reaction can be
regarded as the elastic scattering of the virtual photon off the proton via a
colourless exchange, producing a real photon in the final state lolopic . The
recent measurements from the DESY $ep$ collider HERA at low $x_{Bj}$
($x_{Bj}<0.01$) and large $Q^{2}$ (above 2 GeV2) are thus a decisive
experimental step forward dvcsh1 ; dvcszeus . In the Bjorken scaling regime,
QCD calculations assume that the exchange involves two partons, having
different longitudinal and transverse momenta, in a colourless configuration.
These unequal momenta or skewing are a consequence of the mass difference
between the incoming virtual photon and the outgoing real photon. This
skewedness effect can be interpreted in the context of generalised parton
distributions (GPDs) diehl ; freund2 ; buk or in the dipole model approach
robi ; marquet .
A considerable interest of the high energy situation at HERA is that it gives
the important opportunity to constrain the gluon contribution to GPDs and to
study the evolution in virtuality $Q^{2}$ of the quark and gluon
distributions. On the other hand, recently the color dipole formalism has
provided a simultaneous description of photon induced process. The inclusive
deep inelastic reaction and the photon diffractive dissociation has been
successfully described and the study of other exclusive processes such as DVCS
is an important test of the color dipole approach.
In the following, we examine the most recent data in view of the dipole model
predictions robi ; marquet and their implication on the quarks/gluons imaging
of the nucleon dvcsh1 ; lolotom ; pire .
## 2 The color dipole model
The colour dipole model provides a simple unified picture of inclusive and
diffractive processes and enables hard and soft physics to be incorporated in
a single dynamical framework. At high energies, in the proton’s rest frame,
the virtual photon fluctuates into a hadronic system (the simplest of which is
a $q{\bar{q}}$ dipole) a long distance upstream of the target proton. The
formation time of this hadronic system, and of the subsequent formation of the
hadronic final state, is much longer than the interaction time with the
target.
DVCS is a good probe of the transition between soft and hard regimes in the
dipole model for two reasons. Indeed, the transverse photon wave function can
select large dipoles, even for large $Q^{2}$, and certainly for the $Q^{2}$
range $2<Q^{2}<20$ GeV2. Also, because the final photon is real, DVCS is more
sensitive to large dipoles than DIS at the same $Q^{2}$.
Then, in the colour dipole approach, the DIS (or DVCS) process can be seen as
a succession in time of three factorisable subprocesses: i) the virtual photon
fluctuates in a quark-antiquark pair, ii) this colour dipole interacts with
the proton target, iii) the quark pair annihilates into a virtual (or real)
photon. The imaginary part of the DIS (or DVCS) amplitude at $t=0$ is
expressed in the simple way robi ; marquet
$\displaystyle{\cal I}m\,{\cal
A}\,(W,Q_{1},Q_{2})=\sum_{T,L}\int\limits_{0}^{1}dz\\!\int\limits_{0}^{\infty}d^{2}\mbox{\boldmath$r$}\,\Psi_{T,L}^{*}(z,\,\mbox{\boldmath$r$},\,Q_{1}^{2})\,\sigma_{dip}\,(z,\mbox{\boldmath$r$})\Psi_{T,L}(z,\,\mbox{\boldmath$r$},\,Q_{2}^{2})\,,$
(1)
where $\Psi(z,\mbox{\boldmath$r$},Q_{1,2})$ are the light cone photon wave
functions for transverse and longitudinal photons. The quantity $Q_{1}$ is the
virtuality of the incoming photon, whereas $Q_{2}$ is the virtuality of the
outgoing photon. In the DIS case, one has $Q_{1}^{2}=Q_{2}^{2}=Q^{2}$ and for
DVCS, $Q_{1}^{2}=Q^{2}$ and $Q_{2}^{2}=0$. The relative transverse quark pair
(dipole) separation is labeled by $r$ whilst $z$ (respec. $1-z$) labels the
quark (antiquark) longitudinal momentum fraction.
It should be noticed that the non-forward kinematics for DVCS is encoded in
the colour dipole approach through the different weight coming from the photon
wavefunctions in Eq. (1). The off-diagonal effects, which affect the gluon and
quark distributions in GPDs models, should be included in the parameterisation
of the dipole cross section. At the present stage of the development of the
dipole formalism, we have no accurate theoretical arguments on how to compute
skewedness effects from first principles. A consistent approach would be to
compute the scattering amplitude in the non-forward case, since the non-
forward photon wave function has been recently obtained. In this case, the
dipole cross section,
$\sigma_{dip}\,(x_{1},x_{2},\mbox{\boldmath$r$},\vec{\Delta})$, depends on the
momenta $x_{1}$ and $x_{2}$ carried by the exchanged gluons, respectively, and
on the total transverse momentum transfer $\vec{\Delta}$. In this case,
additional information about the dependence upon $\vec{\Delta}$ is needed for
the QCD Pomeron and proton impact factor. A first attempt in this direction is
done in robi .
## 3 Geometric Scaling
At very small values of the Bjorken scaling variable $x$ the saturation regime
of QCD can be reached. In this domain, the gluon density in the proton is so
large that non-linear effects like gluon recombination tame its growth. In the
dipole model approach, the transition to the saturation regime is
characterised by the so-called saturation scale parametrised here as
$Q_{s}(x)=Q_{0}({{x_{0}}/{x}})^{-\lambda/2}$, where $Q_{0}$, $x_{0}$ and
$\lambda$ are parameters. The transition to saturation occurs when $Q$ becomes
comparable to $Q_{s}(x)$. An important feature of dipole models that
incorporate saturation is that the total cross section can be expressed as a
function of the single variable $\tau$:
$\sigma_{tot}^{\gamma^{\ast}p}(x,Q^{2})=\sigma_{tot}^{\gamma^{\ast}p}(\tau),\;\;\mbox{with}\;\
\ \ \tau=\frac{Q^{2}}{Q_{s}^{2}(x)}.$ (2)
This property, called geometric scaling, has already been observed to hold for
the total $ep$ DIS cross section golec and in diffractive processes marquet .
It has also recently been addressed in the context of exclusive processes
including DVCS and extended to cases with non-zero momentum transfer to the
proton robi . It is therefore interesting to test if the present DVCS
measurements obey the geometric scaling laws predicted by such models.
## 4 Latest News from the experimental Front
Measurements of DVCS cross section have been realised at HERA within the H1
and ZEUS experiments dvcsh1 ; dvcszeus . As mentioned in the introduction,
these results are given in the specific kinematic domain of both experiments,
at low $x_{Bj}$ ($x_{Bj}<0.01$) but they take advantage of the large range in
$Q^{2}$, offered by the HERA kinematics, which covers more than 2 orders of
magnitude. It makes possible to study the transition from the low $Q^{2}$ non-
perturbative region (around $1$ GeV2) towards higher values of $Q^{2}$ where
the higher twists effects are lowered (above $10$ GeV2). The last data on DVCS
cross section as a function of $W\simeq\sqrt{Q^{2}/x}$ are presented on figure
1. They show a strong $W$ dependence with $\sigma_{DVCS}\propto W^{0.7}$,
characteristic of a hard process, which can thus be described in perturbative
QCD as described in previous sections.
Data and model comparisons are presented on figures 2 and 3. They show that
the dipole approach is very efficient in describing the DVCS measurements for
the HERA kinematics. It must be noticed here that the non-forward kinematics
for DVCS is encoded only through the different weights coming from the photon
wavefunctions in Eq. (1). It means that without taking into account non-
diagonal effects that drive all the physics of GPDs, we can get a good
descritption of all existing low $x_{Bj}$ data in the framework of the dipole
model.
Beside $W$ and $Q^{2}$ DVCS cross sections, a major experimental achievement
of H1 dvcsh1 has been the measurement of DVCS cross sections, differential in
$t=(p^{\prime}-p)^{2}$, the momentum transfer (squared) at the proton vertex.
Some results are presente on figure 4: we observe the good description of
$d\sigma_{DVCS}/dt$ by a fit of the form $e^{-b|t|}$. Hence, an extraction of
the $t$-slope parameter $b$ is accessible for different values of $Q^{2}$ and
$W$ (see figures 4 and 5). We exploit these values of $b$ in a section below.
Figure 1: DVCS cross section for positrons/electrons samples as a function of
$W$ for $Q^{2}=8$ GeV2.
Figure 2: DVCS cross section measurements as a function of the scaling
variable $\tau={Q^{2}}/{Q_{s}^{2}(x)}$. Results are shown for the full $t$
range $|t|<$ 1 GeV2 (left) and at four values of $t$ (right). The dashed
curves represent the predictions of the dipole model marquet .
Figure 3: Predictions for the DVCS cross section robi with H1 data. Figure 4:
DVCS cross section, differential in $t$ presented with a fit of the form
$e^{-b|t|}$.
Figure 5: The logarithmic slope of the $t$ dependence for DVCS exclusive
production, $b$ as a function of $Q^{2}$ and $W$, extracted from a fit
$d\sigma/dt\propto\exp(-b|t|)$ where $t=(p-p^{\prime})^{2}$.
## 5 Quantifying skewing effects
Let’s define a basic quantity giving an overall measurement of the skewing
properties lolocompass ; lolocompass2 , which includes both the non-forward
kinematics and the possible non-diagonal effects. Namely, we set the ratio
between the imaginary parts of the DIS and DVCS (forward) scattering
amplitudes at zero momentum transfer:
$R\equiv\frac{Im\,{\cal A}\,(\gamma^{*}+p\to\gamma^{*}+p)|_{t=0}}{Im\,{\cal
A}\,(\gamma^{*}+p\to\gamma+p)|_{t=0}}\ ,$ (3)
where $t$ is the square of the four-momentum exchanged at the proton vertex.
In lolocompass ; lolocompass2 , it has been shown how to extract this quantity
from the DVCS cross sections. The factor $R$ is presented as a function of
energy $W$ in figure 6. An almost flat $W$ dependence is observed within the
present precision. This feature can be easily understood since the $W$
dependence of both the DIS and DVCS cross section is power-like having a
proportional effective power. Namely, $\sigma_{\mathrm{DIS}}\propto
W^{2\lambda}$ and $\sigma_{\mathrm{DVCS}}\propto W^{4\lambda}$. The mean value
$R\simeq 0.5$ is consistent with its early theoretical estimates using the
aligned jet model and the colour dipole model (see figure 6). We also display
a dipole model prediction with an ’ad hoc’ parametrisation for the off-
diagonal effects. It does not improve the quality of the description of data
within the present uncertainties.
Figure 6: The skewing factor $R$ as a function of $W$ at $Q^{2}=8$ GeV2. The
curves represent the theoretical predictions of the dipole approach (see
text).
## 6 Nucleon Tomography
Measurements of the $t$-slope parameters $b$ are key measurements for almost
all exclusive processes, in particular DVCS. Indeed, a Fourier transform from
momentum to impact parameter space readily shows that the $t$-slope $b$ is
related to the typical transverse distance between the colliding objects buk .
At high scale, the $q\bar{q}$ dipole is almost point-like, and the $t$
dependence of the cross section is given by the transverse extension of the
gluons (or sea quarks) in the proton for a given $x_{Bj}$ range. More
precisely, from the generalised parton distribution defined in the
introduction, we can compute a parton density which also depends on a spatial
degree of freedom, the transverse size (or impact parameter), labeled
$R_{\perp}$, in the proton. Both functions are related by a Fourier transform
$PDF(x,R_{\perp};Q^{2})\;\;\equiv\;\;\int\frac{d^{2}\Delta_{\perp}}{(2\pi)^{2}}\;e^{i({\Delta}_{\perp}{R_{\perp}})}\;GPD(x,t=-{\Delta}_{\perp}^{2};Q^{2}).$
Thus, the transverse extension $\langle r_{T}^{2}\rangle$ of gluons (or sea
quarks) in the proton can be written as
$\langle r_{T}^{2}\rangle\;\;\equiv\;\;\frac{\int
d^{2}R_{\perp}\;PDF(x,R_{\perp})\;R_{\perp}^{2}}{\int
d^{2}R_{\perp}\;PDF(x,R_{\perp})}\;\;=\;\;4\;\frac{\partial}{\partial
t}\left[\frac{GPD(x,t)}{GPD(x,0)}\right]_{t=0}=2b$
where $b$ is the exponential $t$-slope. Measurements of $b$ presented in
figure 5 corresponds to $\sqrt{r_{T}^{2}}=0.65\pm 0.02$ fm at large scale
$Q^{2}$ for $x_{Bj}<10^{-2}$. This value is smaller that the size of a single
proton, and, in contrast to hadron-hadron scattering, it does not expand as
energy $W$ increases. This result is consistent with perturbative QCD
calculations in terms of a radiation cloud of gluons and quarks emitted around
the incoming virtual photon. In short, gluons are located at the preiphery of
the proton as measured here and valence quarks are assumed to form the core of
the proton at small value of $\sqrt{r_{T}^{2}}$.
In other words, the Fourier transform of the DVCS amplitude is the amplitude
to find quarks at $R_{\perp}$ in an image plane after focusing by an idealized
lens. The square of the profile amplitude, producing the PDF (in transverse
plane) is positive, real-valued, and corresponds to the image, a weighted
probability to find quarks in the transverse image plane.
## 7 Perspectives at CERN
The complete parton imaging in the nucleon would need to get measurements of
$b$ for several values of $x_{Bj}$, from the low $x_{Bj}<0.01$ till
$x_{Bj}>0.1$. Experimentally, it appears to be impossible. Is it the breakout
of quark and gluon imaging in the proton? In fact, there is one way to recover
$x_{Bj}$ and $t$ correlations over the whole $x_{Bj}$ domain: we need to
measure a Beam Charge Asymmetry (BCA) lolopic ; lolocompass ; lolobca .
A determination of a cross section asymmetry with respect to the beam charge
has been realised by the H1 experiment by measuring the ratio
$(d\sigma^{+}-d\sigma^{-})/(d\sigma^{+}+d\sigma^{-})$ as a function of $\phi$,
where $\phi$ is the azimuthal angle between leptons and proton plane lolopic ;
freund2 . The result is presented on figure 7 with a fit in $\cos\phi$. After
applying a deconvolution method to account for the resolution on $\phi$, the
coefficient of the $\cos\phi$ dependence is found to be $p_{1}=0.17\pm
0.03(stat.)\pm 0.05(sys.)$. This result represents obviously a major progress
in the understanding of the very recent field of the parton imaging in the
proton. We are at the hedge of the giving a new reading on the most
fundamental question to know how the proton is built up by quarks and gluons.
Feasabilities for future BCA measurements at COMPASS have been studied
extensively in the last decade dhose . COMPASS is a fixed target experiment
which can use 100 GeV muon beams and hydrogen targets, and then access
experimentaly the DVCS process $\mu p\rightarrow\mu\gamma p$. The BCA can be
determined when using positive and negative muon beams. One major interest is
the kinematic coverage from $2$ GeV2 till $6$ GeV2 in $Q^{2}$ and $x_{Bj}$
ranging from $0.05$ till $0.1$. It means that it is possible to avoid the
kinematic domain dominated by higher-twists and non-perturbative effects (for
$Q^{2}<1$ GeV2) and keeping a $x_{Bj}$ range which is extending the HERA
(H1/ZEUS) domain.
Figure 7: Beam charge asymmetry as a function of $\phi$ measured by H1 lolopic
; lolobca .
## 8 Summary
DVCS measurements in the HERA kinematics at low $x_{Bj}$ are well described
within a dipole approach, which encodes the non-forward kinematics for DVCS
only through the different weights coming from the photon wavefunctions. Note
that the dipole model presents the great advantage to define a unique
framework that gives a good description of all hard processes accessible at
HERA. For the first time, we have also shown that proton tomography enters
into the experimental domain of high energy physics, with a first experimental
evidence that gluons are located at the periphery of the proton. A new
frontier in understanding this structure would be possible at CERN within the
COMPASS experimental setup.
## References
* (1) L. Schoeffel, proceedings of the 27th International Conference on Physics in Collision (PIC 2007), Annecy, France, 26-29 Jun 2007. arXiv:0707.3706 [hep-ph].
* (2) F. D. Aaron et al. [H1 Collaboration], Phys. Lett. B 659 (2008) 796 [arXiv:0709.4114 [hep-ex]] ; A. Aktas et al. [H1 Collaboration], Eur. Phys. J. C 44, 1 (2005) [hep-ex/0505061] ; C. Adloff et al. [H1 Collaboration], Phys. Lett. B 517 (2001) 47 [arXiv:hep-ex/0107005].
* (3) S. Chekanov et al. [ZEUS Collaboration], Phys. Lett. B 573, 46 (2003) [hep-ex/0305028].
* (4) M. Diehl, AIP Conf. Proc. 842 (2006) 294 [arXiv:hep-ph/0512201].
* (5) A. Freund, Phys. Rev. D 68 (2003) 096006 [hep-ph/0306012].
* (6) M. Burkardt, Int. J. Mod. Phys. A 18 (2003) 173 [hep-ph/0207047].
* (7) C. Marquet, R. B. Peschanski and G. Soyez, Phys. Rev. D 76 (2007) 034011 [arXiv:hep-ph/0702171].
* (8) C. Marquet and L. Schoeffel, Phys. Lett. B 639 (2006) 471 [arXiv:hep-ph/0606079].
* (9) L. Schoeffel, proceedings of the 16th International Workshop on Deep Inelastic Scattering and Related Subjects (DIS 2008), London, England, 7-11 Apr 2008. arXiv:0805.2672 [hep-ph].
* (10) J. P. Ralston and B. Pire, Phys. Rev. D 66 (2002) 111501 [arXiv:hep-ph/0110075].
* (11) K. J. Golec-Biernat and M. Wusthoff, Phys. Rev. D 60 (1999) 114023 [arXiv:hep-ph/9903358].
* (12) L. Schoeffel, Phys. Lett. B 658 (2007) 33 [arXiv:0706.3488 [hep-ph]].
* (13) L. Favart, M. V. T. Machado and L. Schoeffel, Braz. J. Phys. 37 (2007) 798.
* (14) L. Schoeffel, proceedings of the 15th International Workshop on Deep-Inelastic Scattering and Related Subjects (DIS2007), Munich, Germany, 16-20 Apr 2007. arXiv:0705.2925 [hep-ph].
* (15) N. d’Hose, E. Burtin, P. A. M. Guichon, S. Kerhoas-Cavata, J. Marroncle and L. Mosse, Nucl. Phys. A 711 (2002) 160.
| arxiv-papers | 2008-06-18T19:52:45 | 2024-09-04T02:48:56.347265 | {
"license": "Public Domain",
"authors": "Laurent Schoeffel",
"submitter": "Schoeffel Laurent",
"url": "https://arxiv.org/abs/0806.3047"
} |
0806.3312 | # Novel interface-selected waves and their influences on wave competitions
Xiaohua Cui1, Xiaoqing Huang1, Zhoujian Cao2, Hong Zhang3, and Gang Hu1
ganghu@bnu.edu.cn 1Department of physics, Beijing Normal University, Beijing
100875, China 2Institute of Applied Mathematics, Academy of Mathematics and
System Science, Chinese Academy of Sciences, 55, Zhongguancun Donglu, Beijing
100080, China 3Zhejiang Institute of Modern Physics and Department of
Physics, Zhejiang
University, Hangzhou 310027, China
###### Abstract
The topic of interface effects in wave propagation has attracted great
attention due to their theoretical significance and practical importance. In
this paper we study nonlinear oscillatory systems consisting of two media
separated by an interface, and find a novel phenomenon: interface can select a
type of waves (ISWs). Under certain well defined parameter condition, these
waves propagate in two different media with same frequency and same wave
number; the interface of two media is transparent to these waves. The
frequency and wave number of these interface-selected waves (ISWs) are
predicted explicitly. Varying parameters from this parameter set, the wave
numbers of two domains become different, and the difference increases from
zero continuously as the distance between the given parameters and this
parameter set increases from zero. It is found that ISWs can play crucial
roles in practical problems of wave competitions, e.g., ISWs can suppress
spirals and antispirals.
The behaviors of waves around interface between two media have attracted
continual and great interest [1-7]. In linear optics we are familiar with the
problems of wave reflection and refraction, which are predicted analytically.
However, for nonlinear systems the interface-related behaviors become much
more complex and diverse, and much less known.
The problems of wave propagation in linear and nonlinear media have also
attracted considerable attention. Recently, new observations of inwardly
propagating waves have stimulated considerable interest in this field. For
several centuries, scientists have known waves propagating forwardly from wave
source only, called normal waves (NW). In recent years, different types of
waves propagating toward the wave source (called here antiwaves, AW) have been
observed in both linear optics [1, 8] and nonlinear oscillatory systems
[9-14]. These new phenomena introduce completely new topics of the interface
problem. For instance, novel phenomenon of negative refraction has been
reported in both linear optics [1, 8] and nonlinear oscillatory systems [3,7].
In the present paper, we find another completely new nonlinear interface-
phenomenon: interface of two different media can generate waves, called here
interface selected waves (ISWs). In a well defined parameter surface the
frequency and wave number (also wave length) of ISWs are identical in two
media with different parameters, and they can be predicted analytically. Away
but near this surface ISWs still exist, though the above analytical
predictions are no longer available and wave numbers of ISWs in the two
domains are no longer identical. By varying parameter away from this surface
continuously, wave numbers change continuously, and the difference of wave
numbers in two domains also increases from zero continuously. It is found that
ISWs play crucial roles in practical problems of wave competitions in
oscillatory systems, e.g., in suppressing spirals and antispirals.
We consider the following bidomain reaction-diffusion system
$\displaystyle\frac{\partial\mathbf{U}_{1}}{\partial t}$
$\displaystyle=\mathbf{f}(\mathbf{U}_{1},\bm{\mu}_{1},\bm{\nu}_{1})+\mathbf{D}(\bm{\gamma}_{1}):\nabla^{2}\mathbf{U}_{1}$
(1a) $\displaystyle\frac{\partial\mathbf{U}_{2}}{\partial t}$
$\displaystyle=\mathbf{f}(\mathbf{U}_{2},\bm{\mu}_{2},\bm{\nu}_{2})+\mathbf{D}(\bm{\gamma}_{2}):\nabla^{2}\mathbf{U}_{2}$
(1b) $\displaystyle\mathbf{U}_{i}=(U_{i}^{(1)}$
$\displaystyle,\cdots,U_{i}^{(m)}),\quad\bm{\mu}_{i}=(\mu_{i}^{(1)},\cdots,\mu_{i}^{(p)}),$
$\displaystyle\bm{\nu}_{i}=(\nu_{i}^{(1)}$
$\displaystyle,\cdots,\nu_{i}^{(q)}),\quad\bm{\gamma}_{i}=(\gamma_{i}^{(1)};\cdots,\gamma_{i}^{(l)}),$
$\displaystyle i=1,2,\ \ m\geq 2$
where $\mathbf{D}(\bm{\gamma}_{i})$ is a $m\times m$ matrix with constant
elements depending on $\bm{\gamma}_{i}$. The function $\mathbf{f}$ and
diffusion matrix $\mathbf{D}$ in the two domains are identical because the
same reaction-diffusion processes occur in both sides. On the other hand, the
dynamical evolutions in each side may be different due to different control
parameters $(\bm{\mu}_{1},\ \bm{\nu}_{1},\ \bm{\gamma}_{1})$ and
$(\bm{\mu}_{2},\ \bm{\nu}_{2},\ \bm{\gamma}_{2})$. We represent the interface
between the two domains by $I$, then the following boundary conditions are
required on $I$
$\mathbf{U}_{1\\{I\\}}=\mathbf{U}_{2\\{I\\}},\
\frac{\partial\mathbf{U}_{1}}{\partial
n_{\\{I\\}}}=\frac{\partial\mathbf{U}_{2}}{\partial n_{\\{I\\}}}$ (2)
where $\frac{\partial\mathbf{U}_{i}}{\partial n_{\\{I\\}}}$ indicates a
derivative of $\mathbf{U}_{i}$ over space variable along the direction
perpendicular to the interface $I$. We assume that
$\mathbf{U}_{1}=\mathbf{U}_{2}=0$ is a stable point of Eq.(1) and
$\bm{\mu}_{i}(\bm{\nu}_{i})$ controls (not controls) the linear terms of
$\mathbf{U}_{i}$ for the reaction parts. Hopf bifurcation with frequency
$\omega_{0}$ is supposed to occur at parameters
$\bm{\mu}_{1}=\bm{\mu}_{2}=\bm{\mu}_{0}$. Moreover, we assume further that
both $\bm{\mu}_{1}$ and $\bm{\mu}_{2}$ are slightly beyond the Hopf
bifurcation point
$\sum_{j=1}^{p}({\mu_{i}}^{(j)}-{\mu_{0}}^{(j)})^{2}\ll 1,\ i=1,2$ (3)
At $\bm{\mu}_{i}$, $\mathbf{U}_{i}$ performs periodic oscillation of frequency
$\omega_{i}$, and $\omega_{i}$ approaches to $\omega_{0}$ as $\bm{\mu}_{i}$
reduces to $\bm{\mu}_{0}$ ($i=1,\ 2$). Under condition(3) Eq.(1) can be
reduced to amplitude equations, i.e., the bidomain complex Ginzburg-Landau
equations (BCGLE) by the approach standard for the derivative of single-domain
CGLE [15, 16].
$\frac{\partial A_{1}}{\partial
t}=a_{1}(1-i\Omega_{1})A_{1}-b_{1}(1+i\alpha_{1})|A_{1}|^{2}A_{1}+c_{1}(1+i\beta_{1})\nabla^{2}A_{1}$
(4a) $\frac{\partial A_{2}}{\partial
t}=a_{2}(1-i\Omega_{2})A_{2}-b_{2}(1+i\alpha_{2})|A_{2}|^{2}A_{2}+c_{2}(1+i\beta_{2})\nabla^{2}A_{2}$
(4b) $A_{1\\{I\\}}=A_{2\\{I\\}},\ \frac{\partial A_{1}}{\partial
n_{\\{I\\}}}=\frac{\partial A_{2}}{\partial n_{\\{I\\}}}$ (4c)
where $(a_{i},\ \Omega_{i})$, $(b_{i},\ \alpha_{i})$ and $(c_{i},\ \beta_{i})$
are related to $\bm{\mu}_{i}-\bm{\mu}_{0}$, $(\bm{\mu}_{i}-\bm{\mu}_{0},\
\bm{\nu}_{i})$ and $(\bm{\mu}_{i}-\bm{\mu}_{0},\ \bm{\gamma}_{i})$,
respectively. The continuity conditions Eq.(4c) can be derived from condition
(2) because the transformation from $U_{1}$ to $A_{1}$ is exactly the same as
that from $U_{2}$ to $A_{2}$ (in the two domains, amplitude equations are
derived at a common Hopf bifurcation point with an identical linear matrix at
$\bm{\mu}_{0}$), and the transformations from $U_{i}$ to $A_{i}$ ($i=1,2$) are
determined only by this linear matrix. In Eq.(4) $A_{1}$ and $A_{2}$ are
complex variables in each side of the interface. With scaling transformations
we can fix $a_{1},\ b_{1},\ c_{1}$ and $\Omega_{1}$, and the remaining 8
parameters are irreducible for BCGLE systems. In the following study we will
set $a_{1}=b_{1}=c_{1}=1,\ \Omega_{1}=0$ for numerical simulations without
mentioning and all the theoretical formulas are given generally for 12
parameters. Without the interface interaction, the two media have their
single-domain planar wave solutions [2, 17]
$A_{i}(x,t)=\sqrt{(\frac{1}{b_{i}}(a_{i}-c_{i}k_{i}^{2}))}e^{i(k_{i}x-\omega_{i}t)},\
0\leq k^{2}_{i}\leq\frac{a_{i}}{c_{i}}$ (5a)
$\omega_{i}=a_{i}(\alpha_{i}+\Omega_{i})+c_{i}(\beta_{i}-\alpha_{i})k^{2},\
a_{i},b_{i},c_{i}>0$ (5b)
Waves in media $M_{1}$ and $M_{2}$ are classified to NWs and AWs under the
conditions [3, 12, 14]
$\displaystyle NWs:\ \omega_{i}a_{i}(\Omega_{i}+\alpha_{i})<0\ ;\ or,\
\omega_{i}a_{i}(\Omega_{i}+\alpha_{i})>0\ and\
|\omega_{i}|>|a_{i}(\Omega_{i}+\alpha_{i})|$ (6a) $\displaystyle AWs:\
\omega_{i}a_{i}(\Omega_{i}+\alpha_{i})>0\ and\
|\omega_{i}|<|a_{i}(\Omega_{i}+\alpha_{i})|$ (6b)
By AWs we mean waves with negative phase velocity, while both NWs and AWs have
positive group velocities [11, 12, 14]. Now we focus on the interface related
problems, and start from a one-dimensional (1D) BCGLE system. We are
interested in the problem how the interface can significantly influence the
system dynamics. In Fig.1(a)-(c) we study the system evolutions at two
different parameter sets with random initial conditions, and find
characteristically different features in the asymptotic states. The most
significant and new observation is given in Figs.1(a) and 1(b) where we find
homogeneous planar waves moving in both media from right to left with a
constant velocity and transparently crossing the interface. These homogeneous
running waves originate from the interface (see Fig.1(b)), therefore, called
interface selected waves (ISWs). The phenomenon of Fig.1(a) is surprising.
With two media having different control parameters we intuitively expect that
the waves in two media must have different wave numbers (even if both sides
may have the same frequency). This common feature is clearly seen in Fig.1(c)
where we observe uniform bulk oscillation in the right medium and waves
propagating from left to right in the left medium, clearly manifesting the
interface $I$. However, in Fig.1(a) waves propagate seemly in a homogeneous
medium without feeling any difference of $M_{1}$ and $M_{2}$; the interface is
transparent to the waves. Moreover, the realization of these running waves is
stable against different initial perturbations. This characteristic phenomenon
can never exist in linear systems, and has never been observed so far in
nonlinear systems.
It is interesting that we can predict the frequency and wave number of ISWs
explicitly and exactly under certain parameter conditions. For case of
Fig.1(a) we can determine the frequency and wave number by the following
simple requirements:
$\displaystyle\omega_{1}(k_{1})=\omega_{2}(k_{2}),\ k_{1}=k_{2},\
|A_{1}|=|A_{2}|$ (7)
Inserting Eqs.(7) into Eq.(5) we obtain a unique set of solutions
$\omega_{I},k_{I}$
$k_{1}^{2}=k_{2}^{2}=k_{I}^{2}=\frac{a_{1}(\alpha_{1}+\Omega_{1})-a_{2}(\alpha_{2}+\Omega_{2})}{c_{2}(\beta_{2}-\alpha_{2})-c_{1}(\beta_{1}-\alpha_{1})}$
(8a)
$\omega_{1}=\omega_{2}=\omega_{I}=\frac{a_{1}(\alpha_{1}+\Omega_{1})c_{2}(\beta_{2}-\alpha_{2})-c_{1}(\beta_{1}-\alpha_{1})a_{2}(\alpha_{2}+\Omega_{2})}{c_{2}(\beta_{2}-\alpha_{2})-c_{1}(\beta_{1}-\alpha_{1})}$
(8b) And these solutions are exact in the parameter surface $\displaystyle
for\ c_{1}b_{2}-c_{2}b_{1}\neq 0:\ if\ and\ only\ if\
\frac{a_{1}b_{2}-a_{2}b_{1}}{c_{1}b_{2}-c_{2}b_{1}}-\frac{a_{1}(\alpha_{1}+\Omega_{1})-a_{2}(\alpha_{2}+\Omega_{2})}{c_{2}(\beta_{2}-\alpha_{2})-c_{1}(\beta_{1}-\alpha_{1})}=0$
(8c) $\displaystyle for\ c_{1}b_{2}-c_{2}b_{1}=0:\ if\ and\ only\ if\
a_{1}b_{2}-a_{2}b_{1}=0$
which is obtained by inserting Eq.(8a) into Eq.(5a), and by the identifying
$|A_{1}|=|A_{2}|$. It can be easily confirmed that on the parameter surface
Eq.(8c) the planar wave solution Eq.(5) with frequency and wave numbers given
by Eqs.(8a) and (8b) are exact solution of BCGLE. Moreover, the predictions of
Eqs.(7) and (8) agree exactly with the numerical results of Fig.1(a).
In Figs.2(a) and 2(b) we specify the surface of Eq.(8c) in some parameter
planes where solid lines represent the parameters satisfying condition (8c).
In Figs.2(c) and 2(d) we vary parameters along the solid line of Fig.2(a) and
numerically compute 1D BCGLE. we compare the numerical results (empty cycles
and triangles) with theoretical predictions of Eqs.(8a) and (8c) (solid
lines), and find: (i) ISWs exist in a large area of surface Eq.(8c); (ii) the
predictions of Eqs.(8a) and (8b) coincide with numerical results exactly
(within computation precision). In Fig.2(e) we fix a parameter set on the
surface (black disk T) and present the asymptotic pattern evolution of the
BCGLE system. It is clearly shown that ISWs, with the wave number and
frequency predicted by Eqs.(8a) and (8b), asymptotically control the entire
bidomain during their propagation. For demonstrating the possibility of
observation of ISWs in experiments, we study a reaction-diffusion model:
bidomain Brusselator. In Fig.2(f), we show ISWs of this chemical model,
satisfying all conditions of Eqs.(7).
The solutions of Eqs.(8a) and (8b) are exact for BCGLE only on the parameter
surface of Eq.(8c). Slightly away from this surface Eqs.(8a) and (8b) can no
longer predict the wave numbers and the frequencies of ISWs exactly. In
Figs.3(a) and 3(b) we compare the theoretical predictions of Eqs.(8a) and (8b)
with numerical results of frequencies and wave numbers by varying parameters
along the dashed line in Fig.2(a). We find: (i) The solutions of Eqs.(8a) and
(8b) are not exact; the wave numbers in the two domains deviate from each
other (about $7.7\%$ difference in Fig.3(d)); (ii) However, slightly away from
the surface Eq.(8c), the feature that the interface generates waves is still
clearly observed (compare Fig.1(b) with Fig.3(c)), i.e., ISWs still exist;
(iii) By continuously increasing the parameter distance from condition
Eq.(8c), the deviation of the numerical results from the theoretical
predictions Eqs.(8a) and (8b) increases continuously from zero too, and for
small parameter deviation the solutions Eqs.(8a) and (8b) can still be used
for predicting frequency and wave numbers with very good approximation. In
Figs.3(c) and 3(d) we show ISWs for a parameter set away from the surface
Eq.(8c) (Disk Q in Fig.2(a)). It is clear that even away from surface Eq.(8c)
the waves of Fig.3(c) are generated by the interface in the similar way as in
Fig.1(b) though we have $k_{1}=k_{2}$ in Fig.1(b) but $k_{1}\neq k_{2}$ in
Fig.3(c). Thus the waves in Fig.1(a), 1(b) and Figs.3(c), 3(d), have obviously
the same interface-selected nature which is essentially different from the
waves of Fig.1(c). Similar ISWs with $k_{1}\neq k_{2}$ can also be observed in
bidomain Brusselator. In Figs.3(e) and 3(f) we take parameter set far away
from that in Fig.2(f), and can still observe ISWs. Here the wave numbers in
the two sides have slight difference
($\frac{2|k_{1}-k_{2}|}{|k_{1}+k_{2}|}\approx 6.77\%$).
There are some necessary conditions for ISWs to appear. Let us analytically
specify some of these conditions under Eq.(8c) (Eqs.(8a) and (8b) are exact
solutions of $\omega_{I}$ and $k_{I}$). From Eqs.(8a) and (5a) we have an
obvious necessary existence condition for ISWs, i.e.,
$0\leq
k_{I}^{2}=\frac{a_{1}(\alpha_{1}+\Omega_{1})-a_{2}(\alpha_{2}+\Omega_{2})}{c_{2}(\beta_{2}-\alpha_{2})-c_{1}(\beta_{1}-\alpha_{1})}\leq\frac{a_{i}}{c_{i}},\
i=1,2$ (9a) which should be satisfied because wave number $k_{I}$ must be
real. If this condition is violated, there is no physically meaningful
solutions of $k_{I}$ and $\omega_{I}$, and thus no ISWs can be observed. This
is the case of Fig.1(c). ISWs are generated by the interface and propagate
along a certain direction. Therefore, the waves must forwardly propagate in
one domain. Precisely, ISWs are NWs in the left (or right) domain while AWs in
the right (left) domain for waves propagating from right (left) to left
(right), and this requires another parameter condition $\displaystyle AWs:\
\omega_{I}a_{i}(\Omega_{i}+\alpha_{i})>0\ and\
|\omega_{I}|<|a_{i}(\Omega_{i}+\alpha_{i})|\ $ (9b) $\displaystyle NWs:\
\omega_{I}a_{\bar{i}}(\Omega_{\bar{i}}+\alpha_{\bar{i}})<0\ or\
\omega_{I}a_{\bar{i}}(\Omega_{\bar{i}}+\alpha_{\bar{i}})>0\ and\
|\omega_{I}|>|a_{\bar{i}}(\Omega_{\bar{i}}+\alpha_{\bar{i}})|\ $
$\displaystyle i=1,\ \bar{i}=2\ or\ i=2,\ \bar{i}=1$
In order to provide an idea how these conditions influence the existence of
ISWs, we show Fig.4, where one can numerically observe ISWs in the regions
enclosed by disks, called ”ISW” regions. In Fig.4 dashed-dotted lines with
$k^{2}_{I}=0$ and $\alpha_{i}=\beta_{i},\ i=1\ or\ 2$ are the boundaries of
”ISW” theoretically predicted by Eqs.(9a)and (9b), respectively. In ”No ISW”
regions ISWs do not exist due to violations of conditions of Eq.(9a) or (9b).
In the regions ”Unstable”, ISWs exist while waves with wave number $k_{I}$ are
unstable due to Eckhaus instability, and there ISWs cannot be numerically
observed. Figures 4(a) and 4(b) are plotted in a small parameter surface under
the condition of Eq.(8c). Similar structure of distributions ”ISW”, ”No ISW”
and ”Unstable” regions can be observed when parameters are varied slightly
away from the set of Eq.(8c).
Though the above investigations are made for 1D bidomain systems, the
observations of ISWs exist generally for high-dimensional systems. In 2D
oscillatory systems, much richer types of waves, including spirals and
antispirals, can be self-sustained, and wave competitions become an important
issue. Now we explore how ISWs play crucial roles in wave competitions. We
consider a 2D BCGLE system with an interface line $II^{\prime}$ in between.
Without the interface interaction, $M_{1}$ supports normal spirals (Figs.5(a),
(d), (g)) and $M_{2}$ supports antispirals ( Figs.5(b), (e), (h)). With the
interface interaction we find characteristically different results of wave
competitions. Fig.5(c): the antispiral wins the competition and dominates the
system with frequency $\omega_{2}$; Fig.5(f): the spiral wins; Fig.5(i): ISWs
win and dominate the two domains. The reasons why we can observe so diverse
results in similar competitions between spiral and antispiral waves can be
completely understood based on the analysis of ISWs.
For explaining the results of Fig.5 we briefly introduce some known
conclusions on wave competitions in oscillatory systems. If competitions occur
in a homogeneous medium, the results are [3, 18]:
$NWs\ against\ NWs:\ faster\ waves\ win$ (10a) $AWs\ against\ AWs:\ slower\
waves\ win$ (10b) $NWs\ against\ AWs:\ NWs\ win$ (10c)
With the competition rules Eq.(10) and the analytical results of Eqs.(7)-(9)
we can fully understand and predict the diverse results of Fig.5.
Inserting the parameters of Fig.5(c) into Eq.(8) we have $\omega_{I}=0.636$.
Considering conditions Eq.(6) we conclude that ISWs are NWs in $M_{1}$ and AWs
in $M_{2}$ (note, the interface is the source of ISWs). The frequencies of the
spiral in $M_{1}$ and the antispiral in $M_{2}$ are $\omega_{1}=0.288$ and
$\omega_{2}=0.626$, respectively. According to conclusion (10) ISWs win the
competition in $M_{1}$ against the spiral while losing the battle in $M_{2}$
against the AW spiral. Therefore, the antispiral waves of frequency
$\omega_{2}$ finally dominate the whole system. The parameters of Fig.5(f) do
not satisfy condition Eq.(9a), and no ISWs can be generated. In Figs.5(d) and
(e) we observe $\omega_{1}=-0.0567$ and $\omega_{2}=0.151$. According to
Eq.(6) waves of $\omega_{1}(\omega_{2})$ are NWs (AWs) in both $M_{1}$ and
$M_{2}$. On the basis of (10c), the spiral of frequency $\omega_{1}$ wins the
competition. The most interesting observation is given in Fig.5(i) where we
have $\omega_{I}=0.932$. The frequency of the spiral(antispiral) in $M_{1}$
($M_{2}$) is $\omega_{1}=0.293$ ($\omega_{2}=1.037$). Therefore, ISWs win both
competitions against the spiral in $M_{1}$ (condition (10a)) and against the
antispiral in $M_{2}$ (condition (10b)). The asymptotic state is ISWs in 2D
system where ISWs suppress both spiral and antispiral in Fig.5(i).
In conclusion, we investigated the role played by interfaces. A new type of
waves, interface selected waves (ISWs) were found in bidomain systems with one
medium supports AWs and the other NWs. When control parameters are on a well
defined parameter surface ISWs propagate with analytically predictable same
frequency and wave number in two media with different parameters. When the
parameters are away but near this surface, ISWs can be also observed, of which
the frequency and wave numbers can be located approximately. These waves are
selected by interfaces between two media, and some necessary conditions for
observing these ISWs are specified. These ISWs play important roles in wave
competitions. For instance, under certain conditions ISWs can suppress spiral
and antispiral waves in both media. These roles are important in practical
applications. Experimental realizations of ISWs in chemical reaction-diffusion
systems are strongly suggested, based on the well-behaved ISWs of Fig.2(f),
Fig.3(e) and Fig.3(f) computed for a chemical reaction-diffusion model.
###### Acknowledgements.
Z. Cao is supported by the knowledge innovation program of the Chinese Academy
of Sciences.
1. (1) R. A. Shelby, D. R. Smith, and S. Schultz, Science 292, 77 (2001).
2. (2) M. Hendrey, E. Ott, and T. M. Antonsen Jr., Phys. Rev. Lett. 82, 859 (1999); Phys. Rev. E 61, 4943 (2000).
3. (3) Z. Cao, H. Zhang, and G. Hu, Eur. Phys. Lett. 79, 34022(2007).
4. (4) M. Vinson, Physica D 116, 313 (1998).
5. (5) M. Zhan, X. Wang, X. Gong, and C. H. Lai, Phys. Rev. E 71, 036212 (2005).
6. (6) L. B. Smolka, B. Marts, and A. L. Lin, Phys. Rev. E 72, 056205 (2005).
7. (7) R. Zhang, L. Yang, A. M. Zhabotinsky, and I. R. Epstein, Phys. Rev. E 76, 016201 (2007);
8. (8) C. G. Parazzoli, D. B. Brock, and S. Schultz, Phys. Rev. Lett. 90, 107401 (2003).
9. (9) V. K. Vanag and I. R. Epstein, Science 294, 835 (2001); Phys. Rev. Lett. 88, 088303 (2002).
10. (10) L. Yang, M. Dolnik, A. M. Zhabotinsky, and I. R. Epstein, J. Chem. Phys. 117, 7259 (2002).
11. (11) Y. Gong and D. J. Christin, Phys. Rev. Lett. 90, 088302 (2003); Phys. Lett. A 331, 209 (2004).
12. (12) L. Brusch, E. M. Nicola, and M. Bär, Phys. Rev. Lett. 92, 089801 (2004); E. M. Nicola, L.Brusch, and M. Bär, J. Phys. Chem. B 108 14733 (2004).
13. (13) P. Kim, T. Ko, H. Jeong, and H. Moon, Phys. Rev. E 70, R065201 (2004).
14. (14) Z. Cao, P. Li, H. Zhang, and G. Hu, the International Journal of Mordern Physics B 21, 4170 (2007).
15. (15) M. Cross and P. Hohenberg, Rev. Mod. Phys. 65, 851 (1993).
16. (16) I. S. Aranson and L. Kramer, Rev. Mod. Phys. 74, 99 (2002).
17. (17) I. S. Aranson, and L. Aranson, Phys. Rev. A 46, R2992 (1992).
18. (18) K. J. Lee, Phys. Rev. Lett. 79, 2907 (1997).
Figure 1: (a)-(c) Contour patterns of Re$A_{i}$ of a 1D BCGLE system with
interface $I$. Domains $M_{i},\ (i=1,2)$ have length $L$ and parameters
$\alpha_{i},\ \beta_{i}$, and $a_{i}=b_{i}=c_{i}=1,\ \Omega_{i}=0,\ i=1,2$.
Numerical simulations are made with space step $dx=0.5$, time step $dt=0.005$,
and $L=200$. No-flux boundary condition, randomly chosen initial conditions
and the above time and space steps are used in all figures for numerical
simulations unless specified otherwise. (a) $\alpha_{1}=0.1,\ \beta_{1}=-1.4,\
\alpha_{2}=-0.2,\ \beta_{2}=1.4$. Interface-selected waves (ISWs) homogeneous
in $M_{1}$ and $M_{2}$ are observed, and the interface is transparent to ISWs.
(b) The same as (a) with early time evolution plotted. It is clearly shown
that ISWs originate from the interface. (c) $\alpha_{1}=0.1,\ \beta_{1}=-1.4,\
\alpha_{2}=0.5,\ \beta_{2}=1.4$. $M_{1}$ and $M_{2}$ support different regular
waves, clearly manifesting interface $I$.
Figure 2: (a) Surface Eq.(8c) (Solid line) is plotted in $a_{2}-b_{2}$ plane
with $\Omega_{2}=0,\ c_{2}=2.0,\ \alpha_{1}=0.6,\ \beta_{1}=-1.4,\
\alpha_{2}=0.1,\ \beta_{2}=1.2$. (b) The same as (a) with surface (8c) plotted
in $\alpha_{2}-a_{2}$ plane, $b_{2}=1.778$. Point E corresponds to the
boundary $k^{2}=0$. (c) Frequency of ISWs which are selected along the solid
line of (a). Theoretical prediction Eq.(8a) (solid line) coincides with
numerical simulation for 1D BCGLE (empty circles and triangles) perfectly. (d)
The same as (c) with wave numbers $k_{1}=k_{2}$ plotted. Agreement between
theoretical prediction Eq.(8a) and numerical results is also confirmed. (e)
ISW pattern obtained by using the parameter set $a_{2}=1.8,\ b_{2}=1.778$
(point T in (a)). (f) The same as (e) with contour pattern of $v$ variable of
Brusselator RD which is numerically computed for 1D chain. The system is:
$\frac{\partial u_{i}}{\partial
t}=a_{i}-(b_{i}+1+\gamma_{i})u+(1+\sigma_{i})u^{2}v+\delta_{ui}\nabla^{2}u,\
\frac{\partial v_{i}}{\partial
t}=b_{i}u-(1+\sigma_{i})u^{2}v+\delta_{vi}\nabla^{2}v,\ i=1,2,\ a_{1}=1.0,\
b_{1}=2.24,\ \gamma_{1}=\sigma_{1}=0.0,\ \delta_{u1}=2.31,\ \delta_{v1}=2.17;\
a_{2}=1.02,\ b_{2}=2.2624,\ \gamma_{2}=0.02,\ \sigma_{2}=0.01,\
\delta_{u2}=0.95,\ \delta_{v2}=2.47;\ dx=dy=0.5,\ dt=0.0025,\ L=300$. ISWs
with identical $w$ and $k$ are observed.
Figure 3: (a) (b) The same as Figs.2(c) and (d), respectively, with parameters
varied along the dashed line of Fig.2(a). Numerical simulations are made for
1D BCGLE. Now deviations between theoretical results of Eqs.(8a), (8b) and the
numerical results are observed. Deviation increases as parameters vary away
from the surface Eq.(8c). (c) (d) The same as Fig.2(e) with different time
intervals plotted, respectively, in which parameters are taken away from the
solid line of Fig.2(a) (point Q, $a_{2}=1.6,\ b_{2}=1.778$). Now ISWs are
still observed while wave numbers in the two sides are slightly
different($\frac{2|\triangle k|}{|k_{1}+k_{2}|}\approx 7.71\%,\ \triangle
k=|k_{1}-k_{2}|$). (e) (f) The same as (c) and (d), respectively, with 1D
Brusselator chain computed. The parameter set is considerable different from
that of Fig.2(f): $a_{1}=1.0,\ b_{1}=3.2,\ \gamma_{1}=\sigma_{1}=0.0,\
\delta_{u1}=1.0,\ \delta_{v1}=0.5;\ a_{2}=1.1,\ b_{2}=3.2,\ \gamma_{2}=0.1,\
\sigma_{2}=0.0,\ \delta_{u2}=1.0,\ \delta_{v2}=2.5$. Now the wave numbers of
the two sides are also slightly different ($\frac{2|\triangle
k|}{|k_{1}+k_{2}|}\approx 6.77\%$).
Figure 4: The distributions of different types of waves in ($\alpha_{1},\
\alpha_{2}$) parameter planes for different sets of ($\beta_{1},\ \beta_{2}$).
Black disks represent the boundaries of ISW regions (”ISW” regions) identified
by direct numerical simulations of 1D BCGLE. In ”No ISW” regions, ISWs do not
exist due to the violations of condition Eq.(9a) (boundary $k^{2}_{I}=0$,
i.e., $\alpha_{1}=\alpha_{2}$) or condition Eq.(9b) (boundary
$\alpha_{i}=\beta_{i},\ i=1$ or $2$) (both presented in dashed-dotted lines).
In the region ”Unstable”, both conditions Eqs.(9a) and (9b) are satisfied, but
waves with the given $k_{I}$ are unstable due to the Eckhaus instability.
$a_{i}=b_{i}=c_{i}=1,\ \Omega_{i}=0,\ i=1,2.$ (a)
$\beta_{1}=-1.4,\beta_{2}=1.4$, (b) $\beta_{1}=2.5,\beta_{2}=0.5$. Black
triangles A and B in Fig.4(a) represent the parameter sets used in Figs.1(a),
(c), respectively.
Figure 5: Wave competitions between spiral, antispiral and ISWs in 2D BCGLE
with interface $II^{\prime}$. $a_{i}=b_{i}=c_{i}=1,\ \Omega_{i}=0,\ i=1,2$.
(a)(d)(g) Spirals in $M_{1}$ medium. (b)(e)(h) Antispirals in $M_{2}$ medium.
Snapshots in (a)(b), (d)(e) and (g)(h) are used as the initial conditions for
the dynamic evolutions of (c), (f) and (i), respectively. (c)(f)(i) The
asymptotic states of bidomain systems with interface $II^{\prime}$. (a)(b)(c)
$\alpha_{1}=0.2,\ \beta_{1}=2.0,\ \alpha_{2}=1.0,\ \beta_{2}=-0.5$. Antispiral
initially in $M_{2}$ dominates the system in (c). (d)(e)(f) $\alpha_{1}=0.2,\
\beta_{1}=-2.0,\ \alpha_{2}=0.5,\ \beta_{2}=-1.4$. Spiral initially in $M_{1}$
dominates the system in (f). (g)(h)(i) $\alpha_{1}=0.1,\ \beta_{1}=3.2,\
\alpha_{2}=1.2,\ \beta_{2}=0.0$. In (i) ISWs suppress both spiral in $M_{1}$
and antispiral in $M_{2}$, and dominate the whole system.
| arxiv-papers | 2008-06-20T03:30:34 | 2024-09-04T02:48:56.355501 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Xiaohua Cui, Xiaoqing Huang, Zhoujian Cao, Hong Zhang, and Gang Hu",
"submitter": "Zhoujian Cao Dr",
"url": "https://arxiv.org/abs/0806.3312"
} |
0806.3543 | # Ghost imaging in scattering media
Wenlin Gong Pengli Zhang Xia Shen Shensheng Han sshan@mail.shcnc.ac.cn Key
Laboratory for Quantum Optics and Center for Cold Atom Physics, Shanghai
Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, Shanghai
201800, China
###### Abstract
Ghost imaging with thermal light in scattering media is investigated. We
demonstrated both theoretically and experimentally for the first time that the
image with high quality can still be obtained in the scattering media by ghost
imaging. The scattering effect on the qualities of the images obtained when
the object is illuminated directly by the thermal light and ghost imaging is
analyzed theoretically. Its potential applications are also discussed.
###### pacs:
42.50.Ar, 42.68.Mj, 42.50.Dv, 42.62.Be, 42.30.Kq
## I introduction
Multiple scattering has a great effect on the qualities of images and the
transmission of information. The information will be decayed and the images
suffer reduced resolution and contrast because of multiple scattering. For
example, the measurement of the laser radar Kamerman , satellite
communications Booker , the propagation and imaging of light in the atmosphere
Mckechnie , neutron imaging Raine and the imaging and diagnosis in life and
medical science Maher . So the imaging in strong scattering media is always a
great problem and presents a key challenge for the research of better imaging
method and technique.
In clinic applications, the most common imaging modalities include ultrasound
imaging, X-ray computed tomography(CT), and magnetic resonance imaging (MRI)
Maher ; Wang . As the development of imaging technology, optical imaging is
becoming an increasing interesting method for the imaging in biological
tissue. By now most imaging methods are obtained by using the gating
techniques. Such as confocal imaging, spatial filtering, optical coherence
tomography (OCT), Mueller optical coherence tomography, Diffuse optical
tomography (DOT), Photoacoustic tomography (PAT), Ultrasound-Modulated optical
tomograph (UOT) and so on Wang ; Gibson ; Yodh ; Arridge1 ; Arridge2 ; Sachin
; Hebden ; Chance ; Laubscher ; Brian . Although the qualities of the images
have a great increase by these techniques, there is still lots of problems
which are difficult to be done. Because the imaging techniques in scattering
media discussed above mainly are only the first-order effect of light field,
detection and imaging are unseparated. When the information of the object is
distorted by multiple scattering, and the information of both multiple
scattering and the object is unknown, so we can not, in principle, obtain
exactly the images destroyed by the multiple scattering, which leads to be
impossible of the restoration of the qualities of images caused by multiple
scattering.
The first two-photon imaging experiment with entangled source was demonstrated
by Pittman _etal_. in 1995 Pittman , which shew that we could obtain a
nonlocal image by transmitting pairs of photons through a test and a reference
path. Since 2002, the theories and experiments demonstrated that the ghost
imaging could also be obtained with thermal light Cheng ; Gatti1 ; Gatti2 ;
Valencia ; Bennink ; Basano . And the fierce discussion on the essence of
ghost imaging at one time Scarcelli1 ; Gatti3 ; Scarcelli2 . Ghost imaging is
considered as the effect of second-order correlation of light field and is
caused by the undistinguishable relation of identical particles Zhang1 ;
Scarcelli1 ; Scarcelli2 . For the first time, detection and imaging are
separated by ghost imaging. The test path and the reference path are used to
detect the information from the object and imaging for the object,
respectively. Recently we find that the qualities of ghost images are
determined by both the reference path and test path Zhang . Because multiple
scattering only degrades the imaging quality in the test path, whereas there
is no multiple scattering in the reference path. By correlation measurement,
we may get a image with much better quality than the image obtained by
detection in a single path.
## II theory and analysis
Figure 1: Schematic diagram for the light transmitting in a scattering media.
Schematic diagram for the light transmitting in a scattering media is shown in
Fig. 1. In the theory of linear systems Goodman , the light field $E(x)$ on
the plane $x$ is the convolution of the light field $E(x_{0})$ on the plane
$x_{0}$ and the impulse response function $h(x,x_{0})$.
$\displaystyle E(x)=\int{dx_{0}E(x_{0})}h(x,x_{0}).$ (1)
For light transmission in scattering media, the light field on the position
$x$ is the linear superposition the incident light and the scattering light.
$\displaystyle E(x)=\alpha\int{dx_{0}}E(x_{0})h_{in}(x,x_{0})$
$\displaystyle+\beta\int dx_{0}E(x_{0})h_{sca}(x,x_{0}),$ (2)
$\displaystyle\left|\alpha\right|^{2}+\left|\beta\right|^{2}=1.$ (3)
where $h_{in}(x,x_{0})$ is the impulse response function with no scattering
media, and $h_{sca}(x,x_{0})$ is the impulse response function from the plane
$x_{0}$ to the plane $x$ because of the interactions of multiple scattering,
and $\alpha$, $\beta$ are the probability amplitudes of the incident light and
the scattering light, respectively. From Eqs. (1)-(3), we have
$\displaystyle h(x,x_{0})=\alpha h_{in}(x,x_{0})+\beta h_{sca}(x,x_{0}).$ (4)
The probability distribution function in scattering media is called point
scattering function. The impulse response function $h_{sca}(x,x_{0})$ has
close contact with the point scattering function which is Dirac delta function
when there is no scattering media. However, in the scattering media, it is a
spread function with a broadening length, and generally the point scattering
function has two forms: Lorentzian-shaped and Gaussian-shaped distribution
Hassanein ; Segre . In multiple scattering Mie theoryMie ; Toublanc ; Sharma ,
both of probability amplitudes $\alpha$, $\beta$ are depending on the diameter
size of the particle $D$, the wavelength of the incident light $\lambda$, the
concentration of suspended particles $w$ and the effective length of
scattering media $L$. According to the experiments and theories Jessica ;
Milun ; Mustafa ; Faulkner ; wells ; yura ; Hassanein ; Segre , we have
$\displaystyle\alpha=\alpha(D,\lambda,w,L)\propto\frac{{\lambda^{b_{\alpha}}}}{{D^{a_{\alpha}}w^{c_{\alpha}}L^{d_{\alpha}}}},$
(5a)
$\displaystyle\beta=\beta(D,\lambda,w,L)\propto\frac{{D^{a_{\beta}}w^{c_{\beta}}L^{d_{\beta}}}}{{\lambda^{b_{\beta}}}},$
(5b) $\displaystyle h_{sca}(x,x_{0})\propto\int
dx^{\prime}P(x^{\prime},x_{0})_{L_{A}}h(x,x^{\prime})_{(L+d)},$ (5c)
$\displaystyle P(x^{\prime},x_{0})_{L_{A}}=[\frac{2}{{\pi\Delta
x_{L_{A}}^{2}}}]^{1/4}$
$\displaystyle\times\exp\left\\{{-(\frac{{x^{\prime}-x_{0}}}{{\Delta
x_{L_{A}}}})^{2}}\right\\},$ (5d) $\displaystyle\Delta
x_{L_{A}}\propto\frac{{D^{a_{x}}w^{c_{x}}{L_{A}}^{d_{x}}}}{{\lambda^{b_{x}}}},$
(5e)
$\displaystyle\int{\left|{P(x^{\prime},x_{0})_{L_{A}}}\right|}^{2}dx^{\prime}=1.$
(5f)
where $P(x^{\prime},x_{0})_{L_{A}}$ is point scattering probability amplitude.
$\Delta x_{L_{A}}$ is broadening length because of the interactions of
multiple scattering, and it becomes wider with the increase of the scattering
length. With the increase of the broadening length, the frequency spectrum of
the optical transfer function becomes narrower, which is the main reason
leading to the degradation of the quality of information transmission and
images wells ; yura . We suppose point scattering function is Gaussian-shaped
distribution without considering the absorption of scattering media. All the
coefficients in Eq. (5) should be determined by specific experimental
conditions.
The scheme for ghost imaging with thermal light in the scattering media is
shown in Fig. 2. The light source $S$, first propagates through a beam
splitter, then is divided into a test and a reference path. In the test path,
the light propagates through a single lens of focal length $f_{1}$, the
scattering media and then to the detector $D_{t}$. In the reference path, the
light propagates through a single lens of focal length $f$ then to an array of
pixel detector $D_{r}$.
Figure 2: Scheme for ghost imaging with thermal light in the scattering media.
By optical coherence theory Cheng ; Glauber , we can obtain the correlation
function of intensity fluctuations between the detectors:
$\displaystyle\Delta G^{(2,2)}(x_{r},x_{t})=\left\langle{\Delta
I_{r}(x_{r})\Delta
I_{t}(x_{t})}\right\rangle=\left|{\Gamma(x_{r},x_{t})}\right|^{2},$ (6a)
$\displaystyle\Gamma(x_{r},x_{t})=\int{dx_{1}}\int{dx_{2}G^{(1,1)}(x_{1},x_{2})h_{r}^{*}(x_{r},x_{1})}$
$\displaystyle\times h_{t}(x_{t},x_{2}).$ (6b)
where $\Gamma(x_{r},x_{t})$ is the first-order cross-correlation function of
two different points from the test and reference paths.
Suppose the light source is fully spatially incoherent, then
$G^{(1,1)}(x_{1},x_{2})=I_{0}\delta(x_{1}-x_{2}).$ (7)
where $I_{0}$ is a constant, and $\delta(x)$ is the Dirac delta function.
Under the paraxial and small angle approximation, and when the effective
apertures of the lenses in the optical system are large enough, the impulse
response function of the reference system is
$h_{r}(x_{r},x_{1})\propto\exp\left\\{{\frac{{j\pi}}{{\lambda
f}}(1-\frac{z}{f})x_{1}^{2}-\frac{{2j\pi}}{{\lambda f}}x_{r}x_{1}}\right\\}.$
(8)
And $f_{2},z_{2}+L_{2},z_{3}$ obey the Lens Law
$\frac{1}{{z_{2}+L_{2}}}+\frac{1}{{z_{3}}}=\frac{1}{{f_{2}}}.$ (9)
when
$\displaystyle\frac{{f_{1}}}{f}=\frac{{z_{2}+L_{2}}}{{z_{3}}}.$ (10)
The impulse response function of the test system is
$\displaystyle
h(x_{t},x_{2})\propto\int{dx^{\prime}}[\alpha_{1}\exp\left\\{{-\frac{{2j\pi}}{{\lambda
f_{1}}}x^{\prime}x_{2}}\right\\}+\beta_{1}\int dx_{2}^{\prime}$
$\displaystyle\times
P(x^{\prime},x_{2}^{\prime})_{L_{1A}}\exp\left\\{{-\frac{{2j\pi}}{{\lambda
f_{1}}}x_{2}^{\prime}x_{2}}\right\\}]t(x^{\prime})C(x^{\prime})$
$\displaystyle\times\exp\left\\{{\frac{{j\pi}}{{\lambda
f_{1}}}(1-\frac{{z_{1}+L_{1}}}{{f_{1}}})x_{2}^{2}}\right\\},$ (11a)
$\displaystyle
C(x)=[\alpha_{2}\delta(x+\frac{{f_{1}}}{f}x_{t})\exp\left\\{{\frac{{j\pi}}{{\lambda(L_{2}+z_{2})}}x^{2}}\right\\}+\beta_{2}$
$\displaystyle\times
P(-\frac{{f_{1}}}{f}x_{t},x)_{L_{2A}}\exp\left\\{{\frac{{j\pi}}{{\lambda(L_{2}+z_{2})}}(\frac{{f_{1}}}{f}x_{t})^{2}}\right\\}].$
(11b)
Substituting Eqs. (7) and (9)-(11b) into Eq. (6b), we can get the intensity
distribution in the test path
$\displaystyle
I(x_{t})\propto\int{dx^{\prime}}\int{dx^{\prime\prime}}t(x^{\prime})t^{*}(x^{\prime\prime})\\{\left|{\alpha_{1}}\right|^{2}\delta(x^{\prime}-x^{\prime\prime})$
$\displaystyle+2P(x^{\prime},x^{\prime\prime})_{L_{1A}}{\mathop{\rm
Re}\nolimits}[\alpha_{1}^{*}\beta_{1}]+\left|{\beta_{1}}\right|^{2}\int{dx_{2}^{\prime}}$
$\displaystyle\times
P(x^{\prime},x_{2}^{\prime})_{L_{1A}}P(x^{\prime\prime},x_{2}^{\prime})_{L_{1A}}\\}C(x^{\prime})C^{*}(x^{\prime\prime}).$
(12)
Eq. (12) describes the images in the scattering media for the direct
illumination of the thermal light . When
$\displaystyle\frac{{1-\frac{z}{f}}}{f}=\frac{{1-\frac{{z_{1}+L_{1}}}{{f_{1}}}}}{{f_{1}}}.$
(13)
substituting Eqs. (7)-(11b) and (13) into Eq. (6), we can obtain the
correlation function of intensity fluctuations
$\displaystyle\Delta
G^{(2)}(x_{r},x_{t})\propto\left|{\alpha_{1}\alpha_{2}C_{1}+\alpha_{1}\beta_{2}C_{2}}\right.$
$\displaystyle\left.{+\beta_{1}\alpha_{2}C_{3}+\beta_{1}\beta_{2}C_{4}}\right|^{2},$
(14a) $\displaystyle C_{1}=\delta(x_{r}+x_{t})t(-\frac{{f_{1}}}{f}x_{t}),$
(14b) $\displaystyle
C_{2}=t(\frac{{f_{1}}}{f}x_{r})P(-\frac{{f_{1}}}{f}x_{t},\frac{{f_{1}}}{f}x_{r})_{L_{2A}},$
(14c) $\displaystyle
C_{3}=t(-\frac{{f_{1}}}{f}x_{t})P(-\frac{{f_{1}}}{f}x_{t},\frac{{f_{1}}}{f}x_{r})_{L_{1A}},$
(14d) $\displaystyle
C_{4}=\int{dx^{\prime}t(x^{\prime})}P(x^{\prime},\frac{{f_{1}}}{f}x_{r})_{L_{1A}}P(-\frac{{f_{1}}}{f}x_{t},x^{\prime})_{L_{2A}}.$
(14e)
The ghost imaging in the scattering media is described by the Eqs.
(14a)-(14e). And it is a image with high quality for $C_{1}$ and $C_{2}$,
which implies we can still get a image with high quality by ghost imaging in
the scattering media.
From Eqs. (14a)-(14e), if the test detector is an array of pixel detector, and
$x_{t}=-x_{r}$, after some calculation, then
$\displaystyle\Delta
G^{(2)}(x_{r},-x_{r})\propto\left|{\left\\{{\alpha_{1}\alpha_{2}+\alpha_{1}\beta_{2}\left[{\frac{2}{{\pi\Delta
x_{L_{2A}}^{2}}}}\right]^{1/4}}\right.}\right.$
$\displaystyle\left.{+\beta_{1}\alpha_{2}\left.{\left[{\frac{2}{{\pi\Delta
x_{L_{1A}}^{2}}}}\right]^{1/4}}\right\\}t(\frac{{f_{1}}}{f}x_{r})+\beta_{1}\beta_{2}C_{4}}\right|^{2}.$
(15)
if the test detector is a bucket detector, then
$\displaystyle\Delta
G^{(2)}(x_{r})\propto\int{dx_{t}}\left|{\alpha_{1}\alpha_{2}C_{1}+\alpha_{1}\beta_{2}C_{2}}\right.$
$\displaystyle\left.{+\beta_{1}\alpha_{2}C_{3}+\beta_{1}\beta_{2}C_{4}}\right|^{2}.$
(16)
Eqs. (15) and (16) represent ghost imaging when the test detector is an array
of pixel detector or a bucket detector, respectively. From Eqs. (15), if
$\beta_{1}=0$(namely $L_{1}=0$) or $\beta_{2}=0$(namely $L_{2}=0$), it is
clear to find that we can still obtain a image with high quality by ghost
imaging in the scattering media. if $\beta_{1}\neq 0$ and $\beta_{2}\neq 0$
and as the increase of $\beta_{1}$ and $\beta_{2}$, the qualities of ghost
images will reduce, and the term including $C_{4}$ is the main reason leading
to the degradation of the qualities of the images.
For $L_{1}=0$, there is only multiple scattering between the object plane and
the test detector, then we have $\left|\alpha_{1}\right|=1,\beta_{1}=0$. By
Eq. (12), the intensity distribution in the test path is
$\displaystyle
I_{t}(x_{t})\propto(\left|{\alpha_{2}}\right|^{2}+2C_{5t})\left|{t(-\frac{{f_{1}}}{f}x_{t})}\right|^{2}+C_{6t}\left|{\beta_{2}}\right|^{2},$
(17a) $\displaystyle C_{5t}=[\frac{2}{{\pi\Delta
x_{L_{2A}}^{2}}}]^{1/4}{\mathop{\rm Re}\nolimits}[\alpha_{2}^{*}\beta_{2}],$
(17b) $\displaystyle
C_{6t}=\int{dx^{\prime}}\left|{t(x^{\prime})}\right|^{2}\left|{P(-\frac{{f_{1}}}{f}x_{t},x^{\prime})_{L_{2A}}}\right|^{2}$
$\displaystyle=[\frac{2}{{\pi\Delta
x_{L_{2A}}^{2}}}]^{1/2}\int{dx^{\prime}}\left|{t(x^{\prime})}\right|^{2}$
$\displaystyle\times\exp\left\\{{-\frac{2}{{\Delta
x_{L_{2A}}^{2}}}(x^{\prime}+\frac{{f_{1}}}{f}x_{t})^{2}}\right\\}.$ (17c)
the last term in Eq. (17a) is the main reason leading to the decrease of the
quality of the image when there is multiple scattering between the object
plane and the detector. With the increase of $\beta_{2}$ (and the decrease of
the probability amplitude $\alpha_{2}$), the quality of the image will be
further degraded.
Form Eq. (15), when the test detector is an array of pixel detector, after
some calculation, then
$\displaystyle\Delta
G^{(2)}(x_{r},-x_{r})\propto\left|{\alpha_{2}+\beta_{2}\left[{\frac{2}{{\pi\Delta
x_{L_{2A}}^{2}}}}\right]^{1/4}}\right|^{2}$
$\displaystyle\times\left|{t(\frac{{f_{1}}}{f}x_{r})}\right|^{2}.$ (18)
If the test detector is a bucket detector, By Eq. (16), then
$\displaystyle\Delta
G^{(2)}(x_{r})\propto(\left|\alpha_{2}\right|^{2}+2C_{5}+C_{6}\left|\beta_{2}\right|^{2})\left|{t(\frac{{f_{1}}}{f}x_{r})}\right|^{2},$
(19a) $\displaystyle C_{5}=[\frac{2}{{\pi\Delta
x_{L_{2A}}^{2}}}]^{1/4}{\mathop{\rm Re}\nolimits}[\alpha_{2}^{*}\beta_{2}],$
(19b) $\displaystyle
C_{6}=\int{dx_{t}\left|{P(-\frac{{f_{1}}}{f}x_{t},\frac{{f_{1}}}{f}x_{r})_{L_{2A}}}\right|^{2}}\sim
1.$ (19c)
from Eqs. (18)-(19c), we find that whether the test detector is an array of
pixel detector or a bucket detector, the qualities of ghost images can be
obtained even though there is multiple scattering between the object plane and
the test detector.
For $L_{2}=0$, there is only multiple scattering between the source and the
object plane, then we can gain $\left|\alpha_{2}\right|=1,\beta_{2}=0$. By Eq.
(12), the intensity distribution in the test path is
$\displaystyle
I_{t}(x_{t})\propto(\left|{\alpha_{1}}\right|^{2}+2C_{7t}+C_{8t}\left|{\beta_{1}}\right|^{2})\left|{t(-\frac{{f_{1}}}{f}x_{t})}\right|^{2},$
(20a) $\displaystyle C_{7t}=[\frac{2}{{\pi\Delta
x_{L_{1A}}^{2}}}]^{1/4}{\mathop{\rm Re}\nolimits}[\alpha_{1}^{*}\beta_{1}],$
(20b) $\displaystyle
C_{8t}=\int{dx_{2}^{\prime}}\left|{P(-\frac{{f_{1}}}{f}x_{t},x_{2}^{\prime})_{L_{1A}}}\right|^{2}\sim
1.$ (20c)
From Eqs. (20a)-(20c), we find that the multiple scattering between the source
and the object plane has no effect on the quality of the image.
From Eq. (15), when the test detector is an array of pixel detector, after
some calculation, then
$\displaystyle\Delta
G^{(2)}(x_{r},-x_{r})\propto\left|{\alpha_{1}+\beta_{1}\left[{\frac{2}{{\pi\Delta
x_{L_{1A}}^{2}}}}\right]^{1/4}}\right|^{2}$
$\displaystyle\times\left|{t(\frac{{f_{1}}}{f}x_{r})}\right|^{2}.$ (21)
which reveals that we can still obtain a image with high quality when the test
detector is an array of pixel detector even if there is multiple scattering
between the source and the object plane.
If the test detector is a bucket detector, By Eq. (16), then
$\displaystyle\Delta
G^{(2)}(x_{r})\propto(\left|\alpha_{1}\right|^{2}+2C_{7})\left|{t(\frac{{f_{1}}}{f}x_{r})}\right|^{2}+C_{8}\left|\beta_{1}\right|^{2},$
(22a) $\displaystyle C_{7t}=[\frac{2}{{\pi\Delta
x_{L_{1A}}^{2}}}]^{1/4}{\mathop{\rm Re}\nolimits}[\alpha_{1}^{*}\beta_{1}],$
(22b) $\displaystyle
C_{8}=\int{dx_{t}}\left|{t(-\frac{{f_{1}}}{f}x_{t})P(-\frac{{f_{1}}}{f}x_{t},\frac{{f_{1}}}{f}x_{r})_{L_{1A}}}\right|^{2}$
$\displaystyle=\left[{\frac{2}{{\pi\Delta
x_{L_{1A}}^{2}}}}\right]^{1/2}\int{dx_{t}}\left|{t(-\frac{{f_{1}}}{f}x_{t})}\right|^{2}$
$\displaystyle\times\exp\left\\{{-\frac{{2f_{1}^{2}}}{{\Delta
x_{L_{1A}}^{2}f^{2}}}(x_{t}+x_{r})^{2}}\right\\}.$ (22c)
where $C_{8}$ is the main factor leading to the decrease of the qualities of
ghost images. $f_{1}>f$ is helpful to improve the quality of the image.
The second-order degree of coherence $g^{(2)}(x_{1},x_{2})$ at the positions
$x_{1}$ and $x_{2}$ can be defined as follows
$\displaystyle g^{(2)}(x_{1},x_{2})=\frac{{\left\langle{I_{1}\cdot
I_{2}}\right\rangle}}{{\left\langle{I_{1}}\right\rangle\left\langle{I_{2}}\right\rangle}}=1+\frac{{\left\langle{\Delta
I_{1}\cdot\Delta
I_{2}}\right\rangle}}{{\left\langle{I_{1}}\right\rangle\left\langle{I_{2}}\right\rangle}}.$
(23)
for ghost imaging, The second-order degree of coherence
$g^{(2)}(x_{t1},x_{t2})$ obtained only by the correlation measurement in the
test path reveals the characteristic of the source, whereas
$g^{(2)}(x_{r},x_{t})$ describes the correlation between two paths. And the
degradation of $g^{(2)}(x_{r},x_{t})$ will lead to the decay of intensity
fluctuations, which has a great effect on the visibility of ghost images.
Figs. 3-7 present numerical results of imaging a single slit in scattering
media based on Eqs. (12) and (15)-(16) (in which we take $\lambda$=650nm,
$f_{1}$=400mm, $f$=250mm, the single slit width a=0.2mm). From Fig. 3(a), as
the increase of $L_{2}$, the qualities of the images will decrease obviously
when the object is illuminated directly by the thermal light. The quality of
the image also degrade rapidly with the increase of the broadening length
$\triangle x_{2}$ and the decrease of the probability amplitude $\alpha_{2}$
when the object is fixed in the position of $L_{1}$=0mm, $L_{2}$=100mm (Fig.
3(b), (c)), which accord with the results described by the Eqs (17a)-(17c).
Figure 3: Factors which have the effect on the qualities of the images when
the object is illuminated directly by the thermal light. (a). Images of a
single slit at different positions in the scattering media; (b). Images of a
single slit for different probability amplitude $\alpha_{2}$ when $L_{1}$=0mm,
$L_{2}$=100mm and the broadening length $\triangle x_{2}$=5.0; and (c). Images
of a single slit for different probability amplitude $\triangle x_{2}$ when
$L_{1}$=0mm, $L_{2}$=100mm and the broadening length $\alpha_{2}$=0.05.
Fig. 4 shows the numerical results when the position of the object in the
scattering media is shifted. By Fig. 4(a), we can find that if the test
detector is an array of pixel detector, multiple scattering has no effect on
the qualities of ghost images when there is only multiple scattering either
between the object plane and the test detector or between the object plane and
the source. As the object is fixed closing to the middle of scattering media,
the qualities of ghost images will decay. However, if the test detector is a
bucket detector, the qualities of ghost images will decrease sharply with the
degradation of $L_{1}$, which is opposite absolutely to imaging when the
object is illuminated directly by the thermal light (Fig. 3(a)).
Figure 4: Relationship between the qualities of ghost images and the position
of the object in scattering media. (a). the test detector is an array of pixel
detector; and (b). the test detector is a bucket detector.
In Fig. 5, We give the comparison of the qualities of images between imaging
by the direct illumination of the thermal light and ghost imaging when the
object is fixed in the middle of scattering media. It is easy to find that we
can only obtain the same image with low quality as imaging by the direct
illumination of the thermal light when the test detector is a bucket detector
by ghost imaging. But a image with high quality can still be gained if the
test detector is an array of pixel detector.
Figure 5: Comparison between imaging by the direct illumination of the thermal
light and ghost imaging when $L_{1}$=50mm, $L_{2}$=50mm, the broadening length
$\triangle x_{1}$=2.0, $\triangle x_{2}$=2.0 and $\alpha_{1}$=0.1,
$\alpha_{2}$=0.1.
Results shown in Fig. 6 reveals that the qualities of ghost images will reduce
obviously as the degradation of the probability amplitude $\alpha_{1}$ and
$\alpha_{2}$. And the object is still fixed in the middle of scattering media.
Figure 6: Dependence of probability amplitude $\alpha_{2}$ on the qualities of
ghost images when $L_{1}$=50mm, $L_{2}$=50mm and the broadening length
$\triangle x_{1}$=2.0, $\triangle x_{2}$=2.0.
In order to improve the qualities of ghost images further, we investigate a
new method to the effect on the qualities of ghost images by using a circular
aperture gating to change the transverse coherent length near the scattering
media in Fig. 7, which is different from the method to change the transverse
size of the source. But when the space interval $\Delta s$ is greater than the
characteristic scale of the object, the diffraction will emerge.
Figure 7: The effect of the space interval $\Delta s$ on the qualities of
ghost images when a circular aperture gating with different space interval
$\Delta s$ is fixed on the plane $x_{0}$. $L_{1}$=50mm, $L_{2}$=50mm, the
broadening length $\triangle x_{1}$=2.0, $\triangle x_{2}$=2.0 and
$\alpha_{1}$=0.1, $\alpha_{2}$=0.1.
## III experiment
In the experiments, we prepare a suspension liquid which is composed by
emulsion polymerization particles with particle diameter $D$=3.26$\mu$m and
the solution $NaCl$ with the density $\rho$=1.19 $g/cm^{3}$. The vessel used
to put the suspension liquid is designed as 40mm$\times$40mm$\times$20mm. The
liquid can be considered as strong multiple scattering media. And we take
$\lambda$=650nm, $f_{1}$=400mm, $f$=250mm, $z$=211mm, $z_{1}$=300mm,
$z_{2}$=390mm, $z_{3}$=243.8mm. The minimum characteristic scale of the object
(‘zhong’ ring) is 60 $\mu$m and the diameter of the ring is 1.6mm. And the
detectors in both paths are arrays of pixel detectors.
Images shown in Fig. 8 (1) and (2) were the experimental results of the object
(‘zhong’ ring) by direct illumination and ghost imaging when the object was
fixed at different positions in the scattering media, respectively. Form Fig.
8, when the object is illuminated directly by the thermal light, the qualities
of images will reduce as the increase of the length $L_{2}$ of scattering
media. However, when there is only strong multiple scattering between the
object plane and the source or between the object plane and the detector
$D_{t}$, we can both obtain ghost images with high qualities. If the object is
fixed in the middle of the scattering media, the visibility of ghost images
will reduce, but the resolution doesn’t degrade. All discussed above accord
with the theoretical results described by the Eqs. (1).
Figure 8: Images of the aperture (‘zhong’ ring) when the object is fixed in
different position of the scattering media. (a). $L_{1}$=0mm, $L_{2}$=40mm;
(b). $L_{1}$=20mm, $L_{2}$=20mm; and (c). $L_{1}$=40mm, $L_{2}$=0mm. (1). when
the object was illuminated directly by thermal light; (2). ghost imaging.
In Fig. 9, we demonstrated experimentally that the effect of the concentration
of scattering media or the coherent length located at the object plane on
ghost imaging when the object is fixed in the middle of scattering media. The
visibility of ghost images will reduce with the increase of the concentration
of scattering media. When the coherent length located at the object plane
becomes long, the visibility of ghost images will improve. However, the
resolution will degrade.
Figure 9: Effect of the concentration of scattering media and the coherent
length located at the object plane on ghost imaging. (a). $L_{c}$=40.6$\mu$m,
6 drops; (b). $L_{c}$=40.6$\mu$m, 3 drops; and (c). $L_{c}$=192.5$\mu$m, 6
drops. (1). when the object was illuminated directly by thermal light; (2).
ghost imaging.
Carves shown in Figs. 10 and 11 are the effect of concentration of emulsion
polymerization particles on $g^{(2)}$ when $L_{1}$=40mm and $L_{2}$=0mm and
when $L_{1}$=20mm and $L_{2}$=20mm without the object, respectively. We can
find that $g^{(2)}(x_{t1},x_{t2})$ and $g^{(2)}(x_{r},x_{t})$ decay sharply
with the increase of concentration of emulsion polymerization particles. When
there is two drops emulsion polymerization particles put into the solution
$NaCl$, the cross-correlation coefficient will be lower than 1.10.
Figure 10: The effect of concentration of emulsion polymerization particles on
$g^{(2)}$ when $L_{1}$=40mm and $L_{2}$=0mm without the object.
Figure 11: The effect of concentration of emulsion polymerization particles on
$g^{(2)}$ when $L_{1}$=20mm and $L_{2}$=20mm without the object.
When there are five drops emulsion polymerization particles put into the
solution $NaCl$, the intensity distribution on the detector $D_{r}$ is
heterogenous because there is no multiple scattering. However, we can find the
homogeneous intensity distribution on the detector $D_{t}$ (in Fig. 12).
Figure 12: Images from a single frame speckle distribution when $L_{1}$=20mm
and $L_{2}$=20mm without the object and there are five drops emulsion
polymerization particles put into the solution $NaCl$, (a). from the CCD
camera $D_{t}$, (b). from the CCD camera $D_{r}$.
## IV discussion and inclusion
When the object is illuminated directly by thermal light, multiple scattering
between the object plane and the detector $D_{t}$ is the main reason leading
to the degradation of the qualities of images (Fig. 3). In scattering media,
different from images obtained by the first-order correlation of light field,
ghost imaging causes the separation of detection and imaging. Even if the
information from the object is distorted by multiple scattering, there is no
effect on ghost imaging. Because multiple scattering destroys the correlation
of the light field, the visibility of ghost images is degraded but the
resolutions doesn’t degrade. However, the visibility of ghost images can be
enhanced by means of the decrease of the resolution Shen ; Cai1 ; Cai2 ;
Gatti4 . The experimental result in Fig. 9 (c) also demonstrated that.
Transmission information of the light on the object plane (namely
$\alpha_{1}$) has a great effect on the visibility of ghost images. Because
multiple scattering leads to the decay of the correlation between the test and
reference pathes with the decrease of Transmission information of the light
(Figs 8 and 9). The higher the Transmission information is, the better
visibility the ghost image has. When there is no multiple scattering between
the object plane and the source, then ballistic component of the light on the
object plane is equal to 1, so we can always obtain a ghost image with high
quality whether there is multiple scattering between the object plane and the
detector $D_{t}$. Otherwise, because the ballistic component of the light is
impossible to be 0, we can always see the shapes of the object (Figs. 3(b) and
6). The degradation of the qualities of the images is provoked by the decay of
visibility in scattering media, which is different absolutely from the low
resolution caused by diffraction-limited. When the object is fixed in the
middle of the scattering media, as shown in Fig. 11, even if the second-order
degree of coherence $g^{(2)}$ is very low, we can obtain a ghost image with
much better quality than the image gained when the object is illuminated
directly by thermal light. By the characteristics of ghost imaging, the
qualities of ghost images does not depend on the intensity of the light but on
the intensity fluctuations. As the multiple scattering increased, the
intensity fluctuations will reduce (Figs 10 and 11), which leads to the
degradation of the visibility of the image. The expression in the Eq. (15) can
explain basically the experimental results and the fact discussed above, which
shows that the results described by the Eq. (15) are reasonable for the
description of ghost imaging in scattering media. However, when the test
detector is a bucket detector, the qualities of ghost images depend on the
multiple scattering between the object plane and the source, which is opposite
to imaging by the direct illumination of the thermal light (Fig. 3(a)). By the
results in Fig. 8 (a) and (c), maybe we can also obtain a ghost image with
high visibility by means of eliminating the multiple scattering between the
object plane and the source or between the object plane and the detector
$D_{t}$. One is like the technique of DOT. Firstly we obtain a improved image
by iterative recovery method of DOT from a single frame of image collected by
the detector $D_{t}$, then we will get a ghost image with improved visibility
by the correlation measurement of intensity fluctuations between the detector
$D_{r}$ and the improved images. The other is that we get the intensity
distribution on the object plane by the measurement to transmission and
reflection coefficients, then do the correlation measurement of the intensity
fluctuations between the detector $D_{t}$ and intensity distribution obtained
by numerical simulation. The method discussed in Fig. 7 can improve the
visibility of ghost images to some extent. Generally speaking, by ghost
imaging method, we can always obtain a image with much better quality on the
base of the image obtained by a novel conventional imaging technique with
thermal light. For entangled source, because of the entanglement
characteristic of the two photons, maybe we can also improve the visibility of
ghost images. Based on the effect of photon bunching, and the new source with
high $g^{(2)}$ can be obtained, the visibility of ghost images may also be
enhanced obviously.
In medical science, in order to avoid ionizing radiation of X-ray, optical
photons provides nonionizing and safe radiation for medical applications.
Recently there has been increasing interest in the field of the imaging, test
and diagnosis of the biological tissues with the infrared and the near
infrared light Wang ; Sachin ; Brian . Because the near infrared light around
700-nm wavelength can penetrate several centimeters into biological tissue
Brian . But several factors still limit the imaging quality. Because most
biological tissues are characterized by strong optical scattering and hence
are referred to as scattering media or turbid media. The images suffer reduced
resolution and contrast due to multiple scattering, which leads to a low
efficiency and accuracy of diagnosis and a difficulty of analysis in medical
science. So the diffusion-like behavior of light in biological tissue presents
a key challenge for optical imaging. The ghost imaging discussed here may
solve the problems about the low quality of the imaging in biological tissues.
In conclusion, for the first time, detection and imaging of the object
information are separated by ghost imaging, which provides a new way for
imaging in scattering media. we have demonstrated experimentally and
theoretically for the first time that the images with high qualities in a
scattering media can still be obtained for ghost imaging when the test
detector is an array of pixel detector and there is only multiple scattering
between the object plane and the source or between the object plane and the
detector. When the object is fixed in the middle of scattering media, we can
also gain ghost images with much better qualities than the images obtained
when the object is illuminated directly by the thermal light. The probability
amplitudes of the incident light $\alpha_{1}$ and $\alpha_{2}$ have a great
effect on the visibility of ghost images. These results will be very useful
for the imaging and diagnosis in medical science.
The work was partly supported by the Hi-Tech Research and Development Program
of China, Project No. 2006AA12Z115, and Shanghai Fundamental Research Project,
Project No. 06JC14069.
## References
* (1) G. W. Kamerman, Proc. SPIE, 4377, 126-131 (2001).
* (2) H. G. Booker, and D. C. Miller, Joural of Atmospheric and Terrestrial Physics, 42, 3 (1980).
* (3) T. S. Mckechnie, J. Opt. Soc. Am. A, 8,2 (1991).
* (4) Raine D. A. , and Brenizer J. S. Materials evalution, 55, 1174 (1997).
* (5) K. P. Maher, J. F . Malone, and Brain Chance, Contemporary Physics, 38, 131 (1997).
* (6) Lihong V. Wang and Hsin-i Wu, Biomedical optics: principle and imaging, (2007).
* (7) A. P. Gibson, J. C. Hebden and S. R. Arridge, Phys. Med. Biol. 50 R1-R43 (2005).
* (8) A. Yodh, and B. Chance, Phys. Today, 48, 34 (1995).
* (9) S. R. Arridge, M. Cope, and D. T. Deply, Phys. Med. Biol. 37, 1531 (1992).
* (10) S. R. Arridge, Topical Review, R41-93 (1999).
* (11) Sachin V. patwardhan, and Joseph P. Culver, J. Biomed. Opt. 13(1), 011009 (2008).
* (12) J. C. Hebden, S. R. Arridge, M. Cope, and D. T. Deply, Phys. Med. Biol. 42, 825 (1997).
* (13) B. Chance and R. R. Alfano, Proc. SPIE, 2389 (1995).
* (14) M. Laubscher, M. Ducros, B. Karamata, Opt. Exp. 10, 9 (2002)
* (15) Brian C. Wilson, Eva M. Sevick, Michael S. Patterson, and Britton Chance, Proceedings of the IEEE. 80, 6 (1992).
* (16) T. B. Pittman, Y. H. Shih, D. V. Strekalov, and A. V. Sergienko, Phys. Rev. A. 52. R3429 (1995).
* (17) J. Cheng and S. Han, Phys. Rev. Lett. 92. 093903 (2004).
* (18) A. Gatti, E. Brambilla, M. Bache, and L. A. Lugiato, Phys. Rev. A. 70. 013802 (2004).
* (19) A. Gatti, E. Brambilla, M. Bache, and L. A. Lugiato, Phys. Rev. Lett. 93. 093602 (2004).
* (20) A. Valencia, G. Scarcelli, M. D’Angelo, and Y. Shih, Phys. Rev. Lett. 94. 063601 (2005).
* (21) R. S. Bennink, S. J. Bentley, and R. W. Boyd, Phys. Rev. Lett. 89. 113601 (2002).
* (22) L. Basano and P. Ottonello, Appl. Phys.Lett. 89, 091109 (2006).
* (23) Minghui Zhang, Qing Wei, and Shensheng Han, Phys. Lett. A. 366, 569-574 (2007).
* (24) G. Scarcelli, V. Berardi and Y. Shih, Phys. Rev. Lett. 96, 063602 (2006).
* (25) A. Gatti, M. Bondani, L. A.Lugiato, M. G. A. Pairs, and C. Fabre, Phys. Rev. Lett. 98, 039301 (2007).
* (26) G. Scarcelli, V. Berardi and Y. Shih, Phys. Rev. Lett. 98, 039302 (2007).
* (27) Pengli Zhang, Wenlin Gong, Xia Shen, and Shensheng Han, arXiv. Quantum-ph/0804.0575 (2008).
* (28) R. J. Glauber, Phys. Rev. 130, 2529 (1963).
* (29) J. W. Goodman, Introduction to Fourier optics (1976).
* (30) G. Mie, Ann. Phys. 330, 377 (1908).
* (31) D. Toublanc, Appl. Opt. 35, 3270 (1996).
* (32) S. K. Sharma, and A. K. Roy, J. Quant. Spectrosc. Radiat. Transfer. 64, 327 (2000).
* (33) Jessica C. Ramella-Roman, Scott A. Prahl, and Steven L. Jacques, Opt. Exp. 13, 25 (2005).
* (34) Milun J. Rakovic and George W. kattawar, Appl. opt. 37, 15 (1999).
* (35) A. A. M. Mustafa and Daphne F. Jackson, Phys. Med. Biol. 26, 461 (1981).
* (36) J. S. Faulkner and Eva A. Horvath, Phys. Rev. B. 44, 8467 (1991).
* (37) W. H. Wells, J. Opt. Soc. Am. 59, 6 (1969).
* (38) H. T. Yura, Appl. opt. 10, 1 (1971).
* (39) Rene Hassanein, Eberhard Lehmann, and Peter Vontobel, Nucl. Ins. Meth. Phys. Res. A. 542, 353 (2005).
* (40) P. N. Segre and P. N. Pusey, Phys. Rev. Lett. 77, 4 (1996).
* (41) Shen Xia, Bai Yan-Feng, and Han Shen-Sheng, Chin. Phys. Lett. 25, 11 (2007).
* (42) Y. Cai and S. Zhu, Opt. Lett. 29, 2716 (2004).
* (43) Y. Cai and S. Zhu, Opt. Lett. 30, 388 (2005).
* (44) A. Gatti, M. Bache, E. Brambilla, and L. A. Lugiato, J. Mod. Opt. 53, 739 (2006).
| arxiv-papers | 2008-06-22T04:02:35 | 2024-09-04T02:48:56.362295 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Wenlin Gong, Pengli Zhang, Xia Shen, and Shensheng Han",
"submitter": "Peng-Li Zhang",
"url": "https://arxiv.org/abs/0806.3543"
} |
0806.3652 | # Protein Structure Prediction Using Basin-Hopping
Michael C. Prentiss,
Center for Theoretical Biological Physics, University of California at San
Diego,
La Jolla, CA 92093, USA
David J. Wales,
University Chemical Laboratories, Lensfield Road, Cambridge CB2 1EW,
United Kingdom
and Peter G. Wolynes
Center for Theoretical Biological Physics, University of California at San
Diego,
La Jolla, CA 92093, USA
###### Abstract
Associative memory Hamiltonian structure prediction potentials are not overly
rugged, thereby suggesting their landscapes are like those of actual proteins.
In the present contribution we show how basin-hopping global optimization can
identify low-lying minima for the corresponding mildly frustrated energy
landscapes. For small systems the basin-hopping algorithm succeeds in locating
both lower minima and conformations closer to the experimental structure than
does molecular dynamics with simulated annealing. For large systems the
efficiency of basin-hopping decreases for our initial implementation, where the
steps consist of random perturbations to the Cartesian coordinates. We
implemented umbrella sampling using basin-hopping to further confirm when the
global minima are reached. We have also improved the energy surface by
employing bioinformatic techniques for reducing the roughness or variance of
the energy surface. Finally, the basin-hopping calculations have guided
improvements in the excluded volume of the Hamiltonian, producing better
structures. These results suggest a novel and transferable optimization scheme
for future energy function development.
## 1 Introduction
The complexity of the physical interactions that guides the folding of
biomolecules presents a significant challenge for atomistic modeling. Many
current protein models use a coarse grained approach to remove degrees of
freedom, such as non-polar hydrogens, which increases the feasible time step
in molecular dynamics simulations [1, 2]. For a more dramatic improvement of
the computational efficiency solvent degrees of freedom can be reduced [3]. In
this case more severe approximations can prevent the model from reproducing
experimental results. Another option is to reduce the number of degrees of
freedom of the solute. The associative memory Hamiltonian (AMH) [4, 5, 6], is
a coarse-grained molecular mechanics potential inspired by physical models of
the protein folding process, but flexibly incorporates bioinformatic data to
predict protein structure. The AMH is first optimised using the minimal
frustration principle in terms of the $T_{f}/T_{g}$ ratio, which estimates the
separation in energy relative to the variance for the misfolded ensemble.
Along with using the energy of the native structure to estimate $T_{f}$, a
random energy model[7] estimate of the glass transition temperature, $T_{g}$,
is used based on a set of decoy structures. $T_{g}$ represents a
characteristic temperature scale at which kinetic trapping in misfolded states
dominates the dynamics. An improved potential is next obtained that uses
better estimates of the $T_{f}/T_{g}$ ratio obtained by maximizing the
normalized difference between the native state and a sampled set of misfolded
decoys which are self-consistently obtained from the potential itself. The
potential so obtained is transferable for the prediction of structures outside
of the training set. The ratio $T_{f}/T_{g}$ has provided a powerful metric
for the optimisation of both this bioinformatically informed energy function
[8, 9], as well as other types of energy functions incorporating only physical
information [10, 11, 12].
While the optimisation [13] of parameters using a training set of evolved
natural proteins smooths the energy landscape from what it would be for a
random hetero-polymer, the common problem of multiple competing minima
persists even when using a reasonably accurate structure prediction potential,
such as this one. Simulated annealing with molecular dynamics has previously
been used to search the rugged landscapes of optimised structure prediction
potentials [14]. While free energy profiles indicate that better structures
actually are present at low temperatures, the slow kinetics of a glass-like
transition during annealing has prevented these minima from being reached
[15]. To quantitatively investigate the origin of the sampling difficulties it
is desirable to use different search strategies.
Here we implement the basin-hopping global optimisation algorithm [16, 17,
18], which has proved capable of overcoming large energetic barriers in a wide
range of different systems. Basin-hopping is an algorithm where a structural
perturbation is followed by energy minimisation. This procedure effectively
transforms the potential energy surface, by removing high barriers, as shown
in Fig. 1. Moves between local minima are accepted or rejected based upon a
Monte Carlo criterion. Avoiding barriers by employing a numerical minimisation
step not only facilitates movement between local minima, but also broadens
their occupation probability distributions, which overlap over a wider
temperature range, thereby increasing the probability of interconversion [19].
Furthermore, it does not alter the nature of the local minima since the
Hamiltonian itself is not changed, enabling the comparison between molecular
dynamics and basin-hopping generated minima. This method has previously been
applied to find global minima in atomic and molecular clusters [20, 21],
biopolymers [22, 23], and solids [24]. Since the algorithm only requires
coordinates, energies, and gradients, it can be transferred between different
molecular systems such as binary Lennard-Jones clusters, all-atom
biomolecules, or coarse-grained proteins models as in this study.
## 2 Theory and Computational Details
The AMH energy function used in the present work has previously been optimised
over a set of non-homologous $\alpha$ helical proteins, and consists of a
backbone term, $E_{\rm back}$, and an interaction term, $E_{\rm int}$, which
has an additive form [25, 26]. This model is sometimes termed the AMC model
(associative memory contact) to distinguish it from one that uses nonadditive
water mediated interactions termed the AMW model [27, 14]. Since this model
has been described in detail before [15, 28], we will only summarize its form
here. We employ a version of the coarse-grained model where the twenty letter
amino acid code has been reduced to four, and the number of atoms per residue
is limited to three (Cα, Cβ, and O), except for glycine residues. The units of
energy and temperature were both defined during the parameter optimisation.
The interaction energy $\epsilon$ was defined in terms of the native state
energy excluding backbone contributions, $E^{\rm N}_{\rm amc}$, via
$\epsilon=\frac{\left|E^{\rm N}_{\rm amc}\right|}{4N},$ (1)
where $N$ is the number of residues of the protein being considered.
Temperatures are quoted in terms of the reduced temperature ${T_{amc}}=k_{\rm
B}T/\epsilon$. While $E_{\rm back}$ creates self-avoiding peptide-like
stereochemistry, $E_{\rm int}$ introduces the majority of the attractive
interactions that produce folding. The interactions described by $E_{\rm int}$
depend on the sequence separation $\left|i-j\right|$. The interaction between
residues less than 12 amino acids apart were defined by Eqs. (2).
$E_{\rm local}=-\frac{\epsilon}{a}\sum_{\mu=1}^{N_{mem}}\sum_{j-12\leq i\leq
j-3}\gamma(P_{i},P_{j},P_{i^{\prime}}^{\mu},P_{j^{\prime}}^{\mu},x(\left|i-j\right|))\exp\left[-\frac{(r_{ij}-r_{i^{\prime}j^{\prime}}^{\rm\mu})^{2}}{2\sigma^{2}_{ij}}\right],$
(2)
The index $\mu$ runs over all $N_{mem}$ memory proteins to which the protein
has previously been aligned using a sequence-structure threading algorithm
[29] (i.e. each $i$-$j$ pair in the protein has an $i^{\prime}$-$j^{\prime}$
pair associated with it in every memory protein; if, due to gaps in the
alignment, there is no $i^{\prime}$-$j^{\prime}$ pair associated with $i$-$j$
for a particular memory then this memory protein simply gives no contribution
to the interaction between residues $i$ and $j$). The interaction between Cα
and Cβ atoms is a sum of Gaussian wells centred at the separations
$r_{i^{\prime}j^{\prime}}^{\rm\mu}$ of the corresponding memory atoms. The
widths of the Gaussians are given by $\sigma_{ij}=\left|i-j\right|^{0.15}$ Å.
The scaling factor $a$ is used to satisfy Eq. 1. The weights given to each
well are controlled by
$\gamma(P_{i},P_{j},P_{i^{\prime}}^{\mu},P_{j^{\prime}}^{\mu},x(\left|i-j\right|))$,
which depends on the identities $P_{i^{\prime}}$ and $P_{j^{\prime}}$ of the
residues to which $i$ and $j$ are aligned, as well as the identities $P_{i}$
and $P_{j}$ of $i$ and $j$ themselves. The self-consistent optimisation
calculates the $\gamma$ parameter which originates the cooperative folding in
the model. A three well contact potential [Eq. (3)] is used for residues
separated by more than 12 residues as described by,
$E_{\rm
contact}=-\frac{\epsilon}{a}\sum_{i<j-12}\sum_{k=1}^{3}\gamma(P_{i},P_{j},k)c_{k}(N)U(r_{min}(k),r_{max}(k),r_{ij}).$
(3)
Here, the sequence indices $i$ and $j$ sum over all pairs of Cβ atoms
separated by more than 12 residues. The sum over $k$ is over the three wells
which are approximately square wells between $r_{min}(k)$ and $r_{max}(k)$
defined by,
$U(r_{min}(k),r_{max}(k),r_{ij})=\frac{1}{4}\left\\{\left[1+\tanh\left(7[r_{ij}-r_{min}(k)]/\text{\AA}\right)\right]+\left[1+\tanh\left(7[r_{max}(k)-r_{ij}]/\text{\AA}\right)\right]\right\\}.$
(4)
The parameters ($r_{min}(k),r_{max}(k)$), are (4.5Å, 8.0Å) , (8.0Å, 10.0Å),
and (10.0Å, 15.0Å) for $k=1,2$ and 3 respectively. In order to approximately
account for the variation of the probability distribution of pair distances
with number of residues in the protein ($N$) a factor $c_{k}(N)$ has been
included in $E_{\rm long}$. It is given by $c_{1}=1.0$,
$c_{2}=1.0/(0.0065N+0.87)$ and $c_{3}=1.0/(0.042N+0.13)$. The individual wells
are also weighted by $\gamma$ parameters which depend on the identities of the
amino acids involved, using the 4-letter code defined above. In contrast to
the interactions between residues closer in sequence, this part of the
potential does not depend on the database structures that define local-in-
sequence interactions.
To pinpoint the effects of frustration or favorable non-native contacts always
present in any coarse gained protein model, we simulated a perfectly smooth
energy function often called a Gō model [30]. Gō models are an essential tool
for understanding protein folding kinetics [31, 32]. While having the same
backbone terms [33], in this single structure based Hamiltonian [Eq. (5)], all
of the interactions, $E_{\rm int}$ are defined by Gaussians whose minima are
located at the pair distribution found in the experimental structure:
$E_{\rm G{\bar{o}}}^{\rm AM}=-\frac{\epsilon}{a_{\rm G{\bar{o}}}}\sum_{i\leq
j-3}\gamma_{\rm
G{\bar{o}}}(x(\left|i-j\right|))\exp\left[-\frac{(r_{ij}-r_{ij}^{\rm
N})^{2}}{2\sigma^{2}_{ij}}\right].$ (5)
The global minima of such an energy function should be the input structure.
Many have used additional constraining potentials to characterise unsampled
regions of coordinate space while using molecular dynamics [34, 15]. To
characterize the landscape sampled with basin-hopping, we also used a
structure constraining potential to identify ensembles with fixed but varying
fractions of native structure. Using such a potential allows us to access
interesting configurations that are unlikely to be thermally sampled. The
constraining potential also called umbrella potentials are centered on
different values of an order parameter to sample along the collective
coordinates. One of the collective coordinates is $Q$, an order parameter that
measures the sequence-dependent structural similarity of two conformations by
computing the normalized summation of C-alpha pairwise contact differences, as
defined in Eq. (6):[15]
$Q=\frac{2}{(N-1)(N-2)}\sum_{i<j-1}\exp\left[-\frac{(r_{ij}-r_{ij}^{\rm
N})^{2}}{\sigma_{ij}^{2}}\right].$ (6)
The resulting order parameter ranges from zero, where there is no similarity
between structures, to one, which represents an exact overlap. The form of the
potential is $E(Q)=2500\epsilon(Q-Q_{i})^{4}$, where $Q_{i}$ may be varied in
order to sample different regions of the chosen order parameter. As in
equilibrium sampling, simulations were initiated at the native state and the
$Q_{i}$ parameter was reduced throughout the sampling.
We have also studied the potential energy landscape when multiple surfaces are
superimposed upon each other by the use of multiple homologous target
proteins. This manipulation of the energy landscape has been shown to further
reduce local energetic frustration that arises from random mutations in the
sequence away from the consensus optimal sequence for a given structure. By
reducing the number of non-native traps, this averaging often improves the
quality of structure prediction results [35, 36, 37, 38]. As seen in Eq. (7),
the form and the parameters of the energy function are maintained from Eqs.
(2) and (3), but the normalized summation is taken over a set of homologous
sequences:
$E_{\rm
AM}=-\frac{1}{N_{seq}}\sum^{seq}_{k=1}\sum^{N}_{i<j}E_{int}(P^{k}_{i},P^{k}_{j}).$
(7)
Since proteins are not random heteropolymers, the differences in the energy
function for homologous proteins are randomly distributed, therefore the mean
over multiple energy functions should have less energetic variation than the
original function. Indeed, performing this summation is a way of incorporating
to the optimisation of the $T_{f}/T_{g}$ criterion into any energy function.
The target sequences of the homologues can be identified using PSI-Blast with
default parameters [39, 40]. Some classes of proteins have a large number of
sequence homologues, therefore performing a multiple sequence alignment can be
impractical. Removing redundant sequences from within the set of identified
homologues also removes biases that can be introduced where there are few
homologues available. This is done by preventing sequences in the collected
sequences from having greater than 90% sequence identity. The remaining
sequences are aligned in a multiple sequence alignment [41]. Gaps within the
sequence alignment can be addressed within the AMH energy function in a
variety of ways. In the present work, gaps in the target sequence were
removed, while gaps within homologues were completed with residues from the
target protein. While this procedure may introduce small biases toward the
target sequence, it is preferable to ignoring the interactions altogether.
Finally, we made several ad hoc changes to the backbone potential, $E_{\rm
back}$. Eliminating some compromises necessary for rapid molecular dynamics
simulations allowed the AMH potential to be adapted to basin-hopping. Another
goal was to prevent the over-collapse of the proteins by altering the excluded
volume energy term, which should reduce the number of states available during
minimisation. The terms shown in Eq. (8) are used to reproduce the peptide-
like conformations in the original molecular dynamics energy function:
$E_{\rm back}=E_{\rm ev}+E_{\rm harm}+E_{\rm chain}+E_{\rm chi}+E_{\rm Rama}.$
(8)
$E_{\rm ev}$ maintains a sequence specific excluded volume constraint between
the Cα-Cα, Cβ-Cβ, O-O, and Cα-Cβ atoms that are separated by less than $r_{\rm
ev}$. Previously [26], we have seen that modifying $E_{\rm back}$ can produce
a less frustrated energy surface when using thermal equilibrium sampling, but
slow dynamics was often found to result since the local barrier heights became
too large. The ability of basin-hopping to overcome such large, but local
barriers allows us therefore to consider a potential whose dynamics would
otherwise be too slow for molecular dynamics. In the final part of the paper,
we altered the excluded volume term, as shown in Eq. (9) to prevent over-
collapse:
$E_{\rm ev}=\epsilon\lambda_{\rm RV}^{\rm C}\sum_{x,y}\sum_{i<j}\theta(r_{\rm
ev}^{\rm C}(j-i)-r_{{\rm C}^{x}_{i}{\rm C}^{y}_{j}})(r_{\rm ev}^{\rm
C}(j-i)-r_{{\rm C}^{x}_{i}{\rm C}^{y}_{j}})^{2}+\epsilon\lambda_{\rm ev}^{\rm
O}\sum_{i<j}\theta(r_{\rm ev}^{\rm O}-r_{{\rm O}_{i}{\rm O}_{j}})(r_{\rm
ev}^{\rm O}-r_{{\rm O}_{i}{\rm O}_{j}})^{2},$ (9)
by changing the default molecular dynamics parameters, $\lambda_{\rm EV}^{\rm
C}=20$, $\lambda_{\rm EV}^{\rm O}=20$, $r_{\rm ev}^{\rm C}(j-i<5)=3.85\,$Å,
$r_{\rm ev}^{\rm C}(j-i\geq 5)=4.5\,$Å, and $r_{\rm ev}^{\rm O}=3.5\,$Å, to
$\lambda_{\rm EV}^{\rm C}=250$, $\lambda_{\rm EV}^{\rm O}=250$, $r_{\rm
ev}^{\rm C}(j-i<5)=3.85\,$Å, $r_{\rm ev}^{\rm C}(j-i\geq 5)=3.85\,$Å, and
$r_{\rm ev}^{\rm O}=3.85\,$Å. The force constant are over an order of
magnitude larger than those used in the molecular dynamics, and the radii of
the Cα, Cβ, and O atoms are also 10% larger than previously used values. This
increase in excluded volume slows the onset of chain collapse, but improves
steric interactions. The other change to the backbone potential involves terms
which maintain chain connectivity. In molecular dynamics annealing, covalent
bonds are preserved using the SHAKE algorithm [42], which permits an increase
of the molecular dynamics time step. For basin-hopping in all parts of this
paper, we removed the SHAKE method and replaced it with a harmonic potential,
$E_{\rm harm}$, between the Cα-Cα, Cα-Cβ, and Cα-O atoms. This replacement
permits the location of local minima without requiring an internal coordinate
transformation, and avoids discontinuous gradients. When minimised, the
additional harmonic terms typically contribute only only .015 $k_{B}T$ per
bond. The remaining terms of the original backbone potential are maintained.
Depending on the sidechain, the neighbouring residues in sequence sterically
limit the variety of positions the backbone atoms can occupy, as evidenced in
a Ramachandran plot [43]. This distribution of coordinates is reinforced by a
potential, $E_{\rm Rama}$, with artificially low barriers to encourage rapid
local movements. The planarity of the peptide bond is ensured by a harmonic
potential, $E_{\rm chain}$. The chirality of the Cα centres is maintained
using the scalar triple product of neighbouring unit vectors of carbon and
nitrogen bonds, $E_{\rm chi}$.
For basin-hopping simulations, whose algorithm is outlined in Fig. 2 the most
important sampling parameters are the temperature used in the accept/reject
steps for local minima $T_{bh}$, and the maximum step size for perturbations
of the Cartesian coordinates $d$. A higher temperature allows transitions to
an increased energy minima to be accepted, and also creates a larger the
number of iterations typically required to minimise the greater perturbed
configurations. Too high a temperature leads to insufficient exploration of
low-energy regions. The temperature ($T_{bh}$) for these simulations was 10
$T_{amc}$. Lower temperatures resulted in slower escape rates from low energy
traps, while higher temperatures prevented adequate exploration of low energy
regions. The step size needs to be large enough to move the configuration into
the basin of attraction of one local minimum to a neighbouring one, but not be
so large that the new minimum is unrelated to the previous state. Every
Cartesian coordinate was displaced up to a maximum step size($d$) of
$0.75\,$Å, the optimum value determined from preliminary tests. Each run
consisted of 2500 basin-hopping steps saving structures every 5 basin-hopping
steps. The convergence condition ($\delta E_{min}$) on the root-mean-square
(RMS) gradient for each minimisation was set to $10^{-3}\,\epsilon$, and the 5
lowest-lying minima from each run were subsequently converged more tightly
($\delta E_{final}$) to an RMS gradient of $10^{-5}\epsilon$. It is important
to note that basin-hopping does not provide equilibrium thermodynamic
sampling. In structure prediction there, however, is no rigorous need for the
search to obey detailed balance since the global energy minimum is the primary
interest. Basin-hopping provides a means for the optimal global search of the
energy landscape, however other methods must be used when calculating free
energy and entropy.
In previous structure prediction studies with the AMH, low energy structures
were identified using off-lattice Langevin dynamics with simulated annealing,
employing a linear annealing schedule of 10000 steps from a temperature of 2.0
to 0.0, starting from a random configuration [5]. The number and length of
simulations needed in both strategies were determined by the number of
uncorrelated structures encountered. The current basin-hopping method with the
AMC energy function encounters roughly one deep trap per run. In order to
sample 100 independent structures in molecular dynamics 20 separate runs were
needed, because simulated annealing samples about five independent states
before the glass transition temperature is encountered, as measured by the
rapid decay of structural correlations. We compared several $\alpha$ helical
proteins, both from within and outside the training set of the AMH energy
function.
## 3 Results and Discussion
We performed initial calculations with a Gō potential for the 434 repressor
(protein data bank (PDB [44]) ID 1r69). In Fig. 3 we show this model
accurately represents the native basin. Steps where the energy increases are
allowed by the sampling method and are not examples of frustration. Studies on
the Gō model provide a useful benchmark for comparing the computer time
required for the different global optimisation strategies. Using the sampling
parameters used in this report, we compared the time for initial collapse
between the molecular dynamics and basin-hopping runs. The initial collapse
required about 7 minutes for the annealing runs and 31 minutes for basin-
hopping on a desktop computer. However, these values do not reflect the actual
performance of the two approaches in locating global minima, which will depend
upon the move sets, step size, temperature, and convergence criteria.
While using the AMC structure prediction Hamiltonian, we found that basin-
hopping was often able to locate lower energy structures and also identified
minima that have greater structural overlap with the native state than
annealing. These results are produced for structure predictions for proteins
both inside and outside the training set, as demonstrated in Table 1. The
first three proteins (PDB ID 1r69, 3icb, 256b) in Table 1 are in the training
set of the Hamiltonian [25], while the other three are not, and can therefore
be considered as predictions from the algorithm. The minima located with
basin-hopping show an increase in structural overlap with the native state
[Eq. (6)] when compared to the Langevin dynamics approach. $Q$ scores of 0.4
for single domain proteins generally correspond to a low resolution root mean
square deviation (RMSD) of around 5 Åor better. $Q$ scores of 0.5 and higher
have still more accurate tertiary packing and are of comparable quality to the
experimentally derived models. The high quality structures obtained suggest
the form of the backbone terms is appropriate, since the physically correct
stereochemistry is reproduced. Lower energy structures are sampled by basin-
hopping for the non-training set proteins, but the structural overlap
improvement found in these deeper minima was smaller. Larger proteins pose a
greater challenge for basin-hopping with this Hamiltonian due to the random
steps in Cartesian coordinates. Dihedral coordinate move steps would probably
be more efficient, and will be considered in future work.
The distribution of minima encountered from multiple simulations for both
search methods is shown in Fig. 4 where a greater density of high quality
structures is obtained by the basin-hopping algorithm. The potential energy
surface still includes, therefore, significant residual frustration in the
near-native basin in the form of low-lying minima separated by relatively high
barriers. Without the parameter optimisation to reduce frustration, folding
would exhibit more pronounced glassy characteristics. Most of the cooperative
folding occurs during collapse until $Q$ values of around 0.4 are reached.
While the structures from simulated annealing are accurate enough for
functional determination, we see basin-hopping can better overcome barriers
that are created after collapse. The density of the high quality structures is
also important for post-simulation k-means clustering analysis [45]. Another
way of representing the data of a set of independent basin-hopping simulations
is by selecting the lowest energy structures from each simulation of the 434
repressor (PDB ID 1r69) and HDEA (PDB ID 1bg8) proteins and ordering them with
respect to their structural overlap. As shown in Fig. 5, the protein in the
training set (434 repressor) produces better results than the non-training
protein, as expected.
We have decomposed the different energy terms in the Hamiltonian in Table 2,
to examine which interactions are most effectively minimised. The AMC
potential has three different distance classes in terms of sequence
separation, and these are defined as short ($\left|i-j\right|<5$), medium
($5\leq\left|i-j\right|\leq 12$), and long ($\left|i-j\right|>12$). Most
importantly, the long-range AMH interactions are successfully minimised in the
basin-hopping runs, due to the ability of basin-hopping to overcome large
energetic barriers. This term will govern the quality of structures sampled
using an approximately smooth energy landscape. The other terms that define
secondary structure formation are not as well minimised. This result is due to
the disruption of helices by the random Cartesian perturbation move steps.
These move steps benefit favorable steric packing and therefore do well at
minimising the excluded volume energy term of the Hamiltonian. A combined
minimisation approach might be more efficient, where larger dihedral steps
could be made early during minimisation to sample a wider number of
structures, followed by random Cartesian steps to optimise the steric
interactions.
While we sampled high quality structures, we would like to confirm that we
have completely sampled the global minima of the energy surface. To access
these unsampled states we used umbrella potentials. When constraining a set of
simulations to different values of $Q$, we have obtained energy minima for
cytochrome c, roughly $15\,\epsilon$ deeper than those from unconstrained
minimisations starting with a randomized structure, as shown in Fig. 6. For
the 434 repressor the minima obtained from randomized states and those found
with the $Q$ constraints applied differ by only a few $k_{B}T$. This shows
basin-hopping does indeed act as a global optimisation method, by accurately
identifying the global energy minimum from multiple independent unconstrained
simulation. This behavior is predictable from the choices that governed the
design of the Hamiltonian. Low energy barriers between structures are
desirable during a molecular dynamics simulation because they accelerate the
dynamics. However, for basin-hopping these low barriers encourage tertiary
contact formation before secondary structure units condense for sequences
greater than 110 amino acids.
### Superposition of Multiple Energy Landscapes
Constructing a Hamiltonian by calculating the arithmetic average of the
potential over a set of homologous sequences increased the quality of
predictions in both equilibrium and annealing simulations. We have found this
approach also improved the performance in basin-hopping simulations. For two
different proteins, 100 independent basin-hopping runs were performed with
both the standard and sequence-averaged Hamiltonians. By the superposition of
multiple energy landscapes we saw a reduction in the number of competing low
energy traps around $Q$ values of $0.3$ for both the 434 repressor and
uteroglobin (PDB ID 1UTG), as shown in Fig. 7. Improvement of structure
prediction Hamiltonians can be statistically described by the average energy
gap between the native basin and a set of unfolded structures, and by the
roughness of the energy surface, which corresponds to the variance of the
energy. The sequence based energy function summations limited the energetic
variance of the sampled landscapes, thereby reducing the glass transition
temperature. This improvement, even at the low temperatures sampled in basin-
hopping, is predicted from theory, but difficult to observe in conventional
equilibrium simulations due to the emergent glassy dynamics, which slows the
kinetics. The energy gap improvement was smaller than the reduction of the
energetic variation of the Hamiltonian. In terms of the goal of maximizing the
ratio of $T_{f}/T_{g}$, this increase came primarily from to reducing the
glass transition temperature $T_{g}$. In the low energy region we saw fewer
competing states, and an increased correlation between ${E}$ and ${Q}$ for the
sequence-averaged Hamiltonian compared to the original Hamiltonian. For the
434 repressor the lowest energy structure had the highest $Q$ value
encountered.
### Characterisation of Polymer Collapse
When we annealed the Hamiltonian using molecular dynamics we observed some
over-collapse of the polypeptide chain, producing a smaller radius of gyration
than the experimental structure. In basin-hopping runs we also found
structures exhibiting a larger number of contacts than the experimental
structure, as show in Fig. 8, where a contact is defined as a Cα-Cα distance
of less than 8 Å. While the low-energy structures may be native-like, these
structures were more compact than those observed experimentally. To
investigate this behavior, we examined the backbone and interaction terms of
the Hamiltonian separately using the Gō Hamiltonian in Eq. (5). Somewhat
surprisingly, the Gō model also produces over-collapse, as shown in Fig. 9.
Hence the interaction parameters of the structure prediction Hamiltonian were
not responsible for all of the over-collapse. These minimal model-dependent
frustrations were only eliminated in the final stages of minimisation. The
most effective technique for reducing over-collapse was to increase the force
constant and the atomic radius in the excluded volume terms [Eq. (9)]. The
barrier crossing capabilities of basin-hopping steps produce more over-
collapse than do the annealing minimisations without these parameter changes.
The glass-like transition seen in simulated annealing prevents further
collapse in molecular dynamics, as the rearrangement rates slow down
exponentially with temperature. The improved parameter set of Fig. 10 shows
more native-like collapse, but the lowest energy structures had $Q$ values of
0.36 and the best $Q$ value was 0.45, which are worse than basin-hopping
simulations with the original parameters.
## 4 Conclusion
In this report we have demonstrated that minima with lower energy and higher
quality structures can often be located for the AMH potential using basin-
hopping global optimisation compared to annealing. Encouragingly, the long
range in sequence energy contributions are better minimised than with
simulated annealing. Umbrella sampling using basin-hopping can also show when
the global minima are reached for a selected order parameter. Previous
techniques for reducing the energetic variance of the energy surface in
simulated annealing are also applicable to basin-hopping. Using basin-hopping
also permits improvements in certain backbone terms of the Hamiltonian. These
changes would make the kinetics too slow in molecular dynamics annealing runs,
but larger barriers can be easily crossed using basin-hopping.
These results highlight future optimisation strategies where the deep non-
native traps found by basin-hopping could be used as decoys for further
parameter refinement, rather than the higher-lying minima obtained by
quenching with simulated annealing. This re-optimisation of the potential
results in getting a better estimate for $T_{f}/T_{g}$ now possible because of
the efficiency of the basin-hopping algorithm at identifying low energy
decoys. Another future direction would be evaluating the equilibrium
properties of low-lying structures identified by basin-hopping also provides a
means to calculate free energy barriers, which would be difficult to
characterise via conventional simulations.
## Acknowledgements
We thank Dr Joanne Carr, and Dr Justin Bois for helpful comments throughout
this research. The efforts of P.G.W. and M.C.P. are supported through the
National Institutes of Health Grant 5RO1GM44557. Computing resources were
supplied by the Center for Theoretical Biological Physics through National
Science Foundation Grants PHY0216576 and PHY0225630. M.C.P. gratefully
acknowledges the support by the International Institute for Complex Adaptive
Matter (ICAM-I2CAM) NSF Grant DMR-0645461.
## References
* [1] J. C. Phillips et al., J. Comp. Chem. 26, 1781 (2005).
* [2] D. V. D. Spoel et al., J. Comp. Chem. 26, 1701 (2005).
* [3] Y. Levy and J. N. Onuchic, Annu. Rev. Biochem. 35, 389 (2006).
* [4] M. S. Friedrichs and P. G. Wolynes, Science 246, 371 (1989).
* [5] M. S. Friedrichs and P. G. Wolynes, Tet. Comp. Meth. 3, 175 (1990).
* [6] M. S. Friedrichs, R. A. Goldstein, and P. G. Wolynes, J. Mol. Biol. 222, 1013 (1991).
* [7] B. Derrida, Phys. Rev. B. 24, 2613 (1981).
* [8] R. A. Goldstein, Z. A. Luthey-Schulten, and P. G. Wolynes, Proc. Natl. Acad. Sci. USA 89, 4918 (1992).
* [9] P. Barth, J. Schonbrun, and D. Baker, Proc. Natl. Acad. Sci. USA 104, 15682 (2007).
* [10] J. Lee et al., J. Phys. Chem. B 105, 7291 (2001).
* [11] Y. Fujitsuka, S. Takada, Z. A. Luthey-Schulten, and P. G. Wolynes, Proteins 54, 88 (2004).
* [12] Y. Fujitsuka, G. Chikenji, and S. Takada, Proteins 62, 381 (2006).
* [13] In this paper we note, the word optimisation will be used with two similar but distinct ways. The first use is for the calculation of the best set of parameters to define a minimally frustrated energy function. The second use is to find the global energy minima on a surface that has multiple minima of nearly equal energies.
* [14] C. Zong, G. Papoian, J. Ulander, and P. Wolynes, J. Am. Chem. Soc. 128, 5168 (2006).
* [15] M. P. Eastwood, C. Hardin, Z. Luthey-Schulten, and P. G. Wolynes, I.B.M. Systems Research 45, 475 (2001).
* [16] D. Wales and J. Doye, J. Phys. Chem. A 101, 5111 (1997).
* [17] Z. Li and H. A. Scheraga, Proc. Natl. Acad. Sci. USA 84, 6611 (1987).
* [18] D. J. Wales and H. A. Scheraga, Science 285, 1368 (1999).
* [19] J. P. K. Doye and D. J. Wales, Phys. Rev. Lett. 80, 1357 (1998).
* [20] D. Wales, Energy Landscapes, Cambridge University Press, Cambridge, UK, 2003.
* [21] D. J. Wales et al., The Cambridge Cluster Database, URL http://www-wales.ch.cam.ac.uk/CCD.html (2001).
* [22] J. M. Carr and D. J. Wales, J. Chem. Phys. 123, 234901 (2005).
* [23] A. Verma, A. Schug, K. H. Lee, and W. Wenzel, J. Chem. Phys. 124, 044515 (2006).
* [24] T. F. Middleton, J. Hernández-Rojas, P. N. Mortenson, and D. J. Wales, Phys. Rev. B 64, 184201 (2001).
* [25] C. Hardin, M. Eastwood, Z. Luthey-Schulten, and P. G. Wolynes, Proc. Natl. Acad. Sci. USA 97, 14235 (2000).
* [26] M. Eastwood, C. Hardin, Z. Luthey-Schulten, and P. G. Wolynes, J. Chem. Phys. 117, 4602 (2002).
* [27] G. Papoian, J. Ulander, M. Eastwood, Z. Luthey-Schulten, and P. G. Wolynes, Proc. Nat. Acad. Sci. U.S.A. 101, 3352 (2004).
* [28] M. C. Prentiss, C. Hardin, M. Eastwood, C. Zong, and P. G. Wolynes, J. Chem. Ther. Comp. 2:3, 705 (2006).
* [29] K. K. Koretke, Z. Luthey-Schulten, and P. G. Wolynes, Protein Sci. 5, 1043 (1996).
* [30] N. Gō, Ann. Rev. Biophys. Bioeng. 12, 183 (1983).
* [31] J. J. Portman, S. Takada, and P. G. Wolynes, Phys. Rev. Lett. 81, 5237 (1998).
* [32] N. Koga and S. Takada, J Mol Biol 313, 171 (2001).
* [33] M. P. Eastwood and P. G. Wolynes, J. Chem. Phys. 114, 4702 (2002).
* [34] X. Kong and C. L. Brooks III, J. Chem. Phys. 105, 2414 (1996).
* [35] F. R. Maxfield and H. A. Scheraga, Biochemistry 18, 697 (1979).
* [36] C. Keasar, R. Elber, and J. Skolnick, Folding and Design 2, 247 (1997).
* [37] R. Bonneau, C. E. M. Strauss, and D. Baker, Proteins 43, 1 (2001).
* [38] C. Hardin, M. P. Eastwood, M. C. Prentiss, Z. Luthey-Schulten, and P. G. Wolynes, Proc. Natl. Acad. Sci. USA 100, 1679 (2003).
* [39] S. Altschul et al., Nucleic Acids Res. 25, 3389 (1997).
* [40] J. E. Stajich et al., Genome Res. 12, 1611 (2002).
* [41] J. Thompson, D. Higgins, and T. Gibson, Nucleic Acids Res. 22, 4673 (1994).
* [42] J. Ryckaert, G. Ciccotti, and H. Berendsen, J. Comput. Phys. 23, 327 (1977).
* [43] G. Ramachandran and V. Sasisekharan, Adv. Protein. Chem. 23, 283 (1968).
* [44] H. M. Berman et al., Nucleic Acids Res. 28, 235 (2000).
* [45] D. Shortle, K. T. Simons, and D. Baker, Proc. Natl. Acad. Sci. USA 95, 11158 (1998).
## Figure Captions
1. 1.
In the basin-hopping approach the original potential energy surface (solid) is
transformed into a set of plateaus (dashed). The local minima are not changed,
but the transition state regions are removed.
2. 2.
The basin-hopping algorithm is defined by a few parameters that allow its
transfer to different systems.
3. 3.
Variation of the energy of the current minimum as a function of $Q$ for minima
encountered in the Markov chain during a basin-hopping run using a Gō model.
Steps that increase the energy are sometimes allowed by the Monte Carlo
criterion, which employed a temperature of 10 $k_{b}$/$\epsilon$.
4. 4.
Energy as a function of $Q$ for local minima of 434 repressor encountered
during 100 independent basin-hopping optimisations (top) and 20 annealing
simulations (bottom).
5. 5.
The lowest energy structures of the training set protein, 434 repressor (top)
and the blind prediction proteins, HDEA (bottom) identified from 100
independent basin-hopping simulations. Each minimum has values for energy
illustrated by the dots and structural overlap to the native state $Q$
represented by lines. These minima are ordered with respect to their
structural overlap $Q$ with the native state (Index). The data shows
correlations between the energy and $Q$, while the number of high quality
structures is superior for the training protein.
6. 6.
Energy as a function of $Q$ for the 434 repressor and cytochrome c proteins
obtained in basin-hopping calculations with the structure prediction
Hamiltonian. These runs employed an additional umbrella potential that
constrains the simulation to different values of $Q$. The results for the 434
repressor are similar to the unconstrained basin-hopping results, but the
structures for cytochrome c are $15\epsilon$ lower in energy than those found
in unconstrained basin-hopping runs.
7. 7.
Energies of local minima obtained using basin-hopping with the original and a
sequence-averaged Hamiltonian for two training proteins. Importantly for both
the top graph (434 repressor) and the bottom graph (uteroglobin) fewer non-
native states are seen with the sequence averaged (red) Hamiltonian when
compared to standard Hamiltonian (black).
8. 8.
Results of 100 independent basin-hopping runs for the 434 repressor using the
set of backbone parameters that was optimised for molecular dynamics.
Structures were saved every 20 basin-hopping steps. The ratio of contacts to
native state contacts shows that most of the structures are more compact than
the native state.
9. 9.
A Gō potential simulation for the 434 repressor shows a modest amount of over-
collapse during a basin-hopping simulation, which is resolved as the structure
approaches a $Q$ value of 1.0.
10. 10.
Results of 100 independent basin-hopping runs for the 434 repressor using the
set of backbone parameters that was optimised for molecular dynamics.
Structures were saved every 20 basin-hopping steps. An altered set of backbone
parameters produces structures that have similar collapse behavior when
compared to the native state.
## Figures
Figure 1:
* Basin Hopping Algorithm
* Monte Carlo Step ($n$ steps)
* random Cartesian move step with maximum distance ($d$) and temperature ($T_{bh}$)
* Minimisation
* L-BFGS quasi-Newtonian method for optimization
* convergence condition ($\delta E_{min}$) is RMS gradient of $10^{-3}\,\epsilon/r$
* Minimisation with tight convergence (after $n$ steps)
* convergence condition ($\delta E_{final}$) is RMS gradient of $10^{-5}\,\epsilon/r$
Figure 2:
Figure 3:
Figure 4:
Figure 5:
Figure 6:
Figure 7:
Figure 8:
Figure 9:
Figure 10:
## Tables
Table 1: Minima located by molecular dynamics/annealing (MD) and basin-
hopping (BH); the first three proteins are in the training set of the
Hamiltonian, while the results for the second three proteins are predictions.
| | MD | | | | BH | | |
---|---|---|---|---|---|---|---|---|---
PDB ID | length | Lowest $E$ | $Q$ | Highest $Q$ | $E$ | Lowest $E$ | $Q$ | Highest $Q$ | $E$
1r69 | 63 | -428.92 | 0.39 | 0.53 | -307.96 | -435.82 | 0.39 | 0.52 | -408.482
3icb | 75 | -536.98 | 0.47 | 0.52 | -390.54 | -546.57 | 0.40 | 0.49 | -518.92
256b | 106 | -735.02 | 0.42 | 0.65 | -707.51 | -737.31 | 0.37 | 0.40 | -716.51
1uzc | 69 | -457.55 | 0.36 | 0.42 | -383.08 | -458.09 | 0.37 | 0.45 | -433.41
1bg8 | 76 | -469.49 | 0.25 | 0.34 | -465.19 | -468.67 | 0.36 | 0.39 | -461.50
1bqv | 110 | -737.91 | 0.21 | 0.27 | -441.92 | -764.20 | 0.23 | 0.27 | -481.22
Table 2: Contribution of different energy terms in local minima obtained
using molecular dynamics/annealing (MD) and basin-hopping (BH).
PDB | Method | length | Ex Vol | Rama | Short Range | Medium Range | Long Range
---|---|---|---|---|---|---|---
1r69 | MD | 63 | 9.77 | $-$101.64 | $-$128.90 | $-$84.87 | $-$123.28
1r69 | BH | 63 | 2.65 | $-$91.06 | $-$125.04 | $-$84.80 | $-$137.57
3icb | MD | 75 | 11.74 | $-$127.70 | $-$177.21 | $-$90.11 | $-$153.69
3icb | BH | 75 | 4.40 | $-$115.76 | $-$178.47 | $-$83.37 | $-$173.38
1uzc | MD | 69 | 10.10 | $-$118.66 | $-$134.00 | $-$90.75 | $-$124.24
1uzc | BH | 69 | 2.22 | $-$106.20 | $-$137.95 | $-$92.40 | $-$123.77
1bg8 | MD | 76 | 11.68 | $-$136.39 | $-$173.45 | $-$94.40 | $-$76.94
1bg8 | BH | 76 | 2.72 | $-$112.13 | $-$151.95 | $-$94.23 | $-$113.09
| arxiv-papers | 2008-06-23T11:01:38 | 2024-09-04T02:48:56.369703 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Michael C. Prentiss, David J. Wales, Peter G. Wolynes",
"submitter": "Michael Prentiss",
"url": "https://arxiv.org/abs/0806.3652"
} |
0806.3661 | # Full quantum reconstruction of vortex states
I. Rigas Departamento de Óptica, Facultad de Física, Universidad Complutense,
28040 Madrid, Spain L. L. Sánchez-Soto Departamento de Óptica, Facultad de
Física, Universidad Complutense, 28040 Madrid, Spain A. B. Klimov
Departamento de Física, Universidad de Guadalajara, 44420 Guadalajara,
Jalisco, Mexico J. Řeháček Department of Optics, Palacky University, 17.
listopadu 50, 772 00 Olomouc, Czech Republic Z. Hradil Department of Optics,
Palacky University, 17. listopadu 50, 772 00 Olomouc, Czech Republic
###### Abstract
We propose a complete tomographic reconstruction of vortex states carrying
orbital angular momentum. The scheme determines the angular probability
distribution of the state at different times under free evolution. To
represent the quantum state we introduce a bona fide Wigner function defined
on the discrete cylinder, which is the natural phase space for the pair angle-
angular momentum. The feasibility of the proposal is addressed.
###### pacs:
03.65.Wj, 03.75.Lm, 42.50.Dv
Phase singularities of wave fields were brought to attention in a number of
seminal papers Dirac (1931); Nye and Berry (1974); Allen et al. (1992). They
manifest in electron wave packets Bliokh et al. (2007), quantum Hall fluids
Ezawa (2000), supermedia Salomaa and Volovik (1987), ferromagnets Hubert and
Schäfer (1998), Bose-Einstein condensates Leggett (2001), acoustical waves
Hefner and Marston (1999), and light fields Allen et al. (2003), to cite some
relevant examples. Special attention has been paid to the particular case of
vortices, which correspond to helicoidal waves exhibiting a pure screw phase
dislocation along the propagation axis; i.e., an azimuthal phase dependence
$e^{i\ell\phi}$. This means that $\ell$ plays the role of a topological
charge: the phase changes its value in $\ell$ cycles of $2\pi$ in any closed
circuit about this axis, while the amplitude is zero there.
One of the most interesting properties of vortices is that they carry orbital
angular momentum (OAM): indeed, the integer $\ell$ can be seen as the
eigenvalue of the OAM operator and its sign defines the helicity or direction
of rotation. In fact, OAM can be easily transferred, as demonstrated in a
number of recent challenging experiments with optically trapped microparticles
Simpson et al. (1997).
At the microscopic level, quantized vortices are now routinely produced by,
e.g., mechanical stirring of ultracold atomic clouds. Remarkable images of
large lattices containing hundreds of vortices in an Abrikosov-type triangular
configuration have been obtained Abo-Shaeer et al. (2001). However, for a
complete information about the corresponding state, one needs to go beyond
this mere photographic picture and perform a full tomography. Efficient
methods of state reconstruction are of the greatest relevance for quantum
optics. Since the first theoretical proposals, this discipline has witnessed
significant growth Paris and Řeháček (2004) and laboratory demonstrations of
state tomography are numerous and span a broad range of physical systems. The
essence of vortices is their helicoidal structure, while other aspects, such
as energy distribution, may be often ignored for many purposes: in this sense,
a fully quantum reconstruction of the vortex content is pending.
Any reliable quantum tomographical scheme requires three key ingredients
Hradil et al. (2006a): the availability of a tomographically complete
measurement, a suitable representation of the quantum state, and a robust
algorithm to invert the experimental data. When these conditions are not met,
reconstruction becomes difficult, if not impossible: this has been the case so
far for vortex states. The construction of a proper Wigner function for them
(or any other quasiprobability distribution in phase space) is still an open
question. Although some interesting attempts have been published Mukunda
(1979); Bizarro (1994), they seem of difficult application for the problem at
hand. The twofold goal of this paper is precisely to fill this long overdue
gap. We will provide a simple Wigner function with a clear geometrical meaning
and also a universal tomographic reconstruction scheme. One could think in
using the standard Wigner function for the transverse coordinate-momentum
variables instead. However, this is not the right answer in quantum mechanics:
such a representation conveys redundant information, it is impossible to plot,
and it hides all the relevant angular information.
We consider rotations by an angle $\phi$ generated by the OAM operator along
the $z$ axis, which for simplicity we shall denote henceforth as $\hat{L}$. We
do not want to enter in a long and sterile discussion about the possible
existence of an angle operator. For our purposes here the simplest choice is
to use the complex exponential of the angle $\hat{E}=e^{-i\hat{\phi}}$, which
satisfies the commutation relation $[\hat{E},\hat{L}]=\hat{E}$. The action of
$\hat{E}$ on the OAM eigenstates is $\hat{E}|\ell\rangle=|\ell-1\rangle$, and
it possesses then a simple implementation by means of a phase mask removing a
unit charge Hradil et al. (2006b). Since the integer $\ell$ runs from
$-\infty$ to $+\infty$, $\hat{E}$ is a unitary operator whose normalized
eigenvectors $|\phi\rangle$ describe states with well-defined angular
position. In the representation generated by them, $\hat{L}$ acts as
$-i\partial_{\phi}$ (in units $\hbar=1$). Note in passing that one could
intuitively expect a Fourier-like relationship between angle and OAM, which
can be expressed in this context as
$e^{-i\phi^{\prime}\hat{L}}|\phi\rangle=|\phi-\phi^{\prime}\rangle$.
For the standard harmonic oscillator, the phase space is the plane
$\mathbb{R}^{2}$. In the case of angle and OAM the phase space is the discrete
cylinder $\mathbb{S}^{1}\times\mathbb{Z}$ (where $\mathbb{S}^{1}$ is the unit
circle). While quasiprobability distributions on the sphere are commonplace in
quantum optics, to the best of our knowledge their counterparts on the
cylinder have never been used to describe angular variables. Given the key
role played by the displacement operators in defining the Wigner function in
the plane, we introduce a unitary displacement operator $\hat{D}(\ell,\phi)$
on the discrete cylinder as
$\hat{D}(\ell,\phi)=e^{i\alpha(\ell,\phi)}\,\hat{E}{-\ell}e^{-i\phi\hat{L})}\,,$
(1)
where $\alpha(\ell,\phi)$ is an undefined phase factor. Apart from
$2\pi$-periodicity in $\phi$, the requirement of unitarity imposes the
condition $\alpha(\ell,\phi)+\alpha(-\ell,-\phi)=-\ell\phi$. The displacement
operators form a non-Hermitian orthogonal basis on the Hilbert space, in the
sense that
$\mathop{\mathrm{Tr}}\nolimits[\hat{D}(\ell,\phi)\,\hat{D}^{\dagger}(\ell^{\prime},\phi^{\prime})]=2\pi\delta_{\ell\ell^{\prime}}\,\delta_{2\pi}(\phi-\phi^{\prime})\,,$
(2)
where $\delta_{2\pi}$ represents the periodic delta function (or Dirac comb)
of period $2\pi$.
Next, we introduce the Wigner kernel as a kind of double Fourier transform of
the displacement operators
$\hat{w}(\ell,\phi)=\frac{1}{(2\pi)^{2}}\sum_{{\ell^{\prime}}\in\mathbb{Z}}\int_{2\pi}d\phi^{\prime}\exp[-i(\ell^{\prime}\phi-\ell\phi^{\prime})]\hat{D}(\ell^{\prime},\phi^{\prime})\,,$
(3)
where the integral extends to the $2\pi$ interval within which the angle is
defined. The integral of $\hat{w}(\ell,\phi)$ over the whole phase space
yields unity and we may thus regard $\hat{w}(\ell,\phi)$ as equivalent to the
phase-point operators introduced by Wootters Wootters (1987). In addition, one
can check that
$\hat{w}(\ell,\phi)=\hat{D}(\ell,\phi)\,\hat{w}(0,0)\,\hat{D}^{\dagger}(\ell,\phi)\,.$
(4)
Since for the plane and the sphere, the Wigner kernel can be seen as the
transform of the parity by the displacement operators Klimov et al. (2006),
this seems to call for interpreting $\hat{w}(0,0)$ as the parity over our
phase space.
We next define the Wigner function of a quantum state described by the density
matrix $\hat{\varrho}$ as
$W(\ell,\phi)=\mathop{\mathrm{Tr}}\nolimits[\hat{\varrho}\,\hat{w}(\ell,\phi)]\,.$
(5)
Using the previous results for $\hat{w}(\ell,\phi)$ one can show that
$W(\ell,\phi)$ fulfills all the properties required for a reasonable
interpretation as a quasiprobability distribution. Indeed, due to the
hermiticity of the Wigner kernel, $W(\ell,\phi)$ is real. It also provides the
proper marginal distributions and it is covariant, which means that if the
state $\hat{\varrho}^{\prime}$ is obtained from $\hat{\varrho}$ by a
displacement in phase space
$\hat{\varrho}^{\prime}=\hat{D}(\ell^{\prime},\phi^{\prime})\,\hat{\varrho}\,\hat{D}^{\dagger}(\ell^{\prime},\phi^{\prime})$,
then the Wigner function follows along rigidly:
$W^{\prime}(\ell,\phi)=W(\ell-\ell^{\prime},\phi-\phi^{\prime})$. All these
properties are fulfilled independently of the choice of the phase
$\alpha(\ell,\phi)$.
Since the displacement operators constitute a basis, we can write the
expansion
$\hat{\varrho}=\sum_{{\ell}\in\mathbb{Z}}\int_{2\pi}d\phi\,\varrho(\ell,\phi)\hat{D}(\ell,\phi)$
(6)
where
$\varrho(\ell,\phi)=\mathop{\mathrm{Tr}}\nolimits[\hat{\varrho}\,\hat{D}^{\dagger}(\ell,\phi)]/(2\pi)$.
In terms of $\varrho(\ell,\phi)$, the Wigner function has the representation
$W(\ell,\phi)=\frac{1}{2\pi}\sum_{{\ell^{\prime}}\in\mathbb{Z}}\int_{2\pi}d\phi^{\prime}\varrho(\ell^{\prime},\phi^{\prime})\,e^{i(\ell^{\prime}\phi-\ell\phi^{\prime})}\,.$
(7)
To work out explicit examples, one needs to fix once for all the function
$\alpha(\ell,\phi)$. One natural option is to set $\alpha(\ell,\phi)=0$ and
then the Wigner kernel (3) reduces to
$\displaystyle\hat{w}(\ell,\phi)$ $\displaystyle=$
$\displaystyle\displaystyle\frac{1}{2\pi}\sum_{{\ell^{\prime}}\in\mathbb{Z}}e^{-2i\ell^{\prime}\phi}|\ell+\ell^{\prime}\rangle\langle\ell-\ell^{\prime}|$
(8) $\displaystyle+$
$\displaystyle\displaystyle\frac{1}{2\pi^{2}}\sum_{{\ell^{\prime},\ell^{\prime\prime}}\in\mathbb{Z}}\frac{(-1)^{\ell^{\prime\prime}-\ell}}{\ell^{\prime\prime}-\ell+1/2}$
$\displaystyle\times$ $\displaystyle
e^{-i(2\ell^{\prime}+1)\phi}|\ell^{\prime\prime}+\ell^{\prime}+1\rangle\langle\ell^{\prime\prime}-\ell^{\prime}|\,,$
which agrees with the kernel derived by Plebański and coworkers Plebański et
al. (2000) in the context of deformation quantization on the cylinder.
For an OAM eigenstate $|\ell_{0}\rangle$, we obtain
$W_{|\ell_{0}\rangle}(\ell,\phi)=\frac{1}{2\pi}\delta_{\ell\ell_{0}}\,,$ (9)
which is a quite reasonable Wigner function: it is flat in $\phi$ and the
integral over the whole phase space equals unity, reflecting the normalization
of $|\ell_{0}\rangle$.
For an angle eigenstate $|\phi_{0}\rangle$, we get
$W_{|\phi_{0}\rangle}(\ell,\phi)=\frac{1}{2\pi}\,\delta_{2\pi}(\phi-\phi_{0})\,.$
(10)
Now, it is flat in the conjugate variable $\ell$, and thus, the integral over
the whole phase space is not finite, which is a consequence of the fact that
the state $|\phi_{0}\rangle$ is unnormalized.
The coherent states $|\ell_{0},\phi_{0}\rangle$ (parametrized by points on the
cylinder) introduced in Ref. Kowalski et al. (1996) (see also Ref. Kastrup
(2006) for a detailed discussion of the properties of these relevant states)
satisfy
$\displaystyle\langle\ell|\ell_{0},\phi_{0}\rangle$ $\displaystyle=$
$\displaystyle\displaystyle\frac{1}{\sqrt{\vartheta_{3}\left(0\big{|}\frac{1}{e}\right)}}e^{-i\ell\phi_{0}}\,e^{-(\ell-\ell_{0})^{2}/2}\,,$
$\displaystyle\langle\phi|\ell_{0},\phi_{0}\rangle$ $\displaystyle=$
$\displaystyle\displaystyle\frac{e^{i\ell_{0}(\phi-\phi_{0})}}{\sqrt{\vartheta_{3}\left(0\big{|}\frac{1}{e}\right)}}\vartheta_{3}\left(\frac{\phi-\theta}{2}\Big{|}\frac{1}{e^{2}}\right)\,,$
where $\vartheta_{3}$ denotes the third Jacobi theta function. One immediately
finds that
$\varrho(\ell,\phi)=\frac{e^{i\ell(\phi/2-\phi_{0})}}{2\pi\vartheta_{3}\left(0\big{|}\frac{1}{e}\right)}\,e^{-\ell^{2}/2-i\ell_{0}\phi}\,\vartheta_{3}\left(\frac{\phi}{2}+\frac{i\ell}{2}\Big{|}\frac{1}{e}\right)\,,$
(12)
which gives
$W_{|\ell_{0},\phi_{0}\rangle}(\ell,\phi)=W^{(+)}_{|\ell_{0},\phi_{0}\rangle}(\ell,\phi)+W^{(-)}_{|\ell_{0},\phi_{0}\rangle}(\ell,\phi)\,,$
(13)
with
$\displaystyle W^{(+)}_{|\ell_{0},\phi_{0}\rangle}(\ell,\phi)$
$\displaystyle=$
$\displaystyle\frac{1}{2\pi\vartheta_{3}\left(0\big{|}\frac{1}{e}\right)}\,e^{-(\ell-\ell_{0})^{2}}\vartheta_{3}\left(\phi-\phi_{0}\Big{|}\frac{1}{e}\right)\,,$
$\displaystyle W^{(-)}_{|\ell_{0},\phi_{0}\rangle}(\ell,\phi)$
$\displaystyle=$
$\displaystyle\frac{e^{i(\phi-\phi_{0})-1/2}}{2\pi^{2}\vartheta_{3}\left(0\big{|}\frac{1}{e}\right)}\vartheta_{3}\left(\phi-\phi_{0}+i/2\Big{|}\frac{1}{e}\right)$
$\displaystyle\times$
$\displaystyle\sum_{{\ell^{\prime}}\in\mathbb{Z}}\frac{(-1)^{\ell^{\prime}-\ell+\ell_{0}}\,e^{-\ell^{\prime}{}^{2}}-\ell^{\prime}}{\ell^{\prime}+\ell_{0}-\ell+1/2}\,.$
In spite of the fact that these states may be regarded as a basic set to
construct the Wigner kernel, their Wigner function has no simple closed form,
but it splits into two different functions with periods $\pi$ and $2\pi$,
respectively.
Figure 1: Plot of the Wigner function for a coherent state with $\ell_{0}=0$
and $\phi_{0}=0$. The cylinder extends vertically from $\ell=-4$ to $\ell=+4$.
We show the two corresponding marginal distributions.
In Fig. 1 we show the Wigner function for the coherent state
$|\ell_{0}=0,\phi_{0}=0\rangle$ plotted on the discrete cylinder. We can see a
pronounced peak at $\phi=0$ for $\ell=0$ and slightly smaller ones for
$\ell=\pm 1$. A closer look at the picture reveals also a remarkable fact: for
values close to $\phi=\pm\pi$ and $\ell=\pm 1$, the Wigner function takes
negative values. Actually, a numeric analysis suggests the existence of
negativities close to $\phi=\pm\pi$ for odd values of $\ell$. We also plot the
marginals obtained from Eq. (13) by integrating over $\phi$ or summing over
$\ell$.
As our last example, we look at the superposition states
$|\Psi\rangle=\frac{1}{\sqrt{2}}(|\ell_{0}\rangle+e^{i\phi_{0}}\,|-\ell_{0}\rangle)\,,$
(15)
which have been proposed for applications in quantum experiments Vaziri et al.
(2002). The analysis can be carried out for more general superpositions, but
(15) is enough to display the relevant features. The final result is
$W_{|\Psi\rangle}(\ell,\phi)=\frac{1}{4\pi}(\delta_{\ell\ell_{0}}+\delta_{\ell\,-\ell_{0}})+\frac{1}{2\pi}\cos(\phi_{0}-8\phi)\,\delta_{\ell
0}\,.$ (16)
The Wigner function presents then three contributions: two flat slices coming
from the states $|\ell_{0}\rangle$ and $|-\ell_{0}\rangle$ and their
interference located at the origin (see Fig. 2).
Figure 2: Plot of the Wigner function (and the corresponding marginals) for a
superposition state (15) with $\ell_{0}=3$ and $\phi_{0}=\pi$. and its
marginal distributions. The cylinder extends vertically from $\ell=-4$ to
$\ell=+4$.
To complete our theory we also propose a reconstruction scheme for these
sates. A reconstruction of $\hat{\varrho}$ (or, equivalently, of its Wigner
function) is tantamount to finding the coefficients $\varrho(\ell,\phi)$. To
this end, we need a tomographical measurement that allows us to reconstruct
$\varrho(\ell,\phi)$. For $\ell=0$, the coefficients $\varrho(0,\phi)$ read as
$\varrho(0,\phi)=\frac{1}{2\pi}\sum_{{\ell}\in\mathbb{Z}}\langle\ell|\hat{\varrho}|\ell\rangle\,e^{i\ell\phi}\,,$
(17)
where we have made use of the fact that the undetermined phase
$e^{-i\alpha(0,\phi)}$ can be set to 1 for all values of $\phi$, for it
corresponds to displacements along one of the coordinate axes and no
additional phase should be acquired.
For $\ell\neq 0$, we introduce the operator
$\hat{U}_{t}=\exp(-it\hat{L}^{2}/2)$. If we recall that the generic
Hamiltonian of a quantum rotator is $\hat{H}=\hat{L}^{2}/2I$ ($I$ being the
rotational moment of inertia), the operator $\hat{U}_{t}$ is simply the free
propagator of the system (in appropriate units). This quantum rotator
Hamiltonian describes a variety of situations, such as molecular rotations
Mouritzen and Molmer (2006), single-photon OAM Molina-Terriza et al. (2004) or
the azimuthal evolution of optical beams Řeháček et al. (2008a). For all of
them, our proposal for $\omega(\phi,t)$ is precisely to measure the angular
distribution after a free evolution $t$, that is,
$\omega(\phi,t)=\langle\phi(t)|\hat{\varrho}|\phi(t)\rangle=\langle\phi|\hat{U}_{t}^{\dagger}\,\hat{\varrho}\,\hat{U}_{t}|\phi\rangle\,.$
(18)
In other cases, the scheme also works appropriately provided one can
experimentally implement the action of $\hat{L}^{2}$. Using the
representations (6) for $\hat{\varrho}$ and (1) for $\hat{D}$,
$\varrho(\ell,\phi)$ turns out to be
$\varrho(\ell,\phi)=\frac{1}{2\pi}e^{-i\alpha(\ell,\phi)}\int_{2\pi}d\phi^{\prime}\,e^{-i\ell\phi^{\prime}}\,\omega(\phi^{\prime},\phi/\ell)\,.$
(19)
Plugging these coefficients into Eq. (7), we get the reconstruction of the
Wigner function as
$\displaystyle W(\ell,\phi)$ $\displaystyle=$
$\displaystyle\displaystyle\frac{1}{2\pi}\langle\ell|\hat{\varrho}|\ell\rangle+\frac{1}{(2\pi)^{2}}\sum_{\ell^{\prime}\in\mathbb{Z}\atop\ell^{\prime}\neq
0}\int_{2\pi}d\phi^{\prime}\,d\phi^{\prime\prime}e^{-i\alpha(\ell^{\prime},\phi^{\prime})}\,$
(20) $\displaystyle\times$ $\displaystyle
e^{i[\ell\phi^{\prime}-\ell^{\prime}(\phi-\phi^{\prime\prime})]}\,\omega(\phi^{\prime\prime},\phi^{\prime}/\ell^{\prime})\,.$
Consistently, the reconstruction procedure itself does not depend on the
undetermined phase $\alpha(\ell,\phi)$ of the displacement operator, while the
Wigner function does have such a dependence. The phase factor acts as a metric
coefficient for the mapping from Hilbert space onto the phase space. We recall
that for the harmonic oscillator, one can recover the Wigner function via an
inverse Radon transform from the quadrature probability distribution Vogel and
Risken (1989). Equation (20) is then the analogous for our system.
As a rather simple yet illustrative example, let us note that for the vortex
state $|\ell_{0}\rangle$, $\hat{U}_{t}$ is diagonal, so the tomograms
$\omega(\phi^{\prime},\phi^{\prime}/\ell)$ are independent of $\phi$ and
$\ell$ and all of them equal to $1/(2\pi)$. Performing the integration in Eq.
(20) we obtain precisely the Wigner function (9).
Finally, for the feasibility of the proposed scheme, we need the projection
onto the eigenstates $|\phi\rangle$. Since these states correspond to the
measurement of a continuous variable, such a measurement can be done only
approximately. A good approximation seems to be the projection onto the wedge
states, although other experimental schemes are also available Dooley et al.
(2003). According to Eq. (18), this projection, together with the
determination of $t$, are the two main sources of measurement error.
Afterwards, the reconstruction errors for each particular experimental
implementation should be evaluated following the standard recipes developed
in, e.g., Ref. Řeháček et al. (2008b).
In summary, we have carried out a full program for the reconstruction of
generic vortex states, including a complete phase-space description in terms a
bona fide Wigner function. Though the implementation of this scheme may differ
depending on the system under consideration, our formulation provides a common
theoretical framework on the Hilbert space generated by the action of angle
and angular momentum. Experimental demonstrations of the method in terms of
optical beams are presently underway in our laboratory and will be reported
elsewhere.
###### Acknowledgements.
We acknowledge discussions with Hubert de Guise, Jose Gracia-Bondia, Hans
Kastrup, Jakub Rembielinski and Krzysztof Kowalski. This work was supported by
the Czech Ministry of Education, Project MSM6198959213, the Czech Gran Agency,
Grant 202/06/0307, the Spanish Research Directorate, Grant FIS2005-06714 and
the Mexican CONACYT, Grant 45705.
## References
* Dirac (1931) P. Dirac, Proc. R. Soc. London, Ser. A 133, 60 (1931).
* Nye and Berry (1974) J. Nye and M. Berry, Proc. R. Soc. London, Ser. A 336, 165 (1974).
* Allen et al. (1992) L. Allen, M. W. Beijersbergen, R. J. C. Spreeuw, and J. P. Woerdman, Phys. Rev. A 45, 8185 (1992).
* Bliokh et al. (2007) K. Y. Bliokh, Y. P. Bliokh, S. Savel’ev, and F. Nori, Phys. Rev. Lett. 99, 190404 (2007).
* Ezawa (2000) Z. F. Ezawa, _Quantum Hall Effects_ (World Scientific, Singapore, 2000).
* Salomaa and Volovik (1987) M. M. Salomaa and G. E. Volovik, Rev. Mod. Phys 59, 533 (1987).
* Hubert and Schäfer (1998) A. Hubert and R. Schäfer, _Magnetic Domains_ (Springer, New York, 1998).
* Leggett (2001) A. J. Leggett, Rev. Mod. Phys. 73, 307 (2001); E. A. Cornell and C. E. Wieman, _ibid._ 74, 875 (2002).
* Hefner and Marston (1999) B. T. Hefner and P. M. Marston, J. Acoust. Soc. Am. 106, 3313 (1999).
* Allen et al. (2003) L. Allen, S. M. Barnett, and M. J. Padgett, _Optical Angular Momentum_ (Institute of Physics Publishing, Bristol, 2003).
* Simpson et al. (1997) N. B. Simpson, K. Dholakia, L. Allen, and M. J. Padgett, Opt. Lett. 22, 52 (1997); L. Paterson, M. P. MacDonald, J. Arlt, W. Sibbett, P. E. Bryant, and K. Dholakia, Science 292, 912 (2001).
* Abo-Shaeer et al. (2001) J. R. Abo-Shaeer, C. Raman, J. M. Vogels, and W. Ketterle, Science 292, 476 (2001).
* Paris and Řeháček (2004) M. G. A. Paris and J. Řeháček, eds., _Quantum State Estimation_ , vol. 649 of _Lect. Not. Phys._ (Springer, Heidelberg, 2004).
* Hradil et al. (2006a) Z. Hradil, D. Mogilevtsev, and J. Řeháček, Phys. Rev. Lett. 96, 230401 (2006a).
* Mukunda (1979) N. Mukunda, Am. J. Phys. 47, 182 (1979).
* Bizarro (1994) J. P. Bizarro, Phys. Rev. A 49, 3255 (1994).
* Hradil et al. (2006b) Z. Hradil, J. Řeháček, Z. Bouchal, R. Čelechovský, and L. L. Sánchez-Soto, Phys. Rev. Lett. 97 243601 (2006b).
* Wootters (1987) W. K. Wootters, Ann. Phys. (NY) 176, 1 (1987).
* Klimov et al. (2006) A. Klimov, J. Delgado, and L. L. Sánchez-Soto, Opt. Commun. 258, 210 (2006).
* Plebański et al. (2000) J. F. Plebański, M. Prazanowski, J. Tosiek, and F. K. Turrubiates, Acta Phys. Pol. B 31, 561 (2000).
* Kowalski et al. (1996) K. Kowalski, J. Rembieliński, and L. C. Papaloucas, J. Phys. A 29, 4149 (1996).
* Kastrup (2006) H. A. Kastrup, Phys. Rev. A 73, 052104 (2006).
* Vaziri et al. (2002) A. Vaziri, G. Weihs, and A. Zeilinger, J. Opt. B 4, S47 (2002).
* Mouritzen and Molmer (2006) A. S. Mouritzen and K. Molmer, J. Chem. Phys. 124, 244311 (2006).
* Molina-Terriza et al. (2004) G. Molina-Terriza, A. Vaziri, J. Řeháček, Z. Hradil, and A. Zeilinger, Phys. Rev. Lett. 92, 167903 (2004); N. K. Langford, R. B. Dalton, M. D. Harvey, J. L. O’Brien, G. J. Pryde, A. Gilchrist, S. D. Bartlett, and A. G. White, _ibid._ 93, 053601 (2004).
* Řeháček et al. (2008a) J. Řeháček, Z. Bouchal, R. Čelechovský, Z. Hradil, and L. L. Sánchez-Soto, Phys. Rev. A 77, 032110 (2008a).
* Vogel and Risken (1989) K. Vogel and H. Risken, Phys. Rev. A 40, 2847 (1989).
* Dooley et al. (2003) P. W. Dooley, I. V. Litvinyuk, K. F. Lee, D. M. Rayner, M. Spanner, D. M. Villeneuve, and P. B. Corkum, Phys. Rev. A 68, 023406 (2003).
* Řeháček et al. (2008b) J. Řeháček, D. Mogilevtsev, and Z. Hradil, New J. Phys. 10, 043022 (2008b).
| arxiv-papers | 2008-06-23T12:14:10 | 2024-09-04T02:48:56.375194 | {
"license": "Public Domain",
"authors": "I. Rigas, L. L. Sanchez-Soto, A. B. Klimov, J. Rehacek, and Z. Hradil",
"submitter": "Luis L. Sanchez. Soto",
"url": "https://arxiv.org/abs/0806.3661"
} |
0806.3935 | # Spin Glass Order Induced by Dynamic Frustration
E. A. Goremychkin Materials Science Division, Argonne National Laboratory,
Argonne, IL 60439-4845, USA ISIS Pulsed Neutron and Muon Facility, Rutherford
Appleton Laboratory, Chilton, Didcot, Oxfordshire, OX11 0QX, United Kingdom
R. Osborn Materials Science Division, Argonne National Laboratory, Argonne,
IL 60439-4845, USA B. D. Rainford Department of Physics and Astronomy,
University of Southampton, Southampton, SO17 1BJ, United Kingdom R. T.
Macaluso Program of Chemistry and Biochemistry, University of Northern
Colorado, Greeley, CO 80639, USA D. T. Adroja ISIS Pulsed Neutron and Muon
Facility, Rutherford Appleton Laboratory, Chilton, Didcot, Oxfordshire, OX11
0QX, United Kingdom M. Koza Institut Laue Langevin, F-38042 Grenoble Cédex,
France
###### Abstract
Spin glasses are systems whose magnetic moments freeze at low temperature into
random orientations without long-range order. It is generally accepted that
both frustration and disorder are essential ingredients in all spin glasses,
so it was surprising that PrAu2Si2, a stoichiometric compound with a well-
ordered crystal structure, was reported to exhibit spin glass freezing. In
this article, we report on inelastic neutron scattering measurements of the
crystal field excitations, which show that PrAu2Si2 has a singlet ground state
and that the exchange coupling is very close to the critical value to induce
magnetic order. We propose that spin glass freezing results from dynamic
fluctuations of the crystal field levels that destabilize the induced moments
and frustrate the development of long-range magnetic correlations. This novel
mechanism for producing a frustrated ground state could provide a method of
testing the concept of ‘avoided criticality’ in glassy systems.
Frustration arises from competing interactions that favour incompatible ground
states Fischer 1991 ; Ramirez 2003 . For example, in rare earth intermetallic
compounds, the magnetic moments of the f-electrons on each rare earth site
interact with neighbouring moments through RKKY exchange interactions that
oscillate in sign with increasing separation. Except for specific classes of
geometrically frustrated lattices, well-ordered crystal structures produce
either ferromagnetic or antiferromagnetic order depending on the energy
minimization of the interactions between all the neighbouring moments.
However, when there is disorder, either in the site occupations or the
exchange interactions, the additional randomness can prevent a unique ordered
ground state. Instead, these systems may form spin glasses, which possess a
multitude of possible disordered ground states, into one of which the system
freezes below the glass transition temperature, Tg Fischer 1991 . In the
thermodynamic limit, spin glasses display broken ergodicity, preventing
significant fluctuations to any of the other degenerate spin configurations.
In recent years, two stoichiometric intermetallic compounds have displayed
evidence of spin glass order, URh2Ge2 Sullow 1997 and PrAu2Si2 Krimmel 1999 ,
both nominally with the same tetragonal (ThCr2Si2-type) crystal structure. In
both samples, classic spin glass behaviour was observed (Tg = 11 K and 3 K,
respectively), with a frequency dependent peak in the ac susceptibility and
irreversibility in the field-cooled and zero-field-cooled magnetizations. It
was quickly established that spin-glass freezing in the uranium compound
resulted from site disorder on the rhodium and germanium sublattices. Extended
annealing to remove this source of disorder was sufficient to transform
samples into ordered antiferromagnets Sullow 2000 . On the other hand, the
spin glass behaviour of PrAu2Si2 is very robust, and appears in the best
quality samples after extensive annealing. A recent Mössbauer study, combined
with neutron and x-ray diffraction, concluded that interchange of gold and
silicon atoms was less than 1% Ryan 2005 . Furthermore, the intentional
introduction of disorder through the substitution of germanium for silicon
stabilizes long-range antiferromagnetic order at concentrations greater than
12% Krimmel 1999b . Finally, the praseodymium sublattice is face-centred
tetragonal, and does not contains any of the triangular or tetrahedral motifs
normally associated with geometric frustration Ramirez 2003 . This article
seeks to explain the origin of magnetic frustration in PrAu2Si2, given that
conventional models involving static disorder or lattice topology do not seem
to apply.
Firstly, it is essential to establish the nature of the praseodymium
f-electron magnetism. The nine-fold degeneracy of the 4$f^{2}$-electron states
is lifted by the crystal field potential of tetragonal symmetry produced by
the surrounding gold and silicon ions giving a set of singlet and doublet
levels. The doublet states have finite magnetic moments, but the singlet
states are non-magnetic, except in the presence of exchange interactions with
neighbouring sites as discussed below. Transitions between the crystal field
levels can be observed directly by neutron scattering as inelastic peaks,
whose energies and intensities can be used to refine the parameters of the
crystal field potential. The peak widths are inversely proportional to the
lifetimes of the excited states, which, in metals, are limited by conduction
electron scattering.
We have used neutron scattering to study samples of PrAu2(Si1-xGex)2, with $x$
= 0, 0.05, 0.10, 0.15, 0.20, 0.80, and 1, which were synthesized by arc
melting followed by annealing in vacuum for four weeks at 850∘C. Neutron
diffraction measurements on PrAu2Si2 showed no evidence of any additional
phases and measurements of the dc susceptibility in a field of 20 Oe showed
the previously observed difference between field-cooled and zero-field-cooled
susceptibilities confirming a spin glass freezing temperature of 3 K. The
inelastic neutron scattering measurements were carried out at the Institut
Laue-Langevin (Grenoble, France) on the high resolution time-of-flight
spectrometer IN6, using an incident energy of 3.1 meV, at temperatures ranging
from 1.5 K to 30 K.
Figure 1: Crystal field transition in PrAu2Si2. Inelastic neutron scattering
from PrAu2Si2 measured at 1.5 K on IN6 with an incident energy of 3.1 meV. The
solid line is a fit to the singlet-doublet crystal field transition at
$\Delta=0.7$ meV (dashed line) and an elastic resolution-limited peak from
nuclear incoherent scattering. The shaded area represents double-scattering
from the transition at $2\Delta$. Further details of the crystal field
refinement are given in Ref. Goremychkin 2007 . The inset illustrates the
mechanism for induced moment formation, in which interionic exchange coupling,
$J_{ex}$, admixes the excited magnetic doublet into the singlet ground state.
In an earlier report, we determined the phenomenological parameters of the
crystal field potential in PrAu2Si2 Goremychkin 2007 and showed that the
ground state is a non-magnetic singlet with the first excited level a magnetic
doublet at an energy $\Delta=0.7$ meV (see Fig. 1). The remaining levels are
all above 7 meV, so this is effectively a two-level system at temperatures
close to the glass transition.
Figure 2: Phase diagram of PrAu2(Si1-xGex)2. A self-consistent mean-field
calculation of the phase diagram of PrAu2(Si1-xGex)2 as a function of the
interionic exchange energy, $J_{ex}$ for a singlet-doublet transition energy,
$\Delta=0.7$ meV. The light area represents the region in which spontaneous
magnetic moments are induced by the exchange. The red circles are the
antiferromagnetic transition temperatures as a function of germanium
concentration, $x$. The blue circle is the spin glass freezing temperature in
PrAu2Si2.
Before discussing the origin of spin glass behaviour, we review what is known
about magnetic order in singlet ground-state systems. Since the ground state
of the isolated ion is non-magnetic, a magnetic moment can only result from an
admixture of excited crystal field states produced by exchange interactions
with neighbouring praseodymium ions. There is a considerable body of research
on such induced moment systems, both theoretical Grover 1965 ; Cooper 1972
and experimental Birgeneau 1971 ; Houmann 1975 ; Blanco 1997 , showing that,
for any given value of the low-lying crystal field transition energy
($\Delta$), there is a critical value of the exchange energy ($J_{ex}$). Below
this critical value, the system remains paramagnetic (i.e., a Van Vleck
paramagnet) at all temperatures. However, when the exchange is sufficiently
strong, there is a critical temperature below which a ground state moment
forms spontaneously. In a two-level system, this temperature is given in a
mean field model by
$T_{c}=\Delta\bigg{\\{}\ln\left[{\frac{J_{ex}\alpha^{2}+n\Delta}{J_{ex}\alpha^{2}-\Delta}}\right]\bigg{\\}}^{-1}$
(1)
where $\alpha$ is the dipole matrix element coupling the two levels and $n$ is
the degeneracy of the excited state. A calculation of $T_{c}$ as a function of
$J_{ex}$ for $\Delta=0.7$ meV and $n=2$ is shown in Fig. 2.
In most induced moment systems, $T_{c}$ marks the transition to long-range
order. Examples include Pr Houmann 1975 , Pr3Tl Birgeneau 1971 , Pr3In
Heiniger 1975 and PrNi2Si2 Blanco 1997 . However, in the presence of static
disorder, the low-temperature phase could also be a spin glass, as shown by
Sherrington Sherrington 1979 in a mean field model developed to explain
scandium-terbium alloys Sarkissian 1976 and PrP0.9 Yoshizawa 1983 .
Sherrington calculated that a sufficiently large distribution of exchange
interactions, ${\delta}J_{ex}$, could lead to spin glass freezing. There is a
multicritical point at
$zJ_{ex}=\sqrt{z}\;\delta\\!J_{ex}=\Delta/(2\alpha^{2})$ (2)
where $z$ is the number of nearest neighbours. At larger values of
$\delta\\!J_{ex}$, spin glass order would supercede long-range order. In the
case of PrAu2Si2, for which $z=4$, $\delta\\!J_{ex}$ should be more than twice
$J_{ex}$ so Sherrington’s model requires a high degree of disorder to generate
spin glass behaviour.
Figure 3: Crystal field transitions in PrAu2(Si1-xGex)2. Inelastic neutron
scattering from the singlet-doublet transition as a function of germanium
doping at 20K (red circles). The solid line is a fit to two components shown
as dashed lines; the inelastic singlet-doublet transition and quasielastic
scattering within the excited doublet. The fitted peaks have been broadened by
a convolution of Lorenzian and Gaussian lineshapes with the instrumental
resolution.
Although such strong disorder appears unlikely in PrAu2Si2, we have explored
the Sherrington model further by deliberately disordering the silicon
sublattice through germanium substitution. Krimmel et al had already found
that long-range antiferromagnetism in PrAu2(Si1-xGex)2 is stabilized at
concentrations $x>0.12$ Krimmel 1999b . Our inelastic neutron scattering
measurements shows that the energy of the singlet-doublet transition is nearly
independent of $x$ (Fig. 3). What does change is the strength of the
interionic exchange, $J_{ex}$. In Figure 2, we estimate the strength of
$J_{ex}$ vs $x$ subsituting the measured values of $T_{c}$ in equation 1. This
shows that $J_{ex}$ increases by a factor 4 from $x=0$ to $x=1$. Germanium
doping should therefore be an ideal way of enhancing the exchange disorder
that is central to the Sherrington model.
Figure 4: Broadening of the singlet-doublet crystal field transition. (a) The
standard deviation of the Gaussian broadening of the 0.7 meV transition as a
function of germanium doping, $x$, measured at 20 K, i.e., above $T_{c}$. The
red dashed line marks the critical concentration separating spin glass from
antiferromagnetic order. (b) The Lorenzian half-width of the 0.7 meV
transition in PrAu2Si2 as a function of temperature. The blue dashed lines are
guides to the eye.
The most noticeable effect of chemical disorder is the large increase in the
energy width of the singlet-doublet transition (Fig. 3). There are two main
contributions to the overall width; lifetime broadening, which is Lorenzian in
shape and will be discussed later, and inhomogeneous broadening from the
chemical disorder, which has a Gaussian form de Gennes 1958 . If we convolve
both contributions with the instrumental resolution to produce the fits in
Fig. 3, we estimate that the Gaussian width rises approximately linearly with
$x$, for $x\leq 0.2$ (Fig. 4a). This shows that there is no correlation
between the measured inhomogeneous broadening and the transition to long-range
antiferromagnetic order and confirms that static disorder is not the critical
factor determining the occurrence of spin glass freezing.
Figure 5: Dynamic instability of induced moment formation. The mean field
calculation of the phase diagram of PrAu2Si2 and PrAu2Ge2 as a function of
crystal field energy $\Delta$ for the estimated respective values of $J_{ex}$.
The light areas represent regions of induced moment stability. The red circle
marks the measured singlet-doublet transition energy, and the width of the
blue hatched region corresponds to the relaxational broadening of the
transition. This illustrates that the induced moment is stable in PrAu2Ge2
because of the strength of the exchange interactions whereas it is unstable in
PrAu2Si2 because of its proximity to the phase boundary.
If static disorder is not responsible for the frustrated ground state, we must
seek an alternative explanation, one that should take into account the
observation in Figure 2 that spin glass freezing only occurs in proximity to
the induced moment phase boundary. We propose that relaxational broadening of
the singlet-doublet transition provides the key. Figure 4b shows the
temperature dependence of the transition half-width in PrAu2Si2, which
increases from about 0.13 meV at 1.6 K to 0.26 meV at 20 K. The data are
consistent with standard models of $f$-electron relaxation due to fluctuations
in the single-site crystal field populations, which causes the rapid increase
at low temperatures, and conduction electron scattering Jensen 1987 . The
important result is that the width remains substantial down to the lowest
temperature, with the FWHM only falling to a value of 43% of the transition
energy at the glass transition.
This energy broadening will play a significant role in disrupting induced
moment formation. Figure 5 shows the complementary phase diagram to Figure 2,
in which $J_{ex}$ is fixed to the estimated values in PrAu2Si2 and PrAu2Ge2
and $T_{c}$ is plotted against $\Delta$. The induced moments are predicted to
be stable in both compounds, but, if we superimpose the broadened peaks, we
see that the peak traverses the phase boundary significantly in the silicon
compound but not in the germanium compound. Strictly speaking, the width is
proportional to the inverse lifetime of the excited state and not the lifetime
of the induced ground state, but it makes it highly plausible that dynamic
fluctuations would disrupt the stability of such moments in PrAu2Si2, but not
in PrAu2Ge2.
In induced moment systems, a precursor of long-range order is the appearance
in neutron scattering of a quasielastic “central” peak close to the wavevector
of the ordered phase, which is caused by fluctuating regions of short-range
magnetic order Lindgard 1983 . For example, in Pr3Tl, the correlation length
associated with this central peak diverges as the critical temperature is
approached Als-Nielsen 1977 . Our conjecture is that any divergence of the
magnetic correlation length in PrAu2Si2 is suppressed by dynamic fluctuations
that limit the lifetime of induced moments and so introduce magnetic site and
exchange disorder. This scenario is similar to the ‘avoided criticality’
discussed in the theory of structural glass transitions by Tarjus et al Tarjus
2005 . In their review, they argue that many glasses are close to a
conventional second order phase transition, but that frustration prevents the
divergence of the correlation length. This would make the PrAu2(Si1-xGex)2
series promising candidates to test their scaling predictions since the degree
of frustration can be tuned by varying $J_{ex}$ or $\Delta$ with dopant
concentration or pressure.
There is considerable interest in the ways that spin systems respond to the
presence of substantial frustration. Historically, this subject was stimulated
by the observation of spin glass freezing in disordered alloys and compounds,
and has more recently focused on systems where the lattice geometry produces a
macroscopic degeneracy of possible spin configurations. We suggest that
PrAu2Si2 reveals a new avenue to achieving frustration in systems with neither
static disorder nor geometrically frustrated lattices, through dynamic
fluctuations, either thermal or quantum, in proximity to a critical phase
boundary.
###### Acknowledgements.
We thank Amir Murani and Ross Stewart for scientific discussions and
assistance with the ILL experiments. This work was supported by the U.S.
Department of Energy Office of Science, under Contract No. DE-AC02-06CH11357.
## References
* (1) Fischer, K. H. & Hertz, J. A. Spin Glasses (Cambridge University Press, Cambridge, 1991).
* (2) Ramirez, A. P. Geometric frustration: Magic moments. Nature 421, 483 (2003)
* (3) Krimmel, A. et al. Spin-glass behavior in PrAu2Si2. Phys. Rev. B 59, R6604–R6607 (1999).
* (4) Süllow, S. et al. Spin Glass Behavior in URh2Ge2. Phys. Rev. Lett. 78, 354–357 (1997).
* (5) Süllow, S. et al. Disorder to order transition in the magnetic and electronic properties of URh2Ge2. Phys. Rev. B 61, 8878–8887 (2000).
* (6) Ryan, D. H. et al. Site disorder and spin-glass ordering in PrAu2Si2. J. Appl. Phys. 97, 10A908 (2005).
* (7) Krimmel, A., et al. The evolution from long-range magnetic order to spin-glass behaviour in PrAu2(Si1-xGex)2). J. Phys. Cond. Matt. 11, 6991–7003 (1999)
* (8) Goremychkin, E. A., Osborn, R., Rainford, B. D., Adroja, D. T. & Koza, M. Magnetic dynamics of the spin-glass system PrAu2Si2: an inelastic neutron scattering study. J. Magn. Magn. Mater., 310, 1535 (2007).
* (9) Grover, B. Dynamical Properties of Induced-Moment Systems Phys. Rev. 140, A1944–A1951 (1965).
* (10) Cooper, B. R. Magnetic Excitons in Real Singlet-Ground-State Ferromagnets: Application to Pr3Tl and fcc Pr Phys. Rev. B 6, 2730–2743 (1972).
* (11) Birgeneau, R. J., Als-Nielsen, J. & Bucher, E. Magnetic Excitons in Singlet–Ground-State Ferromagnets. Phys. Rev. Lett. 27, 1530–1533 (1971).
* (12) Houmann, J. G. et al. Magnetic Excitations and Magnetic Ordering in Praseodymium. Phys. Rev. Lett. 34, 587–590 (1975).
* (13) Blanco, J. A., Nicklow, R. M. & Schmitt, D. Paramagnetic excitations in singlet ground state PrNi2Si2. Phys. Rev. B 56, 11666–11672 (1997).
* (14) Heiniger, F., Bucher, E., Maita, J. P. & Longinotti, L. D. Thermodynamic properties of superconductors containing impurities in a crystal-field singlet ground state. Phys. Rev. B 12, 1778–1789 (1975).
* (15) Sherrington, D. The induced-moment spin glass: simple mean-field theory. J. Phys.C 12, L929–L933 (1979).
* (16) Sarkissian, B. V. B., and Coles, B. R. Spin-Glass to Overhauser-Alloy Transitions in Yt-Rare-Earth and Sc-Rare-Earth Alloys. Commun. Phys. 1, 17 (1976).
* (17) Yoshizawa, H., Shapiro, S. M., Hasanain, S. K., & Guertin, R. P. Neutron scattering study of the crystal electric field levels in an induced-moment spin-glass PrP0.9. Phys. Rev. B 27, 448–455 (1983).
* (18) In principle, there will also be broadening due to the energy dispersion of the crystal field excitations, which is normally modelled with a Gaussian lineshape [de Gennes, P. G. Inelastic magnetic scattering of neutrons at high temperatures. J. Phys. Chem. Solids 4, 223–226 (1958)]. This could explain the small Gaussian contribution in PrAu2Ge2 ($\sigma=0.13$ meV), where there should be no inhomogeneous broadening. No Gaussian component is required to fit the data from PrAu2Si2, although the Lorentzian widths are very similar in the two compounds. Since the exchange is estimated to be a factor four smaller in the silicide, we believe that its contribution is not significant.
* (19) Jensen, J., McEwen, K. A. & Stirling, W. G. Magnetic excitations in the longitudinally polarized antiferromagnetic phase of praseodymium. Phys. Rev. B 35, 3327–3340 (1987).
* (20) Lindgård, P.-A. Correlation Theory of Planar Magnets. Phys. Rev. Lett. 50, 690–693 (1983).
* (21) Als-Nielsen, J., Kjems, J. K., Buyers, W. J. L. & Birgeneau, R. J. Observation of a central mode in an exchange-coupled singlet-groundstate system. J. Phys. C 10, 2673–2678 (1977).
* (22) Tarjus, G., Kivelson, S. A., Nussinov, Z. & Vlot, P. The frustration-based approach of supercooled liquids and the glass transition: a review and critical assessment. J. Phys. Condens. Matter 17, R1143–R1182 (2005).
| arxiv-papers | 2008-06-24T17:26:03 | 2024-09-04T02:48:56.381962 | {
"license": "Public Domain",
"authors": "E. A. Goremychkin, R. Osborn, B. D. Rainford, R. T. Macaluso, D. T.\n Adroja, and M. Koza",
"submitter": "Ray Osborn",
"url": "https://arxiv.org/abs/0806.3935"
} |
0806.4150 | # A Numerical Model of Hercules A by Magnetic Tower:
Jet/Lobe Transition, Wiggling, and the Magnetic Field Distribution
Masanori Nakamura11affiliation: Theoretical Astrophysics, MS B227, Los Alamos
National Laboratory, NM 87545; nakamura@lanl.gov , Ian L. Tregillis
22affiliation: Applied Physics, MS F699, Los Alamos National Laboratory, NM
87545 , Hui Li11affiliation: Theoretical Astrophysics, MS B227, Los Alamos
National Laboratory, NM 87545; nakamura@lanl.gov , and Shengtai
Li33affiliation: Mathematical Modeling and Analysis, MS B284, Los Alamos
National Laboratory, NM 87545
###### Abstract
We apply magnetohydrodynamic (MHD) modeling to the radio galaxy Hercules A for
investigating the jet-driven shock, jet/lobe transition, wiggling, and
magnetic field distribution associated with this source. The model consists of
magnetic tower jets in a galaxy cluster environment, which has been discussed
in a series of our papers. The profile of underlying ambient gas plays an
important role in jet-lobe morphology. The balance between the magnetic
pressure generated by axial current and the ambient gas pressure can determine
the lobe radius. The jet body is confined jointly by the external pressure and
gravity inside the cluster core radius $R_{\rm c}$, while outside $R_{\rm c}$
it expands radially to form fat lobes in a steeply decreasing ambient thermal
pressure gradient. The current-carrying jets are responsible for generating a
strong, tightly wound helical magnetic field. This magnetic configuration will
be unstable against the current-driven kink mode and it visibly grows beyond
$R_{\rm c}$ where a separation between the jet forward and return currents
occurs. The reversed pinch profile of global magnetic field associated with
the jet and lobes produces projected ${\bm{B}}$-vector distributions aligned
with the jet flow and the lobe edge. AGN-driven shock powered by the expanding
magnetic tower jet surrounds the jet/lobe structure and heats the ambient ICM.
The lobes expand subsonically; no obvious hot spots are produced at the heads
of lobes. Several key features in our MHD modeling may be qualitatively
supported by the observations of Hercules A.
###### Subject headings:
galaxies:individual: Hercules A — galaxies: active — galaxies: jets — methods:
numerical — MHD
††slugcomment: Draft Version Jun. 17, 2008: Accepted for Publication in ApJ
## 1\. INTRODUCTION
In the present paper, we continue our discussion of the dynamics of
extragalactic jets in cluster environments within the framework of the
“magnetic tower” model (Lynden-Bell & Boily, 1994; Lynden-Bell, 1996), which
we have analyzed in a series of papers. Magnetohydrodynamic (MHD) mechanisms
are frequently invoked to model the launching, acceleration and collimation of
astrophysical jets (see, e.g., Ferrari, 1998; Meier et al., 2001, and
references therein). An underlying large-scale (coronal) poloidal field for
producing the magnetically driven jets is almost universally assumed in many
theoretical/numerical models. However, the origin and existence of such a
galactic magnetic field are still poorly understood.
In contrast with the large-scale field models, Lynden-Bell (Lynden-Bell &
Boily, 1994; Lynden-Bell, 1996) examined the expansion of the local force-free
magnetic loops anchored to the star and the accretion disk by using the semi-
analytic approach. Twisted magnetic fluxes due to the disk rotation make the
magnetic loops unstable and splay out at a semi-angle 60° from the rotational
axis of the disk. Global magnetostatic solutions of magnetic towers with
external thermal pressure were also computed by Li et al. (2001) using the
Grad-Shafranov equation in axisymmetry (see also, Lovelace et al., 2002;
Lovelace & Romanova, 2003; Uzdensky & MacFadyen, 2006). Full MHD numerical
simulations of magnetic towers have been performed in two-dimension
(axisymmetric) (Romanova et al., 1998; Turner et al., 1999; Ustyugova et al.,
2000; Kudoh et al., 2002; Kato et al., 2004) and three-dimension (Kato et al.,
2004). Magnetic towers are also observed in laboratory experiments (Hsu &
Bellan, 2002; Lebedev et al., 2005).
The first in our series, Li et al. (2006, hereafter Paper I), described the
basic assumptions and approaches in the numerical modeling of magnetic tower
jets. The evolution of the tower jets in a constant density/pressure
background was examined there. The second in our series, Nakamura et al.
(2006, hereafter Paper II) investigated the global structure of magnetic tower
jets in a gravitationally-stratified atmosphere, in terms of the MHD wave
structure, radial force equilibrium, and collimation. A large current flowing
parallel to the jet bulk flow plays an essential role in the determining the
lobe radius, as does the background pressure profile. The third in our series,
Nakamura et al. (2007, hereafter Paper III) investigated the stability
properties of magnetic tower jets on cluster scales. Current-driven
instabilities are responsible for the non-axisymmetric structures; the
external “kink” ($m=1$) mode grows outside $R_{\rm c}$, while the internal
“double helix” ($m=2$) mode grows predominantly inside $R_{\rm c}$. This is
the forth in our series of papers to examine the nonlinear magnetic tower
jets. Here we examine our model’s applicability to an individual source,
Hercules (Herc) A (3C 348), by performing three-dimensional MHD simulations.
Herc A is one of the most powerful double-lobed radio sources in the sky. Its
total energy content is approximately $3\times 10^{60}$ erg and it is
identified with the central compact diffuse (cD) galaxy of a cluster (Sadun &
Hayes, 1993) at a low redshift of $z=0.154$ (Siebert et al., 1999). The X-ray
luminosity of $4.8\times 10^{44}$ erg s-1 (Gizani & Leahy, 2004) is thought to
be due to bremsstrahlung of the very high temperature intracluster medium
(ICM). Herc A has a peculiar radio structure; the morphology looks like a
classical double-lobed Fanaroff-Riley type II (FR II) radio source (Fanaroff &
Riley, 1974), but it has no compact hotspots, instead showing an unusual jet-
dominated morphology as in an Fanaroff-Riley type I (FR I) radio source, which
was first revealed by Dreher & Feigelson (1984). Furthermore, the two jets in
Herc A are quite different in appearance. The eastern jet appears to have
continuous twisting flows, while the western jet leads to a unique sequence of
“rings”. The jets in Herc A are well-collimated initially but they flare
suddenly at a certain distance from the nucleus. An explanation for the
presence of such different structures in the same source is still unresolved
in the literature. Herc A is not a typical FR I source, either; its jets are
well-collimated and has knot components. Therefore, Herc A might be classified
an intermediate case: FR I/II.
Various models have been proposed for explaining the formation of the jets and
rings in Herc A, including both kinematic (Mason et al., 1988; Morrison &
Sadun, 1996; Sadun & Morrison, 2002) and dynamic (Meier et al., 1991; Saxton
et al., 2002) points of view. By performing axisymmetric, two-dimensional
hydrodynamic simulations, Meier et al. (1991) proposed a model in which an
over-dense and over-pressured jet undergoes a sudden expansion when it becomes
highly over-pressured compared to the background gas at a certain distance
from the core. This causes a subsequent conical expansion, giving rise to the
“ice cream cone” shape of the lobe. However, at present, there is nothing in
the X-ray observations to suggest that the jet becomes thermally over-
pressured as required in this model. In Saxton et al. (2002), the formation of
ring structure in the western jet is interpreted as the result of ribbon-like,
annular shocks propagating through the jet cocoon. Also, these authors
conclude that the absence of hot spots in Herc A results from the dynamics of
turbulent and entraining backflows in the radio lobe.
Recently, Gizani & Leahy (2003) have revealed detailed structures of total
intensity, spectral index, polarization, and projected magnetic field in Herc
A by using multifrequency VLA imaging. The jet disruption in the eastern lobe
is clearly visible, implying flow instabilities. The projected magnetic field,
after correcting for Faraday rotation, closely follows the edges of the lobes,
the jets, and the rings; the two lobes have generally similar field patterns.
This observational picture may disagree with the turbulent structure inside
the lobes suggested by Saxton et al. (2002).
Magnetic field distributions associated with the extragalactic jets and their
comparisons with the observations have been examined by using Stokes
parameters, Faraday rotation measure (RM), and the projected magnetic field
(${\bm{B}}$-vector) (e.g., Laing, 1981; Clarke et al., 1989, 1992; Hardee &
Rosen, 1999, 2002; Laing et al., 2006). In general, helical magnetic fields
produce asymmetric transverse brightness and polarization profiles; they are
symmetrical only if the field is purely toroidal or the jet is at 90∘ to the
line of sight (Laing, 1981). Laing et al. (2006) conclude that, to the first
approximation, the field configuration in FR I jet over 10 kpc scales is a
mixture of toroidal and longitudinal components.
Gizani & Leahy (2004)’s ROSAT X-ray observations of the intracluster gas in
the Herc A cluster have revealed extended X-ray emission coming from a compact
source at the center of cluster, out to a radius of 2.2 Mpc. The azimuthally-
averaged X-ray surface brightness profile is well fitted by a modified King
($\beta$) model (King, 1972; Cavaliere & Fusco-Femiano, 1976), with core
radius $R_{\rm c}=121\pm 10$ kpc and $\beta=0.74\pm 0.03$. Their X-ray
analysis supports the conclusion that the inner jets are confined thermally by
the cluster core gas: thermal pressure in the cluster core is almost 10 times
larger than the equipartition pressure in the eastern jet with the standard
assumptions (including no protons). The radio lobes are largely positioned
beyond the X-ray core radius, allowing for projection, so they are expanding
essentially into a power-law atmosphere with density falling as
$R^{-3\beta}\sim R^{-2.22}$, quite close to the $R^{\\-2}$ profile needed to
give the lobes a self-similar structure (Falle, 1991).
These observational evidence motivate us to apply our dynamical magnetic tower
jet model to Herc A. To our knowledge, no MHD model has been applied to Herc A
before though there are several hydrodynamic models as we have already
referred. Radio observations suggest that the magnetic fields need to be taken
into account to discuss about jet/lobe dynamics in Herc A. We believe that a
key clue to understanding the transition between the jet and the lobe may be
whether or not the inner jets become over-pressured during their propagation
in the X-ray thermal cluster gas. However, the jet internal pressure may be
magnetically, rather than thermally, dominated. This is in contrast to
previous models (Meier et al., 1991). Of particular interest here are the
jet/lobe transition, wiggling, and the magnetic field distribution associated
with the eastern jet of Herc A, which are not yet fully discussed among the
past hydrodynamic models of Herc A. We do not discuss the formation of rings
in the western jet in this paper.
This paper is organized as follows. In §2, we outline our numerical methods.
In §3, we describe our numerical results. Discussions and conclusions are
given in §4 and §5.
## 2\. NUMERICAL METHODS AND MODEL ASSUMPTIONS
We solve the nonlinear system of time-dependent ideal MHD equations
numerically in a 3-D Cartesian coordinate system $(x,\,y,\,z)$. The basic
numerical treatments (including the MHD numerical scheme) is essentially the
same as that in Papers I - III. We assume an initial hydrostatic equilibrium
in the gravitationally stratified medium, adopting an isothermal King model.
The magnetic flux and the mass are steadily injected in a central small volume
during a certain time period. Since the injected magnetic fields are not
force-free, they will evolve as a “magnetic tower” and interact with the
ambient medium. The dimensionless system of MHD equations is integrated in
time by using the TVD upwind scheme Li & Li (2003). Computations were
performed on the parallel Linux clusters at the Los Alamos National
Laboratory.
### 2.1. Numerical Set Up
We normalize physical quantities with the unit length scale $R_{0}$, the unit
density $\rho_{0}$, and the sound speed $C_{\rm s0}$ as the typical speed in
the system. Other quantities are derived from their combinations, e.g., the
typical time $t_{0}$ is $R_{0}/C_{\rm s0}$, etc. In this paper, we use some
observed quantities associated with the source Herc A (Gizani & Leahy, 2004,
hereafter GL04) in order to determine our normalization units. A central
electron density $n_{\rm e}=1.0\times 10^{-2}$ cm-3 is adopted in GL04,
suggesting quite a dense cluster, as such densities in clusters are typically
of order $10^{-3}$ cm-3 (Jones & Forman, 1984). This provides a unit density
$\rho_{0}=1.7\times 10^{-26}$ g cm-3. A single-temperature fit with $kT=4.25$
keV is assumed for the $\beta$-model cluster, giving a unit temperature
$T_{0}=4.9\times 10^{7}$ K, a unit pressure $p_{0}=2.3\times 10^{-10}$ dyn
cm-2, and a unit velocity (sound speed) $C_{s0}(\equiv\gamma
p_{0}/\rho_{0})=1.2\times 10^{8}$ cm s-1. Here, we choose a unit length
$R_{0}=30$ kpc, and therefore, a unit time $t_{0}=2.4\times 10^{7}$ yr. The
unit magnetic field $B_{0}$ is $(4\pi\rho_{0}C_{\rm s0}^{2})^{1/2}=53.4$
$\mu$G. In the King model we use here, we take the cluster core radius $R_{\rm
c}$ to be 120 kpc, corresponding to the normalized core radius 4.0, and take
the exponential slope $\kappa$ to be $1.1$ ($3\beta/2$ with $\beta=0.74$) by
adopting the observations in GL04.
We use $\rho_{0}$ and $p_{0}$ as the initial quantities at the origin
$(x,\,y,\,z)=(0,\,0,\,0)$ and thus, the normalized $\rho$ and $p$ are set to
unity at the origin in these simulations. The total computational domain is
taken to be $|x|,\,|y|,\,|z|\leq 15$. The number of grid points in the
simulations reported here is $360^{3}$, where the grid points are assigned
uniformly in the $x$, $y$, and $z$ directions. The simulation domain is from
$-450$ to $450$ kpc, with $\Delta x=\Delta y=\Delta z\sim 0.083$, which
correspond to $\sim 2.5$ kpc.
The injections of magnetic flux, mass, and the associated energies are the
same as those described in Paper I. The ratio between the toroidal to poloidal
fluxes of the injected fields is characterized by a parameter $\alpha=25$. The
magnetic field injection rate is described by $\gamma_{b}$ and is set to be
$\gamma_{b}=1$. The mass is injected at a rate of $\gamma_{\rho}=0.1$ over a
central volume with a characteristic radius $r_{\rho}=0.5$. Magnetic fluxes
and mass are continuously injected for $t_{\rm inj}=3.0$, after which the
injection is turned off. These parameters correspond to a magnetic energy
injection rate of $\sim 2.7\times 10^{46}$ ergs s-1 and an injection time
$\sim 72$ Myrs. The energy injection rate used here is a similar range to that
of other hydrodynamic simulations with the jet power $2\times 10^{45-46}$ ergs
s-1 (Saxton et al., 2002). We use the outflow boundary conditions at all outer
boundaries. Note that for most of the simulation duration, the waves and
magnetic fields stay within the simulation box, and all magnetic fields are
self-sustained by their internal currents.
### 2.2. Formulation of Synthetic Observation Images
Based on our simulations, we make synthetic observation images to compare with
the real observations of Herc A (Gizani & Leahy, 2003, hereafter GL03). The
synthetic radio telescope observations presented here were produced using a
modified form of the synthetic observation technique first described in
Tregillis et al. (2001). Here we provide an overview of the key points
relevant to the present investigation; a complete description of the technique
is provided in Tregillis (2002).
Our method utilizes the vector magnetic field structures evolved self-
consistently within these MHD calculations and a relativistic-electron
momentum distribution, $f(p)$, chosen by fiat during postprocessing. The
function $f(p)$ is defined over eight logarithmically-spaced momentum bins.
Each bin has an associated slope value, $q$, and within each bin the
distribution is a power law ($f\sim p^{-q}$). Slope values may vary between
bins, however, making it possible to specify non-power law distributions, such
as would be appropriate for cases where radiative aging is significant. The
specified distribution functions are free to vary from location to location
within the simulated flows, as is expected to be the case in real radio
galaxies. We note that although these distributions are not derived self-
consistently within the computed flows, as has been done in previous works
(Tregillis et al., 2001, 2004), imposing it as a postprocessing step does
provide a measure of flexibility in our investigations.
In this paper we choose the sign convention such that $\alpha>0$, and the flux
density $S_{\nu}\propto\nu^{-\alpha}$. For the present investigations, which
do not incorporate the effects of radiative aging and reacceleration, we
applied a spatially-uniform momentum slope value $q=4.5$ throughout the
simulated source. This corresponds to a synchrotron spectral index
$\alpha=0.75$, which is consistent with the observed range of spectral indices
within individual radio galaxies. For example, Gizani & Leahy (2003) mapped
the spectral index distribution in Herc A between 1.3 and 4.8 GHz, finding
$0.6\lesssim\alpha^{4.8}_{1.3}\lesssim 2.0$, with the flattest values within
the western jet and the steepest values along the lobe edges. They also found
$\alpha^{4.8}_{1.3}\approx 0.75$ in portions of the eastern jet. Our choice,
$\alpha=0.75$, is also consistent with typical hotspot spectral index values
$\alpha^{5.0}_{1.4}$ in the sample of 3CR sources studied by Alexander & Leahy
(1987).
The local number density of energetic electrons varied with the local magnetic
energy density (i.e., $n_{e}\propto B^{2}$). This approach represents a
partitioning relationship between the field and particle energies, such as is
posited by minimum-energy arguments (Miley, 1980). However, the overall
scaling was chosen such that the energy density of relativistic electrons was
far less than the magnetic energy density (i.e., we assume our simulated
object is out of equipartition, which is consistent with the lack of dynamical
feedback from the electrons in these calculations).
Given information about the local magnetic field and the local distribution of
energetic electrons, we compute a synchrotron emissivity $j_{\nu}$ in every
zone of the computational grid. As given by Jones et al. (1974), the
emissivity is
$j_{s}(\nu)=j_{\alpha 0}{4\pi e^{2}\over
c}f(p_{s})p_{s}^{q}\left({\nu_{B_{\bot}}\over\nu}\right)^{\alpha}\nu_{B_{\bot}}.$
(1)
The spectral index $\alpha$ is related to the local electron momentum index
$q$ via $\alpha=(q-3)/2$, $\nu_{B_{\bot}}=eB\sin{\Omega}/(2\pi m_{e}c)$, where
$\Omega$ projects the local field onto the sky, and $j_{\alpha 0}$ is an
order-unity dimensionless constant, defined in Jones et al. (1974). For a
selected observing frequency, $\nu$, the distribution, $f(p_{s})$, and the
index, $q$, are determined for each point on the numerical grid by
establishing the relevant electron momentum from the relation
$p_{s}=[2\nu/(3\nu_{B_{\bot}})]^{(1/2)}$, with $p_{s}$ in units $m_{e}c$.
Once the synchrotron emissivity is known for every numerical zone, we compute
surface brightness maps for optically-thin emission via line-of-sight
integrations. Similarly, we can obtain the Stokes $Q$ and $U$ parameters for
the synchrotron emission, as well as the correction for Faraday rotation
through the source, making detailed polarimetric studies possible. We write
the resultant data in fits format, and analyze it using conventional
observational packages (miriad and karma (Gooch, 1995)).
## 3\. RESULTS
Our magnetic tower model has been applied to the Herc A. In particular, our
goal is to reproduce large scale structures in the jets: the jet/lobe
transition, the non-axisymmetric deformation, and the projected magnetic field
distribution associated with the jets and lobes, as observed in GL03.
Figure 1.— Distribution of the gas pressure $p$ (logarithmic scale) with the
velocity field ($V_{x},\,V_{z}$) (arrows) in the $x-z$ plane at 192 Myr. The
length and time scales are converted into the corresponding dimensional
values. The cavities (low density and gas pressure) are formed due to the
magnetic tower expansion as corresponding lobes. The jets are deformed into
the wiggled structures inside the lobes.
Figure 1 shows a snapshot of the gas pressure $p$ in the ($x-z$) cross section
at $t=8.0$ (192 Myr). As we have examined in papers II and III, the global
structure of the magnetic tower consists of a well-collimated “body” and
radially extended “lobes” in the gravitationally stratified atmosphere. A
transition from the narrow jet to the fat lobe can occur at the cluster core
radius. This interesting property can be confirmed in resent observations
(GL03, GL04). Several key features in these magnetic tower jets (the MHD wave
structures, the heating process at the tower front, the cylindrical radial
force equilibrium at the tower edge, and the dynamic collimation process) are
similar to the results of paper II (see also Fig. 1 of paper II for details of
time evolution). The helically-twisted magnetic tower has a closed current
system that includes a pair of current circuits. Each circuit contains a
forward electric current path (the jet flow itself, with its toroidal magnetic
field, toward the lobe) and a return current (along some path back to the AGN
core). The global picture of a current-carrying jet with a closed current
system linking the magnetosphere of the central engine and the hot spots was
introduced by Benford (1978, 2006) and applied to AGN double radio sources. As
seen in Fig. 1, the high-pressure material of the ambient gas seems to follow
the edge of the jet near the bottom of each lobe. However, this is not the
case, but the local pressure enhancement occurs inside the lobes. In the jet
and lobe system, the axial currents drive the current-driven instabilities; it
eventually produces magnetic reconnection process even in our ideal MHD
assumption (the magnetic field will dissipate numerically). As a result, some
local heating may occur, but it never affects the dynamics of jet propagation
dramatically.
For a full 3-D visualization of the magnetic lines of force, the reader is
referred to paper III. The basic behavior of the helically-twisted magnetic
field in the present paper appears similar to that in paper III. The magnetic
tower jet has a well-ordered helical field configuration, with a tightly-wound
central helix going up along the central axis and a loosely -wound helix
coming back at the outer edge of the magnetic tower. The outer edge of the
magnetic tower may be identified as a tangential discontinuity without the
normal field component. The interior of tower (lobe) is separated from the
non-magnetized external gas via this discontinuity. At the tower edge, the
outward-directed magnetic pressure gradient force is roughly balanced with the
inwardly-directed thermal pressure gradient force. On the other hand, at the
core part of jet body, a quasi-force free equilibrium is achieved. In the
context of magnetically-controlled fusion systems, the helical field in the
magnetic tower can be regarded as the reversed field pinch (RFP) profile.
The jet axial current and the ambient gas pressure can together determine the
radius of the magnetically-dominated lobes (paper II). That is, the internal
gas pressure plays a minor role in the lobes (even if the thermal pressure is
greater than the non-thermal pressure (Morganti et al., 1988) as is typically
seen in FR I radio galaxies (Croston et al., 2003). The Alfvén speed becomes
large (about three times the local sound speed), while the plasma
$\beta(\equiv 2p/B^{2})$ becomes small ($\beta\lesssim 0.1$) inside the
density cavities due to the expansion of magnetic fluxes. As we have already
discussed in paper II, it is remarkably noted that the expanding tower
launches a preceding hydrodynamic shock in the ICM, which may be associated
with AGN-driven shocks seen in recent X-ray observations (Forman et al., 2005;
Nulsen et al., 2005; Fabian et al., 2006).
Figure 2 exhibits physical quantities along the $z$-axis at $t=3.0$ ($t=72$
Myrs). The magnetic tower-driven hydrodynamic shock front can be seen around
$z\sim 6.4$ in the profiles of $\rho$, $C_{\rm s}$, and $V_{z}$. The sound
speed jumps by 27 % [the temperature $T(\propto C_{\rm s}^{2})$ increases by
61 %] in the postshocked region due to the shock compression. The magnetic
tower front, which is identified as the lobe front, is located at $z\sim 5.2$.
There is another MHD wave front at $z\sim 4.5$ where an increase in $C_{\rm
s}$ (density/pressure also) is accompanied by a increase in magnetic pressure
(not shown in Fig. 2), indicating the reverse MHD slow mode. In later time,
this can steepen into a MHD slow shock that cause a compression (heating) at
the head of magnetic tower as seen in Fig. 1 (see also Fig. 2 and §3.1 of
paper II for details).
Figure 3 displays the positions of the shock and lobe fronts as a function of
time along the $z$-axis (propagating towards the positive direction). The
shock front has a constant propagation speed $V_{\rm SH}\sim 2.1$ ($2.52\times
10^{8}$ cm s-1) that implies a Mach number $\sim 1.63$. On the other hand, the
magnetically-dominated lobe front expands subsonically. This lobe front also
has a quasi-constant propagation speed $V_{\rm LB}\sim 1.5$ ($1.8\times
10^{8}$ cm s-1) that implies a Mach number $\sim 0.91$ ($C_{\rm s}$ increases
up to $\sim$ 1.64 in shocked ICM; see also Fig. 2).
Figure 2.— Axial profiles of physical quantities along the $z$-axis at
$t=3.0$ ($t=72$ Myrs). Density $\rho$, sound speed $C_{\rm s}$, and the axial
velocity component $V_{z}$. The positions of the expanding shock and lobe
fronts are shown. Figure 3.— The temporal evolution of the shock and lobe
front positions along the $z$-axis. During the time evolution, both fronts
propagate with almost constant speeds. The shock front expands supersonically
with Mach number $\sim 1.63$ (relative to the unshocked ICM), while the lobe
front proceeds subsonically with Mach number $\sim 0.9$ (relative to the
shocked ICM cocoon).
The current-carrying magnetic tower jet, which possesses a highly-wound
helical field, is subject to the current-driven instability (CDI). Although
the destabilization criteria will be modified by the ambient gas and the RFP
configuration of the tower, we find that the propagating magnetic tower jets
can develop the non-axisymmetric CDI modes. As seen in Fig. 4, both the
internal elliptical ($m=2$) mode like the “double helix” and the external kink
($m=1$) mode grow to produce the wiggles at different locations (see also
paper III for details). We see no disruption owing to shear-driven, Kelvin-
Helmholtz modes and/or the pressure-driven interchange modes.
Figure 4.— Distribution of the axial current density $J_{z}$ in the $x-z$
plane at the same time as Fig. 1. Two types of growing current-driven
(external/internal) modes are visible in the lobe and jet.
Figure 5 shows the synthetic radio synchrotron intensity at 5 GHz, along with
the polarization magnetic field vectors. The wiggled structures can be
observed in both the jet and lobe. Also, filamentary structures appear beyond
$R_{\rm c}$. The projected magnetic field is aligned with the jet and along
the lobe edge. The magnetic tower model produces polarization features that
are qualitatively consistent with those seen in the Herc A observations of
GL03. Furthermore, our dynamical solution of magnetic tower’s evolution may
give a self-consistent explanation to a feasible magnetic configuration
discussed in non-dynamical models having “spine” ($B_{\phi}>B_{z}$ inside the
lobe) – “Sheath” ($B_{z}>B_{\phi}$) structure in FR I jet over 10 kpc scale as
discussed in Laing (1981); Laing et al. (2006).
Figure 5.— Synthetic synchrotron surface brightness computed at 5 GHz
(greyscale), overlaid with polarization magnetic field vectors. The field
vectors are corrected for Faraday rotation, and represent the field
orientations at the source. The polarization field is nicely aligned with the
jets and lobe edges.
The image contrast in our synthetic synchrotron surface-brightness map is
approximately 2000:1 between the brightest pixels (beams) near the central
core and the faintest regions inside the lobes. Outside the small bright core
region, the image contrast is roughly 200:1.
When the synthetic 5 GHz synchrotron map is scaled to an angular size
comparable to that of Herc A, the brightest pixel (beam) in the synthetic 5
GHz surface brightness map is 875 mJy/beam. However, away from the small,
bright region near the core, the jets generally exhibit flux densities on the
order of 30-70 mJy/beam, the central sheath exhibits flux densities in the
range 20-30 mJy/beam, and the darkest parts of the lobes exhibit $\lesssim$
1.0 mJy/beam.
Our simulated source exhibits a fractional linear polarization, $m$, at 5 GHz
ranging from a few percent in the most magnetically tangled portions of the
lobes to $m\approx 30-50\%$ on the lobe edges. These values are consistent
with those typically observed in radio galaxies (Bridle & Perly, 1984; Muxlow
& Garrington, 1991) as well as with the $m$ values reported in GL03 for Herc A
at 4848 MHz. Fractional polarization values exceed 60% (roughly twice that
reported in GL03) in small, isolated pockets in the sheath around the central
jet and near the central core, which may indicate insufficient computational
resolution near the core for properly capturing cellular depolarization. In
regions outside the magnetized lobes, where the ambient medium in our
calculational domain has been mildly perturbed by expansion of the magnetic
tower, the degree of polarization approaches 72%. This is the expected result
for a region with a uniform magnetic field and a uniform power-law
distribution of nonthermal electrons with $q=4.5$ (Rybicki & Lightman, 1979).
## 4\. DISCUSSION
### 4.1. Shock and Lobe Expansions
Some features about the shock and lobe expansions in the simulation are
discussed by comparing with observations. A Chandra X-ray image of the Herc A
cluster shows that it has cavities and a shock front associated with the
powerful radio lobes (Nulsen et al., 2005). Unusually, these cavities show no
clear connection to the radio sources, while many cavities are associated with
radio sources in other cluster cases (e.g., McNamara et al., 2000), indicating
that they might be ghost cavities (e.g., McNamara et al., 2001).
Nulsen et al. (2005)’s observation reveals that a shock front surrounds the
Herc A radio source. It is elongated in the direction of the radio lobes and
appears to be a cocoon shock (Scheuer, 1974). Their fitting of a simple
hydrodynamic model to the surface brightness profile gives a Mach number for
the shock front of $\sim 1.65$ and its total energy $\sim 3\times 10^{61}$
ergs (its mean mechanical power $\sim 1.6\times 10^{46}$ ergs s-1) is
estimated. As we examined in §3, the magnetic tower drives a hydrodynamics
shock front in the ambient ICM. Derived Mach number $M\sim 1.63$ matches quite
well to their observation and also our energy inputs are in qualitatively
reasonable range.
As the radio lobes of Herc A lack anything resembling hotspots, it seems
likely that they are expanding at around the sound speed, and so are confined
by the thermal pressure of the ambient ICM, rather than by strong shocks
(Gizani & Leahy, 2004). To our knowledge, there may be no direct result
revealing the lobe expansion speed of Herc A radio lobes, but we discuss
general properties of expanding lobes in radio galaxy based on Kraft et al.
(2006)’s argument. It is believed that the radio lobes are greatly
overpressurized relative to the ambient medium, indicating that sharp X-ray
surface brightness discontinuities or large jumps in gas temperature at the
lobe edges. A high Mach number ($M\sim 8.5$) shock is detected around the SW
radio lobe of Centaurus A (Kraft et al., 2003).
However, there is no evidence especially for FR I or low-power FR II radio
galaxies, including Herc A (Gizani & Leahy, 2004) and 3C 388 (Kraft et al.,
2006). If thermal conduction in the ICM is efficient (an order of the Spitzer
value), then shocks may be nearly isothermal and thus difficult to detect
(Fabian et al., 2006). However, Chandra observations suggest that the thermal
conduction (and viscosity) of the ICM is orders of magnitude below the Spitzer
value (e.g., Vikhlinin et al., 2001; Kraft et al., 2003). Thus, some lobes
including Herc A may be inflating transonically and subsonically (i.e., $M\leq
1$). Based on hydrodynamic arguments of a buoyant bubble, $M\sim 0.6$ in M87
(Churazov et al., 2002) and $M\sim 0.5-0.9$ in 3C 388 (Kraft et al., 2006) are
estimated. Our numerical result indicates that the magnetically-dominated lobe
expands subsonically with $M\sim 0.91$, discussed in §3. This will also match
the observationally preferable picture.
### 4.2. Lobe Size
We next discuss the radial size of the lobes $r_{\rm lobe}$ as based on our
model (Nakamura et al., 2006). The poloidal magnetic fluxes and the poloidal
current $I_{z}$ in the lobes remain well collimated around the central axis
even after the system has evolved fully. This implies that the toroidal
magnetic fields in the lobe region are distributed roughly as
$\displaystyle B_{\phi}\sim I_{z}/r.$ (2)
When the lobe experiences sufficient expansion, we can expect
$|B_{z}|\sim|B_{\phi}|$ at the outer edge of the RFP configuration (Freidberg,
1982). The magnetic and background pressures tend to balance each other at the
tower edge. Thus, we can expect that
$\frac{B_{\phi}^{2}+B_{z}^{2}}{2}\sim B_{\phi}^{2}\sim p_{\rm e}~{},$ (3)
where $p_{\rm e}$ is the external gas pressure at the tower edge. By combining
eqs. (2) and (3), we have
$B_{\phi}^{2}\sim\left(\frac{I_{z}}{r_{\rm lobe}}\right)^{2}\sim p_{\rm
e}~{},$ (4)
which gives
$r_{\rm lobe}\sim I_{z}~{}p_{e}^{-1/2}~{}.$ (5)
Figure 6 shows the distribution of $p$ along a transverse line at $z=8.5$ (255
kpc), when the lobe has a maximum radius (after 192 Myr, as in Figs. 1 and 4).
We can see that the external gas pressure begins to decrease around $x\sim 5$
(150 kpc) toward the central ($z$) axis. The pressure cavity corresponds to
the lobe of our magnetic tower. We can estimate the lobe radius $r_{\rm lobe}$
predicted by eq. (5) from physical quantities in simulation. From Fig. 2, we
take $J_{z}\sim 3$ ($4.5\times 10^{-24}$ A/cm-2) within a radius $r\sim 0.75$
($22.5$ kpc), which makes $I_{z}\sim 1.68$ ( $5\times 10^{17}$ A). As
suggested by Fig. 4, we estimate $p_{\rm e}\sim 0.11$ ($2.5\times 10^{-11}$
dyn cm-2) for the external gas pressure. We therefore obtain $r_{\rm lobe}\sim
5.2$ or roughly 156 kpc. Thus, we find that the depression radius in Fig. 6 is
qualitatively consistent with $r_{\rm lobe}$ and is comparable with the
physical scale seen in radio observations (Gizani & Leahy, 2003). Based on
this numerical result, we estimate that one simulated lobe of Herc-A may
contain a total energy $\sim 1.8\times 10^{61}$ ergs, out of which $\sim
1.0\times 10^{61}$ ergs are in magnetic energy, $\sim 1.4\times 10^{60}$ ergs
are in kinetic energy, and $\sim 6.6\times 10^{60}$ ergs are in thermal
energy, respectively. In order to explain the lobe radius by this magnetically
dominated lobe system, an axial current $\sim 5\times 10^{17}$ A, which is
flowing along the jet central axis, is needed.
Figure 6.— Transverse profile in the $x$-direction of the gas pressure $p$ on
$z=8.5$ (255 kpc). The lobe radius $r_{\rm lobe}$ calculated via eq. (5) is
also shown. The pressure cavity radius and $r_{\rm lobe}$ qualitatively agree
with each other.
## 5\. CONCLUSIONS
By performing 3D MHD simulations, we have investigated the nonlinear dynamics
of magnetic tower jets in a galaxy cluster environment. In this fourth paper
of the series, we have applied our numerical modeling of the magnetic tower to
a specific source, Hercules A.
To our knowledge, this is the first trial of discussing Hercules A jet by
using a magnetohydrodynamic model. Projected magnetic field distributions
associated with the jets and lobes of Hercules A have been revealed in recent
radio observation and therefore, the dynamics including magnetic fields plays
a role in understanding this source. In the present paper, we have
investigated the jet-driven shock, jet/lobe transition, wiggling, and
projected magnetic field distribution of Hercules A and, in general, these
feasutres can be explained by our magnetic tower model.
Our conclusions are summarized as follows:
1. 1.
A transition from the narrow jet to the fat lobe can occur at the cluster core
radius. This interesting feature can also be confirmed in recent observations
(Gizani & Leahy, 2003, 2004). In our models, the tightly collimated helical
field configuration of a magnetic tower jet expands abruptly beyond the core
radius due to the decreasing external gas pressure. Thus, the pressure profile
of the surrounding ICM plays an important role in determining the morphology
of a magnetic tower jet.
2. 2.
The expanding magnetic tower jet produces the preceding hydrodynamic shock
wave in the ambient ICM. This can be interpreted as AGN-driven shock in X-ray
observation and the derived shock Mach number is identical to Nulsen et al.
(2005)’s observation result. The magnetic tower lobes expand subsonically and
thus no hot spots are produced at the end parts of lobes. This may be a
general understanding of FR I or low power FR II radio galaxies (Kraft et al.,
2006).
3. 3.
The size of the magnetic tower lobe can be accurately estimated from the jet
axial current and the external gas pressure at the lobe edge. The estimated
lobe size derived by our numerical simulation is qualitatively comparable with
the physical scale seen in Gizani & Leahy (2003). In our model, if the
background gas has a steep slope ($\beta$) in the King profile, the magnetic
jet can potentially lead to a huge magnetic tower lobe.
4. 4.
Magnetic tower jets, which have a reversed magnetic pinch profile, are subject
to current-driven instabilities. Specifically, the apparent distortions by the
non-axisymmetric kink mode are visible at the jet body inside the lobe rather
than the central core. By expanding the lobe the edge of axial currents could
be free against the exciting external kink mode; this situation may be
suppressed inside the cluster core radius.
5. 5.
The synthetic polarimetry of our magnetic-tower jets is consistent with the
gross polarization features observed in the jets of Hercules A. The magnetic
tower model produces projected ${\bm{B}}$-vector distributions that are
similar to those observed in the global extragalactic jet-lobe system,
especially along the jets and the lobe edges.
6. 6.
Magnetic tower jets may produce a self-consistent picture of the field
configuration discussed in several non-dynamical model of FR I jets. It
consists of toroidally dominated spine inside the lobe and poloidally
dominated sheath at the lobe edge (a closed poloidal magnetic flux with a
toroidal magnetic component, rather than one side directed helical
configuration).
Helpful discussions with Philipp Kronberg and Steven Diehl are gratefully
acknowledged. The authors thank the anonymous referee for helpful suggestions.
This work was carried out under the auspices of the National Nuclear Security
Administration of the U.S. Department of Energy at Los Alamos National
Laboratory under Contract No. DE-AC52-06NA25396. It was supported by the
Laboratory Directed Research and Development Program at LANL and by IGPP at
LANL.
## References
* Alexander & Leahy (1987) Alexander, P. & Leahy, J. P. 1987, MNRAS, 225, 1
* Benford (1978) Benford, G. 1978, MNRAS, 183, 29
* Benford (2006) Benford, G. 2006, MNRAS, 369, 77
* Bridle & Perly (1984) Bridle, A. H., & Perley, R. A. 1984, ARA&A, 22, 319
* Cavaliere & Fusco-Femiano (1976) Cavaliere, A., & Fusco-Femiano, R. 1976, A&A, 49, 137
* Churazov et al. (2002) Churazov, E., Sunyaev, R., Forman, W., & Böhringer, H. 2002, MNRAS, 331, 545
* Croston et al. (2003) Croston, J. H., Hardcastle, M. J., Birkinshaw, M., & Worrall, D. M. 2003, MNRAS, 346, 1041
* Clarke et al. (1989) Clarke, D. A., Norman, M. L., & Burns, J. O. 1989, ApJ, 342, 700
* Clarke et al. (1992) Clarke, D. A., Bridle, A. H., Burns, J. O., Perley, R. A., & Norman, M. L., 1992, ApJ, 385, 173
* Dreher & Feigelson (1984) Dreher, J. W., & Feigelson, E. D. 1984, Nature, 308, 43
* Forman et al. (2005) Forman, W. et al. 2005, ApJ, 635, 894
* Fabian et al. (2006) Fabian, A. et al. 2006, MNRAS, 366, 417
* Falle (1991) Falle, S. A. E. G. 1991, MNRAS, 250, 581
* Fanaroff & Riley (1974) Fanaroff, B & Riley, J. M. 1974, MNRAS, 167,31
* Ferrari (1998) Ferrari, A. 1998, ARA&A, 36, 539
* Freidberg (1982) Freidberg J. P. 1982, Rev. Mod. Phys., 54, 801
* Gizani & Leahy (2003) Gizani, N. A. B., & Leahy, J. P. 2003, MNRAS, 342, 399 (GL03)
* Gizani & Leahy (2004) Gizani, N. A. B., & Leahy, J. P. 2004, MNRAS, 350, 865 (GL04)
* Gooch (1995) Gooch, R. E. 1995, in ASP Conf. Ser. 101, Astronomical Data Analysis Software and Systems V, ed. G. H. Jacoby & J. Barnes (San Francisco: ASP), 80
* Hardee & Rosen (1999) Hardee, P. E., & Rosen, A. 1999, ApJ, 524, 650
* Hardee & Rosen (2002) Hardee, P. E., & Rosen, A. 2002, ApJ, 576, 204
* Hsu & Bellan (2002) Hsu, S. C., & Bellan, P. M. 2002, MNRAS, 334, 257
* Jones et al. (1974) Jones, T. W., O’Dell, S. L., & Stein, W. A. 1974, ApJ, 188, 353
* Jones & Forman (1984) Jones, C., & Forman, W. 1984, ApJ, 276,38
* Kato et al. (2004) Kato, Y., Hayashi, M. R., & Matsumoto, R. 2004, ApJ, 600, 338
* Kato et al. (2004) Kato, Y., Mineshige, S., & Shibata, K. 2004, ApJ, 605, 307
* King (1972) King, I. 1972, ApJ, 174, L123
* Kraft et al. (2003) Kraft R. P., et al. 2003, ApJ, 592, 129
* Kraft et al. (2006) Kraft R. P., Azcona, J., Forman, W. R., Hardcastle M. J., Jones, C., & Murray, S. S. 2006, ApJ, 639, 753
* Kudoh et al. (2002) Kudoh, T., Matsumoto, R., & Shibata, K. 2002, PASJ, 54, 26
* Laing (1981) Laing, R. A. 1981, ApJ, 248, 87
* Laing et al. (2006) Laing, R. A., Canvin, J. R., & Bridle, A. H. 2006, Astron. Nachr. 327, No. 5/6, 523
* Lebedev et al. (2005) Lebedev, S. V. et al. 2005, MNRAS, 361, 97
* Li et al. (2001) Li, H., Lovelace, R. V. E., Finn, J. M., & Colgate, S. A. 2001, ApJ, 561, 915
* Li & Li (2003) Li, S., & Li, H. 2003, Technical Report, LA-UR-03-8935, Los Alamos National Laboratory
* Li et al. (2006) Li, H., Lapenta, G., Finn, J. M., Li, S., & Colgate, S. A. 2006, ApJ, 643, 92 (Paper I)
* Lynden-Bell (1996) Lynden-Bell, D. 1996, MNRAS, 279, 389
* Lynden-Bell & Boily (1994) Lynden-Bell, D., & Boily, C. 1994, MNRAS, 267, 146
* Lovelace et al. (2002) Lovelace, R. V. E., Li, H., Koldoba, A. V., Ustyugova, G. V., & Romanova, M. M. 2002, ApJ, 572, 445
* Lovelace & Romanova (2003) Lovelace, R. V. E., & Romanova, M. M. 2003, ApJ, 596, L159
* Mason et al. (1988) Mason, A., Morrison, P., & Sadun, A. C. 1988, Nature, 333, 640
* McNamara et al. (2000) McNamara, B. R., et al. 2000. ApJ, 534, L135
* McNamara et al. (2001) McNamara, B. R., et al. 2001. ApJ, 562, L149
* Meier et al. (1991) Meier, D. L., Sadun, A. C., & Lind K. R. 1991, ApJ, 379, 141
* Meier et al. (2001) Meier, D. L., Koide, S., & Uchida, Y. 2001, Science, 291, 84
* Miley (1980) Miley, G. 1980, ARA&A, 18, 165
* Morganti et al. (1988) Morganti, R., Fanti, R., Gioia, I. M., Harris, D. E., Parma, P., & Ruiter, H. 1988, A&A, 189, 11
* Morrison & Sadun (1996) Morrison, P., & Sadun, A. C. 1996, MNRAS, 278, 265
* Muxlow & Garrington (1991) Muxlow, T. W. B. & Garrington, S. T. 1991 Beams and Jets in Astrophysics (New York: ambridge University Press), 52–99
* Nakamura et al. (2006) Nakamura, M., Li, H., & Li, S. 2006, ApJ, 652, 1059 (Paper II)
* Nakamura et al. (2007) Nakamura, M., Li, H., & Li, S. 2007, ApJ, 656, 721 (Paper III)
* Nulsen et al. (2005) Nulsen, P. E. J., et al. 2005, ApJ, 625, L9
* Romanova et al. (1998) Romanova, M. M., Ustyugova, G. V., Koldoba, A. V., Chechetkin, V. M., & Lovelace, R. V. E. 1998, ApJ, 500, 703
* Rybicki & Lightman (1979) Rybicki, G. B., & Lightman, A. P. 1979, Radiative Process in Astrophysics (New York: John Wiley and Sons, Inc.)
* Sadun & Hayes (1993) Sadun, A. C., & Hayes, J. J. E. 1993, PASP, 105, 379
* Sadun & Morrison (2002) Sadun, A. C., & Morrison, P. 2002, ApJ, 123, 2312
* Saxton et al. (2002) Saxton, C. J., Bicknell, G. V., & Sutherland, R. S. 2002, ApJ, 579, 176
* Scheuer (1974) Scheuer, P. A. G. 1974, MNRAS, 166, 513
* Siebert et al. (1999) Siebert, J., Kawai, N., & Brinkmann, W. 1999, A&A, 350, 25
* Tregillis et al. (2001) Tregillis, I. L., Jones, T. W., & Ryu, D. 2001, ApJ, 557, 475
* Tregillis (2002) Tregillis, I. L. 2002, Ph.D thesis, U. of Minnesota
* Tregillis et al. (2004) Tregillis, I. L., Jones, T. W., & Ryu, D. 2004, ApJ, 601, 778
* Turner et al. (1999) Turner, N. J., Bodenheimer, P., & Różyczka M. 1999, ApJ, 524, 12
* Ustyugova et al. (2000) Ustyugova, G. V., Lovelace, R. V. E., Romanova, M. M., Li, H., Colgate, S. A. (2000), ApJ, 541, L21
* Uzdensky & MacFadyen (2006) Uzdensky, D. A., & MacFadyen, A. I. 2006, ApJ, 647, 1192
* Vikhlinin et al. (2001) Vikhlinin, A., Markevitch, M., & Murray, S. S. 2001, ApJ, 549, L47
| arxiv-papers | 2008-06-25T17:08:08 | 2024-09-04T02:48:56.391852 | {
"license": "Public Domain",
"authors": "Masanori Nakamura (T-6, LANL), Ian L. Tregillis (X-1, LANL), Hui Li\n (T-6, LANL), and Shengtai Li (T-7, LANL)",
"submitter": "Masanori Nakamura",
"url": "https://arxiv.org/abs/0806.4150"
} |
0806.4166 |
# Beam-Energy and System-Size Dependence of Dynamical Net Charge Fluctuations
Monika Sharma for the STAR Collaboration monika@rcf.rhic.bnl.gov Department of
Physics and Astronomy, Wayne State University, Detroit, Michigan, USA
###### Abstract
We present measurements of net charge fluctuations in $Au+Au$ collisions at
$\sqrt{s_{NN}}=$ 19.6, 62.4, 130, and 200 GeV, $Cu+Cu$ collisions at
$\sqrt{s_{NN}}=$ 62.4, 200 GeV, and $p+p$ collisions at $\sqrt{s}=$ 200 GeV
using the net charge dynamical fluctuations measure $\nu_{+-,dyn}$. The
dynamical fluctuations are non-zero at all energies and exhibit a rather
modest dependence on beam energy. We find that at a given energy and collision
system, net charge dynamical fluctuations violate $1/N_{ch}$ scaling, but
display approximate $1/N_{part}$ scaling. We observe strong dependence of
dynamical fluctuations on the azimuthal angular range and pseudorapidity
widths.
## 1 Introduction
Anomalous transverse momentum and net charge event-by-event fluctuations have
been proposed as indicators of the formation of a quark gluon plasma (QGP) in
the midst of high-energy heavy ion collisions. Entropy conserving
hadronization of a plasma of quarks and gluons could produce a final state
characterized by a dramatic reduction of the net charge fluctuations relative
to that of a hadron gas JeonKoch00 . Measurements of net charge fluctuations
were reported by both PHENIX AdcoxPRC89 and STAR Adams03c collaborations on
the basis of $Au+Au$ data acquired during the first RHIC run at
$\sqrt{s_{NN}}=$ 130 GeV. In this work, we report a systematic study of the
net charge fluctuations as a function of the system size and their energy
using $\nu_{+-{\rm,dyn}}$ fluctuation measure, defined as follows:
$\nu_{+-,dyn}=\frac{{\left\langle{N_{+}(N_{+}-1)}\right\rangle}}{{\left\langle{N_{+}}\right\rangle^{2}}}+\\\
\frac{{\left\langle{N_{-}(N_{-}-1)}\right\rangle}}{{\left\langle{N_{-}}\right\rangle^{2}}}-2\frac{{\left\langle{N_{-}N_{+}}\right\rangle}}{{\left\langle{N_{-}}\right\rangle\left\langle{N_{+}}\right\rangle}}$
(1)
where $N_{\pm}$ are the number of positively and negatively charged particles
in the acceptance of interest. We present measurements of dynamical net charge
fluctuations in $Au+Au$ collisions at $\sqrt{s_{NN}}=$ 19.6, 62.4, 130, and
200 GeV, $Cu+Cu$ collisions at $\sqrt{s_{NN}}=$ 62.4, 200 GeV and in $p+p$
collisions at $\sqrt{s_{NN}}=$ 200 GeV. We study the beam energy, system size
and collision centrality dependencies quantitatively in order to identify
possible signature of the formation of a QGP.
## 2 Experimental Method
The data used in this analysis were measured using the Solenoidal Tracker at
RHIC (STAR) detector during the 2001, 2002, 2004 and 2005 data RHIC runs at
Brookhaven National Laboratory. The $Au+Au$ collisions at $\sqrt{s_{NN}}=$
62.4, 130, and 200 GeV data and $Cu+Cu$ collisions at $\sqrt{s_{NN}}=$ 62.4
and 200 GeV were acquired with minimum bias triggers. For 19.6 GeV data, a
combination of minimum bias and central triggers was used. Technical
descriptions of the STAR detector and its components are published in
technical reports Star03 .
This analysis used tracks from the TPC with transverse momentum in the range
$0.2<p_{T}<5.0$ GeV/c with pseudorapidity $|\eta|<0.5$. The centrality bins
were calculated as a fraction of this multiplicity distribution starting at
the highest multiplicities. The ranges used were 0-5% (most central
collisions), 5-10%, 10-20%, 20-30%, 30-40%, 40-50%, 50-60%, 60-70%, and 70-80%
(most peripheral) for $Au+Au$ collisions. Similarly, collision centrality
slices used in $Cu+Cu$ collisions are 0-10% (most central), 10-20%, 20-30%,
30-40%, 40-50% and 50-60% (most peripheral). Each centrality bin is associated
with an average number of participating nucleons, $N_{part}$, using Glauber
Monte Carlo calculation StarGlauber03 .
## 3 Net Charge Fluctuation Results
We present, in Fig. 1, measurements of the dynamical net charge fluctuations,
$\nu_{+-{\rm,dyn}}$, as a function of collision centrality in $Au+Au$
collisions at $\sqrt{s_{NN}}=$ 19.6, 62.4, 130, and 200 GeV, $Cu+Cu$
collisions at $\sqrt{s_{NN}}=$ 62.4 and 200 GeV.
Figure 1: Net charge dynamical fluctuations, $\nu_{+-{\rm,dyn}}$, of particles
produced within pseudorapidity $|\eta|<0.5$, as function of the number of
participating nucleons.
The dynamical net charge fluctuations, in general, exhibit a monotonic
dependence on the number of participating nucleons. At a given number of
participants the measured fluctuations also exhibit a modest dependence on
beam energy, with $\nu_{+-{\rm,dyn}}$ magnitude being the largest in $Au+Au$
collisions at $\sqrt{s_{NN}}=$ 19.6 GeV. The $\nu_{+-{\rm,dyn}}$ values
measured for $p+p$ collisions at $\sqrt{s}=$ 200 GeV amounts to -0.230 $\pm$
0.019(stat). Dependence on the number of participating nucleons can be seen in
Fig. 2(b), discussed in Section III A.
### 3.1 Collision Centrality Dependence
Fig. 2(a) shows the dynamical fluctuations scaled by the measured particle
multiplicity density in pseudorapidity space ($dN_{ch}/d\eta$). Data from
$Au+Au$ and $Cu+Cu$ collisions are shown with solid and open symbols,
respectively. Values of $dN_{ch}/d\eta$ used for the scaling correspond to
efficiency corrected charged particle multiplicities measured by STAR
StardNdeta and PHOBOS Phobos01 . The magnitude of $\nu_{+-{\rm,dyn}}$ scaled
by $dN_{ch}/d\eta$ for $Au+Au$ 200 GeV data is significantly different from
the results of other systems and energies. The observed
$|\nu_{+-{\rm,dyn}}dN_{ch}/d\eta|$ increases with the increase in collision
centrality. The dashed line in the figure corresponds to charge conservation
effect and the solid line to the prediction for a resonance gas. The data
indicates that dynamical net charge fluctuations, scaled by $dN_{ch}/d\eta$
for the most central collisions are qualitatively consistent with resonance
gas prediction JeonKoch00 .
Fig. 2(b) presents the dynamical fluctuation scaled by the number of
participants, $N_{part}\nu_{+-{\rm,dyn}}$ as a function of the number of
participants. The measured data scaled by the number of participants
($N_{part}$) are thus consistent with no or a very weak collision centrality
dependence. However, a definite system size and energy dependence is observed.
Vertical error bars represent statistical uncertainties.
Figure 2: Net charge dynamical fluctuations, $\nu_{+-{\rm,dyn}}$, of particles
produced with pseudorapidity $|\eta|<0.5$ scaled by (a) the multiplicity,
$dN_{ch}/d\eta$, (b) the number of participants. The dashed line corresponds
to charge conservation effect and the solid line to the prediction for a
resonance gas.
### 3.2 Longitudinal and Azimuthal Dependencies of the Dynamical Fluctuations
We plot in Fig. 4 values of $\nu_{+-{\rm,dyn}}(\eta)$ measured for different
ranges of pseudorapidity, $\eta$, normalized by the magnitude of
$\nu_{+-{\rm,dyn}}(\eta)$ for a pseudorapidity range $|\eta|<1$
($\nu_{+-{\rm,dyn}}(1)$) to enable comparison of different centralities, beam
energies and system size. The magnitude of the normalized correlation is
maximum for the smallest pseudorapidity ranges and decreases monotonically to
unity, at all energies and centralities, with increasing pseudorapidity range.
This shows that the collision dynamics in $p+p$ collisions, 0-10% $Cu+Cu$ and
0-5% $Au+Au$ collisions are significantly different. Indeed, we find the
relative magnitude of the correlations measured for $|\eta|<0.5$ increases by
nearly 25% for $Au+Au$ 200 GeV relative to those in $p+p$. Note in particular
that the slope ($d\nu_{+-{\rm,dyn}}/d\eta$) in $p+p$, $Cu+Cu$ and $Au+Au$
systems depends on the correlation length (in pseudorapidity): the shorter the
correlation, the larger the slope. The observed distributions then indicate
that the correlation length is shorter for central collisions and for larger
systems, in agreement with the observed reduction of the charge balance
function StarBalanceFct .
Figure 3: Dynamical fluctuations $\nu_{+-{\rm,dyn}}$, normalized to their
value for $|\eta|<1$, as function of the integrated pseudorapidity range. Data
for $Au+Au$ collisions at $\sqrt{s_{NN}}=$ 62.4, 200 GeV (0-5%) along with
data for $Cu+Cu$ collisions at $\sqrt{s_{NN}}=$ 62.4, 200 GeV (0-10%), are
compared to inclusive $p+p$ data at $\sqrt{s_{NN}}=$ 200 GeV.
Figure 4: Dynamical fluctuations $\nu_{+-{\rm,dyn}}$, as a function of the
integrated azimuthal range $\phi$ for similar number of participating nucleons
for $Au+Au$ 200 GeV and $Cu+Cu$ 200 GeV.
Fig. 4 displays measured values of $\nu_{+-{\rm,dyn}}$ as a function of
azimuthal angle ranges from 10 to 360 degrees for $Au+Au$ and $Cu+Cu$
collisions at 200 GeV. $Au+Au$ and $Cu+Cu$ collisions are compared at similar
number of participating nucleons. One observes that the magnitude of
$\nu_{+-{\rm,dyn}}$ with respect to azimuthal angle, $\phi$, is similar for
similar number of participating nucleons in both systems with best agreement
for collisions with more than 20 participants.
## 4 Summary and Conclusions
We have presented measurements of dynamical net charge fluctuations in $Au+Au$
collisions at $\sqrt{s_{NN}}=$ 19.6, 62.4, 130, 200 GeV, $Cu+Cu$ collisions at
$\sqrt{s_{NN}}=$ 62.4, 200 GeV and $p+p$ collisions at $\sqrt{s_{NN}}=$ 200
GeV, using the measure $\nu_{+-{\rm,dyn}}$. The fluctuations are non vanishing
at all energies and exhibit a rather modest dependence on beam energy in the
range 19.6 $<$ $\sqrt{s_{NN}}$ $<$ 200 GeV for $Au+Au$ as well as $Cu+Cu$
collisions. Net charge dynamical fluctuations violate the trivial $1/N_{ch}$
scaling expected for nuclear collisions consisting of independent nucleon-
nucleon interactions. Scaled dynamical net charge fluctuations
$|\nu_{+-{\rm,dyn}}dN_{ch}/d\eta|$ grow by up to 40% from peripheral to
central collisions. This centrality dependence may arise, in part due to the
large radial collective flow produced in $Au+Au$ collisions. We also studied
fluctuations as a function of azimuthal angle and pseudorapidity and found
that dynamical fluctuations exhibit a strong dependence on rapidity and
azimuthal angular ranges which could be attributed in part to radial flow
effects.
## References
* (1) S. Jeon and V. Koch, Phys. Rev. Lett. 85, 2076 (2000); H. Heiselberg and A. D. Jackson, Phys. Rev. C63, 064904 (2001); M. Asakawa, U. Heinz, B. Mueller, Phys. Rev. Lett. 85, 2072 (2000).
* (2) K. Adcox et al. (PHENIX Collaboration), Phys. Rev. Lett. 89, 082301 (2002).
* (3) J. Adams et al. (STAR Collaboration), Phys. Rev. C68, 044905 (2003).
* (4) M. Anderson et al., Nucl. Instrum. Meth. A499, 624 (2003); Nucl. Instrum. Meth. A499, 659 (2003).
* (5) J. Adams et al., (STAR Collaboration), [nucl-ex/0311017].
* (6) C. Adler et al., (STAR collaboration), Phys. Rev. Lett. 87, 112303 (2001).
* (7) B. B. Back et al., (PHOBOS Collaboration), Phys. Rev. Lett. 87, 102303 (2001); Phys. Rev. Lett. 88, 022302 (2002); Phys. Rev. C65, 061901R (2002); Phys. Rev. C70, 021902 (2004); Phys. Rev. Lett. 91, 052303 (2003); Phys. Rev. C74, 021901(R) (2006); [nucl-ex/0601026].
* (8) J. Adams et al., (STAR Collaboration), Phys. Rev. Lett. 90, 172301 (2003).
| arxiv-papers | 2008-06-25T19:06:25 | 2024-09-04T02:48:56.397706 | {
"license": "Public Domain",
"authors": "Monika Sharma (for the STAR Collaboration)",
"submitter": "Monika Sharma",
"url": "https://arxiv.org/abs/0806.4166"
} |
0806.4218 | # Electromagnetically-Induced-Transparency-Like Effect in the Degenerate
Triple-Resonant Optical Parametric Amplifier
Chenguang Ye Jing Zhang† The State Key Laboratory of Quantum Optics and
Quantum Optics Devices, Institute of Opto-Electronics, Shanxi University,
Taiyuan 030006, P.R. China
###### Abstract
We investigate experimentally the absorptive and dispersive properties of
triple-resonant optical parametric amplifier OPA for the degenerate
subharmonic field. In the experiment, the subharmonic field is utilized as the
probe field and the harmonic wave as the pump field. We demonstrate that EIT-
like effect can be simulated in the triple-resonant OPA when the cavity line-
width for the harmonic wave is narrower than that for the subharmonic field.
However, this phenomenon can not be observed in a double-resonant OPA. The
narrow transparency window appears in the reflected field. Especially, in the
measured dispersive spectra of triple-resonant OPA, a very steep variation of
the dispersive profile of the subharmonic field is observed, which can result
in a slow light as that observed in atomic EIT medium.
190.4410, 030.1670.
Quantum coherence in atoms has led to a variety of novel phenomena. In
electromagnetically induced transparency (EIT), destructive quantum
interference introduced by a strong coupling laser cancels the absorption from
the ground state to coherent superposing upper states one . The observation of
nonabsorbing resonance through atomic coherence has led to novel concepts,
such as lasing without inversion two . Since the EIT line-width can be made
extremely narrow, the resulting steep linear dispersion has been used to
reduce the velocity of light. This technique has been developed for coherent
optical information storage and to freeze light three ; three1 ; three2 ; four
.
Since EIT results from destructive quantum interference, it has been
recognized that similar coherence and interference effects can also occur in
classical systems, such as plasma five ; five1 , coupled optical resonators
six ; six2 ; six3 ; six4 ; six1 , mechanical or electric oscillators seven .
Particularly, the phenomenology of the EIT and the dynamic Stark effect are
studied theoretically in a dissipative system composed by two coupled
oscillators using quantum optics model in Ref. eight . The optical parametric
amplifier (OPA) is one of the most important nonlinear and quantum optical
devices. Recently, we studied the coherence phenomena in the phase-sensitive
degenerate OPA inside a cavity nine ; ten . This phenomenon results from the
interference between the harmonic pump field and the subharmonic seed field in
OPA. The destructive and constructive interferences correspond to optical
parametric deamplification and amplification respectively. The absorptive
response of an optical cavity for the probe field is changed by optical
parametric interaction in the cavity. Agarwal eleven generalized our
observation of interferences in the classical domain nine to in the quantum
domain and studied the interferences in the quantum fluctuations of the output
fields from a parametric amplifier when the cavity is driven by a quantized
field at the signal frequency.
In the previous works nine ; ten , we studied the absorptive and dispersive
properties of degenerate double-resonant OPA where only the degenerate
subharmonic field resonated inside the cavity but the pump field did not.
Although in the experiment with double-resonant OPA the mode splitting in the
transmission spectra nine and the M shape in the reflection spectra were
observed experimentally ten , the shape of the phase shift of the reflected
field was unchanged ten . Further, we investigated the triple-resonant OPA
theoretically, in which both the subharmonic and harmonic wave ares
simultaneously resonate inside the cavity ten . We found that the EIT-like
effect can be simulated with a triple-resonant OPA. In this Letter, we present
the results of the experimental investigation on the absorptive and dispersive
properties of a triple-resonant OPA for the subharmonic field. The frequency
and polarization of signal and idler fields in the OPA are totally degenerate.
We demonstrate that EIT-like effect can be simulated in OPA when the cavity
line-width for the harmonic wave is narrower than that for the subharmonic
field. A narrow transparency window appears in the reflected field.
Especially, we measured the dispersive spectra of triple-resonant OPA and a
very steep variation in the dispersive profile was observed.
Figure 1: Schematic of the experimental setup. DC: dichroic mirror;
$\lambda/2$: half-wave plate; $D_{1},D_{2},D_{3},D_{4}$: detectors; T-C:
temperature controller, HV-AMP: high voltage amplifier; PZT: piezoelectric
transducer; EOM: electric-optic modulator.
The experimental setup is shown schematically in Fig. 1, which is similar to
that used in our previous work ten . A diode-pumped intra-cavity frequency-
doubled continuous-wave ring Nd:YVO${}_{\text{4}}$/KTP single-frequency laser
severs as the light sources of the pump wave (the second-harmonic wave at 532
nm) and the probe wave (the subharmonic wave at 1064 nm) for OPA. We actively
control the relative phase between the subharmonic and the pump field by
adjusting the phase of the subharmonic beam with a mirror mounted upon a
piezoelectric transducer (PZT). Both beams are combined together by a dichroic
mirror and injected into the OPA cavity. OPA consists of type I periodically
poled KTiOPO4 (PPKTP) crystal (12 mm long) and two external mirrors separated
by 61 mm. Two end faces of crystal are polished and coated with an
antireflection for both wavelengths of 532 nm and 1064 nm. The crystal is
mounted in a copper block, whose temperature is actively controlled at
millidegrees kelvin level around the phase-match temperature for optical
parametric process (31.3∘C). The input coupler $M1$ is a 30 mm radius-of-
curvature mirror with a power reflectivity $99\%$ for 1064 nm in the concave
and $97\%$ for 532nm, which is mounted upon a PZT for adjusting the length os
the optical cavity. The back mirror $M2$ is a 30-mm radius-of-curvature mirror
with a reflectivity $93\%$ for 1064 nm and a high reflectivity coefficient for
532 nm in the concave. This structure forms the triple-resonant OPA, in which
the pump wave and the frequency-degenerated idler and signal waves are all
resonating inside the same cavity. This is distinctly different with our
previous setup. The cavity is a under-coupled resonator for the subharmonic
field due to $\gamma_{in}<\gamma_{c}+\gamma_{l}$, where
$\gamma_{l}$,$\gamma_{c}$ and $\gamma_{in}$ are the decay rates of subharmonic
field resulting from internal losses, input mirror and back mirror
respectively. The cavity line-width for the harmonic wave is narrower than for
the subharmonic field because the total damping of the pump field is much less
than that of the subharmonic field ($\gamma_{b}\ll\gamma)$. Due to the high
nonlinear coefficient of PPKTP, the measured threshold power is $P_{th}=90$
$mW$. In order to measure the phase of the reflected beam, we employ the
Pound-Drever-Hall method (or Frequency-modulation spectrum) thirty ; thirty-
one ; thirty-two ; thirty-three ; thirty-four ; thirty-five , which provides a
way to indirectly measure the phase. The frequency of the injected subharmonic
field is modulated with an electric-optic modulator (EOM), driven by a local
oscillator with a radio frequency of $100$ $MHz$. Here the depth of frequency
modulation is very small, thus we only can observe two sidebands. The
reflected beam is picked off with an optical isolator and send into a
photodetector $D_{3}$ , whose output is compared with the local oscillator’s
signal via mixer. A low-pass filter on the output of the mixer extracts the
low frequency signal, which is called the error signal. When the frequency of
the injected subharmonic field is near resonance and the modulation frequency
is high enough, the sidebands are out of the resonance and the error signal is
the imaginary part of the reflected field. In our scheme, the cavity is an
under-coupled resonator, whose dispersive response results in fast light six1
. When an under-coupled resonator is far from critically-coupled, the
imaginary part of the reflected field approximates to the phase shift.
Figure 2: The spectra of the subharmonic reflection, transmission and phase
shift of the reflected field as a function of the cavity detuning when the
pump field is blocked.
First, we measured the spectra of the subharmonic reflection, transmission and
phase shift of the reflection as the functions of the cavity detuning at the
case without the existence of the pump field, which are shown in Fig.2. Two
small sideband modes generated by frequency modulation can be seen in the
transmission spectrum (See arrows in Fig.2 b). Then we hold the frequencies of
the injected subharmonic and the pump field at $\omega_{p}=2\omega$, and the
phase difference $\varphi=\pi/2$ (i.e. OPA at the state of deamplification),
and mode two optical fields resonate simultaneously in the OPA by adjusting
the cavity length and the temperature of PPKTP. The cavity length is scanned
around the resonant point. In Fig.3, the curves e, f and g show the spectra of
the subharmonic reflection, transmission and phase shift of the reflection
respectively with the pump power below the threshold $P=0.15P_{th}$, and the
curves h, j and k with $P=0.3P_{th}$. We can see that the narrow transparency
window appears in the reflection spectrum and is accompanied by a very steep
variation of the dispersive profile whose dispersive response can result in
slow light as that in EIT medium. This experimental result is in good
agreement with the theoretical predictions in Ref.ten .
Figure 3: The spectra of the subharmonic reflection, transmission and phase
shift of the reflected field as a function of the cavity detuning for
different pump powers. a,b and c correspond to Fig.2 with narrow frequency
detuning. e, f, and g at the pump power $0.15P_{th}$. h, j and k at
$0.3P_{th}$.
The observed phenomenon is easily understood: since the narrow absorption and
dispersion of the pump field are introduced into the subharmonic field via the
parametric interaction in triple-resonant OPA, the reflected field of the
subharmonic field presents the EIT-like effect ten . For the case of EIT in
atomic systems, the analogous condition of $\gamma_{b}\ll\gamma$ is obtained
when a pair of lower energy states have long lifetime. Note that we did not
observe the phenomena under higher pump power as that predicted theoretically
in ten . There are two main reasons. One is that the OPA will produce a large
quantum fluctuation on the subharmonic field under higher pump power. The
other is that the pump field had the strong thermal effect at the high pump
power, which would influence the constructed triple-resonance. When scanning
the cavity length, the thermo-induced effect for the pump field was observed
as shown in Fig.4, from which we can see, it seems like the Kerr nonlinear
effect. The transmission profile of the pump field appears large asymmetry at
higher pump power when cavity length is scanned from shorter to longer and
when the opposite. Thus it is difficult to measure the unusual line shape in
the reflection spectrum and the steep variation of the dispersive profile for
triple-resonant OPA in the case of the strong parametric coupling between two
fields.
Figure 4: The spectra of transmission of the pump field as a function of the
cavity detuning for different pump powers: a at $0.2P_{th}$; b at $0.5P_{th}$;
c at $0.9P_{th}$. The dot line represents the voltage on PZT, which denotes
the cavity length scanned from shorter to longer, then from longer to shorter.
In conclusion, we have investigated experimentally the absorptive and
dispersive response of the reflected subharmonic field from a triple-resonant
phase-sensitive OPA. It is demonstrated that an EIT-like effect is simulated
when the cavity line-width for the harmonic wave is narrower than for the
subharmonic field. The narrow transparency window appears in the reflected
field and is accompanied by a very steep variation of the dispersive profile.
This novel system accompanying parametric gain for the probe signal will be
important for practical optical and photonic applications such as in optical
filters, delay lines, and closely relate to the coherent phenomenon of EIT as
that predicted for quantum systems.
†Corresponding author’s email address: jzhang74@yahoo.com, jzhang74@sxu.edu.cn
###### Acknowledgements.
J. Zhang thanks K. Peng and C. Xie for the helpful discussions. This research
was supported in part by NSFC for Distinguished Young Scholars (Grant No.
10725416), National Basic Research Program of China (Grant No. 2006CB921101),
National Natural Science Foundation of China (Grant No. 60678029). REFERENCES
## References
* (1) S. E. Harris, Phys. Today 50, 37 (1997); J. P. Marangos, J. Mod. Opt. 45, 471 (1998).
* (2) A. S. Zibrov, M. D. Lukin, D. E. Nikonov, L. Hollberg, M. O. Scully, V. L. Velichansky, and H. G. Robinson, Phys. Rev. Lett. 75, 1499 (1995).
* (3) L. V. Hau, S.E. Harris, Z. Dutton and C.H. Behroozi, Nature (London) 397, 594 (1999).
* (4) M. M. Kash, V. A. Sautenkov, A. S. Zibrov, L. Hollberg, G. R. Welch, M. D. Lukin, Y. Rostovtsev, E. S. Fry, and M. O. Scully, Phys. Rev. Lett. 82, 5229 (1999).
* (5) D. Budker, D. F. Kimball, S. M. Rochester, and V. V. Yashchuk, Phys. Rev. Lett. 83, 1767 (1999).
* (6) C. Liu, Z. Dutton, C. Behroozi and L. V. Hau, Nature (London) 409, 490 (2001); D. F. Phillips, A. Fleischhauer, A. Mair, and R. L. Walsworth, Phys. Rev. Lett. 86, 783 (2001).
* (7) S.E. Harris, Phys. Rev. Lett. 77, 5357 (1996).
* (8) A.G. Litvak and M.D. Tokman, Phys. Rev. Lett. 88, 095003 (2002); G. Shvets and J.S Wurtele, Phys. Rev. Lett. 89, 115003 (2002).
* (9) T. Opatrny, and D. G. Welsch, Phys. Rev. A 64, 023805 (2001).
* (10) D. D. Smith, H. Chang, K. A. Fuller, A. T. Rosenberger, and R. W. Boyd, Phys. Rev. A 69, 063804 (2004); L. Maleki, A. B. Matsko, A. A. Savchenkov, and V. S. Ilchenko, Opt. Lett. 29, 626 (2004).
* (11) M. F. Yanik, W. Suh, Z. Wang, and S. Fan, Phys. Rev. Lett. 93, 233903 (2004).
* (12) A. Naweed, G. Farca, S. I. Shopova, and A. T. Rosenberger, Phys. Rev. A 71, 043804 (2005); Q. Xu, S. Sandhu, M. L. Povinelli, J. Shakya, S. Fan, and M. Lipson, Phys. Rev.Lett. 96, 123901 (2006);
* (13) D.D. Smith, H. Chang, J. Mod. Opt. 51, 2503 (2004);
* (14) P.R. Hemmer and M.G. Prentiss, J. Opt. Soc. Am. B 5, 1613 (1988); C. L. Garrido Alzar, M. A. G. Martinez, P. Nussenzveig, Am. J. Phys. 70, 37 (2002).
* (15) M. A. de Ponte, C. J. Villas-Boas, R. M. Serra, and M. H. Y. Moussa, e-print quant-ph/0411087.
* (16) H. Ma, C. Ye, D. Wei, and J. Zhang, Phys. Rev. Lett. 95, 233601 (2005).
* (17) C. Ye and J. Zhang, Phys. Rev. A 73, 023818(2006).
* (18) G. S. Agarwal, Phys. Rev. Lett. 97, 023601 (2006).
* (19) R. W. P. Drever, J. L. Hall, F. V. Kowalski, J. Hough, G. M. Ford, A. J. Munley, and H. Ward, Appl. Phys. B 31, 97 (1983).
* (20) G. C. Bjorklund, Opt. Lett. 5, 15 (1980).
* (21) G. C. Bjorklund, M. D. Levenson, W. Lenth, C. Ortiz, Appl. Phys. B 32, 145 (1983).
* (22) A. Schenzle, R. G. DeVoe, and R. G. Brewer, Phys. Rev. A 25, 2606 (1982).
* (23) J. Zhang, D. Wei, C. Xie and K. Peng, Opt. Express 11, 1338 (2003).
* (24) E. D. Black, Am. J. Phys. 69, 79 (2001).
| arxiv-papers | 2008-06-26T01:11:02 | 2024-09-04T02:48:56.401794 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Chenguang Ye, Jing Zhang",
"submitter": "Jing Zhang",
"url": "https://arxiv.org/abs/0806.4218"
} |
0806.4247 | # convex functions on Grassmannian manifolds and Lawson-Osserman problem
Y. L. Xin and Ling Yang Institute of Mathematics, Fudan University, Shanghai
200433, China and Key Laboratory of Mathematics for Nonlinear Sciences (Fudan
University), Ministry of Education ylxin@fudan.edu.cn
###### Abstract.
We derive estimates of the Hessian of two smooth functions defined on
Grassmannian manifold. Based on it, we can derive curvature estimates for
minimal submanifolds in Euclidean space via Gauss map as [24]. In this way,
the result for Bernstein type theorem done by Jost and the first author could
be improved.
###### 1991 Mathematics Subject Classification:
49Q05, 53A07, 53A10.
The research was partially supported by NSFC and SFECC
## 1\. Introduction
The celebrated theorem of Bernstein [2] says that the only entire minimal
graphs in Euclidean 3-space are planes. Its higher dimensional generalization
was finally proved by J. Simons [19], which says that an entire minimal graph
has to be planar for dimension$\leq 7$, while Bombieri- De Giorgi-Giusti [3]
shortly afterwards produced a counterexample to such an assertion in dimension
8 and higher.
Schoen-Simon-Yau [18] gave us a direct proof for Bernstein type theorems for
$n\leq 5$ dimensional minimal graphs with the aid of curvature estimates for
stable minimal hypersurfaces.
There is a weak version of Bernstein type theorem in arbitrary dimension. It
was J. Moser [15] who proved that the entire solution $f$ to the minimal
surface equation is affine linear, provided $|\nabla f|$ is uniformly bounded.
Afterward Ecker-Huisken [9] obtained curvature estimates by a geometric
approach, as a corollary Moser’s result had been improved for the controlled
growth of $|\nabla f|$.
Moser’s theorem had been generalized to certain higher codimensional cases by
Chern-Osserman [6] for dimension $2$ and Babosa, Fischer-Colbrie for dimension
$3$ [1], [10]. But the counterexample constructed by Lawson-Osserman [13]
prevents us going further. They also raised in the same paper a question for
finding the ”best” constant possible in the theorem.
In contrast, the first author with J. Jost [12] proved the following Bernstein
type theorem without the restriction of dimension and codimension, which is an
improvement of the work done by Hildebrandt-Jost-Widman [11].
###### Theorem 1.1.
Let $z^{\alpha}=f^{\alpha}(x^{1},\cdots,x^{n}),\ \alpha=1,\cdots,m$, be smooth
functions defined everywhere in ${\tenmsb R}^{n}$. Suppose their graph
$M=(x,f(x))$ is a submanifold with parallel mean curvature in ${\tenmsb
R}^{n+m}$. Suppose that there exists a number $\beta_{0}$ with
(1.1) $\beta_{0}<\left\\{\begin{array}[]{cl}2&\mbox{when }m\geq 2,\\\
\infty&\mbox{when }m=1;\end{array}\right.$
such that
(1.2)
$\Delta_{f}=\left[\text{det}\left(\delta_{ij}+\sum_{\alpha}\frac{\partial
f^{\alpha}}{\partial x^{i}}\frac{\partial f^{\alpha}}{\partial
x^{j}}\right)\right]^{\frac{1}{2}}\leq\beta_{0}.$
Then $f^{1},\cdots,f^{m}$ has to be affine linear representing an affine
$n$-plane.
The key point of the proof is to find a geodesic convex set
$B_{JX}(P_{0})=\big{\\{}P\in\mathbb{G}_{n,m}:\mbox{ sum of any two Jordan
angles between }P\mbox{ and }P_{0}<\frac{\pi}{2}\big{\\}}$
in a geodesic polar coordinate of the Grassmannian manifold, where $P_{0}$
denotes a fixed $n-$plane. It is larger than the largest geodesic convex ball
of radius $\frac{\sqrt{2}}{4}\pi$ in $\mathbb{G}_{n,m}$. The geometric meaning
of the condition of the above result is that the image under the Gauss map of
$M$ lies in a closed subset $S\subset B_{JX}(P_{0})$.
Recently, the authors [24] studied complete minimal submanifolds whose Gauss
image lies in an open geodesic ball of radius $\frac{\sqrt{2}}{4}\pi$. They
carried out the Schoen-Simon-Yau type curvature estimates and the Ecker-
Huisken type curvature estimates, and on the basis, the corresponding
Bernstein type theorems with dimension limitation or growth assumption could
be derived.
It is natural to study the situation when $\beta_{0}$ in the condition (1.1)
and (1.2) of the Theorem 1.1 approach to $2$. The present paper will devote to
this problem. We shall follow the main idea of our previous paper [24]. But,
we view now the Grassmannian manifolds as submanifolds in Euclidean space via
Plücker imbedding. The auxiliary functions are constructed from this
viewpoint. As shown before, $B_{JX}(P_{0})$ is defined in a coordinate
neighborhood ${\tenmsb U}$ of the Grassmannian $\mathbb{G}_{n,m}$. We
introduce two functions $v$ and $u$ in ${\tenmsb U}$. Via the Gauss map we can
obtain useful functions on our minimal $n-$submanifold $M$ in ${\tenmsb
R}^{m+n}$ with $m\geq 2$. Then, we can carry out the Schoen-Simon-Yau type
curvature estimates and the Ecker-Huisken type curvature estimates, which
enable us to get the corresponding Bernstein type theorems and other
geometrical conclusions.
In Section 2, we give some facts of a Grassmannian manifold
$\mathbb{G}_{n,m}$, which can be isometric imbedding into a Euclidean space.
There is the height function for a submanifold in Euclidean space. Such a
height function is called $w-$function on $\mathbb{G}_{n,m}$. Then we have an
open domain of ${\tenmsb U}\subset\mathbb{G}_{n,m}$, where the $w-$function is
positive. Every point in ${\tenmsb U}$ has a one-to-one correspondence to an
$n\times m$ matrix. We describe canonical metric and the corresponding
connection on ${\tenmsb U}$ with respect to the coordinate. On the basis, the
Hessian of an arbitrary smooth function could be calculated.
In Section 3 we define $v=\frac{1}{w}$ on ${\tenmsb U}$. We also define
another function $u$ on ${\tenmsb U}$. In the section, we shall show $v$ and
$u$ are convex on $B_{JX}(P_{0})$ and give estimates of the Hessian of them.
The estimates are quite delicate. We use the radial compensation technique to
accurate the estimates.
In Section 4, we define four auxiliary functions, $\tilde{h}_{1}$,
$\tilde{h}_{2}$, $\tilde{h}_{3}$ and $\tilde{h}_{4}$. They are defined on the
minimal submanifolds of ${\tenmsb R}^{n+m}$ whose Gauss image is confined, and
they are expressed in term of $v$ and $u$. We also estimate the Laplacian of
them, which is useful for the next sections.
Later in Section 5, not only we give the Schoen-Simon-Yau type curvature
estimates with the aid of $\tilde{h}_{1}$ and $\tilde{h}_{3}$, but also we
obtain the Ecker-Huisken type curvature estimates with the aid of
$\tilde{h}_{2}$ and $\tilde{h}_{4}$. Our method is completely similar to the
previous paper [24], so we only describe the outline of process. From the
estimates several geometrical conclusions follow, including the following
Bernstein type theorems.
###### Theorem 1.2.
Let $M=(x,f(x))$ be an $n$-dimensional entire minimal graph given by $m$
functions $f^{\alpha}(x^{1},\cdots,x^{n})$ with $m\geq 2,n\leq 4$. If
$\Delta_{f}=\left[\text{det}\left(\delta_{ij}+\sum_{\alpha}\frac{\partial
f^{\alpha}}{\partial x^{i}}\frac{\partial f^{\alpha}}{\partial
x^{j}}\right)\right]^{\frac{1}{2}}<2,$
then $f^{\alpha}$ has to be affine linear functions representing an affine
$n$-plane.
###### Theorem 1.3.
Let $M=(x,f(x))$ be an $n$-dimensional entire minimal graph given by $m$
functions $f^{\alpha}(x^{1},\cdots,x^{n})$ with $m\geq 2$. If
$\Delta_{f}=\left[\text{det}\left(\delta_{ij}+\sum_{\alpha}\frac{\partial
f^{\alpha}}{\partial x^{i}}\frac{\partial f^{\alpha}}{\partial
x^{j}}\right)\right]^{\frac{1}{2}}<2,$
and
(1.3) $\left(2-\Delta_{f}\right)^{-1}=o(R^{\frac{4}{3}}),$
where $R^{2}=|x|^{2}+|f|^{2}$. Then $f^{\alpha}$ has to be affine linear
functions and hence $M$ has to be an affine linear subspace.
Those are what shall be done in Section 6. It is worthy to note that Theorems
1.2-1.3 still hold true when $M$ is a submanifold with parallel mean
curvature. Dong generalized Chern’s result [5] [4] to higher codimension,
which states that a graphic submanifold $M=(x,f(x))$ with parallel mean
curvature has to be minimal if the slope of $f$ is uniformly bounded. Hence
our results improve Theorem 1.1.
It is natural to ask what is the relations between the results here and that
of the previous paper [24]. Since the $v-$function varies in
$\left(\sec^{p}\left(\frac{\pi}{2\sqrt{2p}}\right),\sec\left(\frac{\sqrt{2}}{4}\pi\right)\right)$
on the open geodesic ball of radius $\frac{\sqrt{2}}{4}$ in
$\mathbb{G}_{n,m}$, where $p=\min(n,m)$, the results of the present article do
not generalize those in the previous one. Both results are complementary.
## 2\. Preliminaries on the Grassmannian manifold $\mathbb{G}_{n,m}$
Let $\mathbb{R}^{n+m}$ be an $n+m$-dimensional Euclidean space. All oriented
$n$-subspaces constitute the Grassmannian manifolds $\mathbb{G}_{n,m}$, which
is an irreducible symmetric space of compact type.
Fix $P_{0}\in\mathbb{G}_{n,m}$ in the sequel, which is spanned by a unit
$n-$vector $\varepsilon_{1}\wedge\cdots\wedge\varepsilon_{n}$. For any
$P\in\mathbb{G}_{n,m}$, spanned by a $n-$vector $e_{1}\wedge\cdots\wedge
e_{n}$, we define an important function on $\mathbb{G}_{n,m}$
$w\mathop{=}\limits^{def.}\left<P,P_{0}\right>=\left<e_{1}\wedge\cdots\wedge
e_{n},\varepsilon_{1}\wedge\cdots\wedge\varepsilon_{n}\right>=\det W,$
where $W=(\left<e_{i},\varepsilon_{j}\right>).$ It is well known that
$W^{T}W=O^{T}\Lambda O,$
where $O$ is an orthogonal matrix and
$\Lambda=\begin{pmatrix}\mu_{1}^{2}&&0\cr&\ddots&\cr
0&&\mu_{p}^{2}\end{pmatrix},\qquad p=\min(m,n),$
where each $0\leq\mu_{i}^{2}\leq 1.$ The Jordan angles between $P$ and $P_{0}$
are defined by
$\theta_{i}=\arccos(\mu_{i}).$
Denote
${\tenmsb U}=\\{P\in\mathbb{G}_{n,m}:w(P)>0\\},$
let $\\{\varepsilon_{n+\alpha}\\}$ be $m$-vectors such that
$\\{\varepsilon_{i},\varepsilon_{n+\alpha}\\}$ form an orthornormal basis of
$\mathbb{R}^{m+n}$. Then we can span arbitrary $P\in{\tenmsb U}$ by $n$
vectors $f_{i}$:
$f_{i}=\varepsilon_{i}+z_{i\alpha}\varepsilon_{n+\alpha},$
where $Z=(z_{i\alpha})$ are the local coordinate of $P$ in ${\tenmsb U}$. Here
and in the sequel we use the summation convention and agree the range of
indices:
$1\leq i,j,k,l\leq n;\qquad 1\leq\alpha,\beta,\gamma,\delta\leq m.$
The canonical metric on $\mathbb{G}_{n,m}$ in the local coordinate can be
described as (see [22] Ch. VII)
(2.1)
$g=\mbox{tr}\big{(}(I_{n}+ZZ^{T})^{-1}dZ(I_{m}+Z^{T}Z)^{-1}dZ^{T}\big{)}.$
Let $P\in{\tenmsb U}$ determined by an $n\times m$ matrix
$Z_{0}=\big{(}\lambda_{\alpha}\delta_{i\alpha}\big{)}$, where
$\lambda_{\alpha}=\tan\theta_{\alpha}$ and $\theta_{1},\cdots,\theta_{m}$ be
the Jordan angles between $P$ and $P_{0}$. (Here and in the sequel we assume
$n\geq m$ without loss of generality; for it is similar for $n<m$.) Let
$X,Y,W$ denote arbitrary $n\times m$ matrices. Then (2.1) tells us
(2.2) $\displaystyle\begin{aligned} \langle
X,Y\rangle_{P}&=\mbox{tr}\big{(}(I_{n}+Z_{0}Z_{0}^{T})^{-1}X(I_{m}+Z_{0}^{T}Z_{0})^{-1}Y^{T}\big{)}\\\
&=\sum_{i,\alpha}(1+\lambda_{i}^{2})^{-1}(1+\lambda_{\alpha}^{2})^{-1}X_{i\alpha}Y_{i\alpha}.\end{aligned}$
(Note that if $m+1\leq i\leq n$, $\lambda_{i}=0$.) Furthermore, from
$\displaystyle\begin{aligned}
&\big{(}I_{n}+(Z_{0}+tW)(Z_{0}+tW)^{T}\big{)}^{-1}X\big{(}I_{m}+(Z_{0}+tW)^{T}(Z_{0}+tW)\big{)}^{-1}Y^{T}\\\
&=\big{(}I_{n}+Z_{0}Z_{0}^{T}+t(WZ_{0}^{T}+Z_{0}W^{T})+O(t^{2})\big{)}^{-1}X\\\
&\qquad\big{(}I_{m}+Z_{0}^{T}Z_{0}+t(W^{T}Z_{0}+Z_{0}^{T}W)+O(t^{2})\big{)}^{-1}Y^{T}\\\
&=\big{(}I_{n}+t(I_{n}+Z_{0}Z_{0}^{T})^{-1}(WZ_{0}^{T}+Z_{0}W^{T})+O(t^{2})\big{)}^{-1}(I_{n}+Z_{0}Z_{0}^{T})^{-1}X\\\
&\qquad\big{(}I_{m}+t(I_{m}+Z_{0}^{T}Z_{0})^{-1}(W^{T}Z_{0}+Z_{0}^{T}W)+O(t^{2})\big{)}^{-1}(I_{m}+Z_{0}^{T}Z_{0})^{-1}Y^{T}\\\
&=\big{(}I_{n}-t(I_{n}+Z_{0}Z_{0}^{T})^{-1}(WZ_{0}^{T}+Z_{0}W^{T})+O(t^{2})\big{)}(I_{n}+Z_{0}Z_{0}^{T})^{-1}X\\\
&\qquad\big{(}I_{m}-t(I_{m}+Z_{0}^{T}Z_{0})^{-1}(W^{T}Z_{0}+Z_{0}^{T}W)+O(t^{2})\big{)}(I_{m}+Z_{0}^{T}Z_{0})^{-1}Y^{T}\\\
&=(I_{n}+Z_{0}Z_{0}^{T})^{-1}X(I_{m}+Z_{0}^{T}Z_{0})^{-1}Y^{T}\\\
&\quad-t\big{[}(I_{n}+Z_{0}Z_{0}^{T})^{-1}(WZ_{0}^{T}+Z_{0}W^{T})(I_{n}+Z_{0}Z_{0}^{T})^{-1}X(I_{m}+Z_{0}^{T}Z_{0})^{-1}Y^{T}\\\
&\qquad+(I_{n}+Z_{0}Z_{0}^{T})^{-1}X(I_{m}+Z_{0}^{T}Z_{0})^{-1}(W^{T}Z_{0}+Z_{0}^{T}W)(I_{m}+Z_{0}^{T}Z_{0})^{-1}Y^{T}\big{]}+O(t^{2}),\end{aligned}$
we have
(2.3) $\displaystyle\begin{aligned} W\langle
X,Y\rangle_{P}=&-\mbox{tr}\big{[}(I_{n}+Z_{0}Z_{0}^{T})^{-1}(WZ_{0}^{T}+Z_{0}W^{T})\\\
&\qquad\qquad(I_{n}+Z_{0}Z_{0}^{T})^{-1}X(I_{m}+Z_{0}^{T}Z_{0})^{-1}Y^{T}\\\
&\qquad+(I_{n}+Z_{0}Z_{0}^{T})^{-1}X(I_{m}+Z_{0}^{T}Z_{0})^{-1}\\\
&\qquad\qquad(W^{T}Z_{0}+Z_{0}^{T}W)(I_{m}+Z_{0}^{T}Z_{0})^{-1}Y^{T}\big{]}.\end{aligned}$
We let $E_{i\alpha}$ be the matrix with 1 in the intersection of row $i$ and
column $\alpha$ and 0 otherwise. Denote $g_{i\alpha,j\beta}=\langle
E_{i\alpha},E_{j\beta}\rangle$ and let $\big{(}g^{i\alpha,j\beta}\big{)}$ be
the inverse matrix of $\big{(}g_{i\alpha,j\beta}\big{)}$. Denote by $\nabla$
the Levi-Civita connection with respect to the canonical matric on
$\mathbb{G}_{n,m}$, and by
$\nabla_{E_{i\alpha}}E_{j\beta}=\Gamma_{i\alpha,j\beta}^{k\gamma}E_{k\gamma}.$
Then from (2.2),
(2.4) $\displaystyle
g_{i\alpha,j\beta}(P)=(1+\lambda_{i}^{2})^{-1}(1+\lambda_{\alpha}^{2})^{-1}\delta_{\alpha\beta}\delta_{ij}$
and obviously
(2.5)
$g^{i\alpha,j\beta}(P)=(1+\lambda_{i}^{2})(1+\lambda_{\alpha}^{2})\delta_{\alpha\beta}\delta_{ij}.$
Moreover, a direct calculation from (2.3) and (2.5) shows
(2.6) $\displaystyle\begin{aligned}
\Gamma_{i\alpha,j\beta}^{k\gamma}&=\frac{1}{2}g^{k\gamma,l\delta}\big{(}-E_{l\delta}\langle
E_{i\alpha},E_{j\beta}\rangle+E_{i\alpha}\langle
E_{j\beta},E_{l\delta}\rangle+E_{j\beta}\langle
E_{l\delta},E_{i\alpha}\rangle\big{)}\\\
&=-\lambda_{\alpha}(1+\lambda_{\alpha}^{2})^{-1}\delta_{\alpha
j}\delta_{\beta\gamma}\delta_{ik}-\lambda_{\beta}(1+\lambda_{\beta}^{2})^{-1}\delta_{\beta
i}\delta_{\alpha\gamma}\delta_{jk}.\end{aligned}$
From (2.5), we see that
(2.7)
$(1+\lambda_{i}^{2})^{\frac{1}{2}}(1+\lambda_{\alpha}^{2})^{\frac{1}{2}}E_{i\alpha}\
(1\leq i\leq n,1\leq\alpha\leq m)$
form an orthonormal basis of $T_{P}\mathbb{G}_{n,m}$. Denote its dual basis in
$T_{P}^{*}\mathbb{G}_{n,m}$ by
(2.8) $\omega_{i\alpha}\ (1\leq i\leq n,1\leq\alpha\leq m),$
then
(2.9) $g=\sum_{i,\alpha}\omega_{i\alpha}^{2}$
at $P$.
## 3\. Hessian estimates of two smooths functions on $\mathbb{G}_{n,m}$
On ${\tenmsb U}$, $w>0$, then we can define
(3.1) $v=w^{-1}\qquad\mbox{ on }{\tenmsb U}.$
For arbitrary $Q\in{\tenmsb U}$ determined by an $n\times m$ matrix $Z$, it is
easily seen that
(3.2)
$v(Q)=\big{[}\det(I_{n}+ZZ^{T})\big{]}^{\frac{1}{2}}=\prod_{\alpha=1}^{m}\sec\theta_{\alpha}.$
where $\theta_{1},\cdots,\theta_{m}$ denotes the Jordan angles between $Q$ and
$P_{0}$.
Now we calculate the Hessian of $v$ at $P$ whose corresponding matrix is
$Z_{0}$. At first, by noting that for any $n\times n$ orthogonal matrix $U$
and $m\times m$ orthogonal matrix $V$, $Z\mapsto UZV$ induces an isometry of
${\tenmsb U}$ which keeps $v$ invariant, we can assume
$Z_{0}=(\lambda_{\alpha}\delta_{i\alpha})$ without loss of generality, where
$\lambda_{\alpha}=\tan\theta_{\alpha}$ and $\theta_{1},\cdots,\theta_{m}$
denotes the Jordan angles between $P$ and $P_{0}$. We also need a Lemma as
follows.
###### Lemma 3.1.
Let $M$ be a manifold, $A$ be a smooth nonsingular $n\times n$ matrix-valued
function on $M$, $X,Y$ be local tangent fields, then
(3.3) $\nabla_{X}\log\det A=\text{tr}(\nabla_{X}A\cdot A^{-1})$
and
(3.4) $\nabla_{Y}\nabla_{X}\log\det A=\mbox{tr}(\nabla_{Y}\nabla_{X}A\cdot
A^{-1})-\mbox{tr}(\nabla_{X}A\cdot A^{-1}\cdot\nabla_{Y}A\cdot A^{-1}).$
###### Proof.
Assume that $e_{1},\cdots,e_{n}$ is a standard basis in ${\tenmsb R}^{n}$,
then
$\det A\ e_{1}\wedge\cdots\wedge e_{n}=Ae_{1}\wedge\cdots\wedge Ae_{n}.$
Hence
$\displaystyle\begin{aligned} \nabla_{X}\det A\ e_{1}\wedge\cdots\wedge
e_{n}=&\sum_{i}Ae_{1}\wedge\cdots\wedge Ae_{i-1}\wedge\nabla_{X}Ae_{i}\wedge
Ae_{i+1}\wedge\cdots\wedge Ae_{n}\\\ =&\sum_{i}Ae_{1}\wedge\cdots\wedge
Ae_{i-1}\wedge(\nabla_{X}A\cdot A^{-1})Ae_{i}\wedge Ae_{i+1}\wedge\cdots\wedge
Ae_{n}\\\ =&\mbox{tr}(\nabla_{X}A\cdot A^{-1})Ae_{1}\wedge\cdots\wedge
Ae_{n}\\\ =&\mbox{tr}(\nabla_{X}A\cdot A^{-1})\det A\ e_{1}\wedge\cdots\wedge
e_{n}.\end{aligned}$
Thereby (3.3) immediately follows.
(3.4) follows from (3.3) and
$A\cdot\nabla_{Y}A^{-1}+\nabla_{Y}A\cdot A^{-1}=\nabla_{Y}(AA^{-1})=0.$
∎
Now we let $M={\tenmsb U}$, $A(Z)=I_{n}+ZZ^{T}$, then $\log
v=\frac{1}{2}\log\det A$. A direct calculation shows
$\nabla_{X}A=XZ^{T}+ZX^{T},\qquad\nabla_{Y}\nabla_{X}A=XY^{T}+YX^{T}.$
Hence we compute from Lemma 3.1 that at $P$
$\displaystyle\begin{aligned} \nabla_{X}\log
v&=\frac{1}{2}\mbox{tr}\left((XZ_{0}^{T}+Z_{0}X^{T})(I_{n}+Z_{0}Z_{0}^{T})^{-1}\right)\\\
&=\sum_{\alpha}\lambda_{\alpha}(1+\lambda_{\alpha}^{2})^{-1}X_{\alpha\alpha},\end{aligned}$
$\displaystyle\nabla_{X}\nabla_{Y}\log v$ $\displaystyle=$
$\displaystyle\frac{1}{2}\mbox{tr}\left((XY^{T}+YX^{T})(I_{n}+Z_{0}Z_{0}^{T})^{-1}\right)$
$\displaystyle-\frac{1}{2}\mbox{tr}\left((XZ_{0}^{T}+Z_{0}X^{T})(I_{n}+Z_{0}Z_{0}^{T})^{-1}(YZ_{0}^{T}+Z_{0}Y^{T})(I_{n}+Z_{0}Z_{0}^{T})^{-1}\right)$
$\displaystyle=$
$\displaystyle\sum_{i,\alpha}(1+\lambda_{i}^{2})^{-1}X_{i\alpha}Y_{i\alpha}$
$\displaystyle\qquad-\frac{1}{2}\sum_{i,j}(XZ_{0}^{T}+Z_{0}X^{T})_{ij}(1+\lambda_{j}^{2})^{-1}(YZ_{0}^{T}+Z_{0}Y^{T})_{ji}(1+\lambda_{i}^{2})^{-1}$
$\displaystyle=$ $\displaystyle\sum_{m+1\leq i\leq
n,\alpha}X_{i\alpha}Y_{i\alpha}+\frac{1}{2}\sum_{\alpha,\beta}(1+\lambda_{\alpha}^{2})^{-1}X_{\alpha\beta}Y_{\alpha\beta}+\frac{1}{2}\sum_{\alpha,\beta}(1+\lambda_{\beta}^{2})^{-1}X_{\beta\alpha}Y_{\beta\alpha}$
$\displaystyle\qquad-\frac{1}{2}\sum_{\alpha,\beta}(\lambda_{\beta}X_{\alpha\beta}+\lambda_{\alpha}X_{\beta\alpha})(1+\lambda_{\beta}^{2})^{-1}(\lambda_{\alpha}Y_{\beta\alpha}+\lambda_{\beta}Y_{\alpha\beta})(1+\lambda_{\alpha}^{2})^{-1}$
$\displaystyle\qquad-\sum_{m+1\leq i\leq
n,\alpha}\lambda_{\alpha}^{2}(1+\lambda_{\alpha}^{2})^{-1}X_{i\alpha}Y_{i\alpha}$
$\displaystyle=$ $\displaystyle\sum_{m+1\leq i\leq
n,\alpha}(1+\lambda_{\alpha}^{2})^{-1}X_{i\alpha}Y_{i\alpha}+\sum_{\alpha,\beta}(1+\lambda_{\alpha}^{2})^{-1}(1+\lambda_{\beta}^{2})^{-1}X_{\alpha\beta}Y_{\alpha\beta}$
$\displaystyle\qquad-\sum_{\alpha,\beta}\lambda_{\alpha}\lambda_{\beta}(1+\lambda_{\alpha}^{2})^{-1}(1+\lambda_{\beta}^{2})^{-1}X_{\alpha\beta}Y_{\beta\alpha}.$
Furthermore,
$\displaystyle\begin{aligned} \nabla_{X}v=&v\nabla_{X}\log
v=\left(\sum_{\alpha}\lambda_{\alpha}(1+\lambda_{\alpha}^{2})^{-1}X_{\alpha\alpha}\right)v,\\\
\nabla_{X}\nabla_{Y}v=&v(\nabla_{X}\nabla_{Y}\log v+\nabla_{X}\log
v\cdot\nabla_{Y}\log v)\\\ =&\Big{(}\sum_{m+1\leq i\leq
n,\alpha}(1+\lambda_{\alpha}^{2})^{-1}X_{i\alpha}Y_{i\alpha}+\sum_{\alpha,\beta}(1+\lambda_{\alpha}^{2})^{-1}(1+\lambda_{\beta}^{2})^{-1}X_{\alpha\beta}Y_{\alpha\beta}\\\
&\qquad+\sum_{\alpha,\beta}\lambda_{\alpha}\lambda_{\beta}(1+\lambda_{\alpha}^{2})^{-1}(1+\lambda_{\beta}^{2})^{-1}(X_{\alpha\alpha}Y_{\beta\beta}-X_{\alpha\beta}Y_{\beta\alpha})\Big{)}v\\\
=&\Big{(}\sum_{i,\beta}(1+\lambda_{i}^{2})^{-1}(1+\lambda_{\beta}^{2})^{-1}X_{i\beta}Y_{i\beta}\\\
&\qquad+\sum_{\alpha,\beta}\lambda_{\alpha}\lambda_{\beta}(1+\lambda_{\alpha}^{2})^{-1}(1+\lambda_{\beta}^{2})^{-1}(X_{\alpha\alpha}Y_{\beta\beta}-X_{\alpha\beta}Y_{\beta\alpha})\Big{)}v.\end{aligned}$
In particular,
(3.5)
$\nabla_{E_{i\alpha}}v(P)=\lambda_{\alpha}(1+\lambda_{\alpha}^{2})^{-1}v\delta_{i\alpha}$
and
(3.6)
$\nabla_{E_{i\alpha}}\nabla_{E_{j\beta}}v(P)=\left\\{\begin{array}[]{cl}(1+\lambda_{i}^{2})^{-1}(1+\lambda_{\alpha}^{2})^{-1}v&i=j,\alpha=\beta;\\\
-\lambda_{\alpha}\lambda_{\beta}(1+\lambda_{\alpha}^{2})^{-1}(1+\lambda_{\beta}^{2})^{-1}v&i=\beta,j=\alpha,\alpha\neq\beta;\\\
\lambda_{\alpha}\lambda_{\beta}(1+\lambda_{\alpha}^{2})^{-1}(1+\lambda_{\beta}^{2})^{-1}v&i=\alpha,j=\beta,\alpha\neq\beta;\\\
0&\mbox{otherwise.}\end{array}\right.$
Then, from (2.6), (3.5) and (3.6) we obtain
(3.7) $\displaystyle\begin{aligned}
\mbox{Hess}(v)(E_{i\alpha},E_{j\beta})(P)=&\nabla_{E_{i\alpha}}\nabla_{E_{j\beta}}v-(\nabla_{E_{i\alpha}}E_{j\beta})v\\\
=&\nabla_{E_{i\alpha}}\nabla_{E_{j\beta}}v-\Gamma_{i\alpha,j\beta}^{k\gamma}\nabla_{E_{k\gamma}}v\\\
=&\left\\{\begin{array}[]{cl}(1+\lambda_{i}^{2})^{-1}(1+\lambda_{\alpha}^{2})^{-1}v&i=j,\alpha=\beta,i\neq\alpha;\\\
(1+2\lambda_{\alpha}^{2})(1+\lambda_{\alpha}^{2})^{-2}v&i=j=\alpha=\beta;\\\
\lambda_{\alpha}\lambda_{\beta}(1+\lambda_{\alpha}^{2})^{-1}(1+\lambda_{\beta}^{2})^{-1}v&i=\beta,j=\alpha,\alpha\neq\beta;\\\
\lambda_{\alpha}\lambda_{\beta}(1+\lambda_{\alpha}^{2})^{-1}(1+\lambda_{\beta}^{2})^{-1}v&i=\alpha,j=\beta,\alpha\neq\beta;\\\
0&\mbox{otherwise.}\end{array}\right.\end{aligned}$
In other words
(3.8) $\displaystyle\begin{aligned} \mbox{Hess}(v)_{P}&=\sum_{i\neq\alpha}v\
\omega_{i\alpha}^{2}+\sum_{\alpha}(1+2\lambda_{\alpha}^{2})v\
\omega_{\alpha\alpha}^{2}+\sum_{\alpha\neq\beta}\lambda_{\alpha}\lambda_{\beta}v(\omega_{\alpha\alpha}\otimes\omega_{\beta\beta}+\omega_{\alpha\beta}\otimes\omega_{\beta\alpha})\\\
&=\sum_{m+1\leq i\leq n,\alpha}v\
\omega_{i\alpha}^{2}+\sum_{\alpha}(1+2\lambda_{\alpha}^{2})v\
\omega_{\alpha\alpha}^{2}+\sum_{\alpha\neq\beta}\lambda_{\alpha}\lambda_{\beta}v\
\omega_{\alpha\alpha}\otimes\omega_{\beta\beta}\\\
&\qquad\qquad+\sum_{\alpha<\beta}\Big{[}(1+\lambda_{\alpha}\lambda_{\beta})v\Big{(}\frac{\sqrt{2}}{2}(\omega_{\alpha\beta}+\omega_{\beta\alpha})\Big{)}^{2}\\\
&\hskip
144.54pt+(1-\lambda_{\alpha}\lambda_{\beta})v\Big{(}\frac{\sqrt{2}}{2}(\omega_{\alpha\beta}-\omega_{\beta\alpha})\Big{)}^{2}\Big{]}.\end{aligned}$
(3.8) could be simplified further. Please note (3.5), which also tells us
(3.9) $dv=\sum_{\alpha}\lambda_{\alpha}v\ \omega_{\alpha\alpha};$
then
(3.10) $dv\otimes dv=\sum_{\alpha}\lambda_{\alpha}^{2}v^{2}\
\omega_{\alpha\alpha}^{2}+\sum_{\alpha\neq\beta}\lambda_{\alpha}\lambda_{\beta}v^{2}\
\omega_{\alpha\alpha}\otimes\omega_{\beta\beta}.$
Substituting (3.10) into (3.8) yields
(3.11) $\displaystyle\begin{aligned} \mbox{Hess}(v)_{P}&=\sum_{m+1\leq i\leq
n,\alpha}v\ \omega_{i\alpha}^{2}+\sum_{\alpha}(1+\lambda_{\alpha}^{2})v\
\omega_{\alpha\alpha}^{2}+v^{-1}\ dv\otimes dv\\\
&\qquad+\sum_{\alpha<\beta}\Big{[}(1+\lambda_{\alpha}\lambda_{\beta})v\Big{(}\frac{\sqrt{2}}{2}(\omega_{\alpha\beta}+\omega_{\beta\alpha})\Big{)}^{2}\\\
&\hskip
72.26999pt+(1-\lambda_{\alpha}\lambda_{\beta})v\Big{(}\frac{\sqrt{2}}{2}(\omega_{\alpha\beta}-\omega_{\beta\alpha})\Big{)}^{2}\Big{]}.\end{aligned}$
Note that $\lambda_{\alpha}\geq 0$ and
$1-\lambda_{\alpha}\lambda_{\beta}=1-\tan\theta_{\alpha}\tan\theta_{\beta}=\frac{\cos(\theta_{\alpha}+\theta_{\beta})}{\cos\theta_{\alpha}\cos\theta_{\beta}}$;
which implies that $\mbox{Hess}(v)_{P}$ is positive definite if and only if
$\theta_{\alpha}+\theta_{\beta}<\frac{\pi}{2}$ for arbitrary
$\alpha\neq\beta$, i.e., $P\in B_{JX}(P_{0})$.
By (3.2), $v=\prod_{\alpha}(1+\lambda_{\alpha}^{2})^{\frac{1}{2}}$, then
$\lambda_{\alpha}\lambda_{\beta}\leq\big{[}(1+\lambda_{\alpha}^{2})(1+\lambda_{\beta}^{2})\big{]}^{\frac{1}{2}}-1\leq
v-1,$
the equality holds if and only if $\lambda_{\alpha}=\lambda_{\beta}$ and
$\lambda_{\gamma}=0$ for each $\gamma\neq\alpha,\beta$. Hence, we have
$1-\lambda_{\alpha}\lambda_{\beta}\geq 2-v$. Finally we arrive at an estimate
(3.12) $\mbox{Hess}(v)\geq v(2-v)g+v^{-1}dv\otimes dv.$
Now we introduce another smooth function on ${\tenmsb U}$. For any
$Q\in{\tenmsb U}$,
(3.13) $u\mathop{=}\limits^{def.}\sum_{\alpha}\tan\theta_{\alpha}^{2}.$
where $\theta_{1},\cdots,\theta_{m}$ denotes the Jordan angles between $Q$ and
$P_{0}$. Denote by $Z$ the coordinate of $Q$, then it is easily seen that
(3.14) $u(Q)=\mbox{tr}(ZZ^{T}).$
We can calculate the Hessian of $u$ at $P\in{\tenmsb U}$ whose corresponding
matrix is $Z_{0}$ in the same way. Similar to above, we can assume
$Z_{0}=\big{(}\lambda_{a}\delta_{i\alpha}\big{)}$, where
$\lambda_{\alpha}=\tan\theta_{\alpha}$ and $\theta_{1},\cdots,\theta_{m}$ are
the Jordan angles between $P$ and $P_{0}$.
Obviously
(3.15) $\displaystyle\begin{aligned}
\nabla_{X}u=&\mbox{tr}(XZ^{T})+\mbox{tr}(ZX^{T}),\\\
\nabla_{X}\nabla_{Y}u=&\mbox{tr}(XY^{T})+\mbox{tr}(YX^{T}).\end{aligned}$
Then, at $P$
(3.16) $\displaystyle\begin{aligned}
\mbox{Hess}(u)(E_{i\alpha},E_{j\beta})=&\nabla_{E_{i\alpha}}\nabla_{E_{j\beta}}u-(\nabla_{E_{i\alpha}}E_{j\beta})u\\\
=&\nabla_{E_{i\alpha}}\nabla_{E_{j\beta}}u-\Gamma_{i\alpha,j\beta}^{k\gamma}\nabla_{E_{k\gamma}}u\\\
=&2\delta_{ij}\delta_{\alpha\beta}+\big{(}\lambda_{\alpha}(1+\lambda_{\alpha}^{2})^{-1}\delta_{\alpha
j}\delta_{\beta\gamma}\delta_{ik}+\lambda_{\beta}(1+\lambda_{\beta}^{2})^{-1}\delta_{\beta
i}\delta_{\alpha\gamma}\delta_{jk}\big{)}\\\ &\cdot
2\lambda_{\gamma}\delta_{k\gamma}\\\
=&2\delta_{ij}\delta_{\alpha\beta}+2\lambda_{\alpha}\lambda_{\beta}\big{[}(1+\lambda_{\alpha}^{2})^{-1}+(1+\lambda_{\beta}^{2})^{-1}\big{]}\delta_{\alpha
j}\delta_{\beta i}\\\
=&\left\\{\begin{array}[]{cl}2&i=j,\alpha=\beta,i\neq\alpha;\\\
2+4\lambda_{\alpha}^{2}(1+\lambda_{\alpha}^{2})^{-1}&i=j=\alpha=\beta;\\\
2\lambda_{\alpha}\lambda_{\beta}\big{[}(1+\lambda_{\alpha}^{2})^{-1}+(1+\lambda_{\beta}^{2})^{-1}\big{]}&i=\beta,j=\alpha,\alpha\neq\beta.\end{array}\right.\end{aligned}$
In other words
(3.17) $\displaystyle\begin{aligned}
\mbox{Hess}(u)_{P}=&\sum_{i\neq\alpha}2(1+\lambda_{i}^{2})(1+\lambda_{\alpha}^{2})\omega_{i\alpha}^{2}+\sum_{\alpha}(2+6\lambda_{\alpha}^{2})(1+\lambda_{\alpha}^{2})\omega_{\alpha\alpha}^{2}\\\
&\qquad+\sum_{\alpha\neq\beta}2\lambda_{\alpha}\lambda_{\beta}(2+\lambda_{\alpha}^{2}+\lambda_{\beta}^{2})\omega_{\alpha\beta}\otimes\omega_{\beta\alpha}\\\
=&\sum_{m+1\leq i\leq
n,\alpha}2(1+\lambda_{i}^{2})(1+\lambda_{\alpha}^{2})\omega_{i\alpha}^{2}+\sum_{\alpha}(2+6\lambda_{\alpha}^{2})(1+\lambda_{\alpha}^{2})\omega_{\alpha\alpha}^{2}\\\
&+2\big{[}(1+\lambda_{\alpha}^{2})(1+\lambda_{\beta}^{2})+\lambda_{\alpha}\lambda_{\beta}(2+\lambda_{\alpha}^{2}+\lambda_{\beta}^{2})\big{]}\left[\frac{\sqrt{2}}{2}(\omega_{\alpha\beta}+\omega_{\beta\alpha})\right]^{2}\\\
&+2\big{[}(1+\lambda_{\alpha}^{2})(1+\lambda_{\beta}^{2})-\lambda_{\alpha}\lambda_{\beta}(2+\lambda_{\alpha}^{2}+\lambda_{\beta}^{2})\big{]}\left[\frac{\sqrt{2}}{2}(\omega_{\alpha\beta}-\omega_{\beta\alpha})\right]^{2}\end{aligned}$
By computing,
(3.18)
$2\big{[}(1+\lambda_{\alpha}^{2})(1+\lambda_{\beta}^{2})-\lambda_{\alpha}\lambda_{\beta}(2+\lambda_{\alpha}^{2}+\lambda_{\beta}^{2})\big{]}=2(1-\lambda_{\alpha}\lambda_{\beta})(\lambda_{\alpha}^{2}+\lambda_{\beta}^{2}-\lambda_{\alpha}\lambda_{\beta}+1).$
It is positive if and only if
$1-\lambda_{\alpha}\lambda_{\beta}=1-\tan\theta_{\alpha}\tan\theta_{\beta}=\frac{\cos(\theta_{\alpha}+\theta_{\beta})}{\cos\theta_{\alpha}\cos\theta_{\beta}}\geq
0$, i.e., $\theta_{\alpha}+\theta_{\beta}<\frac{\pi}{2}$. Hence
$\mbox{Hess}(u)_{P}$ is positive definite if and only $P\in B_{JX}(P_{0})$.
Moreover, the right side of (3.18) can be estimated by
$\displaystyle\begin{aligned}
2(1-\lambda_{\alpha}\lambda_{\beta})(\lambda_{\alpha}^{2}+\lambda_{\beta}^{2}-\lambda_{\alpha}\lambda_{\beta}+1)&\geq
2\big{(}1-\frac{\lambda_{\alpha}^{2}+\lambda_{\beta}^{2}}{2}\big{)}\big{(}\frac{\lambda_{\alpha}^{2}+\lambda_{\beta}^{2}}{2}+1\big{)}\\\
&=2\big{(}1-\frac{(\lambda_{\alpha}^{2}+\lambda_{\beta}^{2})^{2}}{4}\big{)}\\\
&\geq 2(1-\frac{1}{4}u^{2})=2-\frac{1}{2}u^{2}.\end{aligned}$
(Here we used the fact
$u=\sum_{\alpha}\tan^{2}\theta_{\alpha}=\sum_{\alpha}\lambda_{\alpha}^{2}$.)
By combining it with (3.17) and (2.9), we arrive that
(3.19) $\mbox{Hess}(u)\geq\left(2-\frac{1}{2}u^{2}\right)g.$
For later applications the estimates (3.12) and (3.19) are not accurate
enough. Using the radial compensation technique we could refine those
estimates which are based on the following lemmas.
###### Lemma 3.2.
Let $V$ be a real linear space, $h$ be a nonnegative definite quadratic form
on $V$ and $\omega\in V^{*}$. $V=V_{1}\oplus V_{2}$, $h$ is positive definite
on $V_{1}$, $h(V_{1},V_{2})=0$ and $\omega(V_{2})=0$. Denote by $\omega^{*}$
the unique vector in $V_{1}$ such that for any $z\in V_{1}$,
$\omega(z)=h(\omega^{*},z).$
Then we have
(3.20) $h\geq\omega(\omega^{*})^{-1}\omega\otimes\omega.$
###### Proof.
For arbitrary $y\in V$, there exist $\lambda\in\mathbb{R}$, $z_{1}\in V_{1}$
and $z_{2}\in V_{2}$, such that $y=\lambda\omega^{*}+z_{1}+z_{2}$ and
$h(\omega^{*},z_{1})=0$. Then
$h(y,y)=\lambda^{2}h(\omega^{*},\omega^{*})+h(z_{1},z_{1})+h(z_{2},z_{2})\geq\lambda^{2}h(\omega^{*},\omega^{*})=\lambda^{2}\omega(\omega^{*})$
and
$\omega(\omega^{*})^{-1}\omega\otimes\omega(y,y)=\omega(\omega^{*})^{-1}\omega(y)^{2}=\lambda^{2}\omega(\omega^{*}).$
Hence (3.20) holds.
∎
###### Lemma 3.3.
Let $\Omega$ be a compact and convex subset of $\mathbb{R}^{k}$, such that for
every $\sigma\in\mathscr{S}(k)$ and $x=(x^{1},\cdots,x^{k})\in\Omega$,
(3.21) $T_{\sigma}(x)=(x^{\sigma(1)},\cdots,x^{\sigma(k)})\in\Omega;$
where $\mathscr{S}(k)$ denotes the permutation group of $\\{1,\cdots,k\\}$. If
$f:\Omega\rightarrow{\tenmsb R}$ is a symmetric $C^{2}$ function, and
$(D^{2}f)$ is nonpositive definite everywhere in $\Omega$, then there exists
$x_{0}=(x_{0}^{1},\cdots,x_{0}^{k})\in\Omega$, such that
$x_{0}^{1}=x_{0}^{2}=\cdots=x_{0}^{k}$ and
(3.22) $f(x_{0})=\sup_{\Omega}f.$
###### Proof.
By the compactness of $\Omega$, there exists
$x=(x^{1},\cdots,x^{k})\in\Omega$, such that $f(x)=\sup_{\Omega}f$.
Furthermore we have
$f\big{(}T_{\sigma}(x)\big{)}=f(x)=\sup_{\Omega}f\qquad\sigma\in\mathscr{S}(k)$
from the fact that $f$ is symmetric. Denote by $C_{\sigma}(x)$ the convex
closure of $\\{T_{\sigma}(x):\sigma\in\mathscr{S}(k)\\}$, then
$C_{\sigma}(x)\subset\Omega$ and
$f(y)\geq\sup_{\sigma\in\mathscr{S}(k)}f\big{(}T_{\sigma}(x)\big{)}=\sup_{\Omega}f$
for arbitrary $y\in C_{\sigma}(x)$, since $(D^{2}f)\leq 0$; which implies
$f\big{|}_{C_{\sigma}(x)}\equiv\sup_{\Omega}f.$
Denote $x_{0}^{1}=\cdots=x_{0}^{k}=\frac{1}{k}\sum_{i=1}^{k}x^{i}$, then
$x_{0}=(x_{0}^{1},\cdots,x_{0}^{k})=\frac{1}{k}\sum_{s=1}^{k}(x^{s},x^{s+1},\cdots,x^{k},x^{1},x^{2},\cdots,x^{s-1})\in
C_{\sigma}(x);$
From which (3.22) follows.
∎
By (3.12),
(3.23) $h\mathop{=}\limits^{def.}\mbox{Hess}(v)-v(2-v)g-v^{-1}dv\otimes dv$
is nonnegative definite on $T_{P}\mathbb{G}_{n,m}$. Denote
(3.24) $V_{1}=\bigoplus_{\alpha}E_{\alpha\alpha},\qquad
V_{2}=\bigoplus_{i\neq\alpha}E_{i\alpha};$
then $T_{P}\mathbb{G}_{n,m}=V_{1}\oplus V_{2}$, and (3.11), (2.9), (3.9) tell
us
$h(V_{1},V_{2})=0,\ dv(V_{2})=0$
and
(3.25) $h|_{V_{1}}=\sum_{\alpha}(v-1+\lambda_{\alpha}^{2})v\
\omega_{\alpha\alpha}^{2}.$
is positive definite. Denote by $\tilde{\nabla}v$ the unique element in
$V_{1}$ such that for any $X\in V_{1}$,
$h(\tilde{\nabla}v,X)=dv(X).$
From (3.25) and (3.9), it is not difficult to obtain
$\tilde{\nabla}v=\sum_{\alpha}\frac{\lambda_{\alpha}(1+\lambda_{\alpha}^{2})}{v-1+\lambda_{\alpha}^{2}}E_{\alpha\alpha}$
and
(3.26)
$dv(\tilde{\nabla}v)=\sum_{\alpha}\frac{\lambda_{\alpha}^{2}}{v-1+\lambda_{\alpha}^{2}}v.$
Then Lemma 3.2 and (3.23) tell us
(3.27) $\mbox{Hess}(v)\geq
v(2-v)g+\Big{[}1+\big{(}\sum_{\alpha}\frac{\lambda_{\alpha}^{2}}{v-1+\lambda_{\alpha}^{2}}\big{)}^{-1}\Big{]}v^{-1}dv\otimes
dv.$
It is necessary to estimate the upper bound of
$\sum_{\alpha}\frac{\lambda_{\alpha}^{2}}{v-1+\lambda_{\alpha}^{2}}$. Denote
(3.28) $\nu_{\alpha}=\log(1+\lambda_{\alpha}^{2}),$
then $\lambda_{\alpha}^{2}=-1+e^{\nu_{\alpha}}$; since
$v=\prod_{\alpha}(1+\lambda_{\alpha}^{2})^{\frac{1}{2}}$,
$\log
v=\frac{1}{2}\sum_{\alpha}\log(1+\lambda_{\alpha}^{2})=\frac{1}{2}\sum_{\alpha}\nu_{\alpha}$
and
$\sum_{\alpha}\frac{\lambda_{\alpha}^{2}}{v-1+\lambda_{\alpha}^{2}}=\sum_{\alpha}\frac{-1+e^{\nu_{\alpha}}}{v-2+e^{\nu_{\alpha}}}.$
Now we define
(3.29) $\Omega=\big{\\{}(\nu_{1},\cdots,\nu_{m})\in{\tenmsb
R}^{m}:\nu_{\alpha}\geq 0,\sum_{\alpha}\nu_{\alpha}=2\log v\big{\\}},$
and $f:\Omega\rightarrow{\tenmsb R}$ by
$(\nu_{1},\cdots,\nu_{m})\mapsto\sum_{\alpha}\frac{-1+e^{\nu_{\alpha}}}{v-2+e^{\nu_{\alpha}}}.$
Then obviously $\Omega$ is compact and convex, $T_{\sigma}(\Omega)=\Omega$ for
every $\sigma\in\mathscr{S}(m)$ (cf. Lemma 3.3), $f$ is a symmetric function
and a direct calculation shows
$\frac{\partial^{2}f}{\partial\nu_{\alpha}\partial\nu_{\beta}}=\frac{(v-1)e^{\nu_{\alpha}}(v-2-e^{\nu_{\alpha}})}{(v-2+e^{\nu_{\alpha}})^{3}}\delta_{\alpha\beta};$
i.e.,
$(D^{2}f)\leq 0\qquad\mbox{when }v\in(1,2].$
Then from Lemma 3.3,
$\sup_{\Omega}f=f\big{(}\frac{2\log v}{m},\cdots,\frac{2\log
v}{m}\big{)}=\frac{m(-1+v^{\frac{2}{m}})}{v-2+v^{\frac{2}{m}}};$
which is an upper bound of
$\sum_{\alpha}\frac{\lambda_{\alpha}^{2}}{v-1+\lambda_{\alpha}^{2}}$.
Substituting it into (3.27) gives
$\mbox{Hess}(v)\geq
v(2-v)g+\Big{(}\frac{v-1}{mv(v^{\frac{2}{m}}-1)}+\frac{m+1}{mv}\Big{)}dv\otimes
dv.$
In summary, we have the following Proposition.
###### Proposition 3.1.
$v$ is a convex function on $B_{JX}(P_{0})\subset{\tenmsb
U}\subset\mathbb{G}_{n,m}$, and
(3.30) $\mbox{Hess}(v)\geq
v(2-v)g+\Big{(}\frac{v-1}{pv(v^{\frac{2}{p}}-1)}+\frac{p+1}{pv}\Big{)}dv\otimes
dv$
on $\\{P\in{\tenmsb U}:v(P)\leq 2\\}$, where $p=min(n,m)$.
Similarly, we consider
(3.31)
$\tilde{h}\mathop{=}\limits^{def.}\mbox{Hess}(u)-(2-\frac{1}{2}u^{2})g;$
which is nonnegative definite on $T_{P}\mathbb{G}_{n,m}$. The definition of
$V_{1}$ and $V_{2}$ is similar to above. It is easily seen from (3.17) and
(2.9) that
$\tilde{h}(V_{1},V_{2})=0$
and
(3.32)
$\tilde{h}|_{V_{1}}=\sum_{\alpha}(8\lambda_{\alpha}^{2}+6\lambda_{\alpha}^{4}+\frac{1}{2}u^{2})\omega_{\alpha\alpha}^{2}$
is positive definite. By (3.15),
(3.33)
$du=\sum_{\alpha}2\lambda_{\alpha}(1+\lambda_{\alpha}^{2})\omega_{\alpha\alpha},$
then
$du(V_{2})=0.$
Hence Lemma 3.2 can be applied for us to obtain
(3.34)
$\mbox{Hess}(u)\geq(2-\frac{1}{2}u^{2})g+\big{(}du(\tilde{\nabla}u)\big{)}^{-1}du\otimes
du.$
where $\tilde{\nabla}u$ denotes the unique element in $V_{1}$ such that for
arbitrary $X\in V_{1}$,
$\tilde{h}(\tilde{\nabla}u,X)=du(X).$
From (3.32) and (3.33), we can derive
$\tilde{\nabla}u=\sum_{\alpha}\frac{2\lambda_{\alpha}(1+\lambda_{\alpha}^{2})^{2}}{8\lambda_{\alpha}^{2}+6\lambda_{\alpha}^{4}+\frac{1}{2}u^{2}}E_{\alpha\alpha},$
and hence
(3.35)
$du(\tilde{\nabla}u)=\sum_{\alpha}\frac{2\lambda_{\alpha}^{2}(1+\lambda_{\alpha}^{2})^{2}}{3\lambda_{\alpha}^{4}+4\lambda_{\alpha}^{2}+\frac{1}{4}u^{2}}.$
(3.34) tells us it is necessary for us to estimate the upper bound of the
right side of (3.35).
Define $\Omega=\\{(\nu_{1},\cdots,\nu_{m})\in{\tenmsb
R}^{m}:\sum_{\alpha}\nu_{\alpha}=u\\}$ and $f:\Omega\rightarrow{\tenmsb R}$
$(\nu_{1},\cdots,\nu_{m})\mapsto\sum_{\alpha}\frac{2\nu_{\alpha}(1+\nu_{\alpha})^{2}}{3\nu_{\alpha}^{2}+4\nu_{\alpha}+C}\qquad\mbox{
where }C=\frac{1}{4}u^{2}.$
Then it is easy to see that $\sup f$ is an upper bound of
$du(\tilde{\nabla}u)$, since
$u=\sum_{\alpha}\tan^{2}\theta_{\alpha}=\sum_{\alpha}\lambda_{\alpha}^{2}.$
Obviously $\Omega$ is compact and convex, $T_{\sigma}(\Omega)=\Omega$ for
every $\sigma\in\mathscr{S}(m)$, $f$ is a symmetric function and a direct
calculation shows
$\frac{\partial^{2}f}{\partial\nu_{\alpha}\partial\nu_{\beta}}=\frac{-4\big{[}(3C-1)\nu_{\alpha}^{3}+6C\nu_{\alpha}^{2}+(9C-3C^{2})\nu_{\alpha}+4C-2C^{2}\big{]}}{(3\nu_{\alpha}^{2}+4\nu_{\alpha}+C)^{3}}\delta_{\alpha\beta}.$
To show $(D^{2}f)\leq 0$ when $u\in(0,2]$, it is sufficient to prove
$F:[0,u]\rightarrow{\tenmsb R}$
$t\mapsto(3C-1)t^{3}+6Ct^{2}+(9C-3C^{2})t+4C-2C^{2}$
is a nonnegative function, where $C=\frac{u^{2}}{4}\in(0,1]$. If $F$ attains
its minimum at $t_{0}\in(0,u)$, then
(3.36) $\displaystyle
0=F^{\prime}(t_{0})=3(3C-1)t_{0}^{2}+12Ct_{0}+9C-3C^{2},$ (3.37)
$\displaystyle 0\leq F^{\prime\prime}(t_{0})=6(3C-1)t_{0}+12C.$
On the other hand, when $3C-1\geq 0$, we have $F^{\prime}(t_{0})\geq
9C-3C^{2}>0$, which causes a contradiction; when $3C-1<0$, from (3.37),
$t_{0}\leq\frac{2C}{1-3C}$, then $F^{\prime}(t_{0})\geq
F^{\prime}(0)=9C-3C^{2}>0$, which also causes a contradiction. Therefore
$\min_{[0,u]}F=\min\big{\\{}F(0),F(u)\big{\\}}.$
In conjunction with
$\displaystyle F(0)$ $\displaystyle=$ $\displaystyle 4C-2C^{2}>0$
$\displaystyle F(u)$ $\displaystyle=$
$\displaystyle(3C-1)u^{3}+6Cu^{2}+(9C-3C^{2})u+4C-2C^{2}$ $\displaystyle=$
$\displaystyle\frac{9}{16}u^{5}+\frac{11}{8}u^{4}+\frac{5}{4}u^{3}+u^{2}>0,$
$F$ is a nonnegative function. Thereby applying Lemma 3.3 we have
(3.38) $du(\tilde{\nabla}u)\leq\sup
f=f(\frac{u}{m},\cdots,\frac{u}{m})=\frac{2(u+m)^{2}}{(3+\frac{1}{4}m^{2})u+4m}.$
Substituting (3.38) into (3.34) gives
$\mbox{Hess}(u)\geq\left(2-\frac{1}{2}u^{2}\right)g+\frac{(3+\frac{1}{4}m^{2})u+4m}{2(u+m)^{2}}du\otimes
du.$
We rewrite the conclusion as the following Proposition.
###### Proposition 3.2.
$u$ is a convex function on $B_{JX}(P_{0})\subset{\tenmsb
U}\subset\mathbb{G}_{n,m}$ and
(3.39)
$\mbox{Hess}(u)\geq\left(2-\frac{1}{2}u^{2}\right)g+\frac{(3+\frac{1}{4}p^{2})u+4p}{2(u+p)^{2}}du\otimes
du$
on $\\{P\in{\tenmsb U}:u(P)\leq 2\\}$, where $p=min(n,m)$.
## 4\. The construction of auxiliary functions
Let
(4.1) $h_{1}=v^{-k}(2-v)^{k},$
where $k>0$ to be chosen, then
$\displaystyle\begin{aligned}
h^{\prime}_{1}=&-kv^{-k-1}(2-v)^{k}-kv^{-k}(2-v)^{k-1}\\\
=&-2kv^{-k-1}(2-v)^{k-1},\\\
h^{\prime\prime}_{1}=&2k(k+1)v^{-k-2}(2-v)^{k-1}+2k(k-1)v^{-k-1}(2-v)^{k-2}\\\
=&4kv^{-k-2}(2-v)^{k-2}(k+1-v).\end{aligned}$
Here ′ denotes derivative with respect to $v$. Hence, from (3.30)
(4.2) $\displaystyle\begin{aligned}
&\mbox{Hess}(h_{1})=-2kv^{-k-1}(2-v)^{k-1}\mbox{Hess}(v)+4kv^{-k-2}(2-v)^{k-2}(k+1-v)dv\otimes
dv\\\ &\hskip 43.36243pt\leq-2kv^{-k}(2-v)^{k}g-\\\
&\quad-2kv^{-k-2}(2-v)^{k-2}\Big{[}\frac{(v-1)(2-v)}{p(v^{\frac{2}{p}}-1)}+\frac{p+1}{p}(2-v)-2(k+1-v)\Big{]}dv\otimes
dv.\end{aligned}$
Please note that $\frac{v-1}{v^{\frac{2}{p}}-1}$ is an increasing function on
$[1,2]$: it is easily seen when $p$ is even, since
$\frac{v-1}{v^{\frac{2}{p}}-1}=1+v^{\frac{2}{p}}+v^{\frac{4}{p}}+\cdots+v^{1-\frac{2}{p}};$
otherwise, when $p$ is odd,
$\frac{v-1}{v^{\frac{2}{p}}-1}=\frac{v^{1-\frac{1}{p}}-1}{v^{\frac{2}{p}}-1}+\frac{v-v^{1-\frac{1}{p}}}{v^{\frac{2}{p}}-1}=1+v^{\frac{2}{p}}+v^{\frac{4}{p}}+\cdots+v^{1-\frac{3}{p}}+\frac{v^{1-\frac{1}{p}}}{v^{\frac{1}{p}}+1}$
it follows from
$\Big{(}\frac{v^{1-\frac{1}{p}}}{v^{\frac{1}{p}}+1}\Big{)}^{\prime}=\frac{1-\frac{2}{p}+(1-\frac{1}{p})v^{-\frac{1}{p}}}{(v^{\frac{1}{p}}+1)^{2}}\geq
0.$
Hence
$\frac{v-1}{v^{\frac{2}{p}}-1}\geq\frac{p}{2},$
and moreover
$\displaystyle\begin{aligned}
\frac{(v-1)(2-v)}{p(v^{\frac{2}{p}}-1)}&+\frac{p+1}{p}(2-v)-2(k+1-v)\\\
\geq&\big{(}\frac{1}{2}+\frac{p+1}{p}\big{)}(2-v)-2(k+1-v)\\\
=&\big{(}\frac{1}{2}-\frac{1}{p}\big{)}v+\big{(}3+\frac{2}{p}\big{)}-2(k+1)\\\
\geq&\frac{3}{2}+\frac{1}{p}-2k.\end{aligned}$
Now we take
(4.3) $k=\frac{3}{4}+\frac{1}{2p},$
then $\frac{(v-1)(2-v)}{p(v^{\frac{2}{p}}-1)}+\frac{p+1}{p}(2-v)-2(k+1-v)\geq
0$ and then (4.2) becomes
(4.4) $\mbox{Hess}(h_{1})\leq-2kh_{1}\
g=-\left(\frac{3}{2}+\frac{1}{p}\right)h_{1}\ g.$
Denote
(4.5) $h_{2}=h_{1}^{-\frac{6p}{3p+2}}=v^{\frac{3}{2}}(2-v)^{-\frac{3}{2}},$
then
(4.6) $\displaystyle\begin{aligned}
\mbox{Hess}(h_{2})=&-\frac{6p}{3p+2}h_{1}^{-\frac{6p}{3p+2}-1}\mbox{Hess}(h_{1})+\frac{6p}{3p+2}\big{(}\frac{6p}{3p+2}+1\big{)}h_{1}^{-\frac{6p}{3p+2}-2}dh_{1}\otimes
dh_{1}\\\ \geq&3h_{1}^{-\frac{6p}{3p+2}}\
g+\big{(}\frac{3}{2}+\frac{1}{3p}\big{)}h_{1}^{\frac{6p}{3p+2}}dh_{2}\otimes
dh_{2}\\\ =&3h_{2}\
g+\big{(}\frac{3}{2}+\frac{1}{3p}\big{)}h_{2}^{-1}dh_{2}\otimes
dh_{2}.\end{aligned}$
Let
(4.7) $h_{3}=(u+\alpha)^{-1}(2-u),$
where $\alpha>0$ to be chosen. A direct calculation shows
$\displaystyle\begin{aligned}
h^{\prime}_{3}=&-(u+\alpha)^{-2}(2-u)-(u+\alpha)^{-1}\\\
=&-(2+\alpha)(u+\alpha)^{-2},\\\
h^{\prime\prime}_{3}=&2(2+\alpha)(u+\alpha)^{-3}.\end{aligned}$
Here ′ denotes derivative with respect to $u$. Combining with (3.39), we have
(4.8) $\displaystyle\begin{aligned}
\mbox{Hess}(h_{3})=&-(2+\alpha)(u+\alpha)^{-2}\mbox{Hess}(u)+2(2+\alpha)(u+\alpha)^{-3}du\otimes
du\\\ \leq&-\frac{(2+\alpha)(u+2)}{2(u+\alpha)}h_{3}\ g\\\
&-(2+\alpha)(u+\alpha)^{-3}\Big{[}\frac{(u+\alpha)\big{(}(3+\frac{1}{4}p^{2})u+4p\big{)}}{2(u+p)^{2}}-2\Big{]}du\otimes
du.\end{aligned}$
Choose
(4.9) $\alpha=p,$
then
$\displaystyle\frac{(u+\alpha)\big{(}(3+\frac{1}{4}p^{2})u+4p\big{)}}{2(u+p)^{2}}-2=\frac{\big{(}3+\frac{1}{4}p^{2}\big{)}u+4p}{2(u+p)}-2\geq
2-2\geq 0,$
and
$\frac{(2+\alpha)(u+2)}{2(u+\alpha)}\geq\frac{2+p}{p}=1+\frac{2}{p}.$
Thereby (4.8) becomes
(4.10) $\mbox{Hess}(h_{3})\leq-\big{(}1+\frac{2}{p}\big{)}h_{3}\ g.$
Denote
(4.11)
$h_{4}=h_{3}^{-\frac{3p}{p+2}}=(u+p)^{\frac{3p}{p+2}}(2-u)^{-\frac{3p}{p+2}},$
then
(4.12) $\displaystyle\begin{aligned}
\mbox{Hess}(h_{4})=&-\frac{3p}{p+2}h_{3}^{-\frac{3p}{p+2}-1}\mbox{Hess}(h_{3})+\frac{3p}{p+2}\big{(}\frac{3p}{p+2}+1\big{)}h_{3}^{-\frac{3p}{p+2}-2}dh_{3}\otimes
dh_{3}\\\ \geq&3h_{3}^{-\frac{3p}{p+2}}\
g+\big{(}\frac{4}{3}+\frac{2}{3p}\big{)}h_{3}^{\frac{3p}{p+2}}dh_{4}\otimes
dh_{4}\\\ =&3h_{4}\
g+\big{(}\frac{4}{3}+\frac{2}{3p}\big{)}h_{4}^{-1}dh_{4}\otimes
dh_{4}.\end{aligned}$
Let $M$ be an $n$-dimensional submanifold in ${\tenmsb R}^{n+m}$ with $m\geq
2.$ The Gauss map $\gamma:M\to\mathbb{G}_{n,m}$ is defined by
$\gamma(x)=T_{x}M\in\mathbb{G}_{n,m}$
via the parallel translation in $\mathbb{R}^{m+n}$ for arbitrary $x\in M$. The
energy density of the Gauss map (see [21] Chap.3, §3.1) is
$e(\gamma)=\frac{1}{2}\left<\gamma_{*}e_{i},\gamma_{*}e_{i}\right>=\frac{1}{2}|B|^{2}.$
Ruh-Vilms proved that the mean curvature vector of $M$ is parallel if and only
if its Gauss map is a harmonic map [17].
If the Gauss image of $M$ is contained in $\\{P\in{\tenmsb
U}\subset\mathbb{G}_{n,m}:v(P)<2\\}$, then the composition function
$\tilde{h}_{1}=h_{1}\circ\gamma$ of $h_{1}$ with the Gauss map $\gamma$
defines a function on $M$. Using composition formula, we have
(4.13) $\displaystyle\Delta\tilde{h}_{1}$
$\displaystyle=\mbox{Hess}(h_{1})(\gamma_{*}e_{i},\gamma_{*}e_{i})+dh_{1}(\tau(\gamma))$
$\displaystyle\leq-\big{(}\frac{3}{2}+\frac{1}{p}\big{)}|B|^{2}\tilde{h}_{1},$
where $\tau(\gamma)$ is the tension field of the Gauss map, which is zero,
provided $M$ has parallel mean curvature by the Ruh-Vilms theorem mentioned
above. Similarly, for $\tilde{h}_{2}=h_{2}\circ\gamma$ defined on $M$, we have
(4.14) $\displaystyle\Delta\tilde{h}_{2}$
$\displaystyle=\mbox{Hess}(h_{2})(\gamma_{*}e_{i},\gamma_{*}e_{i})+dh_{2}(\tau(\gamma))$
$\displaystyle\geq 3\
\tilde{h}_{2}|B|^{2}+\big{(}\frac{3}{2}+\frac{1}{3p}\big{)}\tilde{h}_{2}^{-1}|\nabla\tilde{h}_{2}|^{2}.$
If the Gauss image of $M$ is contained in $\\{P\in{\tenmsb
U}\subset\mathbb{G}_{n,m}:u(P)<2\\}$, we can defined composition function
$\tilde{h}_{3}=h_{3}\circ\gamma$ and $\tilde{h}_{4}=h_{4}\circ\gamma$ on $M$.
Again using composition formula, we obtain
(4.15)
$\Delta\tilde{h}_{3}\leq-\Big{(}1+\frac{2}{p}\Big{)}|B|^{2}\tilde{h}_{3}$
and
(4.16) $\Delta\tilde{h}_{4}\geq 3\
\tilde{h}_{4}|B|^{2}+\Big{(}\frac{4}{3}+\frac{2}{3p}\Big{)}\tilde{h}_{4}^{-1}|\nabla\tilde{h}_{4}|^{2}.$
With the aid of $\tilde{h}_{1}$ and $\tilde{h}_{3}$, we immediately have the
following lemma.
###### Lemma 4.1.
Let $M$ be an $n$-dimensional minimal submanifold of $\mathbb{R}^{n+m}$ ($M$
needs not be complete), if the Gauss image of $M$ is contained in
$\\{P\in{\tenmsb U}\subset\mathbb{G}_{n,m}:v(P)<2\\}$ (or respectively,
$\\{P\in{\tenmsb U}\subset\mathbb{G}_{n,m}:u(P)<2\\}$), then we have
(4.17) $\displaystyle\int_{M}|\nabla\phi|^{2}$
$\displaystyle*1\geq\big{(}\frac{3}{2}+\frac{1}{p}\big{)}\int_{M}|B|^{2}\phi^{2}*1$
$\displaystyle\Big{(}\mbox{or respectively, }$
$\displaystyle\int_{M}|\nabla\phi|^{2}*1\geq\big{(}1+\frac{2}{p}\big{)}\int_{M}|B|^{2}\phi^{2}*1\Big{)}$
for any function $\phi$ with compact support $D\subset M$.
###### Remark 4.1.
For a stable minimal hypersurface there is the stability inequality, which is
one of main ingredient for Schoen-Simon-Yau’s curvature esimates for stable
minimal hypersurfaces. For minimal submanifolds with the Gauss image
restriction we have stronger inequality as shown in (4.17). Our proof is
similar to [23] and [24], so we omit the detail of it.
## 5\. Curvature estimates
We are now in a position to carry out the curvature estimates of Schoen-Simon-
Yau type.
Let $M$ be an $n$-dimensional minimal submanifold in $\mathbb{R}^{n+m}$.
Assume that the estimate
(5.1) $\int_{M}|\nabla\phi|^{2}*1\geq\lambda\int_{M}|B|^{2}\phi^{2}*1$
holds for arbitrary function $\phi$ with compact support $D\subset M$, where
$\lambda$ is a positive constant.
Replacing $\phi$ by $|B|^{1+q}\phi$ in (5.1) gives
(5.2) $\displaystyle\int_{M}|B|^{4+2q}$
$\displaystyle\phi^{2}*1\leq\lambda^{-1}\int_{M}\big{|}\nabla(|B|^{1+q}\phi)\big{|}^{2}*1$
$\displaystyle=\lambda^{-1}(1+q)^{2}\int_{M}|B|^{2q}\big{|}\nabla|B|\big{|}^{2}\phi^{2}*1+\lambda^{-1}\int_{M}|B|^{2+2q}|\nabla\phi|^{2}*1$
$\displaystyle+2\lambda^{-1}(1+q)\int_{M}|B|^{1+2q}\nabla|B|\cdot\phi\nabla\phi*1.$
Using Bochner technique, the estimate done in [14][7], and the Kato-type
inequality derived in [24], we obtain
(5.3) $\Delta|B|^{2}\geq
2\big{(}1+\frac{2}{mn}\big{)}\big{|}\nabla|B|\big{|}^{2}-3|B|^{4}.$
(For the detail, see [24] Section 2.) It is equivalent to
(5.4)
$\frac{2}{mn}\big{|}\nabla|B|\big{|}^{2}\leq|B|\Delta|B|+\frac{3}{2}|B|^{4}.$
Multiplying $|B|^{2q}\phi^{2}$ with both sides of (5.4) and integrating by
parts, we have
(5.5) $\displaystyle\frac{2}{mn}$
$\displaystyle\int_{M}|B|^{2q}\big{|}\nabla|B|\big{|}^{2}\phi^{2}*1$
$\displaystyle\qquad\leq-(1+2q)\int_{M}|B|^{2q}\big{|}\nabla|B|\big{|}^{2}\phi^{2}*1$
$\displaystyle\qquad\qquad-2\int_{M}|B|^{1+2q}\nabla|B|\cdot\phi\nabla\phi*1+\frac{3}{2}\int_{M}|B|^{4+2q}\phi^{2}*1.$
By multiplying $\frac{3}{2}$ with both sides of (5.2) and then adding up both
sides of it and (5.5), we have
(5.6)
$\displaystyle\big{(}\frac{2}{mn}+1+2q-\frac{3}{2}\lambda^{-1}(1+q)^{2}\big{)}\int_{M}|B|^{2q}\big{|}\nabla|B|\big{|}^{2}\phi^{2}*1$
$\displaystyle\qquad\leq\frac{3}{2}\lambda^{-1}\int_{M}|B|^{2+2q}|\nabla\phi|^{2}*1+\big{(}3\lambda^{-1}(1+q)-2\big{)}\int_{M}|B|^{1+2q}\nabla|B|\cdot\phi\nabla\phi*1.$
By using Young’s inequality, (5.6) becomes
(5.7)
$\displaystyle\big{(}\frac{2}{mn}+1+2q-\frac{3}{2}\lambda^{-1}(1+q)^{2}-\varepsilon\big{)}\int_{M}|B|^{2q}\big{|}\nabla|B|\big{|}^{2}\phi^{2}*1$
$\displaystyle\hskip 72.26999pt\leq
C_{1}(\varepsilon,\lambda,q)\int_{M}|B|^{2+2q}|\nabla\phi|^{2}*1.$
If
(5.8) $\lambda>\frac{3}{2}\big{(}1-\frac{2}{mn}\big{)},$
then
$\frac{2}{mn}+1+2q-\frac{3}{2}\lambda^{-1}(1+q)^{2}>0$
whenever
(5.9)
$q\in\Big{[}0,-1+\frac{2}{3}\lambda+\frac{1}{3}\sqrt{4\lambda^{2}-6\big{(}1-\frac{2}{mn}\big{)}\lambda}\
\Big{)}.$
Thus we can choose $\varepsilon$ sufficiently small, such that
(5.10) $\int_{M}|B|^{2q}\big{|}\nabla|B|\big{|}^{2}\phi^{2}*1\leq
C_{2}\int_{M}|B|^{2+2q}|\nabla\phi|^{2}*1$
where $C_{2}$ only depends on $n$, $m$, $\lambda$ and $q$.
Combining with (5.2) and (5.10), we can derive
(5.11) $\int_{M}|B|^{4+2q}\phi^{2}*1\leq
C_{3}(n,m,\lambda,q)\int_{M}|B|^{2+2q}|\nabla\phi|^{2}*1$
by again using Young’s inequality.
By replacing $\phi$ by $\phi^{2+q}$ in (5.11) and then using Hölder
inequality, we have
(5.12) $\displaystyle\int_{M}|B|^{4+2q}\phi^{4+2q}*1\leq
C\int_{M}|\nabla\phi|^{4+2q}*1.$
where $C$ is a constant only depending on $n$, $m$, $\lambda$ and $q$.
Similarly, replacing $\phi$ by $\phi^{1+q}$ in (5.11) and then again using
Hölder inequality yields
(5.13) $\displaystyle\int_{M}|B|^{4+2q}\phi^{2+2q}*1\leq
C^{\prime}\int_{M}|B|^{2}|\nabla\phi|^{2+2q}*1.$
where $C^{\prime}$ is a constant only depending on $n$, $m$, $\lambda$ and
$q$.
Let $r$ be a function on $M$ with $|\nabla r|\leq 1$. For any $R\in[0,R_{0}]$,
where $R_{0}=\sup_{M}r$, suppose
$M_{R}=\\{x\in M,\quad r\leq R\\}$
is compact.
(5.12) and Lemma 4.1 enable us to prove the following results by taking
$\phi\in C_{c}^{\infty}(M_{R})$ to be the standard cut-off function such that
$\phi\equiv 1$ in $M_{\theta R}$ and $|\nabla\phi|\leq
C(1-\theta)^{-1}R^{-1}$.
###### Theorem 5.1.
Let $M$ be an $n$-dimensional minimal submanifolds of $\mathbb{R}^{n+m}$. If
the Gauss image of $M_{R}$ is contained in $\\{P\in{\tenmsb
U}\subset\mathbb{G}_{n,m}:v(P)<2\\}$, then we have the estimate
(5.14) $\big{\|}|B|\big{\|}_{L^{s}(M_{\theta R})}\leq
C(n,m,s)(1-\theta)^{-1}R^{-1}\text{Vol}(M_{R})^{\frac{1}{s}}$
for arbitrary $\theta\in(0,1)$ and
$s\in\left[4,4+\frac{4}{3p}+\frac{2}{3}\sqrt{\big{(}3+\frac{2}{p}\big{)}\big{(}\frac{6}{mn}+\frac{2}{p}\big{)}}\
\right).$
If $p\leq 4$, and the Gauss image of $M_{R}$ is contained in $\\{P\in{\tenmsb
U}\subset\mathbb{G}_{n,m}:u(P)<2\\}$, then (5.14) still holds for arbitrary
$\theta\in(0,1)$ and
$s\in\left[4,2+\frac{4}{3}+\frac{8}{3p}+\frac{2}{3}\sqrt{\big{(}1+\frac{2}{p}\big{)}\big{(}\frac{12}{mn}+\frac{8}{p}-2\big{)}}\
\right).$
We can also fulfil the curvature estimates of Ecker-Huisken type.
Assume that $h$ is a positive function on $M$ satisfying the following
estimate
(5.15) $\Delta h\geq 3h\ g+c_{0}h^{-1}dh\otimes dh,$
where
$c_{0}>\frac{3}{2}-\frac{1}{mn}$
is a positive constant.
We compute from (5.15) and (5.3):
$\displaystyle\begin{aligned} &\Delta\big{(}|B|^{2s}h^{q}\big{)}\\\
\geq&3(q-s)|B|^{2s+2}h^{q}+2s(2s-1+\frac{2}{mn})|B|^{2s-2}\big{|}\nabla|B|\big{|}^{2}h^{q}+q(q+c_{0}-1)|B|^{2s}h^{q-2}|\nabla
h|^{2}\\\ &+4sq|B|^{2s-1}\nabla|B|\cdot h^{q-1}\nabla h.\end{aligned}$
By Young’s inequality, when $2s(2s-1+\frac{2}{mn})\cdot
q(q+c_{0}-1)\geq(2sq)^{2}$, i.e.,
(5.16) $q\geq
s\geq\frac{1}{2}-\frac{1}{mn}+\frac{1}{c_{0}-1}\big{(}\frac{1}{2}-\frac{1}{mn}\big{)}q,$
the inequality
(5.17) $\Delta\big{(}|B|^{2s}h^{q}\big{)}\geq 3(q-s)|B|^{2s+2}h^{q}$
holds. Especially,
(5.18)
$\Delta\big{(}|B|^{s-1}h^{\frac{s}{2}}\big{)}\geq\frac{3}{2}|B|^{s+1}h^{\frac{s}{2}}$
whenever
(5.19)
$s\geq\frac{2-\frac{2}{mn}}{1-\frac{1}{c_{0}-1}(\frac{1}{2}-\frac{1}{mn})}.$
Let $\eta$ be a smooth function with compact support. Integrating by parts in
conjunction with Young’s inequality lead to
(5.20) $\displaystyle\begin{aligned}
\int_{M}|B|^{2s}h^{s}\eta^{2s}*1\leq&\frac{2}{3}\int_{M}|B|^{s-1}h^{\frac{s}{2}}\eta^{2s}\Delta\big{(}|B|^{s-1}h^{\frac{s}{2}}\big{)}*1\\\
=&-\frac{2}{3}\int_{M}\Big{|}\nabla\big{(}|B|^{s-1}h^{\frac{s}{2}}\big{)}\Big{|}^{2}\eta^{2s}*1\\\
&-\frac{2}{3}\int_{M}|B|^{s-1}h^{\frac{s}{2}}\cdot
2s\eta^{2s-1}\nabla\eta\cdot\nabla\big{(}|B|^{s-1}h^{\frac{s}{2}}\big{)}*1\\\
\leq&\frac{2}{3}s^{2}\int_{M}|B|^{2s-2}h^{s}\eta^{2s-2}|\nabla\eta|^{2}*1.\end{aligned}$
By Hölder inequality,
(5.21)
$\int_{M}|B|^{2s-2}h^{s}\eta^{2s-2}|\nabla\eta|^{2}*1\leq\Big{(}\int_{M}|B|^{2s}h^{s}\eta^{2s}*1\Big{)}^{\frac{s-1}{s}}\Big{(}\int_{M}h^{s}|\nabla\eta|^{2s}*1\Big{)}^{\frac{1}{s}}.$
Substituting (5.21) into (5.20), we finally arrive at
(5.22)
$\Big{(}\int_{M}|B|^{2s}h^{s}\eta^{2s}*1\Big{)}^{\frac{1}{s}}\leq\frac{2}{3}s^{2}\Big{(}\int_{M}h^{s}|\nabla\eta|^{2s}*1\Big{)}^{\frac{1}{s}}.$
Take $\eta\in C_{c}^{\infty}(M_{R})$ to be the standard cut-off function such
that $\eta\equiv 1$ in $M_{\theta R}$ and $|\nabla\eta|\leq
C(1-\theta)^{-1}R^{-1}$; then from (5.22) we have the following estimate.
###### Theorem 5.2.
Let $M$ be an $n$-dimensional minimal submanifolds of $\mathbb{R}^{n+m}$. If
there exists a positive function $h$ on $M$ satisfying (5.15), then there
exists $C_{1}=C_{1}(n,m,c_{0}),$ such that
(5.23) $\big{\|}|B|^{2}h\big{\|}_{L^{s}(M_{\theta R})}\leq
C_{2}(s)(1-\theta)^{-2}R^{-2}\big{\|}h\big{\|}_{L^{s}(M_{R})}$
whenever $s\geq C_{1}$ and $\theta\in(0,1)$.
By (4.14) and (4.16), if the Gauss image of $M$ is contained in
$\\{P\in{\tenmsb U}\subset\mathbb{G}_{n,m}:v(P)<2\\}$, or $p\leq 4$ and the
Gauss image of $M$ is contained in $\\{P\in{\tenmsb
U}\subset\mathbb{G}_{n,m}:u(P)<2\\}$, there exists a positive function on $M$,
which is $\tilde{h}_{2}$ or respectively $\tilde{h}_{4}$, satisfying (5.15).
Hence the estimate (5.23) holds for both cases.
Furthermore, the mean value inequality for any subharmonic function on minimal
submanifolds in $\mathbb{R}^{m+n}$ (ref. [8], [16]) can be applied to yield an
estimate of the upper bound of $|B|^{2}$. We write the results as the
following theorem without detail of proof, for it is similar to [24]. Please
note that $B_{R}(x)\subset\mathbb{R}^{m+n}$ denotes a ball of radius $R$
centered at $x\in M$ and its restriction on $M$ is denoted by
$D_{R}(x)=B_{R}(x)\cap M.$
###### Theorem 5.3.
Let $x\in M$, $R>0$ such that the image of $D_{R}(x)$ under the Gauss map lies
in $\\{P\in{\tenmsb U}\subset\mathbb{G}_{n,m}:v(P)<2\\}$. Then, there exists
$C_{1}=C_{1}(n,m)$, such that
(5.24) $|B|^{2s}(x)\leq
C(n,s)R^{-(n+2s)}(\sup_{D_{R}(x)}\tilde{h}_{2})^{s}\text{Vol}(D_{R}(x)),$
for arbitrary $s\geq C_{1}$.
If $p\leq 4$, the image of $D_{R}(x)$ under the Gauss map lies in
$\\{P\in{\tenmsb U}\subset\mathbb{G}_{n,m}:u(P)<2\\}$, then there exists
$C_{2}=C_{2}(n,m)$ such that
(5.25) $|B|^{2s}(x)\leq
C(n,s)R^{-(n+2s)}(\sup_{D_{R}(x)}\tilde{h}_{4})^{s}\text{Vol}(D_{R}(x)),$
holds for any $s\geq C_{2}$.
## 6\. Bernstein type theorems and related results
If $M$ is a submanifold in $\mathbb{R}^{n+m}$, then the function $w$ defined
on $\mathbb{G}_{n,m}$ (see Section 2) and the Gauss map $\gamma$ could be
composed, yielding a smooth function on $M$, which is also denoted by $w$. By
studying the properties of $w$-function, we can obtain:
###### Proposition 6.1.
[24] Let $M$ be a complete submanifold in $\mathbb{R}^{n+m}$. If the
$w-$function is bounded below by a positive constant $w_{0}$. Then $M$ is an
entire graph with Euclidean volume growth. Precisely,
(6.1) $\text{Vol}(D_{R}(x))\leq\frac{1}{w_{0}}C(n)R^{n}.$
Now we let $M$ be a complete minimal submanifold in ${\tenmsb R}^{n+m}$ whose
Gauss image lies in $\\{P\in{\tenmsb U}\subset\mathbb{G}_{n,m}:v(P)<2\\}$.
Then $w=v^{-1}>\frac{1}{2}$ on $M$ and Proposition 6.1 tells us $M$ is an
entire graph. Precisely, the immersion $F:M\to\mathbb{R}^{m+n}$ is realized by
a graph $(x,f(x))$ with
$f:\mathbb{R}^{n}\to\mathbb{R}^{m}.$
At each point in $M$ its image $n$-plane $P$ under the Gauss map is spanned by
$f_{i}=\varepsilon_{i}+\frac{\partial f^{\alpha}}{\partial
x^{i}}\varepsilon_{\alpha}.$
Hence the local coordinate of $P$ in ${\tenmsb U}$ is
$Z=\Big{(}\frac{\partial f^{\alpha}}{\partial x^{i}}\Big{)}.$
By (3.2),
$v(P)=\left[\text{det}\left(\delta_{ij}+\sum_{\alpha}\frac{\partial
f^{\alpha}}{\partial x^{i}}\frac{\partial f^{\alpha}}{\partial
x^{j}}\right)\right]^{\frac{1}{2}}.$
Hence
(6.2) $\left[\text{det}\left(\delta_{ij}+\sum_{\alpha}\frac{\partial
f^{\alpha}}{\partial x^{i}}\frac{\partial f^{\alpha}}{\partial
x^{j}}\right)\right]^{\frac{1}{2}}<2$
at each $x\in{\tenmsb R}^{n}$. Conversely, if $M=(x,f(x))$ is a minimal graph
given by $f:{\tenmsb R}^{n}\rightarrow{\tenmsb R}^{m}$ which satisfy (6.2),
then the Gauss image of $M$ lies in $\\{P\in{\tenmsb
U}\subset\mathbb{G}_{n,m}:v(P)<2\\}$.
Let $P\in{\tenmsb U}$ such that $u(P)=\sum_{\alpha}\tan^{2}\theta_{\alpha}<2$,
then $\cos^{2}\theta_{\alpha}=(1+\tan^{2}\theta_{\alpha})^{-1}>\frac{1}{3}$
and
$w(P)=\prod_{\alpha}\cos\theta_{\alpha}>3^{-\frac{p}{2}}.$
Hence Proposition 6.1 could be applied when $M$ is a complete minimal
submanifold in ${\tenmsb R}^{n+m}$ whose Gauss image lies in $\\{P\in{\tenmsb
U}\subset\mathbb{G}_{n,m}:v(P)<2\\}$; which is hence a minimal graph given by
$f:{\tenmsb R}^{n}\rightarrow{\tenmsb R}^{m}$. Thereby (3.14) shows
(6.3) $\sum_{i,\alpha}\left(\frac{\partial f^{\alpha}}{\partial
x^{i}}\right)^{2}<2.$
And vice versa.
Theorem 5.1 and Proposition 6.1 give us the following Bernstein-type theorem.
###### Theorem 6.1.
Let $M=(x,f(x))$ be an $n$-dimensional entire minimal graph given by $m$
functions $f^{\alpha}(x^{1},\cdots,x^{n})$ with $m\geq 2,n\leq 4$. If
$\Delta_{f}=\left[\text{det}\left(\delta_{ij}+\sum_{\alpha}\frac{\partial
f^{\alpha}}{\partial x^{i}}\frac{\partial f^{\alpha}}{\partial
x^{j}}\right)\right]^{\frac{1}{2}}<2$
or
$\Lambda_{f}=\sum_{i,\alpha}\left(\frac{\partial f^{\alpha}}{\partial
x^{i}}\right)^{2}<2,$
then $f^{\alpha}$ has to be affine linear functions representing an affine
$n$-plane.
###### Proof.
If $\Delta_{f}<2$, then the Gauss image of $M$ is contained in
$\\{P\in{\tenmsb U}\subset\mathbb{G}_{n,m}:v(P)<2\\}$. We choose
$s=4+\frac{4}{3p}>4.$
Fix $x\in M$ and let $r$ be the Euclidean distance function from $x$ and
$M_{R}=D_{R}(x)$. Hence, letting $R\to+\infty$ in (5.14) yields
$\big{\|}|B|\big{\|}_{L^{s}(M)}=0.$
i.e., $|B|=0$. $M$ has to be an affine linear subspace.
For the case $\Lambda_{f}<2$, the proof is similar. ∎
Theorem 5.3 and Proposition 6.1 yield Bernstein type results as follows.
###### Theorem 6.2.
Let $M=(x,f(x))$ be an $n$-dimensional entire minimal graph given by $m$
functions $f^{\alpha}(x^{1},\cdots,x^{n})$ with $m\geq 2$. If
$\Delta_{f}=\left[\text{det}\left(\delta_{ij}+\sum_{\alpha}\frac{\partial
f^{\alpha}}{\partial x^{i}}\frac{\partial f^{\alpha}}{\partial
x^{j}}\right)\right]^{\frac{1}{2}}<2,$
and
(6.4) $\left(2-\Delta_{f}\right)^{-1}=o(R^{\frac{4}{3}}),$
where $R^{2}=|x|^{2}+|f|^{2}$. Then $f^{\alpha}$ has to be affine linear
functions and hence $M$ has to be an affine linear subspace.
###### Theorem 6.3.
Let $M=(x,f(x))$ be an $n$-dimensional entire minimal graph given by $m$
functions $f^{\alpha}(x^{1},\cdots,x^{n})$ with $p=\min\\{n,m\\}\leq 4$. If
$\Lambda_{f}=\sum_{i,\alpha}\left(\frac{\partial f^{\alpha}}{\partial
x^{i}}\right)^{2}<2,$
and
(6.5) $\left(2-\Lambda_{f}\right)^{-1}=o(R^{\frac{2(p+2)}{3p}})$
where $R^{2}=|x|^{2}+|f|^{2}$. Then $f^{\alpha}$ has to be affine linear
functions and hence $M$ has to be an affine linear subspace.
###### Proof.
Here we only give the proof of Theorem 6.2, for the proof of Theorem 6.3 is
similar.
From (4.5), it is easily seen that
$h_{2}\leq C(2-v)^{-\frac{3}{2}},$
where $C$ is a positive constant. Thus, for any point $q\in M$, by Theorem 5.3
and Proposition 6.1, we have
$|B|^{2s}(q)\leq C(n,s)R^{-2s}\left(2-v\circ\gamma\right)^{-\frac{3}{2}s}$
Letting $R\to+\infty$ in the above inequality forces $|B(q)|=0$.
∎
###### Remark 6.1.
If $n=2$ or $3$, the conclusion of Theorem 6.1-6.3 could be inferred from the
work done by Chern-Osserman [6], Babosa [1] and Fischer-Colbrie [10].
From (5.13) it is easy to obtain the following result.
###### Theorem 6.4.
Let $M=(x,f(x))$ be an $n$-dimensional entire minimal graph given by $m$
functions $f^{\alpha}(x^{1},\cdots,x^{n})$. Assume $M$ has finite total
curvature. If $\Delta_{f}<2$, or $p\leq 4$ and $\Lambda_{f}<2$, then $M$ has
to be an affine linear subspace.
There are other applications of the strong stability inequalities (4.17),
besides its key role in S-S-Y’s estimates. We state following results, whose
detail proof can be found in the previous paper of the first author [23].
###### Theorem 6.5.
Let $M=(x,f(x))$ be an $n$-dimensional entire minimal graph given by $m$
functions $f^{\alpha}(x^{1},\cdots,x^{n})$. If $\Delta_{f}<2$ or
$\Lambda_{f}<2$, then any $L^{2}$-harmonic $1$-form vanishes.
###### Theorem 6.6.
Let $M$ be one as in Theorem 6.5, $N$ be a manifold with non-positive
sectional curvature. Then any harmonic map $f:M\to N$ with finite energy has
to be constant.
## References
* [1] J.L.M.Babosa: An extrinsic rigidity theorem for minimal immersion from $S^{2}$ into $S^{n}$. J. differntial Geometry 14(3) (1980), 355-368.
* [2] S. Bernstein: Sur un théorème de géométrie et ses application aux équations aux dérivés partielles du type elliptique. Comm. de la Soc Math. de Kharkov (2é sér.) 15 (1915-1917), 38-45.
* [3] E. Bombieri, E. De Giorgi and E. Guusti: Minimal cones and Bernstein problem. Invent. Math. 7(1969), 243-268.
* [4] Yuxin Dong: On graphic submanifolds with parallel mean curvature in Euclidean space. Preprint.
* [5] S. S. Chern: On the curvature of a piece of hypersurfaces in Euclidean space, Abh. Math. Sem. Univ. Hamberg 29 (1965), 77-91.
* [6] S. S. Chern and R. Osserman: Complete minimal surfaces in Euclidean $n-$space. J. d’Anal. Math. 19(1967), 15-34.
* [7] Qing Chen and Senlin Xu: Rigidity of compact minimal submanifolds in a unit sphere. Geom. Dedicata 45 (1)(1993), 83-88.
* [8] S. Y. Cheng, P. Li and S. T. Yau: Heat equations on minimal submanifols and their applications. Amer. J. Math. 103(1981), 1021-1063.
* [9] K. Ecker and G. Huisken: A Bernstein result for minimal graphs of controlled growth. J. Diff. Geom. 31(1990), 397-400.
* [10] D. Fischer-Colbrie: Some rigidity theorems for minimal submanifolds of the sphere. Acta math. 145(1980), 29-46.
* [11] S. Hildebrandt, J. Jost, J and K. O. Widman: Harmonic mappings and minimal submanifolds. Invent. math. 62 (1980), 269-298.
* [12] J. Jost and Y. L. Xin: Bernstein type theorems for higher codimension. Calculus. Var. PDE 9 (1999), 277-296.
* [13] H. B. Lawson and R. Osserman: Non-existence, non-uniqueness and irregularity of solutions to the minimal surface system. Acta math. 139(1977), 1-17.
* [14] An-Min Li and Jimin Li: An intrinsic rigidity theorem for minimal submanifolds in a sphere. Arch. Math.58 (1992), 582-594.
* [15] J. Moser: On Harnack’s theorem for elliptic differential equations. Comm. Pure Appl. Math. 14 (1961), 577-591.
* [16] Lei Ni: Gap theorems for minimal submanifolds in $\mathbb{R}^{n+1}$. Comm. Analy. Geom. 9 (3)(2001), 641-656.
* [17] E. A. Ruh and J. Vilms: The tension field of Gauss map. Trans. Amer. Math. 149(1970), 569-573.
* [18] R. Schoen, L. Simon and S. T. Yau: Curvature estimates for minimal hypersurfaces. Acta Math. 134 (1975), 275-288.
* [19] J. Simons: Minimal varieties in Riemannian manifolds. Ann. Math. 88 (1968), 62-105.
* [20] K. Smoczyk, Guofang Wang and Y. L. Xin: Bernstein type theorems with flat normal bundle. Calc. Var. and PDE. 26(1)(2006), 57-67.
* [21] Y. L. Xin: Geometry of harmonic maps. Birkhäuser PNLDE 23, (1996).
* [22] Yuanlong Xin: Minimal submanifolds and related topics. World Scientific Publ., (2003).
* [23] Y. L. Xin: Bernstein type theorems without graphic condition. Asia J. Math. 9(1)(2005), 31-44.
* [24] Y. L. Xin and Ling Yang: Curvature estimates for minimal submanifolds of higher codimension. arXiv:0709.3686.
| arxiv-papers | 2008-06-26T06:58:03 | 2024-09-04T02:48:56.406317 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Y. L. Xin and Ling Yang",
"submitter": "Yuanlong Xin",
"url": "https://arxiv.org/abs/0806.4247"
} |
0806.4628 | # Quantum construction of a unitary SU(2/1) model of the electro-weak
interactions with 2 Higgs doublets.
Jean Thierry-Mieg
NCBI. NLM. NIH Bldg 38A,
8600 Rockville Pike,
Bethesda, MD 20894, USA.
Tel 1 (301) 435 49 21 Fax 1 (301) 480 92 41 E-mail:
and
Laboratoire de Physique Théorique et d’Astroparticules,
CNRS, Montpellier, France mieg@ncbi.nlm.nih.gov
###### Abstract:
The interactions and even the number of the Higgs scalar fields are not fixed
in the $SU(2)U(1)$ standard model of the electro-weak interactions and the
intrinsically chiral nature of the weak interactions is not explained.
Embedding $SU(2)U(1)$ into the Lie superalgebra $SU(2/1)$ fills these gaps.
The 2 smallest representations of $SU(2/1)$ adequately describe the electron,
neutrino, up and down quarks and correlate their chiralities with their $U(1)$
charges, and the Higgs fields have the quantum numbers of the odd generators.
But so far, there was an apparent conflict with unitarity, because the quark
representation is not Hermitian and the super-Killing metric is not positive
definite. We solve this paradox by assuming the existence of 2 complex Higgs
doublets minimally coupled to the Fermions via the chiral projections of the
odd generators of $SU(2/1)$. We find that Lagrangian induced by the Fermion
loops is unitary, thanks to the balance between the leptons and the quarks
needed to cancel the triangle anomaly and that the super-Jacobi identity
guarantees that the photon remains massless after symmetry breaking. In
addition, the Lagrangian has a classical geometric interpretation in terms of
the curvature of the corresponding Hermitian algebra. Assuming that the
relative strength of the scalar and vector couplings does not depend on the
number of families constrains the mass of the Higgs to
$M^{2}_{H^{0}_{1}}+M^{2}_{H^{0}_{2}}=32/9\;M^{2}_{W}=2\;(107.2GeV)^{2}$.
Contrary to grand-unified or Wess-Zumino supersymmetric models, the $SU(2/1)$
internal superunification does not predict any unobserved particle besides the
2 Higgs doublets.
SU(2/1), superalgebra , chiral Yang Mills , renormalization, Hermitian
algebra, super symmetry, Higgs mass , standard model , chirality , neutrino ,
quark
††preprint: PTA/07-058
## 1 Background
The Weinberg-Salam $SU(2).U(1)$ model of the electro-magnetic and weak
interactions [Wei67, Sa68] is extremely accurate. At high energy, the Fermions
are described as massless $SU(2)$ doublets of left spinors: the electron
neutrino left doublet $(e_{L},\nu_{L})$ with $U(1)$ charge (-1), and the up
and down left quark doublet $(u_{L},d_{L})$ with charge (2/3), accompanied by
right singlets: the right electron $(e_{R})$ with charge (-2) and the right
quarks $(u_{R}),(d_{R})$ with respective charges (4/3) and (-1/3). The right
neutrino is $SU(2)$ and $U(1)$ neutral and drops out from the model.
The corresponding $SU(2)$ and $U(1)$ Yang-Mills gauge fields [YM54] are the
$(W^{+},W^{-})$, the $Z^{0}$ and the photon. The existence and interactions of
the $W^{+}$, the $W^{-}$ and the $Z^{0}$ were beautifully confirmed over the
last 30 years. The only component of the model that has not yet been observed
is the scalar Higgs field. In the minimal approach, it is predicted to be a
scalar $SU(2)$ doublet $\phi$ with $U(1)$ charge (-1). The Higgs potential is
assumed to be of the form $V(\phi)=(\phi^{2}-v^{2})^{2}$ So, in the vacuum the
Higgs field does not vanish. The gauge group breaks down to the electro-
magnetic $U(1)$, the maximal subgroup leaving $v$ invariant. The photon and
the left neutrino remain massless. All other particles acquire a mass
proportional to $v$ (see for example [AL73]).
This construction is superb, but the number of arbitrary parameters is very
large. Since the left and right chiral states are independent, we need to
choose 10 charges to describe the leptons and the quarks. For example, a
massless electrically charged particle, which experimentally does not exist,
would be acceptable at the classical level and is only ruled out by the study
of the quantum anomalies. Also, since the Higgs field plays such a central
role, one would wish to derive it from first principle. The Yang-Mills gauge
fields are well understood, but the Higgs potential and even the number of
Higgs fields needed in the standard model remain arbitrary.
In 1979, Ne’eman [N79] and Fairlie [F79] have independently proposed to
consider the $SU(2/1)$ Lie superalgebra as a basis for describing the weak
interactions. Indeed, the fundamental representation, recalled in section 3,
exactly corresponds to the lepton triplet if we grade it by chirality. The
representation is Hermitian (3.2) and scaling the trace of all the even
matrices to a common value seemed to predict a electro-weak angle
$sin^{2}\theta_{W}=1/4$, a good value at the time, although higher than
today’s experimental value $.22$. In addition, the odd generators have the
quantum numbers of the Higgs fields. The quarks were either left out [F79] or
mentioned as a counterargument [N79]. But remarkably, it was found a few weeks
later [DJ79, NT80] that $SU(2/1)$ also admits a 4 dimensional representation,
recalled in section 4, exactly fitting the quarks, with 2 right singlets and a
left doublet with electric charges $2/3$ and $-1/3$.
These results, summarized in a recent Physics Report [NSF05], raised some
interest but also some doubts [S92]. A superalgebraic structure seems in
direct conflict with unitarity. The supertrace, which is the natural invariant
of a superalgebra, yields a negative sign for the propagator of the $U(1)$
vector, the quark representation (4.3) is neither Hermitian nor anti-
Hermitian, and even the Hermitian nature of the lepton representation is
artefactual, since the odd generators of antilepton representation are anti-
Hermitian (3.8). It was therefore often argued that the $SU(2/1)$ superalgebra
could not play a role in the quantum field theory. The present paper reverses
the situation.
## 2 Results
By following literally the paradigm of the minimal couplings and turning
immediately to the renormalization theory, we predict 2 Higgs doublets $H$ and
$K$. This solves the problem of the metric and allows to adjust freely the
electro-weak angle. We prove that in the symmetry-breaking vacuum the photon
remains massless because of the super-Jacobi identity. For a particular choice
of the scalar coupling constants, we find a scalar Ward identity insuring that
the 1-loop counterterms do not depend on the number of families and derive the
mass relation
$M^{2}_{H^{0}_{1}}+M^{2}_{H^{0}_{2}}=32/9\;M^{2}_{W}=2\;(107.2\;GeV)^{2}$. The
whole construction plays on balance between the leptons and the quarks which
ensures the cancellation of the triangle anomaly [BIM72]. The theory is
however incomplete, because the scalar loops and the gluon loops disturb the
superalgebraic structure.
We start in section 5 from the assumption that the adjoint representation of
the $SU(2/1)$ superalgebra directly describes the Bosons of the theory, and
that the quark and lepton representations, graded by chirality, describe their
interactions to the Fermions. The even generators are gauged as usual as Yang-
Mills vector fields. But since the odd generators are not Hermitian, we
propose (5.3) to separate the left and right interactions and construct 2 sets
of scalar fields, $\Phi_{L}$ and $\Phi_{R}$ by projecting out the relevant
chiral part of the odd matrices, and to define the interactions
${\overline{\Psi}_{L}}\;\Phi_{L}\;\Psi_{R}+{\overline{\Psi}_{R}}\;\Phi_{R}\;\Psi_{L}$.
We then construct the rest of the Lagrangian by studying the counterterms
induced by the Fermion loops.
As shown in (5.4), the left-trace of the odd matrices induces the propagator
of the scalar fields. It is antisymmetric and proportional to the super-
Killing metric, as befits a superalgebra. But since the $\Phi_{L}$ and
$\Phi_{R}$ are distinct, the propagator does not vanish by symmetrization and
can be rediagonalized (5.9) in the form of 2 Higgs scalar complex doublets,
$H$ and $K$. $H$ is coupled to the negatively charged right particles
(electron and down quark) and $K$ to the positively charged right up quark.
This re-shuffling depends on the same balance between the leptons and the
quarks which cancels the triangle anomaly and the equations are similar. In
the later case, one computes the triple vector Fermion loop, which is possibly
spoiled by the impossibility to regularize the chiral Fermions in a gauge
invariant way, and depends on a supertrace condition (4.5). In the present
paper we compute the 2 scalars and 4 scalars Fermion loops, which are a priori
spoiled by the non Hermitian nature of the $SU(2/1)$ scalar-Fermion couplings.
In all cases, we find that the respective contributions of the leptons and the
quarks cancel each other and induce a canonical counterterm.
In section 6, we construct the Higgs potential. A priori, since we have 2
Higgs doublets, there are 3 possible $SU(2)U(1)$ invariant quartic terms
$V_{1}=(H^{2})^{2}+(K^{2})^{2}$ which controls the mass of the neutral Higgs
fields, a kind of vector product squared $V_{2}(H,K)$ which controls the mass
of the charged Higgs fields, and a scalar product squared $V_{3}=(H.K)^{2}$.
We show that $V_{2}$ vanishes if $H$ and $K$ are parallel as a consequence of
the super-Jacobi identity, and that $V_{3}$ is absent, because of the lepton
quark compensation. The potential can then be rewritten in terms of a simple
quadratic form $\widetilde{G}$ (6.2) which plays the role of the Killing
metric and contains a free parameter $\alpha$ later associated (10.3) to the
electro-weak angle. This construction gives the structure of the scalar
potential, but not the strength of the couplings.
In section 7, we show that for a particular value of the scalar Fermion
coupling, we have a Ward identity insuring that the Fermion loops do not
modify the relative strength of the scalar and vector couplings and that the
scalar potential renormalizes like $g^{2}$. This indicates that the mass of
the neutral Higgs fields should satisfy the relation
$M^{2}_{H^{0}_{1}}+M^{2}_{H^{0}_{2}}=32/9\;M^{2}_{W}=(151.6\;GeV)^{2}$. In the
symmetric situation, we would have $M_{H^{0}_{1}}=M_{H^{0}_{2}}=107.2\;GeV$.
However, we also observe that other quantum corrections disrupt the
universality of the scalar-Fermion coupling, indicating that the theory is
still incomplete.
In section 9, we show that the chiral decomposition of the Higgs scalars, in
conjunction with the Dirac 1-forms $\Gamma$ of [TM06], naturally constructs a
covariant differential corresponding to the SU(2/1) Hermitian algebra first
defined by Sternberg and Wolf [SW78]. The chirality operator and the complex
Higgs structure conspire so that the corresponding curvature 2-forms is well
defined and indeed valued in the adjoint representation of the $SU(2)U(1)$
even Lie algebra.
In section 10, we use the quadratic form (6.2) and the 2 CP conjugated
curvatures (9.12-13) to construct the vector Lagrangian and show that the
square of the H-algebra curvature tensor reproduces the Higgs quartic
potential induced by renormalization. Within this $SU(2/1)$ framework, the
electro-weak angle is free.
The theory is not complete and several important problems are presented in the
discussion. But even considering these difficulties, the results so far are
new and unexpected. Close to 30 years after the initial proposal [N79, F79],
this is the first time that $SU(2/1)$ is shown to play a role in the quantized
theory, and the first time that the super-Killing metric is used, both in the
construction of the propagators (5.9 and 10.1) and of the potential (6.2,6.4),
while explicitly respecting unitarity.
## 3 The $SU(2/1)$ lepton representation
The smallest nontrivial simple Lie superalgebra is $SU(2/1)$, also called
$A(1/0)$ in the Cartan-Kac classification. The Lie sub-algebra
$\mathcal{A}_{0}$ is $SU(2)U(1)$, and the odd $\mathcal{A}_{0}$ module
$\mathcal{A}_{1}$ is a complex doublet with $U(1)$ charge $-1$. The
fundamental representation [SNR77] is of superdimension (2/1) and fits the
leptons [N79, F79]. In the $(\nu_{L},\;e_{L};\;e_{R})$ basis, the 4 even
generators read:
$\lambda_{1}=\pmatrix{0&1&0\cr 1&0&0\cr
0&0&0}\,,\;\;\lambda_{2}=\pmatrix{0&-i&0\cr i&0&0\cr
0&0&0}\,,\;\;\lambda_{3}=\pmatrix{1&0&0\cr 0&-1&0\cr
0&0&0}\,,\;\;\lambda_{8}=\pmatrix{1&0&0\cr 0&1&0\cr 0&0&2}\,.$ (1)
the 4 odd generators read:
$\lambda_{4}=\pmatrix{0&0&1\cr 0&0&0\cr
1&0&0}\,,\;\;\lambda_{5}=\pmatrix{0&0&-i\cr 0&0&0\cr
i&0&0}\,,\;\;\lambda_{6}=\pmatrix{0&0&0\cr 0&0&1\cr
0&1&0}\,,\;\;\lambda_{7}=\pmatrix{0&0&0\cr 0&0&-i\cr 0&i&0}\,,$ (2)
and the chirality operator is
$\begin{array}[]{c}\chi=diag(1,1,-1)\,.\end{array}$ (3)
By inspection, we find that the anticommutator of the odd matrices close on
the even matrices, defining the symmetric structure constants $d^{a}_{ij}$ and
that the commutators of the odd matrices close on $\chi$ times the even
matrices, defining the skew-symmetric structure constants $f^{a}_{ij}$:
$\begin{array}[]{c}\lambda_{i}\,\lambda_{j}+\lambda_{j}\,\lambda_{i}=d^{a}_{ij}\,\lambda_{a}\;,\\\
\lambda_{i}\,\lambda_{j}-\lambda_{j}\,\lambda_{i}=i\;\chi\;f^{a}_{ij}\,\lambda_{a}\;,\\\
a=1,2,3,8\,\,;\;\;\;i,j=4,5,6,7\end{array}$ (4)
in these conventions, the $f^{a}_{ij}$ and $d^{a}_{ij}$ constants are real.
Finally, the vacuum $v$ is chosen along the $\lambda_{6}$ direction, whose
centralizer is the photon:
$\hbox{photon}=-(\lambda_{6})^{2}={1\over
2}\;(\lambda_{3}-\lambda_{8})=\pmatrix{0&0&0\cr 0&-1&0\cr 0&0&-1}\,.$ (5)
There is no loss of generality in this choice of $v$, it just corresponds to
the way we named the particles. If we rotate $v$ we just have to rotate the
name of the electron and the neutrino and keep the photon in the direction of
$v^{2}$. Notice that, with respect to the $SU(2/1)$ supermetric :
$\begin{array}[]{c}Str(.)=Tr(\chi\;.)\\\
\widehat{g}_{MN}=1/2\;STr(\lambda_{M}\lambda_{N})\;,\;\;M,N=1,2...8\\\
\widehat{g}_{ab}=diag(1,1,1,-1)\,,\qquad\widehat{g}_{45}=-\widehat{g}_{54}=\widehat{g}_{67}=-\widehat{g}_{76}=i\end{array}$
(6)
the photon is on the light-cone of the superalgebra.
The even generator of antilepton representation, in the
$({\overline{(e_{R})}}_{L};{\overline{(e_{L})}}_{R},{\overline{(\nu_{L})}}_{R})$
basis, are:
$\lambda_{1}=\pmatrix{0&0&0\cr 0&0&1\cr
0&1&0}\,,\;\;\lambda_{2}=\pmatrix{0&0&0\cr 0&0&-i\cr
0&i&0}\,,\;\;\lambda_{3}=\pmatrix{0&0&0\cr 0&1&0\cr
0&0&-1}\,,\;\;\lambda_{8}=\pmatrix{-2&0&0\cr 0&-1&0\cr 0&0&-1}\,.$ (7)
the 4 odd generators:
$\lambda_{4}=\pmatrix{0&0&-1\cr 0&0&0\cr
1&0&0}\,,\;\;\lambda_{5}=\pmatrix{0&0&i\cr 0&0&0\cr
i&0&0}\,,\;\;\lambda_{6}=\pmatrix{0&1&0\cr-1&0&0\cr
0&0&0}\,,\;\;\lambda_{7}=\pmatrix{0&-i&0\cr-i&0&0\cr 0&0&0}\,,$ (8)
and the chirality operator is
$\begin{array}[]{c}\chi=diag(1,-1,-1)\,.\end{array}$ (9)
Equations (3.4-5) are again true in this representation. Notice however that
the odd matrices (3.8) are anti-Hermitian, this is necessary to maintain (3.5)
since the electric charge of the antielectron is positive. Notice also that
the charge and parity (CP) are linked. The doublet and antidoublet must be of
opposite parity with respect to $\chi$ (3.3) and (3.9) in order to maintain
the commutator (3.4b). This switches the sign of the supermetric (3.6), as
befits a CP invariant theory.
## 4 The $SU(2/1)$ quark representation
In the standard model [Wei67, Sa68, GIM70, BIM72], each family of quarks
consists of 4 states, a left doublet and two right singlets, i.e.
$(u_{R}/(u_{L},d_{L})/d_{R})$ for the ’electron family’. Amazingly, the second
smallest irreducible representation of the $SU(2/1)$ superalgebra corresponds
to this chiral decomposition, it contains a free parameter $n$ which allows us
to fix the charge of the up quark. From the mathematical point of view
[SNR77], the existence of this 4 dimensional representation is a simple
consequence of the isomorphism between $SU(2/1)$ and $OSp(2/2)$, which is a
generalization of the well known isomorphisms between the first members of the
infinite families of simple Lie algebras. But from the physical point of view,
it came as a great surprise. In the original papers [F79, N79], the quarks
were respectively left out and listed as a counterargument. The incorporation
a few weeks later of the quarks [DJ79, NT80] really appeared as a significant
confirmation of the model. The 4 even generators read:
$\lambda_{1}=\pmatrix{0&0&0&0\cr 0&0&1&0\cr 0&1&0&0\cr
0&0&0&0}\,,\;\;\lambda_{2}=\pmatrix{0&0&0&0\cr 0&0&-i&0\cr 0&i&0&0\cr
0&0&0&0}\,,\;\;$ (10) $\lambda_{3}=\pmatrix{0&0&0&0\cr 0&1&0&0\cr 0&0&-1&0\cr
0&0&0&0}\,,\;\;\lambda_{8}={1\over n}\;\pmatrix{-n-1&0&0&0\cr 0&-1&0&0\cr
0&0&-1&0\cr 0&0&0&n-1}\,.$ (11)
The 4 odd generators are not Hermitian, for any value of $n$, they read:
$\begin{array}[]{c}\lambda_{4}={1\over\sqrt{2n}}\pmatrix{0&0&-\sqrt{n+1}&0\cr
0&0&0&\sqrt{n-1}\cr\sqrt{n+1}&0&0&0\cr 0&\sqrt{n-1}&0&0}\,,\\\ \\\
\lambda_{5}={1\over\sqrt{2n}}\pmatrix{0&0&i\sqrt{n+1}&0\cr
0&0&0&-i\sqrt{n-1}\cr i\sqrt{n+1}&0&0&0\cr 0&i\sqrt{n-1}&0&0}\,,\\\ \\\
\lambda_{6}={1\over\sqrt{2n}}\pmatrix{0&\sqrt{n+1}&0&0\cr-\sqrt{n+1}&0&0&0\cr
0&0&0&\sqrt{n-1}\cr 0&0&\sqrt{n-1}&0}\,,\\\ \\\
\lambda_{7}={1\over\sqrt{2n}}\pmatrix{0&-i\sqrt{n+1}&0&0\cr-i\sqrt{n+1}&0&0&0\cr
0&0&0&-i\sqrt{n-1}\cr 0&0&i\sqrt{n-1}&0}\,.\end{array}$ (12)
Finally the chirality operator is
$\begin{array}[]{c}\chi=diag(-1,1,1,-1)\,.\end{array}$ (13)
We have normalized the quark matrices (4.1-4.4) so that they have exactly the
same supercommutators as the lepton matrices (3.1,3.2). As desired, the super-
Killing metric (3.6) is also identical in the lepton and the quark
representation. Notice however that the commutators of the odd quark matrices
do not close on $\chi$ times the even matrices.
Since $n$ is a free parameter, $SU(2/1)$ does not predict the charge of the
quarks, or the number of colors. However, the superalgebra does relate the
$U(1)$ charges of the singlets and a number of desired relations (post-
dictions) are automatically satisfied.
First, for $n=-1$ (or $n=1$), we recover the lepton (or antilepton)
representation with only 3 states. In other words, $SU(2/1)$ implies that if
the charge of the electron is equal to the charge of the $W$ vector, the
$U(1)$ neutral right-neutrino is also neutral with respect to the odd
generators.
Then, we should impose the cancellation of the triangle anomalies.
$\begin{array}[]{c}d_{abc}=STr_{f}(\lambda_{a}(\lambda_{b}\lambda_{c}+\lambda_{c}\lambda_{b}))=0\,,\;\;\;a,b,c=1,2,3,8\;,\end{array}$
(14)
If we consider a lepton family, with one lepton of $U(1)$ charge $1$, and
$n^{\prime}$ quarks of $U(1)$ charge $-1/n$, the $SU(2)^{2}U(1)$ anomaly
vanishes if and only if we choose $n=n^{\prime}$. But although there are no
other adjustable parameters, $SU(2/1)$ then implies that the $U(1)^{3}$
anomaly also vanishes
$\begin{array}[]{c}STr_{f}\;(\lambda_{8}^{3})=2-8+n\;((n+1)^{3}-2-(n-1)^{3})/n^{3}=0\end{array}$
(15)
for any value of $n$. The vacuum and the photon direction were already chosen
in the lepton representation. In the same conventions (3.5) where the electric
charge of the electron is -1, we find:
$\hbox{photon}=-(\lambda_{6})^{2}={1\over
2}\;(\lambda_{3}-\lambda_{8})={1\over 2n}\;\pmatrix{n+1&0&0&0\cr 0&n+1&0&0\cr
0&0&1-n&0\cr 0&0&0&1-n}\,.$ (16)
Choosing the correct number of colors $n=3$ implies the correct electric
charges for the $up$ quark $2/3$ and the $down$ quark $-1/3$ . Finally, the
odd sector antisymmetric structure constants and trace metric also vanish:
$\begin{array}[]{c}f_{aij}=Tr_{f}(\lambda_{a}(\lambda_{i}\lambda_{j}-\lambda_{j}\lambda_{i}))=0\,,\;\;\;a=1,2,3,8\,,\\\
g_{ij}={1\over
2}Tr_{f}(\lambda_{i}\lambda_{j})=0\,,\;\;\;i,j=4,5,6,7\,.\end{array}$ (17)
We will now study in the next sections some consequences of these identities.
## 5 The scalar propagator
We would like to construct a quantum field theory, based on the $SU(2/1)$
superalgebra, which would extend the Yang-Mills theory [YM54] associated to
its maximal $SU(2)U(1)$ Lie sub-algebra and incorporate in some way the odd
matrices. In [TM06], we have shown that we can construct an associative
covariant exterior differential, mixing left and right chiral Fermions, if and
only if the Fermion form a representation of a Lie superalgebra, graded by the
chirality. As we have just seen, these conditions are met exactly by the
existing fundamental Fermions: the leptons and the quarks. The connection is
then defined as the sum of the usual Yang-Mills 1-form $A_{\mu}\;dx^{\mu}$
plus the constant Dirac 1-form $\gamma_{\mu}\;dx^{\mu}$ multiplied by scalar
field valued in the odd sector of the superalgebra.
The natural idea would therefore be to introduce a scalar-field $\Phi$ with a
minimal coupling to the Fermions of the form
$\Psi^{\dagger}\;\Phi^{i}\lambda_{i}\;\Psi$. However, this is not directly
possible since the $\lambda_{i}$ are not Hermitian. To fix this problem, we
postulate that the field $\Phi^{i}$ is composed of 2 parts, each with
intrinsically chiral couplings. Considering the chirality operator $\chi$
(3.3,4.4) with eigenvalue $1$ $(-1)$ on the left (right) Fermions, we define
the left and right projectors $p_{L}=(1+\chi)/2$ and $p_{R}=(1-\chi)/2$, and
define, for any matrix M, the chiral traces.
$\begin{array}[]{c}Tr_{L}(M)=Tr(p_{L}\;M)\;,\;\;Tr_{R}(M)=Tr(p_{R}\;M)\;,\;\;Str(M)=Tr(\chi\;M)\;.\end{array}$
(18)
We then decompose the odd $\lambda$ matrices as
$\begin{array}[]{c}\lambda_{i}=\lambda_{iL}+\lambda_{iR}\\\
\lambda_{iL}=p_{L}\;\lambda_{i}=\lambda_{i}\;p_{R}\;,\;\;\lambda_{iR}=p_{R}\;\lambda_{i}=\lambda_{i}\;p_{L}\;\;,\end{array}$
(19)
and assume the existence of 2 scalar fields $\Phi_{L}$ which absorbs right
Fermions and emits left ones and $\Phi_{R}$ which absorbs left Fermions and
emits right ones, according to the coupling
$\begin{array}[]{c}\Psi^{\dagger}_{L}\;\Phi^{i}_{L}\lambda_{iL}\;\Psi_{R}\;+\Psi^{\dagger}_{R}\;\Phi^{i}_{R}\lambda_{iR}\;\Psi_{L}\;.\end{array}$
(20)
To find the correct propagator of the $\Phi$ field, we compute the quantum
corrections due to the insertion of a Fermion loop between 2 $\Phi$ states.
This induces a scalar counterterm of the form
$\begin{array}[]{c}\mathcal{L}_{s}=\eta_{ij}\;\partial_{\mu}\Phi^{i}_{L}\;\partial_{\mu}\Phi^{j}_{R}\end{array}$
(21)
where
$\begin{array}[]{c}\eta_{ij}={1\over
n+1}Tr_{L}(\lambda_{i}\;\lambda_{j})={1\over
n+1}Tr_{R}(\lambda_{j}\;\lambda_{i})\,.\end{array}$ (22)
There are two contributions, one coming from the lepton loop and one from the
quark loop
$\begin{array}[]{c}(\eta_{ij})_{lepton}={1\over n+1}\delta_{ij}+{i\over
n+1}\;\epsilon_{ij}\;,\\\ (\eta_{ij})_{quark}=-{1\over
n(n+1)}\delta_{ij}+{i\over n+1}\;\epsilon_{ij}\,,\end{array}$ (23)
where $\epsilon_{45}=-\epsilon_{54}=\epsilon_{67}=-\epsilon_{76}=1$, all other
components being zero, which coincides with the odd part of (3.6). Summing
over a whole family $f$ with $n$ colored quarks, we find that the Fermion loop
is proportional to the canonical supermetric of $SU(2/1)$:
$\begin{array}[]{c}(\eta_{ij})_{f}=i\;\epsilon_{ij}\;=\widehat{g}_{ij}\end{array}$
(24)
For the moment, we have a weird scalar model, with an antisymmetric propagator
and non Hermitian couplings. But now a miracle occurs. If we change variables
as follows, the Lagrangian becomes canonical. Consider the two $SU(2/1)$ odd
valued scalar fields $H$ and $K$:
$\begin{array}[]{c}\Phi_{iL}=a_{i}+i\epsilon_{ij}b_{j}\;,\;\;a_{i}={1\over
2}\;(H_{i}+i\;K_{i})\;,\\\ \\\
\Phi_{iR}=a_{i}-i\epsilon_{ij}b_{j}\;,\;\;b_{i}={1\over
2}\;(H_{i}-i\;K_{i})\;,\end{array}$ (25)
where $a_{i}$ and $b_{i}$ are convenient intermediate variables which are not
reused below. Substituting (5.8) in (5.4) we obtain the canonical propagator
$\begin{array}[]{c}\mathcal{L}_{s}=i\;\epsilon_{ij}\;\partial_{\mu}\Phi^{i}_{L}\;\partial_{\mu}\Phi^{j}_{R}\;={1\over
2}\delta_{ij}\;(\partial_{\mu}H^{i}\;\partial_{\mu}H^{j}\;+\partial_{\mu}K^{i}\;\partial_{\mu}K^{j}\;)\,.\end{array}$
(26)
We then define the $\mu^{\pm}$ Hermitian matrices:
$\begin{array}[]{c}\lambda_{i}=(\mu^{-}_{i}-i\;\mu^{+}_{i})\,,\qquad\mu^{-}_{i}={1\over
2}\;(\lambda_{i}+\lambda_{i}^{\dagger})\,,\\\
\lambda_{i}^{\dagger}=(\mu^{-}_{i}+i\;\mu^{+}_{i})\,,\qquad\mu^{+}_{i}={i\over
2}\;(\lambda_{i}-\lambda_{i}^{\dagger})\,,\end{array}$ (27)
and rewrite the scalar-Fermion interaction in terms of these new variables:
$\Psi^{\dagger}\;(\Phi^{i}_{L}\;\lambda_{iL}+\Phi^{i}_{R}\;\lambda_{iR})\;\Psi=\Psi^{\dagger}\;(H^{i}\;\mu^{-}_{i}+K^{i}\;\mu^{+}_{i})\;\Psi\,.$
(28)
The matrices $\mu^{\pm}_{i}$ split according to the electric charge and are
Hermitian. The field $H^{i}$ interacts only with the negatively charged right
singlets (electron, down quark) and the field $K^{i}$ just with the positively
charged right singlets (up quark). In the lepton representation, the
$\mu^{-}_{i}$ matrices are equal to the Hermitian matrices given in (3.2)
$\begin{array}[]{c}\mu^{-}_{4}=\pmatrix{0&0&1\cr 0&0&0\cr
1&0&0}\,,\;\;\mu^{-}_{5}=\pmatrix{0&0&-i\cr 0&0&0\cr i&0&0}\,,\\\ \\\
\mu^{-}_{6}=\pmatrix{0&0&0\cr 0&0&1\cr
0&1&0}\,,\;\;\mu^{-}_{7}=\pmatrix{0&0&0\cr 0&0&-i\cr 0&i&0}\,.\end{array}$
(29)
In the quark representation, the $\mu^{-}_{i}$ are equal to the right and low
Hermitian corner of the matrices (4.3)
$\begin{array}[]{c}\mu^{-}_{4}=\sqrt{n-1\over 2n}\;\pmatrix{0&0&0&0\cr
0&0&0&1\cr 0&0&0&0\cr 0&1&0&0}\,,\;\;\mu^{-}_{5}=\sqrt{n-1\over
2n}\;\pmatrix{0&0&0&0\cr 0&0&0&-i\cr 0&0&0&0\cr 0&i&0&0}\,,\\\ \\\
\mu^{-}_{6}=\sqrt{n-1\over 2n}\;\pmatrix{0&0&0&0\cr 0&0&0&0\cr 0&0&0&1\cr
0&0&1&0}\,,\;\;\mu^{-}_{7}=\sqrt{n-1\over 2n}\;\pmatrix{0&0&0&0\cr 0&0&0&0\cr
0&0&0&-i\cr 0&0&i&0}\,.\end{array}$ (30)
The top left corner of the $\mu^{+}_{i}$ matrices are proportional to the
antielectron matrices (3.8)
$\begin{array}[]{c}\mu^{+}_{4}={\sqrt{n+1\over 2n}}\;\pmatrix{0&0&i&0\cr
0&0&0&0\cr-i&0&0&0\cr 0&0&0&0}\,,\;\;\mu^{+}_{5}={\sqrt{n+1\over
2n}}\;\pmatrix{0&0&1&0\cr 0&0&0&0\cr 1&0&0&0\cr 0&0&0&0}\,,\\\ \\\
\mu^{+}_{6}={\sqrt{n+1\over 2n}}\;\pmatrix{0&-i&0&0\cr i&0&0&0\cr 0&0&0&0\cr
0&0&0&0}\,,\;\;\mu^{+}_{7}={\sqrt{n+1\over
2n}}\;\pmatrix{0&-1&0&0\cr-1&0&0&0\cr 0&0&0&0\cr 0&0&0&0}\,.\end{array}$ (31)
If we sum over a complete family, we find that the matrices associated to $H$
and $K$ have the same trace-metric:
$Tr(\mu^{+}_{i}\mu^{-}_{j})=0\;,\;\;Tr(\mu^{+}_{i}\mu^{+}_{j})=Tr(\mu^{-}_{i}\mu^{-}_{j})=(n+1)\;\delta_{ij}\,,\qquad
i,j=4,5,6,7\,.$ (32)
This identity, like (4.8), is a consequence of the vanishing of the sum of the
charges of all the right singlets (electron + n up quarks + n down quarks) of
a family. It shows that one can compute the renormalization of the scalar
propagator either using the $\Phi_{L},\;\Phi_{R}$ fields (5.4) or using the
$H,\;K$ fields (5.8), and obtain the same result (5.9).
## 6 The scalar potential
From now on, we restrict our attention to the case $n=3$ colors. Consider the
structure of the scalar potential induced by a Fermion loop. The cubic terms
vanish, since each interaction with a scalar changes the chirality of the
Fermions. The term of degree 6 or higher converge. So we only have to compute
the quartic potential. If we work in terms of the $\Phi$ fields, the
counterterm is proportional to
$\begin{array}[]{c}\Phi^{i}_{L}\;\Phi^{j}_{R}\;\Phi^{k}_{L}\;\Phi^{l}_{R}\;\;Tr_{L}(\lambda_{i}\;\lambda_{j}\;\lambda_{k}\;\lambda_{l}\;)\;.\end{array}$
(33)
Generalizing (5.6), we find that many unfriendly terms, for example
$\Phi^{4}_{L}\;\Phi^{4}_{R}\;\Phi^{4}_{L}\;\Phi^{5}_{R}$, are eliminated by
the balance between the leptons and the quarks. If we introduce the metric:
$\widetilde{G}_{ab}=\pmatrix{2\delta_{ab}-\alpha\;\widehat{g}_{ab}&-\widehat{g}_{ab}\cr-\widehat{g}_{ab}&2\delta_{ab}-\alpha\;\widehat{g}_{ab}}\;,$
(34)
and the products
$\begin{array}[]{c}Z^{a}_{L}=(f^{a}_{ij}-i\;d^{a}_{ij})\;\Phi^{i}_{L}\,\Phi^{j}_{R}\;,\\\
Z^{a}_{R}=(f^{a}_{ij}+i\;d^{a}_{ij})\;\Phi^{i}_{L}\,\Phi^{j}_{R}\;.\end{array}$
(35)
where the $d^{a}_{ij}$ and the $f^{a}_{ij}$ are defined in (3.4), the direct
calculation of the Fermion loop induces, up to an infinite multiplicative
constant, a scalar potential of the form
$\begin{array}[]{c}V={1\over
8}\;\pmatrix{Z^{a}_{L}&Z^{a}_{R}}\;\widetilde{G}_{ab}\;\pmatrix{Z^{a}_{L}\cr
Z^{a}_{R}}\;.\end{array}$ (36)
The $\alpha$ parameter is arbitrary in (6.4), because we have the identity
$Z^{a}_{L}\;\widehat{g}_{ab}\;Z^{b}_{L}=0\;.$ (37)
If we work with the $\mu^{\pm}$ Hermitian matrices (5.12-14), we find,
generalizing (5.15), that the symmetrized quartic traces are identical
$\begin{array}[]{c}Tr((\mu^{+}_{i}\mu^{+}_{j}+\mu^{+}_{j}\mu^{+}_{i})(\mu^{+}_{k}\mu^{+}_{l}+\mu^{+}_{l}\mu^{+}_{k}))\\\
=Tr((\mu^{-}_{i}\mu^{-}_{j}+\mu^{-}_{j}\mu^{-}_{i})(\mu^{-}_{k}\mu^{-}_{l}+\mu^{-}_{l}\mu^{-}_{k}))\\\
=32/9\;(\delta_{ij}\delta_{kl}+\delta_{ik}\delta_{jl}+\delta_{il}\delta_{jk})\;.\end{array}$
(38)
Finally the mixed traces satisfy, for all $i,j,k,l$:
$\begin{array}[]{c}Tr(\mu^{+}_{i}\mu^{-}_{j})=Tr(\mu^{+}_{i}\mu^{-}_{j}\mu^{-}_{k}\mu^{-}_{l})=Tr(\mu^{+}_{i}\mu^{+}_{j}\mu^{+}_{k}\mu^{-}_{l})=Tr(\mu^{+}_{i}\mu^{-}_{j}\mu^{+}_{k}\mu^{-}_{l})=0\,,\\\
\\\ Tr(\mu^{+}_{i}\mu^{+}_{i}\mu^{-}_{i}\mu^{-}_{i})=0\;,\qquad\hbox{for any
i, no sum implied.}\end{array}$ (39)
If we now introduce the symmetric product
$\begin{array}[]{c}(X*X)^{a}=d^{a}_{ij}\;X^{i}\,X^{j}\;;\;\;\;X=H,\;K\;,\end{array}$
(40)
the direct calculation of the Fermion loop induces, up to the same infinite
multiplicative constant, a scalar potential of the form
$V={1\over
16}\;\pmatrix{(H*H)^{a}&(K*K)^{a}}\pmatrix{\widetilde{G}_{ab}}\pmatrix{(H*H)^{b}\cr(K*K)^{b}}\;.$
(41)
The $\alpha$ parameter (6.2) is arbitrary in (6.9), because in any Lie
superalgebra, we have the identity
$\widehat{g}_{ab}(X*X)^{a}\;(X*X)^{b}=0$ (42)
Notice that for the same reason, the mixed $H\;K$ term vanishes if and only if
$H$ and $K$ are parallel. This identity insures that the centralizer of the
true vacuum is nontrivial, or in other words, that the photon remains
massless.
If we change variable again and define
$\begin{array}[]{c}u_{1}=(H_{4}+iH_{5})/\sqrt{2}\;,\\\
u_{2}=(H_{6}+iH_{7})/\sqrt{2}\;,\\\ v_{1}=(K_{4}+iK_{5})/\sqrt{2}\;,\\\
v_{2}=(K_{6}+iK_{7})/\sqrt{2}\;,\end{array}$ (43)
and substitute in (5.9) and (6.9), we recover up to a scale factor $\sigma$
the propagator and potential of Fayet [F74, F75]
$\begin{array}[]{c}\mathcal{L}_{s}+V=\partial_{\mu}\overline{u}\;\partial_{\mu}u+\partial_{\mu}\overline{v}\;\partial_{\mu}v\;+\\\
+{\sigma}((\overline{u}u)^{2}+(\overline{v}v)^{2}+(\overline{u}_{1}\overline{v}_{2}-\overline{u}_{2}\overline{v}_{1})(u_{1}v_{2}-u_{2}v_{1}))\;.\end{array}$
(44)
In these variable, it is immediately clear that, assuming a symmetry breaking
negative mass term for the scalars, the potential is minimal when $u$ and $v$
are parallel, and therefore the photon remains massless. The scalar potential
(6.9) is invariant under the infinitesimal transformation
$\begin{array}[]{c}\Delta\;H_{i}=(H^{k}\;\delta_{kl}\;K^{l})\;\widehat{g}_{ij}\;K^{j}-(H^{k}\;\widehat{g}_{kl}\;K^{l})\;\delta_{ij}\;K^{j}\;\;,\\\
\Delta\;K_{i}=(H^{k}\;\delta_{kl}\;K^{l})\;\widehat{g}_{ij}\;H^{j}-(H^{k}\;\widehat{g}_{kl}\;K^{l})\;\delta_{ij}\;H^{j}\;\;,\end{array}$
(45)
or, if we use the variables $u$ and $v$:
$\begin{array}[]{c}\Delta\;u=(u.{\overline{v}})\;v\;\;,\;\;\;\Delta\;v=-(v.{\overline{u}})\;u\;\;,\\\
\Delta\;{\overline{v}}=(u.{\overline{v}})\;{\overline{u}}\;\;,\;\;\;\Delta\;{\overline{u}}=-(v.{\overline{u}})\;{\overline{v}}\;\;,\end{array}$
(46)
whereas the term $(u.v)({\overline{u}}.{\overline{v}})$ which could give a
mass to the photon and is absent from our Lagrangian, is not invariant under
$\Delta$.
## 7 Scalar Ward identities
The study of the counterterms induced by the Fermion loop has given in the
previous section the structure of the propagator and of the potential, but not
their scale. If we write our Lagrangian as
$\begin{array}[]{c}\mathcal{L}=-{1\over 4}(F^{a}_{\mu\nu})^{2}-{1\over
2}((D_{\mu}H)^{2}+(D_{\mu}K)^{2})+\lambda\;{g^{2}}\;V+{\overline{\Psi}}\;{g\over
2}\;(\gamma^{\mu}\;A_{\mu}+\xi\Phi\;)\;\Psi\;,\end{array}$ (47)
The factor $2$ in $g/2$ is needed because our $SU(2)$ matrices (3.1) are
normalized as usual in mathematics, not as the standard spin $1/2$ isospin
matrices of physics. The weak angle and the coefficients $\lambda$ and $\xi$
are not yet known. We will now fix them by requiring that they do not depend
on the number of lepton families.
The only contribution of the other families to the coupling of the vectors and
scalars to the Fermion of the first family is through the renormalization of
the wave function of the vector and the scalar. The counterterms are
proportional to
$\begin{array}[]{c}Z(A_{a}^{2})={2\over 3}\;{g^{2}\over
4}\;Tr_{f}\;\lambda_{a}^{2}\;\;,\;\;\;Z(A_{8}^{2})=tg^{2}\theta\;{2\over
3}{g^{2}\over
4}\;Tr_{f}\;\lambda_{a}^{2}\;\;,\;\;\;Z(H^{2})=\xi^{2}\;{g^{2}\over
4}\;Tr_{f}\;\mu_{i}^{2}\;\;.\end{array}$ (48)
The factor $2/3$ in the vectors comes from the kinematic contribution of the
Dirac matrices. We count $2/3$, not the usual $4/3$ of a massive Fermion,
since here the left and right spinors contribute separately to the trace term.
Over a full family, the $SU(2)$ even matrices (3.1,4.1) are normalized to
$Tr_{f}(\lambda_{a}\;\lambda_{b})=8\delta_{ab}\;$. The $U(1)$ matrix is
normalized to $Tr_{f}(\lambda_{8}\;\lambda_{8})=40/3$, and the odd matrices to
$Tr_{f}(\mu^{\pm}_{i}\;\mu^{\pm}_{j})=4\;\delta_{ij}$. To integrate these 3
counterterms in the renormalization of the coupling constant $g$, and insure
that $tg\theta$ and $\xi$ are not renormalized, we fix
$\begin{array}[]{c}tg^{2}\theta={3\over 5}\;\;,\;\;\;sin^{2}\theta={3\over
8}\;\;,\;\;\;\xi={2\over\sqrt{3}}\;\;.\end{array}$ (49)
The weak angle coincides with the value derived from the $SU(5)$ grand unified
theory [GG74]. We then look at the renormalization of the scalar potential.
The standard Lie algebra Ward identities ensure that the contribution of the
Fermions to the self coupling of the vectors, the $g^{2}\;A^{4}$ term, can be
reabsorbed in the renormalization of $g^{2}$. A similar condition holds for
the scalar potential $\lambda\;g^{2}\;H^{4}$.
$\begin{array}[]{c}Z(V)={1\over 3}\,{(g\xi)^{4}\over\lambda g^{2}}\end{array}$
(50)
The contribution of the Fermions is reabsorbed in the renormalization of
$g^{2}$ and therefore $\lambda$ is not renormalized if we choose
$\begin{array}[]{c}Z(V)=Z(H^{2})==>\lambda={\xi^{2}\over 3}={4\over
9}\end{array}$ (51)
However, we note that the Bosonic counterterms, induced by the vector and
scalar 1-loops perturb all the scalar couplings, except the very structure of
the scalar potential which guarantees the survival of an unbroken $U(1)$ and
the existence of the massless photon. Furthermore the strong interactions
gluon vector fields interact with the quarks, but not with the leptons. The
photon-quark coupling is protected by a Ward identity and is not renormalized
by the gluon-quark interaction, but the scalar-quark vertex is not protected.
Hence, the balance between quarks and leptons, which is central to our
analysis is broken by these counterterms. This problem can either be
interpreted as a fatal flaw of the $SU(2/1)$ model or as sign that the quantum
theory is still incomplete.
## 8 The mass of the Higgs
Since we have now fixed the scale of the potential we can derive the mass of
the physical scalars, the two neutral fields $H^{0}_{1or2}$ and the charged
field $H^{\pm}$. Suppressing the Fermions, the Lagrangian reads:
$\begin{array}[]{c}\mathcal{L}=-{1\over 4}(F^{a}_{\mu\nu})^{2}-{1\over
2}((D_{\mu}H)^{2}+(D_{\mu}K)^{2})+{\lambda}\;V\end{array}$ (52)
where $V$ is given in (6.9). If we develop the potential and show explicitly
the occurrences of the component fields $H_{4}$. We find:
$\begin{array}[]{c}{\lambda}\;V={\lambda\over
4}\;(H^{4}_{4}+H^{2}_{4}\;(2\;(H^{2}_{5}+H^{2}_{6}+H^{2}_{7})+K^{2}_{6}+K^{2}_{7})+\\\
+2\;H_{4}H_{6}\;(K_{4}K_{6}+K_{5}K_{7})+2\;H_{4}H_{7}\;(K_{4}K_{7}+K_{5}K_{6}))+...\end{array}$
(53)
Three points should be noticed. The $\lambda$ parameter is scaled as usual to
give $\lambda/4\;H^{4}_{4}$. There are no terms $H^{2}_{4}\;K^{2}_{4}$ or
$H^{2}_{4}\;K^{2}_{5}$, finally the coefficients of $H_{4}$ are invariant
under a $\lambda_{8}$ rotation of the $K$ fields, which rotates $K_{4}$ into
$K_{5}$ and $K_{6}$ into $K_{7}$.
We now add by hand in the Lagrangian a quadratic symmetry breaking ’imaginary
mass’ term
$\begin{array}[]{c}-{\lambda\over
2}\;v^{2}\;(sin^{2}\beta\;H^{2}+cos^{2}\beta\;K^{2})\;.\end{array}$ (54)
We assume that the vacuum expectation value of $H$ is in the direction
$H_{6}$. As discussed in section 3, this choice just fixes the name of the
particles. To minimize the mixed $H^{2}\;K^{2}$ potential, the $K$ vacuum
expectation value must be in the plane $(K_{6},K_{7})$. The vacuum correctly
preserves the photon (3.5, 4.7). We then rotate $<K>$ to $<K_{6}>$. This
’phase’ invariance corresponds to the fact that one of the 4 neutral states in
the hyper-plane $(H_{6},H_{7},K_{6},K_{7})$, usually called the ’$A_{0}$’
pseudo-scalar, is absent from the quartic potential and its mass can be fixed
arbitrarily in the symmetry-breaking term. We minimize the potential and find:
$\begin{array}[]{c}<H_{6}>=v\;sin\beta\;\;,\;\;\;<K_{6}>=v\;cos\beta\;\;,\;\;\;M_{W}=v\;g/2\;,\end{array}$
(55)
where $\beta$ is defined modulo $\pi/2$. We then expand the potential around
the vacuum to second order in $v$. We find 4 massless states which produce the
$A_{0}$ and the longitudinal components of the massive vector fields $Z$ and
$W^{\pm}$
$\begin{array}[]{c}H_{7}\;,\;\;K_{7}\;,\;\;(sin\beta\;H_{4}+cos\beta K_{4})\pm
i(sin\beta\;H_{5}+cos\beta K_{5})\;,\end{array}$ (56)
two charged physical scalars
$\begin{array}[]{c}H^{\pm}=(cos\beta\;H_{4}-sin\beta K_{4})\pm
i(cos\beta\;H_{5}-sin\beta K_{5})\;,\\\
M_{H^{\pm}}={\sqrt{2\;\lambda}}\;M_{W}=75.8\;GeV\end{array}$ (57)
and two neutral scalars in the $(H_{6},K_{6})$ plane, usually called
$H^{0}_{1}$ and $H^{0}_{2}$, with masses
$\begin{array}[]{c}M_{H^{0}_{1}}={2\;\sqrt{2\;\lambda}}\;cos\beta\;M_{W}=({\sqrt{2}}cos\beta)\;107.2\;GeV\;,\\\
M_{H^{0}_{2}}={2\;\sqrt{2\;\lambda}}\;sin\beta\;M_{W}=({\sqrt{2}}sin\beta)\;107.2\;GeV\;.\end{array}$
(58)
In the symmetric case $\beta=\pi/4$ the 2 neutral scalar are at $107.2\;GeV$.
## 9 The $SU(2/1)$ hyper-curvature
Since the scalar fields (5.3) play the role of a gauge field associated to the
odd generators of the superalgebra, it seems natural to introduce them as part
of the covariant differential and curvature 2-form. Such a construction was
proposed in [TM06], where we found that the covariant differential is
associative if and only if the Fermions sit in a representation of the
superalgebra graded by the chirality. As shown in sections 2 and 3, this
condition is met by the observed leptons and quarks. But we found in section 3
the additional complication that the quark representation is non Hermitian, so
the scalar field is complex, so effectively, the number of odd generators in
the quantum field theory is twice the number of odd generators of the
$SU(2/1)$ superalgebra.
We are going to show that the construction of the curvature can be generalized
and yields a structure we would like to call a H-curvature, by reference to
Fayet and Iliopoulos [F74, F75], who were the first to consider, in the
context of Hyper-Symmetry, a doubling of the Higgs field introduced by
Weinberg [Wei67] and to Sternberg and Wolf [SW78] who introduced the Hermitian
algebras.
Following [TM06], we define the Dirac 1-forms
$\begin{array}[]{c}\Gamma=dx^{\mu}\;\sigma_{\mu}\
+dx^{\mu}\;\overline{\sigma}_{\mu}\,.\end{array}$ (59)
where $\sigma$ maps the right Fermions on the left Fermions and vice versa. We
then define the chiral Yang-Mills 1-form $\widetilde{A}$ and covariant
differential $\widetilde{D}$
$\begin{array}[]{c}\widetilde{A}=A+\Phi_{L}+\Phi_{R}=dx^{\mu}\;A^{a}_{\mu}\;\lambda_{a}+\Gamma\;\Phi^{i}_{L}\lambda_{iL}+\Gamma\;\Phi^{i}_{R}\lambda_{iR}\qquad,\widetilde{D}=d+\widetilde{A}\;.\end{array}$
(60)
Notice that $\widetilde{A}$ is an exterior form of homogeneous degree 1, in
contradistinction with the Ne’eman Quillen connection [TMN82, Q85] sometimes
discussed in the $SU(2/1)$ literature ([NS90, NSF05] and references therein).
The curvature is defined as the square of the covariant differential
$\begin{array}[]{c}\widetilde{F}=\widetilde{D}\;\widetilde{D}=d\widetilde{A}+\widetilde{A}\;\widetilde{A}\;.\end{array}$
(61)
Since the $dx^{\mu}$ anticommute, the classical term
$\begin{array}[]{c}A\;A=dx^{\mu}dx^{\nu}\;A^{a}_{\mu}\;A^{b}_{\nu}\;\lambda_{a}\lambda_{b}\end{array}$
(62)
is antisymmetric in $ab$ and is internal provided the even matrices
$\lambda_{a}$ close by commutation
$\begin{array}[]{c}\lambda_{a}\lambda_{b}-\lambda_{b}\lambda_{a}=i\;f^{c}_{ab}\;\lambda_{c}\;.\end{array}$
(63)
Similarly, the scalar vector term in $\Gamma\;D\Phi$ generate the covariant
differential of the scalars provided the even odd sector close by commutation
$\begin{array}[]{c}\lambda_{a}\lambda_{i}-\lambda_{i}\lambda_{a}=i\;f^{j}_{ai}\;\lambda_{j}\;.\end{array}$
(64)
However, the structure of the scalar-scalar term is more involved. Since
$\Phi_{L}$ and $\Phi_{R}$ are independent, the $\lambda$ matrices are neither
symmetrized nor skew symmetrized and we get
$\begin{array}[]{c}{1+\chi\over
2}\;\Phi^{i}_{L}\Phi^{j}_{R}\;\lambda_{i}\lambda_{j}+{1-\chi\over
2}\;\Phi^{i}_{R}\Phi^{j}_{L}\;\lambda_{i}\lambda_{j}\;.\end{array}$ (65)
We now decompose the matrix product into the symmetric and skew symmetric
parts
$\begin{array}[]{c}\lambda_{i}\lambda_{j}={1\over
2}(\\{\lambda_{i},\lambda_{j}\\}+[\lambda_{i},\lambda_{j}])\end{array}$ (66)
and we find that the scalar-scalar term can be expressed as a linear
combination of even generators provided the anticommutator of the odd matrices
close on the even matrices and the commutators of the odd matrices close on
$\chi$ times the even matrices, defining the skew-symmetric structure
constants $f^{a}_{ij}$:
$\begin{array}[]{c}\lambda_{i}\,\lambda_{j}+\lambda_{j}\,\lambda_{i}=d^{a}_{ij}\,\lambda_{a}\;,\\\
\lambda_{i}\,\lambda_{j}-\lambda_{j}\,\lambda_{i}=i\;\chi\;f^{a}_{ij}\,\lambda_{a}\;.\end{array}$
(67)
Therefore the unsymmetrized product of the odd matrices satisfy
$\begin{array}[]{c}\lambda_{i}\,\lambda_{j}={1\over
2}(d^{a}_{ij}+i\;\chi\;f^{a}_{ij})\lambda_{a}\;.\end{array}$ (68)
By inspection (3.4), all these conditions are satisfied by the lepton
representation (3.1-3). In fact, the construction works with the fundamental
representation of any $SU(m/n)$ superalgebra, because all the generators
coincide with the generators of $SU(m+n)$ except for the supertraceless $U(1)$
but $\chi\;U(1)$ is traceless and coincides with the $U(1)$ of $SU(m+n)$.
The structure (9.9) was introduced by Sternberg and Wolf [SW78], who call it a
Hermitian algebra and discussed by Ne’eman and Sternberg in several subsequent
papers [NS90, NSF05], but its exact relevance to quantum field theory remained
unclear. The unusual presence of the chirality operator $\chi$ on the right
hand side of the commutator (9.9) was not explained, because they did not
consider the possibility of splitting the odd generators and doubling the
number of Higgs fields. But in our modified framework, if we compare (9.7) and
(9.10) and use $\chi^{2}=1$, we see that the $\chi$ operator present in (9.9)
disappears from the curvature 2-form which is now correctly valued on the even
matrices and can be written as
$\begin{array}[]{c}\widetilde{F}_{L}=\widehat{F}^{a}_{L}\;\lambda_{a}+\Gamma\;D\Phi^{i}_{L}\;\lambda_{iL}+\Gamma\;D\Phi^{i}_{R}\;\lambda_{iR}\;,\\\
\widehat{F}^{a}_{L}=dA^{a}+{i\over 2}f^{a}_{bc}\;A^{b}A^{c}+{i\over
2}\;\Gamma\Gamma\;\Phi^{i}_{L}\Phi^{j}_{R}\;(f^{a}_{ij}-i\;d^{a}_{ij})\;.\end{array}$
(69)
The Bianchi identity
$\widetilde{D}\;\widetilde{F}=(\widetilde{D}\widetilde{D})\;\widetilde{D}-\widetilde{D}\;(\widetilde{D}\widetilde{D})=0$
follows from the associativity of the matrix product and implies the Hermitian
Jacobi identity of [SW78]. We can also construct a second solution
$\widehat{F}_{R}$ by switching the sign of the $d$ constants changing the
chirality:
$\begin{array}[]{c}\widehat{F}^{a}_{R}=dA^{a}+{i\over
2}f^{a}_{bc}\;A^{b}A^{c}+{i\over
2}\;\Gamma\Gamma\;\Phi^{i}_{L}\Phi^{j}_{R}\;(f^{a}_{ij}+i\;d^{a}_{ij})\;.\end{array}$
(70)
We recover in (9.11-12) the object $Z_{L/R}$ introduced in (6.3)
$\begin{array}[]{c}\widehat{F}_{L/R}=F+{i\over
2}\;\Gamma\Gamma\;\;Z_{L/R}\end{array}$ (71)
## 10 The electro-weak angle
From the definition of the classical curvature 2-form given in the previous
section, we can construct the Lagrangian in the usual way, using the Hodge
adjunction ∗ and the quadratic form $\widetilde{G}_{ab}$ (6.2) which
reproduces the scalar potential induced by the Fermion loops. The Lagrangian
can be written as:
${\mathcal{L}}_{F}={1\over
8\;g^{2}}Tr\;\pmatrix{{}^{*}\widehat{F}^{a}_{L}&{}^{*}\widehat{F}^{a}_{R}}\widetilde{G}_{ab}\pmatrix{\widehat{F}^{b}_{L}\cr\widehat{F}^{b}_{R}}$
(72)
Since the new bilinear scalar term in the H-curvature (9.13) coincides with
(6.3), the scalar potential is proportional to (6.4) which can be rewritten as
(6.9). It is curious to see that the combined effect of the leptons and quarks
in (6.9) is mimicked in the classical construction (10.1) by combined effect
of the 2 chiral ways (9.11-12) of extending the superalgebra $SU(2/1)$
curvature of [TM06] to the H-algebra.
The interesting point is that, using (6.2) and (3.6), the effective metric for
the Yang-Mill curvature $F^{a}$ comes out as
$2\;g_{ab}-(1+\alpha)\;\widehat{g}_{ab}=\hbox{diag}(1-\alpha,1-\alpha,1-\alpha,3+\alpha)\,,\qquad
a,b=1,2,3,8$ (73)
and we learn the choice of the $\alpha$ parameter in the definition of the
$\widetilde{G}$ hyper-metric corresponds to the choice of the electro-weak
angle in the classical Lagrangian according to:
$\begin{array}[]{c}tg^{2}\theta=(1-\alpha)/(3+\alpha)<=>\alpha=1-4\;sin^{2}\theta\;.\end{array}$
(74)
Substituting (7.3), we find $\alpha=-1/2$.
## 11 Discussion
The $SU(2/1)$ model appeared in 1979 as very compelling, but so far no setting
had been found that could explain how to extend the Yang-Mills quantum field
theory from gauging an internal Lie algebra symmetry to the case of a
superalgebra. We believe that the present paper represents a step in this
direction.
The original observation [N79, F79] is that, provided we grade the Fermions by
their chirality, the smallest Lie superalgebra $SU(2/1)$ exactly corresponds
to the standard model. The leptons and the quarks fit the 2 smallest
irreducible representations (3.1-2) and (4.1-4) of the superalgebra $SU(2/1)$.
The $U(1)$ hyper-charge, which is supertraceless, coincides with the original
choice of Weinberg [Wei67] who wanted to avoid a current coupled to the lepton
number, and explains the non existence of any massless charged particle. The
difficulty is to understand the role of the odd generators.
Our new idea is to take the $SU(2/1)$ matrices at face value, despite the fact
that they are not Hermitian (4.3), and to construct the whole Lagrangian by
studying the counterterms induced by the Fermions. The caveat is to remember
the triangle anomaly (4.5). It is avoided if and only if we always consider as
our building block a whole family, constituted of one lepton triplet and three
quark quadruplets, for example the electron, the left neutrino and 3 colored
copies of the up and down quarks, counting independently the left and right
chiral states. If we associate a single real scalar to each odd generator the
induced scalar propagator vanishes (4.8). But if we split the odd matrices
into their chiral parts (5.2), and consider 2 sets of scalars $\Phi_{L}$ and
$\Phi_{R}$ (5.3) the leptons and the quarks conspire to induce a the scalar
propagator (5.7) proportional to the odd part of the skew-symmetric Killing
metric of the superalgebra (3.6). It is then possible to change variables
(5.8) from $\Phi_{L},\;\Phi_{R}$ to $H,\;K$ and recover two Higgs scalar
doublets, $H$ coupled to the electron (5.12) and down quark (5.13), and $K$
coupled to the up quark (5.14). If we compute in the same way the quartic
scalar counterterms we recover (6.12) the potential first considered by Fayet
[F74], a building stone of what later became the Minimal Standard
Supersymmetric Model (MSSM). This is rather surprising, because we have not
introduced any Wess-Zumino supersymmetry in our construction, but may be
attributed to the near unicity of the $SU(2)U(1)$ quartic invariants.
Some interesting results follow from the analysis of the quantum stability of
the symmetry breaking pattern of the theory. In the Fayet potential, the
crucial element is the form of the mixed term which insures that, in the
vacuum, the 2 fields $H$ and $K$ are parallel, and therefore leave an unbroken
$U(1)$ symmetry, gauged by a massless photon. In our model, this condition is
a consequence of the super-Jacobi identity (6.10) and is stable against the
Fermionic 1-loop quantum corrections, since this is the way we derived the
scalar potential. Furthermore, if fix (7.3) the relative strength of the 2
vectors coupling to $tg^{2}\theta=5/3$ (as in $SU(5)$ grand unified theory
[GG74]), the strength of the scalar to $2/{\sqrt{3}}$, and the quartic self
coupling to $\lambda=4/9$, these coefficients are stable against the quantum
correction induced by the Fermion loops for any number of families, thanks to
the Ward identities (7.5). These identities are typical of a Yang-Mills
minimal coupling and states that scalar Fermion coupling and the scalar
quartic self coupling are indeed proportional to $g$ and $g^{2}$ where $g$ is
the Yang-Mills $SU(2)$ coupling constant. This fixes (8.7) the mass of the
neutral Higgs at
$M^{2}_{H^{0}_{1}}+M^{2}_{H^{0}_{2}}=2\;(4/3\;M_{W})^{2}=2\;(107.2\;GeV)^{2}$.
The mass of the charged Higgs corresponds to the mixed $H$ $K$ term in the
potential and gives (8.6): $M_{H^{\pm}}=2{\sqrt{2}}/3\;M_{W}=75.8\;GeV$, a
value a few $GeV$ too low relative to the minimum given by the particle data
group [Yao06]. Finally the mass of the ’$A_{0}$’ pseudo-scalar is not
constrained.
On the other hand, we clearly have a difficulty with the mass of the quarks.
If we assume that the 2 Higgs have the same mass and that the $SU(2/1)$
minimal coupling (5.3) directly gives the mass of the heaviest quark, we
predict $M_{top}=M_{H^{0}}=4/3\;M_{W}=107.2\;GeV$ which falls way under the
current experimental value $M_{top}=174.3\pm 1.8\;GeV$, deduced from direct
observations [Yao06]. But a worse problem is that the $charm$ and $up$ quarks,
in the two other families should have the same mass, when actually they are
very light ($1.25\;GeV$ and $2\;MeV$). This problem is possibly solved by the
$SU(2/1)$ indecomposable representations which have been shown by Haussling
and Scheck to correctly reproduce the Masakawa quark mixing phenomenology
[HS94]. Assuming that the masses get redistributed, we are not so far from the
relation $M^{2}_{top}+M^{2}_{charm}+M^{2}_{up}=3\;M^{2}_{H}$. This question
needs further studies.
A classical geometric interpretation of the theory is also possible. When
expressed in terms of the $\Phi_{L}$ and $\Phi_{R}$ fields, we found that the
potential involves both the symmetric $d^{a}_{ij}$ structure constants of the
superalgebra and the skew symmetric $f^{a}_{ij}$ structure constants of
$SU(3)$, times the chirality operator (3.4). This gives the idea of revisiting
the Hermitian algebras introduced by Sternberg and Wolf [SW78]. Our new result
is to show that this structure exactly fits the chiral decomposition (5.3) of
the $\Phi$ fields, allowing us to construct a classical curvature 2-forms
(9.12), where the unexpected apparition of the chirality operator $\chi$ in
the definition of the Hermitian Lie bracket (9.9) exactly compensates the
signs implied when adapting our chiral connection 1-form $\widetilde{A}$
[TM06] to the doubling of the $\Phi$ fields. There is also a probable relation
with non commutative differential geometry, as discussed in [TM06].
We are now in a position to give a classical geometrical construction of the
vector/scalar part of our Lagrangian as the square of the curvature of the
$SU(2/1)$ Hermitian algebra. In the earlier work of Ne’eman and others, the
central problem was the choice of the quadratic form needed to compute the
square of the curvature. The natural choice was the supertrace metric (3.6),
but this yields a negative propagator for the $U(1)$ field and breaks
unitarity. The phenomenological choice seemed to be the trace of the lepton
representation (3.1) which yields $sin^{2}\theta=.25$. In 1979, this value was
acceptable, but is now too high. Also the trace metric is not a good invariant
for a superalgebra, and this choice was often criticized. Our new result is to
deduce the relevant quadratic form from the renormalization theory. We found
the very elegant hyper metric $\widetilde{G}$ (6.2) which is a combination of
the trace and the supertrace metric and, thanks to the super-Jacobi identity,
contains a free parameter $\alpha$. The 2 rows and columns of $\widetilde{G}$
are used to combine the 2 $CP$ conjugated versions of the H-curvature (10.1),
and this combination restores unitarity, because in the combined Lagrangian we
can switch between the native $SU(2/1)$ geometric but non-Hermitian $\Phi_{L}$
and $\Phi_{R}$ (6.4) variables and the good Hermitian quantum field theory
scalars $H$ and $K$ (6.9). When we inject this metric in our Lagrangian
(10.1), we find that the freedom in the choice of $\alpha$ reflects a freedom
in the choice of the electro-weak angle $\alpha=1-4\;sin^{2}\theta$ (10.3) and
that the simplest choice $\alpha=0$ corresponds to the trace-metric choice,
$\alpha=-1/2$ corresponds to $sin^{2}\theta=3/8$.
Let us now recall the distinction between this model and the minimal
supersymmetric standard model (MSSM). The geometrical space behind Yang-Mills
theory is the principal fiber bundle. The base is space-time, the fiber is the
gauge group. In both cases, a superalgebra is introduced but not in the same
way. In the MSSM, the Poincare superalgebra is introduced in the base space,
acts on Bose/Fermi supermultiplets and naturally leads to supergravity. In our
approach, the superalgebra is introduced in the fiber, by embedding the
Glashow-Weinberg-Salam $SU(2)U(1)$ Lie algebra in $SU(2/1)$, it acts of
left/right Fermionic chiral supermultiplets and groups the Higgs scalars with
the Yang-Mills vectors. This construction is not incompatible with
supergravity at much higher energy or with superstrings but does not lead to
these theories. In both models, (5.11) and [F74], we are led to introduce 2
sets of scalar doublets, $H$ coupled to the electron and down-quark right
singlets, and $K$ to the up-quark singlet. The quartic scalar potential has
the same structure (6.9-12) and implies the survival of an unbroken $U(1)$ and
a massless photon. So the scalar spectrum is similar, 3 states become the
longitudinal components of the $W$ and $Z$ vector fields, 5 survive. In both
models the scale of the potential is fixed and the values are in the same
range. But of course the main difference is that the MSSM predicts for every
known particle the existence of supersymmetric partner, like the winos,
gluinos, squarks and sleptons, whereas $SU(2/1)$ fits exactly the known
particles but does not predict any new ones, except the Higgs fields. The
existence or non existence of these particles below $1\;TeV$ will allow us to
choose between the 2 kinds of supersymmetry.
The $SU(2/1)$ theory is testable, but incomplete in several directions. At the
experimental level, the model predicts 2 complex doublets of Higgs fields,
rather than 1 as in the original Weinberg paper, and constrains their masses
within the range of the new CERN collider which opens this year. At the
phenomenological level, the mixing of the quark families is not understood in
details, but may be linked to the indecomposable representations of $SU(2/1)$.
At the theoretical level, the balance between the quarks and leptons is broken
by the quantum counterterms induced by the gluons, and all the scalar
couplings, except those insuring the existence of a massless photon, are
perturbed by the Boson loops. We believe that this problem could indicate the
incompleteness of the current quantization procedure and will discuss it
elsewhere.
But even if these difficulties cannot be solved immediately, the 1-loop
stability of the structure of the scalar potential, ensuring a massless
photon, and the resulting prediction of the number and masses of the Higgs
fields seem worth reporting as a first step towards the construction of
$SU(2/1)$ Quantum Astheno Dynamics.
###### Acknowledgments.
In many ways, this paper is a concretization of our long lasting collaboration
with Yuval Ne’eman on SU(2/1). It is also a pleasure to thank Pierre Fayet and
Victor Kac for many discussions, Florian Scheck and Shlomo Sternberg for
communications, David Lipman for always prompting new ideas and Danielle
Thierry-Mieg for important suggestions.
## References
* [AL73] E.S. Abers and B.W. Lee, “Gauge Theories”, Physics Reports, 9C no. 1 1973
* [BIM72] C.Bouchiat, J.Iliopoulos and P.Meyer, An anomaly free version of Weinberg’s model Phys. Let. A38, 519 (1972)
* [CL90] A. Connes and J. Lott, “Particle models and noncommutative geometry”, Nucl. Phy. 18B (Proc. Suppl.) 29 (1990)
* [DJ79] P.H. Dondi and P.D. Jarvis, “A supersymmetric Weinberg-Salam model”, Physics Letters 84B Erratum 87B 403 (1979)
* [F79] D.B. Fairlie, “Higgs Fields and the Determination of the Weinberg Angle”, Phys. Lett. B82 97 (1979)
* [F74] P.Fayet, Nucl. Phy. B78, 14 (1974)
* [F75] P.Fayet, Nucl. Phy. Lett. B90, 104 (1975)
* [GG74] H.Georgi and S.Glashow, Phys. Rev. Lett. 32, (1974)
* [GIM70] S.Glashow, J.Iliopoulos and L.Maiani, Phys. Rev. D2, 1285-1292 (1970)
* [BIM72] C.Bouchiat, J.Iliopoulos and Ph.Meyer Phys. Lett. B38, 519-523 (1972)
* [HS94] R.Haussling and F.Scheck, “Quark mass matrices and generation mixing in the standard model with non-commutative geometry”, Phys. Lett. B336, 477-486 (1994)
* [N79] Y. Ne’eman, “Irreducible Gauge Theory of a Consolidated Salam-Weinberg Model”, Phys. Lett. B81, 190 (1979)
* [NS90] Y. Ne’eman, S. Sternberg, “Superconnections and internal supersymmetry dynamics”, Proc. Nat. Acad. Sci. USA 87 7875 (1990)
* [NSF05] Y. Ne’eman, S.Sternberg and D.Fairlie “Superconnections electroweak su(2/1) and extensions, and the mass of the Higgs”, Physics Reports 406, 303-377 (2005)
* [NT80] Y. Ne’eman and J. Thierry-Mieg, “Geometrical Gauge Theory of Ghost and Goldstone Fields and of Ghost Symmetries”, Proc. Nat. Acad. Sci., USA 77 720-723 (1980)
* [Q85] D. Quillen, “Superconnections and the Chern character”, Topology 24 89 (1985)
* [Sa68] A. Salam, in Elementary Particle Theory, N. Svartholm, ed. Almquist Verlag A.B., Stockholm (1968)
* [S92] F.Scheck, “Anomalies, Weinberg angle and a non-commutative geometric description of the standard model”, Phys. Lett. B284, 303-308 (1992)
* [SNR77] M. Scheunert, W. Nahm & V. Rittenberg, “Irreducible Representations of the osp(2/1) and sp1(2/1) graded Lie algebras”, J. Math. Phys. 18 156-167 (1977)
* [SW78] S. Sternberg, J. Wolf, “Hermitian Lie algebras and metaplectic representations”, Trans. Amer. Math. Soc. 231 1 (1978)
* [TMN82] J. Thierry-Mieg and Y. Ne’eman, “Exterior Gauging of an Internal Supersymmetry and $SU(2/1)$ Quantum Asthenodynamics”, in Proc. of Nat. Acad. of Sciences, USA 79 (1982) p. 7068-7072.
* [TM06] J. Thierry-Mieg, “Chiral Yang-Mills theory, non commutative differential geometry, and the need for a Lie superalgebra”, JHEP 2006 038
* [Wei67] S. Weinberg, “A Model of Leptons”, Phys. Rev. Lett. 19 (1967) 1264.
* [Yao06] W.-M. Yao et al, “Particle Data Group //pdg.lbl.gov” J. Phys. G 33 1 (2006)
* [YM54] C.N. Yang and R. Mills, “Conservation of isotopic spin and isospin gauge invariance,” Phys. Rev. 96 191-195 (1954)
| arxiv-papers | 2008-06-29T18:06:52 | 2024-09-04T02:48:56.417947 | {
"license": "Public Domain",
"authors": "Jean Thierry-Mieg",
"submitter": "Jean Thierry-Mieg",
"url": "https://arxiv.org/abs/0806.4628"
} |
0806.4639 | # Topological spin-current in non-centrosymmetric superconductors
Yukio Tanaka1, Takehito Yokoyama1, Alexander V. Balatsky2 and Naoto Nagaosa3,4
1Department of Applied Physics, Nagoya University, Nagoya, 464-8603, Japan
2 Theoretical Division, Los Alamos National Laboratory, Los Alamos, New Mexico
87545, USA
3 Department of Applied Physics, University of Tokyo, Tokyo 113-8656, Japan
4 Cross Correlated Materials Research Group (CMRG), ASI, RIKEN, WAKO 351-0198,
Japan
###### Abstract
We study the spin transport properties of the non-centrosymmetric
superconductor with time-reversal-symmetry where spin-triplet $(p_{x}\pm
ip_{y})$-wave and spin-singlet $s$-wave pair potential can mix each other. We
show that when the amplitude of $(p_{x}\pm ip_{y})$-wave pair potential is
larger than that of $s$-wave one, the superconducting state belongs to the
topologically nontrivial class analogous to the quantum spin Hall system, and
the resulting helical edge modes as Andreev bound states are topologically
protected. We find that the incident angle dependent spin polarized current
flows through the interface due to the presence of the helical edge modes.
With a weak magnetic field, also the angle-integrated current is strongly spin
polarized.
###### pacs:
74.45.+c, 74.50.+r, 74.20.Rp
††preprint: Helical edge
The topological properties of the electronic states have been attracting
intensive interests in condensed matter physics. Especially, it was
highlighted by the discovery of the quantum Hall system (QHS) showing the
accurate quantization of the Hall conductance $\sigma_{H}$ which is related to
the topological integer Girvin ; Thouless .
Recently, the concept of the QHS has been generalized to the time-reversal
($T$) symmetric system, i.e., the quantum spin Hall system (QSHS) Mele ;
Bernevig ; Fu . QSHS could be regarded as the two copies of QHS for up and
down spins with the opposite chiralities. In generic case, however, the
mixture of up and down spins occurs due to the spin-orbit interaction, which
necessitates the new topological number to characterize QSHS Mele ; Fu . In
QSHS, there exist the helical edge modes, i.e., the time-reversal pair of
right- and left-going one-dimensional modes, which has been experimentally
demonstrated for the quantum well of HgTe system by the measurement of the
charge conductance Konig .
In the field of superconductivity, the chiral $p$-wave superconductors such as
Sr2RuO4 Maeno can be considered as an analogue of the QHS, and novel
phenomena such as one-dimensional Majorana fermions (real fermions) modes at
the edge MatsumotoSigrist ; ReadGreen and the non-Abelian statistics of the
vortex ReadGreen ; Ivanov has been proposed there. Beside these issues, the
non-centrosymmetric (NCS) superconductors such as CePt3Si are a central topic
Bauer ; Frigeri . Also the two-dimensional NCS superconductors are expected at
the interfaces and/or surfaces due to the strong potential gradient. An
interesting example is the superconductivity at LaAlO3/SrTiO3 interface
Interface . In NCS superconductors, the spin-orbit interaction comes into
play. Especially, Frigeri et al. Frigeri have shown that $(p_{x}\pm
ip_{y})$-pairing state has the highest $T_{c}$ within the triplet-channel in
CePt3Si. However, the singlet ($s$-wave) and triplet ($p$-wave) pairings are
mixed, and several novel associated properties such as the large upper
critical field beyond the Pauli limit have been focused on Frigeri . On the
other hand, the pure $(p_{x}\pm ip_{y})$-pairing state has been studied from
the viewpoint of the superconducting analogue of QSHS Qi . Therefore, it is an
important and urgent issue to study the spin transport properties of the NCS
superconductors from the topological viewpoint.
In this Letter, we study the spin transport properties of the non-
centrosymmetric (NCS) superconductor Bauer with $T$-symmetry, where
$(p_{x}\pm ip_{y})$-wave and spin-singlet $s$-wave pair potential can mix each
other. We show that when the amplitude of $(p_{x}\pm ip_{y})$-wave pair
potential is larger than that of $s$-wave one, the superconducting state
belongs to the topologically nontrivial class analogous to the quantum spin
Hall system, and the resulting helical edge modes as Andreev bound states(ABS)
are topologically protected. We study Andreev reflection Andreev at low
energy, which is determined mostly by the helical edge modes, and find the
incident angle dependent spin polarized current flowing through the interface.
When the magnetic field is applied, even the angle-integrated current is spin
polarized.
We start with the Hamiltonian of NCS superconductor
$\check{H}=\left({\begin{array}[]{*{20}c}{\hat{H}\left({\bf
k}\right)}&{\hat{\Delta}\left({\bf k}\right)}\\\
{-\hat{\Delta}^{*}\left({-{\bf k}}\right)}&{-\hat{H}^{*}\left({-{\bf
k}}\right)}\\\ \end{array}}\right)$
with $\hat{H}({\bm{k}})=\xi_{{\bm{k}}}+\bm{V}(\bm{k})\cdot\hat{\bm{\sigma}}$,
$\bm{V}({\bm{k}})=\lambda(\hat{\bm{x}}k_{y}-\hat{\bm{y}}k_{x})$,
$\xi_{\bm{k}}=\hbar^{2}{\bm{k}}^{2}/(2m)-\mu$. Here, $\mu$, $m$,
$\hat{\bm{\sigma}}$ and $\lambda$ denote chemical potential, effective mass,
Pauli matrices and coupling constant of Rashba spin-orbit interaction,
respectively Frigeri . The pair potential $\hat{\Delta}(\bm{k})$ is given by
$\hat{\Delta}(\bm{k})=[\bm{d}(\bm{k})\cdot\hat{{\bm{\sigma}}}]i\hat{\sigma}_{y}+i\psi(\bm{k})\hat{\sigma}_{y}.$
(1)
We choose $(p_{x}\pm ip_{y})$-wave pair for spin-triplet component with
$\bm{d}(\bm{k})=\Delta_{p}(\hat{\bm{x}}k_{y}-\hat{\bm{y}}k_{x})/\mid{\bm{k}}\mid$
Frigeri and $s$-wave one with $\psi(\bm{k})=\Delta_{s}$ with $\Delta_{p}\geq
0$ and $\Delta_{s}\geq 0$. The superconducting gaps
$\Delta_{1}=\Delta_{p}+\Delta_{s}$ and
$\Delta_{2}=\mid\Delta_{p}-\Delta_{s}\mid$ open for the two spin-splitted
band, respectively, in the homogeneous state Iniotakis .
However, as seen below, surface states are crucially influenced by the
relative magnitude between $\Delta_{p}$ and $\Delta_{s}$. Let us consider wave
function including ABS localized at the surface. Consider a two-dimensional
semi-infinte superconductor on $x>0$ where the surface is located at $x=0$.
The corresponding wave function is given by Yokoyama
$\displaystyle\Psi_{S}(x)=\exp(ik_{y}y)[c_{1}\psi_{1}\exp(iq^{+}_{1x}x)+c_{2}\psi_{2}\exp(-iq^{-}_{1x}x)$
$\displaystyle+c_{3}\psi_{3}\exp(iq^{+}_{2x}x)+c_{4}\psi_{4}\exp(-iq^{-}_{2x}x)],$
(2) $\displaystyle
q^{\pm}_{1(2)x}=k^{\pm}_{1(2)x}\pm\frac{k_{1(2)}}{k^{\pm}_{1(2)x}}\sqrt{\frac{E^{2}-\Delta_{1(2)}^{2}}{\lambda^{2}+2\hbar^{2}\mu/m}},$
with $k^{+}_{1(2)x}=k^{-}_{1(2)x}=k_{1(2)x}$ for $\mid k_{y}\mid\leq k_{1(2)}$
and $k^{+}_{1(2)x}=-k^{-}_{1(2)x}=k_{1(2)x}$ for $\mid k_{y}\mid>k_{1(2)}$.
Here, $k_{1}$ and $k_{2}$ are Fermi momentum of the small and large magnitude
of Fermi surface given by
$-m\lambda/\hbar^{2}+\sqrt{(m\lambda/\hbar^{2})^{2}+2m\mu/\hbar^{2}}$ and
$m\lambda/\hbar^{2}+\sqrt{(m\lambda/\hbar^{2})^{2}+2m\mu/\hbar^{2}}$,
respectively. $k_{1(2)x}$ denotes the $x$ component of the Fermi momentum
$k_{1(2)}$, with $k_{1(2)x}=\sqrt{k_{1(2)}^{2}-k_{y}^{2}}$. The wave functions
are given by
${}^{T}\psi_{1}=\left(u_{1},-i\alpha_{1}^{-1}u_{1},i\alpha_{1}^{-1}v_{1},v_{1}\right)$,
${}^{T}\psi_{2}=\left(v_{1},-i\tilde{\alpha}_{1}^{-1}v_{1},i\tilde{\alpha}_{1}^{-1}u_{1},u_{1}\right)$,
${}^{T}\psi_{3}=\left(u_{2},i\alpha_{2}^{-1}u_{2},i\gamma\alpha_{2}^{-1}v_{2},-\gamma
v_{2}\right)$, and
${}^{T}\psi_{4}=\left(v_{2},i\tilde{\alpha}_{2}^{-1}v_{2},i\gamma\tilde{\alpha}_{2}^{-1}u_{2},-\gamma
u_{2}\right)$, with $\gamma=\rm{sgn}(\Delta_{p}-\Delta_{s})$. In the above,
$u_{1(2)}$ and $v_{1(2)}$ are given as
$\sqrt{\frac{1}{2}\left({1+\frac{{\sqrt{E^{2}-\Delta_{1(2)}^{2}}}}{E}}\right)}$,
and
$\sqrt{\frac{1}{2}\left({1-\frac{{\sqrt{E^{2}-\Delta_{1(2)}^{2}}}}{E}}\right)}$.
Here we have introduced $\alpha_{1}=(k^{+}_{1x}-ik_{y})/k_{1}$,
$\alpha_{2}=(k^{+}_{2x}-ik_{y})/k_{2}$,
$\tilde{\alpha}_{1}=(-k^{-}_{1x}-ik_{y})/k_{1}$, and
$\tilde{\alpha}_{2}=(-k^{-}_{2x}-ik_{y})/k_{2}$. $E$ is the quasiparticle
energy measured from the Fermi energy.
By postulating $\Psi_{S}(x)=0$ at $x=0$, we can determine the ABS. The bound
state condition can be expressed by
$\displaystyle\sqrt{(\Delta_{1}^{2}-E^{2})(\Delta_{2}^{2}-E^{2})}=\frac{1-\zeta}{1+\zeta}(E^{2}+\gamma\Delta_{1}\Delta_{2}),$
(3)
$\displaystyle\displaystyle\zeta=\left\\{\begin{array}[]{ll}\frac{\sin^{2}[\frac{1}{2}(\phi_{1}+\phi_{2})]}{\cos^{2}[\frac{1}{2}(\phi_{1}-\phi_{2})]}&\mid\phi_{2}\mid\leq\phi_{C}\\\
1&\phi_{C}<\mid\phi_{2}\mid\leq\pi/2,\end{array}\right.$ (6)
with $\zeta\leq 1$, $\cos\phi_{1}=k_{1x}/k_{1}$ and
$\cos\phi_{2}=k_{2x}/k_{2}$. The critical angle $\phi_{C}$ is defined as
$\sin^{-1}(k_{1}/k_{2})$. For $\lambda=0$, eq. (3) reproduces the previous
result Iniotakis . As seen from eq. (3), the ABS including zero energy state
is only possible for $\mid\phi_{2}\mid\leq\phi_{C}$ and $\gamma=1$, $i.e.$,
$\Delta_{p}>\Delta_{s}$. The present ABS is just the edge state, where the
localized quasiparticle can move along the edge. The energy level of the edge
state depends crucially on the direction of the motion of the quasiparticle.
The inner gap edge modes are absent for large magnitude of $k_{y}$, $i.e.$
$\phi_{2}$. The parameter regime where the edge modes survive is reduced with
the increase of the magnitude of $\lambda$. However, as far as we concentrate
on the perpendicular injection, the edge modes survive as the mid gap ABS ABS
; TK95 irrespective of the strength of $\lambda$. If we focus on the low
energy limit, ABS can be written as
$\displaystyle
E=\pm\Delta_{p}(1-\frac{\Delta^{2}_{s}}{\Delta^{2}_{p}})\frac{k_{1}+k_{2}}{2k_{1}k_{2}}k_{y},$
(7)
with $\Delta_{s}<\Delta_{p}$ for any $\lambda$ with small magnitude of
$k_{y}$. For $\Delta_{s}\geq\Delta_{p}$, the presnet ABS vanishes since the
value of right side of eq. (3) becomes negative due to the negative sign of
$\gamma$ for $\mid E\mid<\Delta_{1}$ and $\mid E\mid<\Delta_{2}$. It should be
remarked that the present ABS do not break the time reversal symmetry, since
the edge current carried by each Kramers doublet flows in the opposite
direction. Thus they can be regarded as helical edge modes, where two modes
are connected to each other by time reversal operation.
Now we give an argument why the superconducting state with
$\Delta_{p}>\Delta_{s}$ has the ABS from the viewpoint of $Z_{2}$
(topological) class Mele . We commence with the pure $(p_{x}\pm ip_{y})$-wave
state without the spin-orbit interaction $\lambda$. Spin Chern number Fu for
the Bogoliubov-de Gennes (BdG) Hamiltonian is 2. Turning on $\lambda$
adiabatically, which leaves the $T$-symmetry intact and keeps the gap open,
one can arrive at the BdG Hamiltonian of interest. Upon this adiabatic change
of $\lambda$, the number of the helical edge mode pairs does not change. Then
we increase the magnitude of $\Delta_{s}$ from zero. As far as
$\Delta_{p}>\Delta_{s}$ is satisfied, the number of helical edge modes does
not change, since it is a topological number. However, if $\Delta_{s}$ exceeds
$\Delta_{p}$, the helical mode disappears. In this regime, the topological
nature of superconducting state belongs to pure $s$-wave state without
$\lambda$. It is remarkable, just at $\Delta_{s}=\Delta_{p}$, one of the
energy gap of the quasiparticle in the bulk closes, where a quantum phase
transition occurs.
Now we turn to the spin transport property governed by the ABS in the NCS
superconductors Eschrig . First, we point out that the spin Hall effect, i.e.,
the appearance of the spin Hall voltage perpendicular to the superconducting
current is suppressed by the compressive nature of the superconducting state
by the factor of $(k_{F}\lambda_{m})^{-2}$ ($k_{F}$: Fermi momentum,
$\lambda_{m}$: penetration depth) Furusaki . Instead, we will show below that
the spin transport through the junction between the ballistic normal metal at
$x<0$ and NCS superconductor, i.e., (N/SC) junction, can be enhanced by the
Doppler effect at the Andreev reflection. We assume an insulating barrier at
$x=0$ expressed by a delta-function potential $U\delta(x)$. The wave function
for spin $\sigma$ in the normal metal $\Psi_{N}(x)$ is given by
$\displaystyle\Psi_{N}(x)\\!\\!$ $\displaystyle=$
$\displaystyle\\!\\!\exp(ik_{Fy}y)[(\psi_{i\sigma}+\sum_{\rho=\uparrow,\downarrow}a_{\sigma,\rho}\psi_{a\rho})\exp(ik_{Fx}x)$
(8)
$\displaystyle+\sum_{\rho=\uparrow,\downarrow}b_{\sigma,\rho}\psi_{b\rho}\exp(-ik_{Fx}x)]$
with ${}^{T}\psi_{i\uparrow}$$=$ ${}^{T}\psi_{b\uparrow}$
$=$$\left(1,0,0,0\right)$,
${}^{T}\psi_{i\downarrow}$=${}^{T}\psi_{b\downarrow}$
$=$$\left(0,1,0,0\right)$, ${}^{T}\psi_{a\uparrow}$ $=$$\left(0,0,1,0\right)$,
and ${}^{T}\psi_{a\downarrow}$ $=$$\left(0,0,0,1\right)$. The corresponding
$\Psi_{S}(x)$ is given by eq. (2). The coefficients $a_{\sigma,\rho}$ and
$b_{\sigma,\rho}$ are determined by postulating the boundary condition between
$\Psi_{N}(x)$ and $\Psi_{S}(x)$ given by $\Psi_{N}(0)=\Psi_{S}(0)$, and
$\frac{d}{dx}\Psi_{S}(0)-\frac{d}{dx}\Psi_{N}(0)=\frac{2mU}{\hbar^{2}}\Psi_{S}(0)$.
We assume that the Fermi momentum $k_{F}$ of normal metal and NCS
superconductor before putting $\lambda$ are the same. The quantities of
interest are the angle resolved spin conductance $f_{S}(\phi)$ and charge
conductance $f_{C}(\phi)$ defined by Kashiwaya99
$\displaystyle f_{S}(\phi)=[(\mid a_{\uparrow,\uparrow}\mid^{2}-\mid
a_{\uparrow,\downarrow}\mid^{2}-\mid b_{\uparrow,\uparrow}\mid^{2}+\mid
b_{\uparrow,\downarrow}\mid^{2})$ $\displaystyle-(\mid
a_{\downarrow,\downarrow}\mid^{2}-\mid a_{\downarrow,\uparrow}\mid^{2}-\mid
b_{\downarrow,\downarrow}\mid^{2}+\mid
b_{\downarrow,\uparrow}\mid^{2})]\frac{\cos\phi}{2},$ $\displaystyle
f_{C}(\phi)=[2+\sum_{\sigma,\rho}(\mid a_{\sigma,\rho}\mid^{2}-\mid
b_{\sigma,\rho}\mid^{2})]\frac{\cos\phi}{2},$ (9)
where $\phi$ denotes the injection angle measured from the normal to the
interface. First we consider pure $(p_{x}\pm ip_{y})$-wave state. In Fig. 1,
the angle resolved spin conductance is plotted as a function of injection
angle $\phi$ and bias voltage $V$ with $E=eV$. Note here that the $k_{y}$ is
related to $\phi$ as $k_{y}=k_{F}\sin\phi$. It is remarkable that spin
conductance has a non zero value although the NCS superconductor does not
break time reversal symmetry. $f_{S}(\phi)$ has a peak when the angle $\phi$
or $k_{y}$ gives the energy $E$ in the energy dispersion of ABS. With this
condition, the spin-dependent Andreev reflection occurs to result in the spin
current. Besides this property, we can show that $f_{S}(\phi)=-f_{S}(-\phi)$
is satisfied. By changing the sign of $eV$, $f_{S}(\phi)$ changes sign as seen
in Fig. 1(a). Next, we look at the case where $s$-wave component coexists. We
can calculate spin current similar to the pure $(p_{x}\pm ip_{y})$-wave case.
For $\Delta_{s}<\Delta_{p}$, where helical edge modes exist, $f_{S}(\phi)$ has
a sharp peak and $f_{S}(\phi)=-f_{S}(-\phi)$ is satisfied [see Fig. 1(b)].
These features are similar to those of pure $(p_{x}\pm ip_{y})$-wave case. On
the other hand, for $\Delta_{s}>\Delta_{p}$, where the helical edge modes are
absent, sharp peaks of $f_{S}(\phi)$ as shown in Fig. 1 are absent.
We have checked that there is negligible quantitative change by taking
$\lambda=0$ limit compared to Fig. 1, $e.g.$, less than 0.5% change of the
peak height. In this limit, for pure $(p_{x}\pm ip_{y})$-wave state,
$f_{S}(\phi)$ is given simply as follows
$\frac{-8\sigma_{N}^{2}(1-\sigma_{N})\sin 2\phi\sin 2\varphi\cos\phi}{\mid
4(\sin^{2}\phi-\sin^{2}\varphi)+\sigma_{N}[\exp(-2i\varphi)(\sigma_{N}-2)+2\cos
2\phi]\mid^{2}},$
for $\mid E\mid<\Delta_{p}$ and $f_{S}(\phi)=0$ for $\mid E\mid>\Delta_{p}$
with $\sin\varphi=E/\Delta_{p}$ Transparency of the interface $\sigma_{N}$ is
given by $4\cos^{2}\phi/(4\cos^{2}\phi+Z^{2})$ with a dimensionless constant
$Z=2mU/\hbar^{2}k_{F}$. The magnitude of $f_{S}(\phi)$ is largely enhanced at
$E=\pm\Delta_{p}\sin\phi$ corresponding to the energy dispersion of ABS. The
origin of nonzero $f_{S}(\phi)$ even without $\lambda$ is due to the spin-
dependent ABS. We have checked that even if we take into account the spatial
dependence of the $(p_{x}\pm ip_{y})$-wave pair potential explicitly, the
resulting $f_{S}(\phi)$ does not qualitatively change Eschrig .
Summarizing these features, we can conclude that the presence of the helical
edge modes in NCS superconductor is the origin of the large angle resolved
spin current through N/NCS superconductor junctions. However, the magnitude of
the angle averaged normalized spin conductance becomes zero since
$f_{S}(\phi)=-f_{S}(-\phi)$ is satisfied.
Figure 1: (Color online) Angle resolved spin conductance for $Z=5$. a:
$eV=0.1\Delta_{p}$, b: $eV=-0.1\Delta_{p}$ and c: $eV=0.6\Delta_{p}$ with
$\lambda k_{F}=0.1\mu$. (a)pure $(p_{x}\pm ip_{y})$-wave case with
$\Delta_{s}=0$. (b)$\Delta_{s}=0.3\Delta_{p}$.
Magnetic field offers an opportunity to observe the spin current in a much
more accessible way, where $T$-symmetry is broken by the shielding current at
the interface. Here we consider the angle averaged normalized spin conductance
$\sigma_{S}$ and charge conductance $\sigma_{C}$ as a function of magnetic
field which are given by TK95 ; Kashiwaya99
$\sigma_{S}=\frac{\int^{\pi/2}_{-\pi/2}f_{S}(\phi)d\phi}{\int^{\pi/2}_{-\pi/2}f_{NC}(\phi)d\phi},\
\
\sigma_{C}=\frac{\int^{\pi/2}_{-\pi/2}f_{C}(\phi)d\phi}{\int^{\pi/2}_{-\pi/2}f_{NC}(\phi)d\phi},$
(10)
where $f_{NC}(\phi)$ denotes the angle resolved charge conductance in the
normal state with $\Delta_{p}=\Delta_{s}=0$. Now we consider the magnetic
field $H$ applied perpendicular to the two-dimensional plane, which induces a
shielding current along the N/NCS superconductor interface. When the
penetration depth of the NCS superconductor is much longer than coherence
length, the vector potential can be approximated as
$\bm{A}(\bm{r})=(0,A_{y}(x),0)$ with
$A_{y}(x)=-\lambda_{m}H\exp(-x/\lambda_{m})$ with the penetration depth
$\lambda_{m}$. Here we consider the situation where the quantization of the
Landau level can be neglected. Then quasiclassical approximation becomes
available. The applied magnetic field shifts the quasiparticle energy $E$ in
wave function of $\Psi_{S}(x)$ to $E-H\Delta_{p}\sin\phi/H_{0}$ with
$H_{0}=h/(2e\pi^{2}\xi\lambda_{m})$ and $\xi=\hbar^{2}k_{F}/(\pi m\Delta_{p})$
Doppler . For typical values of $\xi\sim 10$nm, $\lambda_{m}\sim 100$nm, the
magnitude of $H_{0}$ is of the order of 0.2Tesla. Here the order of the energy
of Doppler shift is given by $H\Delta_{p}/H_{0}$. Since the Zeeman energy is
given by $\mu_{B}H$, the order of the energy of Doppler shift is
$k_{F}\lambda_{m}$ times larger than that of Zeeman energy. Thus, we can
neglect the Zeeman effect in the present analysis. This is in sharp contrast
to QSHS where the Zeeman energy is the main effect of $H$, which opens the gap
in the helical edge modes and modulates the transport properties Konig . The
enhanced spin current due to the Doppler shift is specific to superconducting
state not realized in QSHS.
Figure 2: (Color online) Angle averaged spin conductance and charge
conductance as a function of $eV$ with bias voltage $V$ with $\lambda
k_{F}=0.1\mu$. a: $H=0$, b: $H=0.2H_{0}$, c: $H=-0.2H_{0}$, and d:
$H=0.4H_{0}$. Curves $b$ and $c$ of the right panel are identical.
As shown in Fig. 1, to discuss topological nature of the helical edge modes,
it is sufficient to consider pure $(p_{x}\pm ip_{y})$-wave state. In the
following, we choose $(p_{x}\pm ip_{y})$-wave case. In Fig. 2, the spin
conductance $\sigma_{S}$ and charge conductance $\sigma_{C}$ normalized by the
charge conductance in the normal state are plotted. It should be noted that
$\sigma_{S}$ becomes nonzero in the presence of the magnetic field $H$ (see
curves $b$, $c$ and $d$), since $f_{S}(\phi)$ is no more odd function of
$\phi$ due to the imbalance of the helical edge modes. For $\lambda=0$ limit,
the corresponding helical edge modes are given by
$E=\Delta_{p}(1+H/H_{0})\sin\phi$ and $E=-\Delta_{p}(1-H/H_{0})\sin\phi$. As
seen from the curves $b$ and $c$, the sign of $\sigma_{S}$ is reversed by
changing the direction of the applied magnetic field. On the other hand, the
resulting charge conductance has different features. For $H=0$, the resulting
line shape of $\sigma_{C}$ is the same as that of chiral $p$-wave
superconductor (see curve $a$ of right panel) Yokoyama ; Iniotakis ; Linder .
As seen from curves $b$ and $c$ of right panel, $\sigma_{C}$ does not change
with the change of the direction of the magnetic field $H$.
Figure 3: (Color online) Angle averaged spin and charge conductance for $eV=0$
as a function of $H$ with $\lambda k_{F}=0.1\mu$. a: spin conductance and b:
charge conductance.
Finally, we show in Fig. 3 the zero-voltage $\sigma_{S}$ and $\sigma_{C}$.
$\sigma_{S}$ is nearly linearly proportional to $H$. Note that with a small
magnetic field $H\cong 0.4H_{0}\sim 100{\rm Oe}$, $\sigma_{S}$ is already of
the order of 1. Meanwhile, $\sigma_{C}$ is almost independent of $H$.
In conclusion, we have studied the spin transport property of non-
centrosymmetric (NCS) superconductor from the viewpoint of topology and
Andreev bound state (ABS). We have found the incident angle dependent spin
polarized current flowing through the interface. When the weak magnetic field
is applied, even the angle-integrated current is largely spin polarized. As
the analogy to quantum spin Hall system (QSHS), the ABS in NCS superconductor
corresponds to the helical edge modes in QSHS. The Andreev reflection via
helical edge modes produces the enhanced spin current specific to NCS
superconductor.
This work is partly supported by the Grant-in-Aids from under the Grant No.
20654030, and NAREGI Nanoscience Project from the Ministry of Education,
Culture, Sports, Science, and Technology, Japan, NTT basic research
laboratories, DOE BES and by LDRD.
## References
* (1) See for e.g., The Quantum Hall effect, edited by R.E. Prange and S.M. Girvin, (Springer-Verlag, 1987), and references therein.
* (2) D. J. Thouless, $et$ $al.$, Phys. Rev. Lett. 49, 405 (1982).
* (3) C. L. Kane and E. J. Mele, Phys. Rev. Lett. 95, 146802 (2005); C. L. Kane and E. J. Mele, Phys. Rev. Lett. 95, 226801 (2005).
* (4) B. A. Bernevig, and S. C. Zhang, Phys. Rev. Lett. 96, 106802 (2006).
* (5) L. Fu and C. L. Kane, Phys. Rev. B 74, 195312 (2006); L. Fu and C. L. Kane, Phys. Rev. B 76, 045302 (2007).
* (6) M. König et al., Science, 318, 766 (2007).
* (7) Y. Maeno et al., Nature 394, 532 (1994)
* (8) M. Matsumoto and M. Sigrist, J. Phys. Soc. Jpn. 68, 994 (1999).
* (9) N. Read and D. Green, Phys. Rev. B61, 10267 (2000).
* (10) D. A. Ivanov, Phys. Rev. Lett. 86, 268 (2001).
* (11) E. Bauer, $et.$ $al.$, Phys. Rev. Lett. 92, 027003 (2004).
* (12) P. A. Frigeri, $et.$ $al.$, Phys. Rev. Lett. 92, 097001 (2004).
* (13) N. Reyren et al., Science 317, 1196 (2007).
* (14) X.L. Qi, $et.$ $al.$, arXiv:0803.3614; M. Sato, arXiv:0806.0426; R. Roy, arXiv:cond-mat/0608064.
* (15) A. F. Andreev, Sov. Phys. JETP 19, 1228 (1964).
* (16) C. Iniotakis, $et.$ $al.$, Phys. Rev. B 76, 012501 (2007).
* (17) T. Yokoyama, Y. Tanaka and J. Inoue, Phys. Rev. B 72 220504(R) (2005).
* (18) L. J. Buchholtz and G. Zwicknagl, Phys. Rev. B 23, 5788 (1981); C. R. Hu, Phys. Rev. Lett. 72, 1526 (1994).
* (19) Y. Tanaka and S. Kashiwaya, Phys. Rev. Lett. 74, 3451 (1995). S. Kashiwaya and Y. Tanaka, Rep. Prog. Phys. 63, 1641 (2000).
* (20) A.B. Vorontsov, I. Vekhter, M. Eschrig, Phys. Rev. Lett. 101, 127003 (2008).
* (21) J. Goryo and K. Ishikawa, J. Phys. Soc. Jpn. 67, 3006 (1998); A. Furusaki, M. Matsumoto, and M. Sigrist, Phys. Rev. B 64 054514 (2001).
* (22) S. Kashiwaya, $et.$ $al.$, Phys. Rev. B, 60 3572 (1999).
* (23) M. Fogelström, D. Rainer and J. A. Sauls, Phys. Rev. Lett. 79 281 (1997).
* (24) J. Linder and A. Sudbø, Phys. Rev. B 76, 054511 (2007).
| arxiv-papers | 2008-06-28T01:37:28 | 2024-09-04T02:48:56.424426 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Yukio Tanaka, Takehito Yokoyama, Alexander V. Balatsky and Naoto\n Nagaosa",
"submitter": "Yukio Tanaka",
"url": "https://arxiv.org/abs/0806.4639"
} |
0806.4691 | # Observation of an ultracold plasma instability
X. L. Zhang, R. S. Fletcher, and S. L. Rolston Joint Quantum Institute and
Department of Physics, University of Maryland, College Park, MD 20742
###### Abstract
We present the first observation of an instability in an expanding ultracold
plasma. We observe periodic emission of electrons from an ultracold plasma in
weak, crossed magnetic and electric fields, and a strongly perturbed electron
density distribution in electron time-of-flight projection images. We identify
this instability as a high-frequency electron drift instability due to the
coupling between the electron drift wave and electron cyclotron harmonic,
which has large wavenumbers corresponding to wavelengths close to the electron
gyroradius.
###### pacs:
52.35.-g, 52.35.Kt, 32.80.Fb
Ultracold plasmas (UCPs), produced by photoionizing a sample of laser-cooled
and trapped atoms, have extended neutral plasma parameters by over two orders
of magnitude, to electron temperatures below 1 K killian1999 . Studies of UCPs
have primarily concentrated on temperature measurements roberts2004 ;
gupta2007 ; fletcher2007 , and expansion dynamics kulin2000 ; bergeson2003 ;
robicheaux2003 , and recent work has identified a stable collective mode
fletcher2006 . A signature of the collective and nonlinear nature of plasmas
is the existence of plasma instabilities, perturbations that grow
exponentially to large amplitudes and dominate plasma dynamics. Much of the
quest for fusion energy involves control and suppression of plasma
instabilities conway2000 . This universal dynamics occurs in all kinds of
plasmas, including space plasmas oppenheim2003 ; farley1963 ; buneman1963 ,
dusty plasmas Merlino1998 , magnetically confined plasmas krall1971 , and Hall
thruster plasmas for spacecraft propulsion litvak2004 ; lazurenko2005 ;
adam2004 ; ducrocq2006 .
In this work, we present the first observation of a plasma instability in an
expanding UCP. By applying a small magnetic field ($\sim$ 2 G) perpendicular
to an applied electric field ($\sim$20 mV/cm), we observe periodic pulsed
emission of electrons from an expanding UCP, with a frequency range from 50 to
500 KHz. Using a time-of-flight electron projection imaging technique
zhang2008 , we image the electron spatial distribution by extracting them with
a high-voltage pulse and accelerating them onto a position-sensitive detector.
We observe that electron projection images split into two or three lobes in
the $E\times B$ direction, coincident with the observation of periodic
electron emission signals. This provides strong evidence for a plasma
instability in the expanding UCP due to the electrons drifting relative to the
ions across the magnetic field. A high-frequency electron drift instability
adam2004 ; ducrocq2006 quantitatively matches our observation, which has a
frequency lower than the electron gyrofrequency and a short wavelength on the
order of the electron gyroradius, due to the coupling between the electron
drift wave and a harmonic of the electron cyclotron frequency.
Our production of UCPs is similar to our previous work killian1999 , which we
briefly summarize. We cool and trap about a few million metastable Xenon atoms
in a magneto-optical trap (MOT). The neutral atom cloud has a temperature of
about 20 $\mu$K, peak density of about 2 x 1010 cm-3, and Gaussian spatial
density distribution with an rms radius of about 300 $\mu$m. We then produce
the plasma by a two-photon excitation process, ionizing up to 30$\%$ of the
MOT population. One photon for this process is from the cooling laser at 882
nm, and the other is from a pulsed dye laser at 514 nm (10-ns pulse). We
control the ionization fraction with the intensity of the photoionization
laser, while the initial electron temperature $T_{e0}$ (typically around 1-500
K) is controlled by tuning the 514-nm photon energy with respect to the
ionization limit.
The ionized cloud rapidly loses a few percent of the electrons, resulting in a
slightly attractive potential for the remaining electrons, and quickly reaches
a quasineutral plasma state. It then expands with an asymptotic velocity
(50-100 m/s) caused by outward electron pressure kulin2000 , and maintains a
roughly Gaussian density profile during the lifetime of the UCP zhang2007 .
There are two grids about 1.5 cm above and below the plasma to provide a small
electric field E ($\sim$5-50 mV/cm) so that electrons leaving the plasma are
guided to a microchannel plate detector. We can apply a longitudinal magnetic
field $B_{\parallel}$ parallel to E, or a transverse magnetic field
$B_{\perp}$ perpendicular to E. The applied magnetic fields are turned on
before the two-photon excitation process. The black curve in Fig. 1 is the
typical electron emission signal from a freely expanding UCP for $T_{e0}=100$
K. The signal consists of a prompt peak (initially escaped electrons) followed
by a region of little electron loss where the quasineutral plasma state forms.
This is followed by a long $\sim$150 $\mu$s loss of electrons, interpreted as
the decay of the plasma as electrons spill out of the potential well, which
gets shallower as the plasma expands. The electron signal with a small
longitudinal $B_{\parallel}$ ($\sim 12$ G) looks similar to that without a
magnetic field except for small enhancement and some changes in the expansion
dynamics (expansion dynamics in a large $B_{\parallel}$ is studied by time-of-
flight projection imaging technique zhang2007 ).
Figure 1: Electron emission signals for different E and $B_{\perp}$. (a)-(c)
constant E and different $B_{\perp}$; (d)-(f) constant $B_{\perp}$ and
different E. The red curves (with large periodic emission) are single shot;
the blue curve (with much less periodic emission) is the average of 40 shots.
the black curve (with large prompt peak) is the electron signal without a
magnetic field.
The electron emission signal with a small transverse magnetic field
$B_{\perp}$ is dramatically different from that with a small $B_{\parallel}$.
We observe periodic pulsed electron emission even with a very small
$B_{\perp}$, as shown in Fig. 1a-1c. As we increase $B_{\perp}$ from zero, the
electron signal starts to have oscillations at about 0.8 G. The periodic
emission appears at about 30-50 $\mu$s after the formation of the plasma with
frequency of several hundred KHz. The three traces in each panel of Fig. 1
correspond to individual single realizations of the UCP. Note that the
emissions have similiar character for each shot, although the phases are
random. As we continue to increase $B_{\perp}$ to about 2-2.5 G (Fig. 1b), the
amplitude of the emissions gets larger, comparable to the prompt peak in the
absence of a magnetic field. The frequency decreases to 50-100 KHz and the
prompt peak gets even smaller. The background electron emission signal almost
vanishes except for few peaks at about 3-3.5 G (Fig. 1c).
Figure 2: The emission frequency as a function of $E/B_{\perp}$; each symbol
for varying $B_{\perp}$ from 1 to 3.2 G and constant E; different symbol for
varying E from 7 to 33 mV/cm and constant $B_{\perp}$. The emission frequency
is linearly dependent on the electron drift velocity $V_{d}$.
The frequency and amplitude of the periodic electron signal also depends on
the applied electric field E (Fig. 1d-1f). The frequency increases and the
amplitude of the emissions decreases as we increase E, which is similar to the
case of decreasing $B_{\perp}$. Using fast Fourier transform (FFT) of the
electron signals in Fig. 1, we can extract the emission frequency as a
function of $E/B_{\perp}$ (electron drift velocity $V_{d}$), as shown in Fig.
2. The emission frequency depends linearly on the drift velocity with a slope
of $10.2\pm 0.2$ $m^{-1}$.
Figure 3: Electron projection images in weak crossed magnetic field and
electric field. All the units of the images here are in pixels, and one pixel
is about 150 $\mu$m.
By using a time-of-flight projection imaging technique zhang2008 , we image
the electron spatial distribution onto a position-sensitive detector (a
phosphor screen attached to a multichannel plate detector), and record images
with a CCD camera. Figure 3 shows the electron projection images at different
delay times after the formation of the UCP for different $B_{\perp}$ and E.
All the images are single shot, with little variation from shot to shot. In
the absence of a magnetic field (Fig. 3a), the electron images show a Gaussian
density profile. Note that the electron image size slowly decreases during the
lifetime of the UCP because of the strong Coulomb force of the ion cloud on
the electrons, electron loss and electron Coulomb explosion effects, unlike
ion images (whose sizes decrease in the first 20 $\mu$s due to the strong
Coulumb explosion effect of the dense ion cloud, and then increase afterward,
reflecting the real size of the plasma) zhang2008 . As we increase $B_{\perp}$
to about 1 G (Fig. 3b), the electron images start to split into two lobes at
about 30 $\mu$s, coincident with the observation of large periodic electron
emission signals. At about 2 G, we observe up to 3 lobes in the electron
images between 30 $\mu$s and 70 $\mu$s (Fig. 3a and 3d). The extra electron
lobes are in the E$\times B_{\perp}$ direction, and they show up in the other
side of the main electron cloud if we change the sign of $B_{\perp}$. We did
not observe any changes in the ion images and ion current in the cross-field
configuration compared to those without a magnetic field. This is quite
surprising, as one typically expects the electron distribution to closely
follow the ion distribution for a cold neutral plasma. In order to confirm
that the extra lobes of the electron images are coming from the plasma itself,
rather than dynamics during the flight to the detector of the projection
imaging technique, we took a series of images at different extraction
parameters such as the high-voltage amplitude and width and the accelerating
voltages on the middle and front grids (located between the plasma and the
detector, and strongly affect the sizes of electron images). We always observe
the similiar results as Fig. 3, implying this is a good measure of the
electron distribution in the plasma. Using a simple model of the electron
space charge effects during transit to the detector, we find the separation
between the main cloud and the extra lobes before time of flight to be about
0.5-1$\sigma_{t}$, where $\sigma_{t}$ is the plasma size at delay time t. That
is, the extra lobes are located inside the plasma.
There are a multitude of different instabilities in plasmas, but given our
parameters we can constrain possible choices to a small number. We identify
the periodic electron emission signals as well as the extra lobes in the
electron projection images as a signature of a plasma instability due to
electrons drifting relative to ions across the magnetic field. The crossed
magnetic and electric fields will drive the electrons to drift with a velocity
$V_{d}$ ($V_{d}=E/B_{\perp}$) in the $E\times B_{\perp}$ direction. The ions
are unmagnetized, not affected by the small $B_{\perp}$ (several G), due to
their large mass (the ion gyroradius is much larger than the UCP size and the
ion gyro-period is much longer than the UCP lifetime). In our UCP the
electrons are unmagnetized in the first $\sim$30 $\mu$s because the electron
gyroradius is about the same order of magnitude as the UCP size and the
electron collision rate is higher than the electron cyclotron frequency. As
the UCP expands, the plasma size gets larger, the plasma density gets lower
and the electron temperature gets smaller (due to various cooling mechanisms)
fletcher2007 . The electrons become magnetized at about 30-50 $\mu$s after the
formation of the UCP because the electron collsion rate starts to be less than
the electron gyrofrequency and the plasma size at that time is about a factor
of 10 larger than the electron gyroradius, which is consistent with our
observation that the periodic emissions start at the same time. The frequency
is in the range from 50 KHz up to a few hundred KHz, which is much larger than
the ion cyclotron frequency $f_{ci}$ ($\sim$12 Hz at 1 G) and much less than
the electron cyclotron frequency $f_{ce}$ ($\sim$2.8 MHz at 1 G) and the
electron plasma wave $f_{pe}$ which is about 10-20 MHz at the delay time of
about 30-50 $\mu$s. The periodic emission signal is roughly independent of
plasma density (or plasma frequency) as shown in Fig. 1 (note that the plasma
density drops by a factor of 8 from 50 $\mu$s to 100 $\mu$s). There are
periodic emissions with different frequencies (Fig. 1b and 1d), which also
indicates that we have plasma instability (mode switching) in the cross-field
configuration.
A candidate instability is the high-frequency electron drift instability,
which has been studied in Hall thrusters (a type of plasma-based propulsion
systems for spacecrafts) experimentally litvak2004 ; lazurenko2005 and
theoretically adam2004 ; ducrocq2006 in the past few years to explain the
transport of electrons across the magnetic field lines. A 2D fully kinetic
model of the Hall thruser developed in ref. adam2004 has demonstated that the
large drift velocity at the exhaust of the thruster was the source of an
instability that gives rise to plasma turbulence and could induce a
significant current across the magnetic field. It is a high-frequency electron
drift instability with frequency $f_{r}$ lower than the electron cyclotron
frequency $f_{ce}$ and short wavelength close to the electron gyroradius
$r_{ce}$, which is studied in ref. ducrocq2006 .
The theory of high-frequency electron drift instability is developed from the
dispersion equation of electrostatic waves in a hot magnetized electron beam
drifting across a magnetic field with unmagnetized cold ions, and is closely
related to the ion-acoustic-wave instabilities in other plasma systems, such
as non-specular radar meteor trails oppenheim2003 , the ionosphere of the
earth farley1963 ; buneman1963 , dusty plasmas Merlino1998 and magnetic
pulses krall1971 , but they usually restrict the analysis to the cases where
the drift velocity $V_{d}$ is much smaller than the electron thermal velocity
$V_{eth}$. For high-frequency electron drift instabilities in Hall thrusters
and here, $V_{d}$ is about the same order of magnitude as the $V_{eth}$, and
much larger than the magnetic field gradient drift velocity and the density
gradient drift velocity.
Figure 4: Numerical solutions of the 1D dispersion relation of equation (1)
as in ducrocq2006 , but for our condition. (a) represents the envelopes of the
real (frequency) and the imaginary (growth rate) as a function of
$k_{y}V_{d}/\omega_{pe}$. (b) the corresponding ones for small wavenumbers.
The electron cyclotron frequency is equal to $0.2\omega_{pe}$, and
$V_{eth}/V_{d}=0.5$.
Assuming an electrostatic field perturbation
$\phi=\phi_{0}\times\rm{exp}(ik\cdot r-i\omega t)$ and a Maxwellian electron
distribution function with mean velocity $V_{d}$ and temperature $T_{e}$, the
dispersion equation can be written in the limit $k_{z}=0$ (where $k_{z}$ is
the wavenumber along $B_{\perp}$) ducrocq2006 :
$\displaystyle
k^{2}_{xy}\lambda^{2}_{D}\left(1-\frac{m}{M}\frac{\omega^{2}_{pe}}{\omega^{2}}\right)$
$\displaystyle=$ $\displaystyle\sum^{n=\infty}_{n=1}\frac{2{(\omega-
k_{y}V_{d})}^{2}I_{n}(b)e^{-b}}{{(\omega-
k_{y}V_{d})}^{2}-{(n\Omega_{ce})}^{2}}$ (1) $\displaystyle+(I_{0}(b)e^{-b}-1)$
where $k_{xy}$ is the wavenumber in the plane perpendicular to $B_{\perp}$;
$\lambda^{2}_{D}=k_{B}T_{e}/m\omega^{2}_{pe}$ is the Debye length;
$\Omega_{ce}$ is the electron cyclotron frequecy; $\omega_{pe}$ is the
electron plasma frequency; $k_{y}$ is the wavenumber in the $E\times
B_{\perp}$ direction, and $b=k^{2}_{y}V^{2}_{eth}/\Omega^{2}$. The functions
$I_{n}$ are modified Bessel functions of order n.
In order to find the dependence of $\omega_{r}$ and $\omega_{i}$ on the
wavenumber $k_{y}$, we have solved numerically equation (1) by taking a
further approximation to neglect the perturbations perpendicular to the drift
motion ($k_{x}=0$, the wavenumber along E) because the electron drift velocity
$V_{d}$ is about an order of magnitude larger than the density gradient drift
velocity. Figure 4 shows the frequency and the corresponding growth rate of
the unstable modes as a function of $k_{y}$. Figure 4a and 4b are the
envelopes of the solutions and the first 5 modes, respectively. The values of
$f_{r}$ for the maximum growth rate correspond to frequencies ranging from 90
KHz to 140 KHz for $f_{pe}=21$ MHz, $B_{\perp}=2$ G, $E=20$ mV/cm and
$T_{e}=2.7$ K at delay time of about 30 $\mu$s (where $T_{e}$ is from
fletcher2007 ), which agrees with the measured frequencies at $V_{d}=10^{4}$
m/s in Fig. 4. The ratio of the frequency to the drift velocity is about 9-14
$m^{-1}$, in agreement with the measured slope of Fig. 2 (10.2 $m^{-1}$). The
maximum of the growth rate is reached for $k_{y}V_{d}/\omega_{pe}=1.2$, and
the corresponding wavelength is about 0.5 mm, close to the electron gyroradius
(about 0.2 mm for $B_{\perp}=2$ G and $T_{e}=2.7$ K). We can also see the
transitions from stability to instability whenever $k_{y}V_{d}/\omega_{pe}$ is
close to a cyclotron harmonic $n\Omega_{ce}/\omega_{pe}$. The growth rate
reaches a maximum and then decreases sharply between each cyclotron harmonic,
and is separated by stable regions. The frequency is several orders of
magnitude below the growth rate except in the vicinity of the maximum.
The frequency roughly linearly depends on the drift velocity (below
$k_{y}V_{d}/\omega_{pe}=1$), but is independent of the plasma frequency (i.e.
plasma density) before the frequency reaches the maximum value (both axis are
scaled by electron plasma frequency in Fig. 4a), which explains the lack of
density dependence seen in Fig. 1. We measure the growth rate by suddenly
applying the electric field at different delay times, and find the periodic
emission occurs within 1-2 $\mu$s after application of the field,
corresponding to a growth rate of $\sim$ 100-150 KHz for $B_{\perp}=2$ G and
$E=20$ mV/cm, which is about the same as the frequency $f_{r}$.
Although we see large scale density structures in the electron images, we do
not have a model of the microscopic electron motion that produces the periodic
pulses. Note that the applied magnetic field should suppress electron
detection, since it is transverse to the detector direction. This is seen in
the strong suppression of the prompt electron emission peak. Nevertheless, we
detect large pulses of electrons, presumably due to large electron
trajectories that extend past the grid (1.5 cm below the plasma) where the
large acceleration field can direct the electrons to the detector, even in the
presence of the transverse magnetic field.
In summary, we have observed large periodic electron emission and splitting of
the electron distribution into two or three lobes from an expanding UCP in the
presence of crossed magnetic and electric fields. We identify them as a
signature of high-frequency electron drift instability due to the electrons
drifting relative to the ions across the magnetic field. We calculate the
unstable mode frequencies and growth rates by solving 1D dispersion equation
and find good agreement. The large scale changes in the electron spatial
distribution remain a mystery, as does the exact mechanism that leads to the
emission of electrons. This work shows that UCPs will continue to provide an
interesting place to study fundamental plasma physics phenomena.
###### Acknowledgements.
This work was supported by the National Science Foundation PHY-0714381.
## References
* (1) T. C. Killian et al., Phys. Rev. Lett. 83, 4776 (1999).
* (2) J. L. Roberts et al., Phys. Rev. Lett. 92, 253003 (2004).
* (3) P. Gupta et al., Phys. Rev. Lett. 99, 075005 (2007).
* (4) R. S. Fletcher et al., Phys. Rev. Lett. 99, 145001 (2007).
* (5) S. Kulin et al., Phys. Rev. Lett. 85, 318 (2000).
* (6) S. D. Bergeson et al., Phys. Rev. E 67, 026414 (2003).
* (7) F. Robicheaux et al., J. Phys. Plasmas 10, 2217 (2003).
* (8) R. S. Fletcher et al., Phys. Rev. Lett. 96, 105003 (2006).
* (9) G. D. Conway et al., Phys. Rev. Lett. 84, 1463 (2000).
* (10) M. Oppenheim et al., J. Geophys. Res. 108, 0163 (2003).
* (11) D. T. Farley, Phys. Rev. Lett. 10, 279 (1963).
* (12) O. Buneman, Phys. Rev. Lett. 10, 7 (1963).
* (13) R. L. Merlino et al., Phys. Plasma 5, 1607 (1998).
* (14) N. A. Krall and P. C. Liewer, Phys. Rev. A 4, 5 (1971).
* (15) A. A. Litvak et al., Phys. Plasma 11, 1701 (2004).
* (16) A. Lazurenko et al., Phys. Plasma 12, 013501 (2005).
* (17) J. C. Adam et al., Phys. Plasma 11, 295 (2004).
* (18) A. Ducrocq et al., Phys. Plasma 13, 102111 (2006).
* (19) X. L. Zhang et al., Preprint arXiv:0806.1511. (2008).
* (20) X. L. Zhang, et al., Phys. Rev. Lett. 100, 235002 (2008).
| arxiv-papers | 2008-06-29T12:15:35 | 2024-09-04T02:48:56.429772 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "X. L. Zhang, R. S. Fletcher, and S. L. Rolston",
"submitter": "Xianli Zhang",
"url": "https://arxiv.org/abs/0806.4691"
} |
0806.4706 | # Measuring the intrinsic charge transfer gap using
K-edge X-ray absorption spectroscopy
C. Gougoussis, M. Calandra, A. Seitsonen, Ch. Brouder, A. Shukla and F. Mauri
CNRS and Institut de Minéralogie et de Physique des Milieux condensés, case
115, 4 place Jussieu, 75252, Paris cedex 05, France
###### Abstract
Pre-edge features in X-ray absorption spectroscopy contain key information
about the lowest excited states and thus on the most interesting physical
properties of the system. In transition metal oxides they are particularly
structured but extracting physical parameters by comparison with a calculation
is not easy due to several computational challenges. By combining core-hole
attraction and correlation effects in first principles approach, we calculate
Ni K-edge X-ray absorption spectra in NiO. We obtain a striking, parameter-
free agreement with experimental data and show that dipolar pre-edge features
above the correlation gap are due to non-local excitations largely unaffected
by the core-hole. We show that in charge transfer insulators, this property
can be used to measure the correlation gap and probe the intrinsic position of
the upper-Hubbard band.
###### pacs:
The description of electronic excitations in correlated materials is a
challenge since many relevant phenomena, such as magnetism in transition metal
(TM) compounds or high Tc superconductivity in doped Mott-insulators are a
consequence of strong electron-electron interaction. Recently, core-hole
spectroscopy unveiled unexpected electronic excitations in correlated
antiferromagnetic insulators Hasan ; Hayashi as well as in high Tc
superconductors Collart ; Kim_Hill_PRL . In La2CuO4, K$\alpha$ resonant
inelastic X-ray scattering (RIXS) and theoretical calculationsShukla
demonstrated that dipolar pre-edge features just above the correlation gap are
due to intersite Cu 4p-3d hybridization. Since in correlated insulators the TM
empty d-states form the upper Hubbard bands, it is crucial to see (i) if off-
site pre-edge excitations occur in other correlated compounds Vanko , (ii) to
what extent their energy depends on the presence of a core-hole in the final
state.
The presence of a core-hole substantially complicates the interpretation of
X-ray absorption (XAS) spectra. Since core-hole attraction shifts the empty
d-states much more than the empty p-states, typically by several eV, it is
unclear to what extent the energy of the pre-edge features can provide useful
information on the position of the corresponding excitations in the material
in the absence of a core-hole. Furthermore, in correlated insulators core-hole
attraction and Hubbard repulsion partially compensate each other. Thus the use
of fitting parameters to describe these interactions, as is commonly done in
the literature Kotani_RMP , does not allow to distinguish between the two
effects. A treatment of core-hole attraction and Hubbard-repulsion from first
principles is needed.
In the case of K-edge XAS spectra, theoretical calculations are difficult
since excitations in a large energy window above the Fermi level need to be
described. Thus methods dealing successfully with correlation effects in NiO
in an energy window close to the Fermi level (such as DMFT KunesPRL07 or
cluster calculations TaguchiPRL08 ) cannot be used to describe pre-edge and
near-edge structures in K-edge XAS spectra. This is relevant since (i) low
energy pre-edge features provide relevant information on the local environment
of the absorbing atom Uozumi92 ; Joly99 ; Uozumi ; Shukla ; Vanko and (ii) it
has been suggested that in correlated oxides “shakedown” excitations occur as
near-edge structures Tolentino .
In this work we present a first-principles approach to describe the K-edge XAS
spectra in correlated insulators. The method is based on a recently developed
DFT+U method and includes core-hole attractiontaillefumier . Differently from
other DFT+U schemes Anisimov , the U parameter is obtained by linear-response
cococcioni1 ; cococcioni2 . Consequently, U is not a fitting parameter but an
intrinsic linear response property since it measures the spurious curvature of
the energy functional as a function of occupation cococcioni1 . We apply the
method to the K-edge XAS of NiO, which is the prototype antiferromagnetic
correlated insulator and whose ground state and excitations are incorrectly
described by standard DFT approaches. We demonstrate that the pre-edge dipolar
features above the correlation gap are due to non-local excitations to second
nearest-neighboring Ni atoms reflecting the superexchange interaction. We show
that, due to its non-local nature, the dipolar pre-edge feature is unshifted
by core-hole attraction and it is, thus, a measure of the upper-Hubbard band
in the absence of a core-hole in the final state. Finally we exploit the
insensitivity of the pre-edge dipolar features on core-hole attraction to show
how to measure the charge transfer gap using K-edge XAS.
We use the NiO experimental crystal structure. The paramagnetic-cell group-
space is then Fm3̄m, Ni occupies the 4a position and O the 4b. The cubic
lattice parameter is $a=4.1788\AA$ for NiO srivastava . DFT calculations are
performed using the Quantum-ESPRESSO code PWSCF . We use Troullier-Martins
Troullier pseudopotentials, the spin-polarized generalized gradient
approximation (GGA) PBE and the recently developed DFT+U method of ref.
cococcioni1 ; cococcioni2 . The wave functions are expanded using a 140 Ry
energy cutoff for NiO. The calculated value of U for NiO is U=7.6 eV.
XAS spectra are computed in a supercell approach including core-hole effects
in the pseudopotential of the absorbing atom. From the self-consistent charge
density the XAS spectra are obtained using the continued fraction taillefumier
and the PAW method of ref. Blochl . We considered $2\times 2\times 2$
supercells of the magnetic cell (32 atoms). We used $4\times 4\times 4$
k-points grid both for the charge density and the continued fraction
calculation.
Figure 1: (Color online) Calculated and measured vedrinskii Ni K-edge XAS of
NiO including partial fluorescence yield (PFY) (this work). The dashed (solid)
curves correspond to $e_{g}$ ($t_{2g}$) orientation.
In fig. 1 we show calculated NiO XAS spectra and their decomposition into
dipolar and quadrupolar parts using GGA and GGA+U, as compared to experimental
data vedrinskii for single crystals. We consider the following two sets of
polarization ${\bm{\epsilon}}$ and wavevector ${\bf k}$: (i)
${\bm{\epsilon}}\parallel[110]$ and ${\bf k}\parallel[-110]$ , denoted by
$e_{g}$ orientation (dashed lines) and (ii) ${\bm{\epsilon}}\parallel[100]$
and ${\bf k}\parallel[010]$, denoted by $t_{2g}$ orientation (solid lines).
The directions are in terms of the paramagnetic cubic crystal cell. We note
that U has no effect on high-energy near-edge and far-edge features (labeled
C,D,E). On the contrary the energy and angular dependence of pre-edge
structures (see insets in Fig. 1) are incorrect in the framework of GGA. Peak
B is at too low energy and a too large mixing occurs between structures A and
B. In the GGA+U calculation the dipolar peak B is shifted to $\approx 2$ eV
higher energies and the angular dependence is in excellent agreement with
experimental data. The CGA+U calculation also shows that a very small dipolar
component is present in peak A, commonly interpreted as purely quadrupolar.
Detailed analysis of the angular dependence of the experimental NiO XAS
spectra vedrinskii shows that even for the $t_{2g}$ orientation, where no
quadrupolar transition occurs because the t2g states are occupied, a small
peak is present ApeakNiO , in agreement with our findings. Thus, in correlated
insulators, the use of the DFT+U approximation is mandatory to obtain a
correct description of the pre-edge features.
To complete the understanding of the pre-edge features in NiO we resolve the
XAS spectrum in its spin dependence and calculate the density of states
projected over atomic orbitals using Löwdin projections. Without loss of
generality we have considered the absorbing Ni to be spin-up polarized. As can
be seen in Fig. 2, the quadrupolar part of peak A is mostly due to intra-site
excitations to d-states lowered by core-hole attraction. However, since the
hybridization between Ni and O is very strong, core-hole attraction lowers a
small portion of the O 2p states generating a small dipolar component in peak
A. In a atomic orbital picture, peak A is due to direct dipole transitions
from Ni 1s states to O 2p states. In our calculation the intensity of this
excitations is of the order of $1\%$ of the edge jump. In experiments it is
somewhat smaller because of the larger linewidth. This estimate is crucial for
the quantitative description of this off-site excitation in multiplet
calculation.
The spin-resolved dipolar spectrum shows that peak B is mostly due to
transition to up spin-polarized states. Since the absorbing atom has 5 up
electrons in d-states, the B excitation must have an off-site component.
Löwdin projections demonstrate that it is due to transition to on-site Ni
empty 4p-states hybridized to empty 3d-states of next-to-nearest neighbours Ni
atoms. This off-site excitation is then a fingerprint of the hopping process
leading to superexchange in NiO and a direct probe of the upper Hubbard band
in NiO.
Figure 2: (Color online) Comparison between GGA+U calculated K-edge Ni XAS and
Löwdin projected density of states.
To understand the role of core-hole effects on the A and B features we have
performed calculations with and without a core hole in the final state. We
found that peak A is shifted by 4 eV with respect to the edge by core-hole
attraction. On the contrary peak B is essentially unshifted. The difference in
the behavior of the two peaks is due to the non-local nature of peak B.
Figure 3: (Color online) Calculated pre-edge features of Ni K-edge XAS spectra
of LixNi1-xO with (a) and without (b) core-hole effects using a rigid doping
(solid line) approach or adding a compensating charged background (red-dashed
lines). Peaks labeling is the same as in Fig. 1, VB stands for Valence Band.
(c) Band structure of NiO along $\Gamma$X.
We now demonstrate that the non-local nature of the B excitation and its weak
dependence on core-hole effects can be used to measure the correlation gap
between the $|3d^{8}\underline{L}\rangle$ and the $|3d^{9}L\rangle$ states. We
consider hole-doped NiO. In practice this can be achieved through Li doping
ReinertZPHYSB1995 , as in the case of LixNi1-xO. A large range of doping can
be experimentally obtained with $0<x<0.7$. It has been shown KuiperPRL89 ;
vanElp that, when NiO is doped with holes, the holes mainly reside on O
atoms, due to the charge transfer character of NiO. Thus the top of the NiO
valence band has mainly p character. This suggests that in Ni K-edge XAS of
LixNi1-xO, in the low doping regime, an additional dipolar peak should occur
in the pre-edge, resulting from the holes entering the top of the valence
band. This is confirmed by oxygen K-edge absorption where the top of the
valence band is seen even at dopings as low as $x=0.05$)KuiperPRL89 . At
somewhat larger doping, a similar peak due to Oxygen p-holes should also occur
in the dipolar part of Ni K-edge XAS. Since no experimental data are available
in the literature, we directly simulate Ni K-edge XAS in LixNi1-xO.
We calculate hole-doping of NiO in two different ways, namely (i) by rigid-
band doping of NiO and (ii) adding a compensating charge background and
recalculating self-consistently the charge-density and the XAS spectra. The
results of the Ni K-edge XAS in LixNi1-xO in the pre-edge region with and
without a core-hole in the final state are shown in Fig. 3. We find that the
top of the valence band should be visible in Ni K-edge XAS at dopings of
$x\approx 15-20\%$. In the calculated spectra, with or without a core-hole in
the final state, the energy position of the top of the valence band and of
peak B are independent on doping. The agreement between the rigid band picture
and the compensating-charge background calculation validates the rigid band
picture for $0<x<0.2$
In NiO the experimentally measurable gap is due to the excitation between the
top of the valence band and the empty d-states. In DFT+U Bengone and GW
Faleev calculations the lowest single particle excitation is between the
valence band and an s band occurring at energies lower than the empty d states
(see Fig. 3). This band is invisible in experiments since its optical matrix
elements are extremely weak Bengone . Moreover its also invisible in K-edge
XAS since it is highly dispersive and has mostly Ni s and O s components. Thus
in the absence of core-hole effects, the distance between the top of the
valence band and peak B is a measure of the correlation gap, being the
dependence of the energy position on doping very weak (see Fig. 3, right
panel). When core-hole attraction is considered (see Fig. 3, left panel), we
find that, despite the occurrence of a core-hole exciton (peak A), the
distance between the top of the valence band and peak B is very weakly
affected ($\approx 5\%$). Consequently in lightly hole-doped NiO the distance
between the top of the valence band and peak B is a measure of the correlation
gap even in the presence of a core-hole in the final state.
In our calculation the charge-transfer gap is 5 eV (peak-to-peak distance). In
NiO optical absorption starts at $3.1$ eV and reaches its maximum at $4.0$ eV
Sawatzky84 ; HuefnerReview . This value is reduced respect to our due to
excitonic effects. In photoemission and inverse photoemission Huefner86 the
gap (peak-to-peak distance) is roughly 5.5 eV, in fairly good agreement with
our value. Thus our proposal allows for an independent estimate of the optical
gap in NiO.
The procedure outlined for NiO can be used to measure the charge-transfer gap
and the upper Hubbard band in other charge-transfer insulators. In these
systems the top of the valence band is due to Oxygen p-states and it is thus
visible in dipolar TM K-edge XAS upon doping. If the doping necessary to
detect this feature is low enough not to affect substantially the electronic
structure then a rigid-band doping picture applies and the top of the valence
band is at the same energy as in the undoped system. Moreover, hole states
being dipolar in nature, the effects of core-hole attraction are weak. The top
of the valence band of the doped system can then be used as a reference energy
for the measurement of the excitations seen in TM K-edge XAS. The second
necessary condition is the occurrence of non-local dipolar features Shukla ;
Vanko (peak B) in the pre-edge region due to transitions to d-states of
neighboring TM atoms promoted by hybridization with the TM absorbing atom
p-states. If these excitations are visible, we have shown that they represent
the upper Hubbard band of the material and the energy difference between this
excitation and the top of the valence band in the weakly hole-doped system is
a measure of the charge transfer gap. Thus, TM K-edge XAS in weakly hole-doped
charge-transfer insulators is unbiased experimental tool to measure the
charge-transfer gap.
In this work we have developed a new first-principles parameter-free method to
calculate K-edge XAS spectra including core-hole effects in the final state
and electronic correlation at the DFT+U level. We have shown that the method
provides spectra in excellent agreement with experimental data for NiO, the
prototype correlated insulator. We have interpreted all the pre-edge and near-
edge features in the experimental data. In particular, we have identified a
dipolar pre-edge feature as due to non-local excitations to d-states of the
second nearest-neighboring Ni atoms, namely the upper Hubbard band. We have
shown that, due to its non-local nature, this excitation is unaffected by the
presence of a core-hole in the final state. Starting from this result we have
proposed a new way to measure the intrinsic correlation gap in charge-transfer
insulators based on TM K-edge XAS. This method is complementary to optical
measurements and more straightforward than a combined photoemission-inverse
photoemission experiment.
We acknowledge useful conversation with A. Kotani, G. Dräger, O.Šipr , Ph.
Sainctavit, M. Arrio, D. Cabaret, A. Juhin, H. Kulik, N. Marzari and L.
Reining. Calculations were performed at the IDRIS supercomputing center
(project 071202).
## References
* (1) M. Z. Hasan, E. D. Isaacs, Z.-X. Shen, L. L. Miller, K. Tsutsui, T. Tohyama, S. Maekawa, Science 288,5472 (2000)
* (2) H. Hayashi, Y. Udagawa, W. A. Caliebe, and C. -C. Kao, Phys. Rev. B 66, 033105 (2002)
* (3) E. Collart, A. Shukla, J.-P. Rueff, P. Leininger, H. Ishii, I. Jarrige, Y. Q. Cai, S.-W. Cheong, and G. Dhalenne, Phys. Rev. Lett. 96, 157004 (2006)
* (4) Y. J. Kim, J. P. Hill, C. A. Burns, S. Wakimoto, R. J. Birgeneau, D. Casa, T. Gog, and C. T. Venkataraman, Phys. Rev. Lett. 89, 177003 (2002)
* (5) A. Shukla, M. Calandra, M. Taguchi, A. Kotani, György Vanko, and S.-W. Cheong, Phys. Rev. Lett. 96, 077006 (2006)
* (6) G. Vanko, F. M. F. de Groot, S. Huotari, R. J. Cava, T. Lorenz, M. Reuther, arXiv:0802.2744
* (7) A. Kotani and S. Shin, Rev. Mod. Phys. 73, 203 (2001)
* (8) J. Kunes, V. I. Anisimov, S. L. Skornyakov, A. V. Lukoyanov, D. Vollhardt, Phys. Rev. Lett. 99, 156404 (2007)
* (9) M. Taguchi, M. Matsunami, Y. Ishida, R. Eguchi, A. Chainani, Y. Takata, M. Yabashi, K. Tamasaku, Y. Nishino, T. Ishikawa, Y. Senba, H. Ohashi, S. Shin, Physical Review Letters 100, 206401 (2008)
* (10) T. Uozumi, K. Okada, A. Kotani, O. Durmeyer, J. P. Kappler, E. Beaurepaire, and J. C. Parlebas, Europhys. Lett. 18, 85 (1992)
* (11) Y. Joly, D. Cabaret, H. Renevier and C. R. Natoli, Phys. Rev. Lett. 82, 2398 (1999)
* (12) T. Uozumi, K. Okada, A. Kotani, O. Durmeyer, J. P. Kappler, E. Beaurepaire and J. C. Parlebas, Europhys. Lett. 18, 85 (1992)
* (13) H. Tolentino, M. Medarde, A. Fontaine, F. Baudelet, E. Dartyge, D. Guay, and G. Tourillon, Phys. Rev. B 45, 8091 (1992)
* (14) M. Taillefumier, D. Cabaret, A-M. Flank, and F. Mauri, Phys. Rev. B 66, 195107 (2002)
* (15) V. I. Anisimov, J. Zaanen, and O. K. Andersen, Phys. Rev. B 44, 943 (1991), V. I. Anisimov, F. Aryasetiawan, and A. I. Liechtenstein, J. Phys. Condens. Matter 9, 767 (1997).
* (16) M. Cococcioni and S. de Gironcoli, Phys. Rev. B 71, 035105 (2005)
* (17) H.J. Kulik, M.Cococcioni, D.A. Scherlis and N. Marzari , Phys. Rev. Lett. 97, 103001 (2006)
* (18) S.P. Srivastava, R.C. Srivastava, I.D. Singh, S.D. Pandey and P.L. Gupta , J. Phy. Soc. Jap. 43 , 885 (1977)
* (19) http://www.pwscf.org, S. Baroni, et al., Rev. Mod. Phys. 73, 515-562 (2001)
* (20) N. Troullier and J. L. Martins, Phys. Rev. B 43, 1993 (1991)
* (21) J.P.Perdew, K.Burke, M.Ernzerhof, Phys. Rev. Lett. 77, 3865 (1996)
* (22) P. Blöchl, Phys. Rev. B 50, 17 953 (1994)
* (23) R.V. Vedrinskii, V.L. Kraizman, A.A. Novakovich, S.M. Elyafi, S.Bocharov, T.Kirchner, G. Dräger, Phys. Stat. Sol. 226, 203 (2001)
* (24) In fig. 1 of this work and in fig. 3 of ref. vedrinskii a very small peak is present for t2g orientation.
* (25) F. Reinert, P. Steiner, S. Hüfner, H. Schmitt, J. Fink, M. Knupfer, P. Sandl, E. Bertel, Z. Phys. B 97, 83 (1995)
* (26) P. Kuiper, G. Kruizinga, J. Ghijsen, G. A. Sawatzky, and H. Verweij, Phys. Rev. Lett. 62, 221 (1989)
* (27) J. van Elp, H. Eskes, P. Kuiper, and G. A. Sawatzky, Phys. Rev. B 45, 1612 (1992).
* (28) O. Bengone, M. Alouani, P. Blöchl, and J. Hugel, Phys. Rev. B 62, 16392 (2000)
* (29) S. V. Faleev, M. van Schilfgaarde, and T. Kotani, Phys. Rev. Lett. 93, 126406 (2004)
* (30) G. A. Sawatzky and J. W. Allen, Phys. Rev. Lett. 53, 2339 (1984)
* (31) S. Hüfner, Advances in Physics 43, 183, (1994)
* (32) S. Hüfner and T. Riesterer, Phys. Rev. B, 33 7267 (1986)
| arxiv-papers | 2008-06-28T16:35:28 | 2024-09-04T02:48:56.433695 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "C. Gougoussis, M. Calandra, A. Seitsonen, Ch. Brouder, A. Shukla and\n F. Mauri",
"submitter": "Matteo Calandra",
"url": "https://arxiv.org/abs/0806.4706"
} |
0806.4790 | 2010119-130Nancy, France 119
Vladimir Braverman
Kai-Min Chung
Zhenming Liu
Michael Mitzenmacher
Rafail Ostrovsky
# AMS Without $\bf{4}$-Wise Independence on Product Domains111This paper is a
merge from the work of [7, 9, 10]
V. Braverman University of California Los Angeles. Supported in part by NSF
grants 0716835, 0716389, 0830803, 0916574 and Lockheed Martin Corporation.
vova@cs.ucla.edu http://www.cs.ucla.edu/~vova , K. Chung Harvard School of
Engineering and Applied Sciences. Supported by US-Israel BSF grant 2006060 and
NSF grant CNS-0831289. kmchung@fas.harvard.edu
http://people.seas.harvard.edu/~kmchung/ , Z. Liu Harvard School of
Engineering and Applied Sciences. Supported in part by NSF grant CNS-0721491.
The work was finished during an internship in Microsoft Research Asia.
zliu@fas.harvard.edu http://people.seas.harvard.edu/~zliu/ , M. Mitzenmacher
Harvard School of Engineering and Applied Sciences. Supported in part by NSF
grant CNS-0721491 and research grants from Yahoo!, Google, and Cisco.
michaelm@eecs.harvard.edu http://www.eecs.harvard.edu/~michaelm/ and R.
Ostrovsky University of California Los Angeles. Supported in part by IBM
Faculty Award, Lockheed-Martin Corporation Research Award, Xerox Innovation
Group Award, the Okawa Foundation Award, Intel, Teradata, NSF grants 0716835,
0716389, 0830803, 0916574 and U.C. MICRO grant. rafail@cs.ucla.edu
http://www.cs.ucla.edu/~rafail
###### Abstract.
In their seminal work, Alon, Matias, and Szegedy introduced several sketching
techniques, including showing that $4$-wise independence is sufficient to
obtain good approximations of the second frequency moment. In this work, we
show that their sketching technique can be extended to product domains
$[n]^{k}$ by using the product of $4$-wise independent functions on $[n]$. Our
work extends that of Indyk and McGregor, who showed the result for $k=2$.
Their primary motivation was the problem of identifying correlations in data
streams. In their model, a stream of pairs $(i,j)\in[n]^{2}$ arrive, giving a
joint distribution $(X,Y)$, and they find approximation algorithms for how
close the joint distribution is to the product of the marginal distributions
under various metrics, which naturally corresponds to how close $X$ and $Y$
are to being independent. By using our technique, we obtain a new result for
the problem of approximating the $\ell_{2}$ distance between the joint
distribution and the product of the marginal distributions for $k$-ary
vectors, instead of just pairs, in a single pass. Our analysis gives a
randomized algorithm that is a $(1\pm\epsilon)$ approximation (with
probability $1-\delta$) that requires space logarithmic in $n$ and $m$ and
proportional to $3^{k}$.
###### Key words and phrases:
Data Streams, Randomized Algorithms, Streaming Algorithms, Independence,
Sketches
###### 1991 Mathematics Subject Classification:
F.2.1, G.3
## 1\. Introduction
In their seminal work, Alon, Matias and Szegedy [4] presented celebrated
sketching techniques and showed that $4$-wise independence is sufficient to
obtain good approximations of the second frequency moment. Indyk and McGregor
[12] make use of this technique in their work introduce the problem of
measuring independence in the streaming model. There they give efficient
algorithms for approximating pairwise independence for the $\ell_{1}$ and
$\ell_{2}$ norms. In their model, a stream of pairs $(i,j)\in[n]^{2}$ arrive,
giving a joint distribution $(X,Y)$, and the notion of approximating pairwise
independence corresponds to approximating the distance between the joint
distribution and the product of the marginal distributions for the pairs.
Indyk and McGregor state, as an explicit open question in their paper, the
problem of whether one can estimate $k$-wise independence on $k$-tuples for
any $k>2$. In particular, Indyk and McGregor show that, for the $\ell_{2}$
norm, they can make use of the product of $4$-wise independent functions on
$[n]$ in the sketching method of Alon, Matias, and Szegedy. We extend their
approach to show that on the product domain $[n]^{k}$, the sketching method of
Alon, Matias, and Szegedy works when using the product of $k$ copies of
$4$-wise independent functions on $[n]$. The cost is that the memory
requirements of our approach grow exponentially with $k$, proportionally to
$3^{k}$.
Measuring independence and $k$-wise independence is a fundamental problem with
many applications (see e.g., Lehmann [13]). Recently, this problem was also
addressed in other models by, among others, Alon, Andoni, Kaufman, Matulef,
Rubinfeld and Xie [1]; Batu, Fortnow, Fischer, Kumar, Rubinfeld and White [5];
Goldreich and Ron [11]; Batu, Kumar and Rubinfeld [6]; Alon, Goldreich and
Mansour [3]; and Rubinfeld and Servedio [15]. Traditional non-parametric
methods of testing independence over empirical data usually require space
complexity that is polynomial to either the support size or input size. The
scale of contemporary data sets often prohibits such space complexity. It is
therefore natural to ask whether we will be able to design algorithms to test
for independence in streaming model. Interestingly, this specific problem
appears not to have been introduced until the work of Indyk and McGregor.
While arguably results for the $\ell_{1}$ norm would be stronger than for the
$\ell_{2}$ norm in this setting, the problem for $\ell_{2}$ norms is
interesting in its own right. The problem for the $\ell_{1}$ norm has been
recently resolved by Braverman and Ostrovsky in [8]. They gave an
$(1\pm\epsilon,\delta)$-approximation algorithm that makes a single pass over
a data stream and uses polylogarithmic memory.
### 1.1. Our Results
In this paper we generalize the “sketching of sketches” result of Indyk and
McGregor. Our specific theoretical contributions can be summarized as follows:
_Main Theorem._
Let ${\vec{v}}\in\mathrm{R}^{(n^{k})}$ be a vector with entries
${\vec{v}}_{{\mathbf{p}}}\in R$ for ${\mathbf{p}}\in[n]^{k}$. Let
$h_{1},\dots,h_{k}:[n]\rightarrow\\{-1,1\\}$ be independent copies of 4-wise
independent hash functions; that is, $h_{i}(1),\dots,h_{i}(n)\in\\{-1,1\\}$
are $4$-wise independent hash functions for each $i\in[k]$, and
$h_{1}(\cdot),\dots,h_{k}(\cdot)$ are mutually independent. Define
$H({\mathbf{p}})=\prod_{i=1}^{k}h_{j}(p_{j})$, and the sketch
$Y=\sum_{p\in[n]^{k}}\vec{v}_{\mathbf{p}}H(p)$.
We prove that the sketch $Y$ can be used to give an efficient approximation
for $\|{\vec{v}}\|^{2}$; our result is stated formally in Theorem 4.2. Note
that $H$ is not $4$-wise independent.
As a corollary, the main application of our main theorem is to extend the
result of Indyk and McGregor [12] to detect the dependency of $k$ random
variables in streaming model.
###### Corollary 1.1.
For every $\epsilon>0$ and $\delta>0$, there exists a randomized algorithm
that computes, given a sequence $a_{1},\dots,a_{m}$ of $k$-tuples, in one pass
and using $O(3^{k}\epsilon^{-2}\log\frac{1}{\delta}(\log m+\log n))$ memory
bits, a number $Y$ so that the probability $Y$ deviates from the $\ell_{2}$
distance between product and joint distribution by more than a factor of
$(1+\epsilon)$ is at most $\delta$.
### 1.2. Techniques and a Historical Remark
This paper is merge from [7, 9, 10], where the same result was obtained with
different proofs. The proof of [10] generalizes the geometric approach of
Indyk and McGregor [12] with new geometric observations. The proofs of [7, 9]
are more combinatorial in nature. These papers offer new insights, but due to
the space limitation, we focus on the proof from [9] in this paper. Original
papers are available on line and are recommended to the interested reader.
## 2\. The Model
We provide the general underlying model. Here we mostly follow the notation of
[7, 12].
Let $S$ be a stream of size $m$ with elements $a_{1},\dots,a_{m}$, where
$a_{i}\equiv(a^{1}_{i},\dots,a^{k}_{i})\in[n]^{k}$. (When we have a sequence
of elements that are themselves vectors, we denote the sequence number by a
subscript and the vector entry by a superscript when both are needed.) The
stream $S$ defines an _empirical_ distribution over $[n]^{k}$ as follows: the
frequency $f(\omega)$ of an element $\omega\in[n]^{k}$ is defined as the
number of times it appears in $S$, and the empirical distribution is
$\Pr[\omega]=\frac{f(\omega)}{m}\quad\mbox{for any $\omega\in[n]^{k}$.}$
Since $\omega=(\omega_{1},\dots,\omega_{k})$ is a vector of size $k$, we may
also view the streaming data as defining a joint distribution over the random
variables $X_{1},\dots,X_{k}$ corresponding to the values in each dimension.
(In the case of $k=2$, we write the random variables as $X$ and $Y$ rather
than $X_{1}$ and $X_{2}$.) There is a natural way of defining marginal
distribution for the random variable $X_{i}$: for $\omega_{i}\in[n]$, let
$f_{i}(\omega_{i})$ be the number of times $\omega_{i}$ appears in the $i$th
coordinate of an element of $S$, or
$f_{i}(\omega_{i})=\left|\\{a_{j}\in S:a^{i}_{j}=\omega_{i}\\}\right|.$
The empirical marginal distribution $\Pr_{i}[\cdot]$ for the $i$th coordinate
is defined as
$\mathrm{Pr}_{i}[\omega_{i}]=\frac{f_{i}(\omega_{i})}{m}\quad\mbox{for any
$\omega_{i}\in[n]$.}$
Next let $\vec{v}$ be the vector in ${\mathrm{R}}^{[n]^{k}}$ with
$\vec{v}_{\omega}=\Pr[\omega]-\prod_{1\leq i\leq k}\Pr_{i}[\omega_{i}]$ for
all $\omega\in[n]^{k}$. Our goal is to approximate the value
$\|\vec{v}\|\equiv\left(\sum_{\omega\in[n]^{k}}\left|\Pr[\omega]-\prod_{1\leq
i\leq k}\mathrm{Pr}_{i}[\omega_{i}]\right|^{2}\right)^{\frac{1}{2}}.$ (1)
This represent the $\ell_{2}$ norm between the tensor of the marginal
distributions and the joint distribution, which we would expect to be close to
zero in the case where the $X_{i}$ were truly independent.
Finally, our algorithms will assume the availability of 4-wise independent
hash functions. For more on 4-wise independence, including efficient
implementations, see [2, 16]. For the purposes of this paper, the following
simple definition will suffice.
###### Definition 2.1.
_(4-wise independence)_ A family of hash functions $\mathcal{H}$ with domain
$[n]$ and range $\\{-1,1\\}$ is _4-wise independent_ if for any distinct
values $i_{1},i_{2},i_{3},i_{4}\in[n]$ and any
$b_{1},b_{2},b_{3},b_{4}\in\\{-1,1\\}$, the following equality holds,
$\Pr_{h\leftarrow\mathcal{H}}\left[h(i_{1})=b_{1},h(i_{2})=b_{2},h(i_{3})=b_{3},h(i_{4})=b_{4}\right]=1/16.$
###### Remark 2.2.
In [12], the family of 4-wise independent hash functions $\mathcal{H}$ is
called 4-wise independent random vectors. For consistencies within our paper,
we will always view the object $\mathcal{H}$ as a hash function family.
## 3\. The Algorithm and its Analysis for $k=2$
We begin by reviewing the approximation algorithm and associated proof for the
$\ell_{2}$ norm given in [12]. Reviewing this result will allow us to provide
the necessary notation and frame the setting for our extension to general $k$.
Moreover, in our proof, we find that a constant in Lemma 3.1 from [12] that we
subsequently generalize appears incorrect. (Because of this, our proof is
slightly different and more detailed than the original.) Although the error is
minor in the context of their paper (it only affects the constant factor in
the order notation), it becomes more important when considering the proper
generalization to larger $k$, and hence it is useful to correct here.
In the case $k=2$, we assume that the sequence
$(a^{1}_{1},a^{2}_{1}),(a^{1}_{2},a^{2}_{2}),\dots,(a^{1}_{m},a^{2}_{m})$
arrives an item by an item. Each $(a^{1}_{i},a^{2}_{i})$ (for $1\leq i\leq m$)
is an element in $[n]^{2}$. The random variables $X$ and $Y$ over $[n]$ can be
expressed as follows:
$\left\\{\begin{array}[]{lllll}\Pr[i,j]&=&\Pr[X=i,Y=j]&=&|\\{\ell:(a^{1}_{\ell},a^{2}_{\ell})=(i,j)\\}|/m\\\
\Pr_{1}[i]&=&\Pr[X=i]&=&|\\{\ell:(a^{1}_{\ell},a^{2}_{\ell})=(i,\cdot)\\}|/m\\\
\Pr_{2}[j]&=&\Pr[Y=j]&=&|\\{\ell:(a^{1}_{\ell},a^{2}_{\ell})=(\cdot,j)\\}|/m.\end{array}\right.$
We simplify the notation and use $p_{i}\equiv\Pr[X=i]$, $q_{j}\equiv\Pr[Y=j]$,
$r_{i,j}=\Pr[X=i,Y=j]$. and $s_{i,j}=\Pr[X=i]\Pr[Y=j]$.
Indyk and McGregor’s algorithm proceeds in a similar fashion to the streaming
algorithm presented in [4]. Specifically let $s_{1}=72\epsilon^{-2}$ and
$s_{2}=2\log(1/\delta)$. The algorithm computes $s_{2}$ random variables
$Y_{1},Y_{2},\dots,Y_{s_{2}}$ and outputs their median. The output is the
algorithm’s estimate on the norm of $v$ defined in Equation 1. Each $Y_{i}$ is
the average of $s_{1}$ random variables $Y_{ij}$: $1\leq j\leq s_{1}$, where
$Y_{ij}$ are independent, identically distributed random variables. Each of
the variables $D=D_{ij}$ can be computed from the algorithmic routine shown in
Figure 1.
* $\textnormal{2-D Approximation}\left((a^{1}_{1},a^{2}_{1}),\dots,(a^{1}_{m},a^{2}_{m})\right)$
1Independently generate 4-wise independent random functions $h_{1},h_{2}$ from
$[n]$ to $\\{-1,1\\}$. 2for $c\leftarrow 1$ to $m$ 3 doLet the $c$th item
$(a^{1}_{c},a^{2}_{c})=(i,j)$ 4 $t_{1}\leftarrow t_{1}+h_{1}(i)h_{2}(j)$,
$t_{2}\leftarrow t_{2}+h_{1}(i)$, $t_{3}\leftarrow t_{3}+h_{2}(j)$. 5Return
$Y=(t_{1}/m-t_{2}t_{3}/m^{2})^{2}$.
Figure 1. The procedure for generating random variable $Y$ for $k=2$.
By the end of the process 2-D Approximation, we have
$t_{1}/m=\sum_{i,j\in[n]}h_{1}(i)h_{2}(j)r_{i,j}$,
$t_{2}/m=\sum_{i\in[n]}h_{1}(i)p_{i}$, and
$t_{3}/m=\sum_{i\in[n]}h_{2}(i)q_{i}$. Also, when a vector is in
${\mathrm{R}}^{(n^{2})}$, its indices can be represented by
$(i_{1},i_{2})\in[n]^{2}$. In what follows, we will use a bold letter to
represent the index of a high dimensional vector, e.g., $v_{\mathbf{i}}\equiv
v_{i_{1},i_{2}}$. The following Lemma shows that the expectation of $Y$ is
$\|v\|^{2}$ and the variance of $Y$ is at most $8(\mathrm{E}[Y])^{2}$ because
$\mathrm{E}[Y^{2}]\leq 9\mathrm{E}[Y]^{2}$.
###### Lemma 3.1.
([12]) Let $h_{1},h_{2}$ be two independent instances of 4-wise independent
hash functions from $[n]$ to $\\{-1,1\\}$. Let $v\in\mathrm{R}^{n^{2}}$ and
$H(\mathbf{i})(\equiv H\big{(}(i_{1},i_{2})\big{)}=h_{1}(i_{i})\cdot
h_{2}(i_{2})$. Let us define
$Y=\left(\sum_{\mathbf{i}\in[n]^{2}}H(\mathbf{i})v_{\mathbf{i}}\right)^{2}$.
Then $\mathrm{E}[Y]=\sum_{\mathbf{i}\in[n]^{2}}\vec{v}_{\mathbf{i}}^{2}$ and
$\mathrm{E}[Y^{2}]\leq 9(\mathrm{E}[Y])^{2}$, which implies
$\mathrm{Var}[Y]\leq 8E^{2}[Y]$.
###### Proof 3.2.
We have
$\mathrm{E}[Y]=\mathrm{E}[(\sum_{\mathbf{i}}H(\mathbf{i})\vec{v}_{\mathbf{i}})^{2}]=\sum_{\mathbf{i}}\vec{v}^{2}_{\mathbf{i}}\mathrm{E}[H^{2}(\mathbf{i})]+\sum_{\mathbf{i}\neq\mathbf{j}}\vec{v}_{\mathbf{i}}\vec{v}_{\mathbf{j}}\mathrm{E}[H(\mathbf{i})H(\mathbf{j})]$.
For all $\mathbf{i}\in[n]^{2}$, we know $h^{2}(\mathbf{i})=1$. On the other
hand, $H(\mathbf{i})H(\mathbf{j})\in\\{-1,1\\}$. The probability that
$H(\mathbf{i})H(\mathbf{j})=1$ is
$\Pr[H(\mathbf{i})H(\mathbf{j})=1]=\Pr[h_{1}(i_{1})h_{1}(j_{1})h_{2}(i_{2})h_{2}(j_{2})=1]=1/16+\binom{4}{2}1/16+1/16=1/2$.
The last equality holds is because
$h_{1}(i_{1})h_{1}(j_{1})h_{2}(i_{2})h_{2}(j_{2})=1$ is equivalent to saying
either all these variables are 1, or exactly two of these variables are -1, or
all these variables are -1. Therefore,
$\mathrm{E}[h(\mathbf{i})h(\mathbf{j})]=0$. Consequently,
$\mathrm{E}[Y]=\sum_{\mathrm{i}\in[n]^{2}}(\vec{v}_{\mathrm{i}})^{2}$.
Now we bound the variance. Recall that
$\mathrm{Var}[Y]=\mathrm{E}[Y^{2}]-\mathrm{E}[Y]^{2}$, we bound
$\mathrm{E}[Y^{2}]=\sum_{\mathbf{i},\mathbf{j},\mathbf{k},\mathbf{l}\in[n]^{2}}\mathrm{E}[H(\mathbf{i})H(\mathbf{j})H(\mathbf{k})h(\mathbf{l})]\vec{v}_{\mathbf{i}}\vec{v}_{\mathbf{j}}\vec{v}_{\mathbf{k}}\vec{v}_{\mathbf{l}}\leq\sum_{\mathbf{i},\mathbf{j},\mathbf{k},\mathbf{l}\in[n]^{2}}\left|\mathrm{E}[H(\mathbf{i})H(\mathbf{j})H(\mathbf{k})H(\mathbf{l})]\right|\cdot|\vec{v}_{\mathbf{i}}\vec{v}_{\mathbf{j}}\vec{v}_{\mathbf{k}}\vec{v}_{\mathbf{l}}|.$
Also
$\left|\mathrm{E}[H(\mathbf{i})H(\mathbf{j})H(\mathbf{k})H(\mathbf{l})]\right|\in\\{0,1\\}$.
The quantity
$\mathrm{E}[H(\mathbf{i})H(\mathbf{j})H(\mathbf{k})H(\mathbf{l})]\neq 0$ if
and only if the following relation holds,
$\forall
s\in[2]:\left((i_{s}=j_{s})\wedge(k_{s}=l_{s})\right)\vee\left((i_{s}=k_{s})\wedge(j_{s}=l_{s})\right)\vee\left((i_{s}=l_{s})\wedge(k_{s}=j_{s})\right).$
(2)
Denote the set of 4-tuples $(\mathbf{i},\mathbf{j},\mathbf{k},\mathbf{l})$
that satisfy the above relation by $\mathcal{D}$. We may also view each
4-tuple as an ordered set that consists of 4 points in $[n]^{2}$. Consider the
unique smallest axes-parallel rectangle in $[n]^{2}$ that contains a given
4-tuple in $\mathcal{D}$ (i.e. contains the four ordered points). Note this
could either be a (degenerate) line segment or a (non-degenerate) rectangle,
as we discuss below. Let $M:\mathcal{D}\rightarrow\\{A,B,C,D\\}$ be the
function that maps an element $\sigma\in\mathcal{D}$ to the smallest rectangle
$ABCD$ defined by $\sigma$. Since a rectangle can be uniquely determined by
its diagonals, we may write
$M:\mathcal{D}\rightarrow(\chi_{1},\chi_{2},\varphi_{1},\varphi_{2})$, where
$\chi_{1}\leq\chi_{2}\in[n]$, $\varphi_{1}\leq\varphi_{2}\in[n]$ and the
corresponding rectangle is understood to be the one with diagonal
$\\{(\chi_{1},\varphi_{1}),(\chi_{2},\varphi_{2})\\}$. Also, the inverse
function $M^{-1}(\chi_{1},\chi_{2},\varphi_{1},\varphi_{2})$ represents the
pre-images of $(\chi_{1},\chi_{2},\varphi_{1},\varphi_{2})$ in $\mathcal{D}$.
$(\chi_{1},\chi_{2},\varphi_{1},\varphi_{2})$ is degenerate if either
$\chi_{1}=\chi_{2}$ or $\varphi_{1}=\varphi_{2}$, in which case the rectangle
(and its diagonals) correspond to the segment itself, or $\chi_{1}=\chi_{2}$
and $\varphi_{1}=\varphi_{2}$, and the rectangle is just a single point.
###### Example 3.3.
Let $\mathbf{i}=(1,2)$, $\mathbf{j}=(3,2)$, $\mathbf{k}=(1,5)$, and
$\mathbf{l}=(3,5)$. The tuple is in $\mathcal{D}$ and its corresponding
bounding rectangle is a non-degenerate rectangle. The function
$M(\mathbf{i},\mathbf{j},\mathbf{k},\mathbf{l})=(1,3,2,5)$.
###### Example 3.4.
Let $\mathbf{i}=\mathbf{j}=(1,4)$ and $\mathbf{k}=\mathbf{l}=(3,7)$. The tuple
is also in $\mathcal{D}$ and minimal bounding rectangle formed by these points
is an interval $\\{(1,4),(3,7)\\}$. The function
$M(\mathbf{i},\mathbf{j},\mathbf{k},\mathbf{l})=(1,3,4,7)$.
To start we consider the non-degenerate cases. Fix any
$(\chi_{1},\chi_{2},\varphi_{1},\varphi_{2})$ with $\chi_{1}<\chi_{2}$ and
$\phi_{1}<\phi_{2}$. There are in total $\binom{4}{2}^{2}=36$ tuples
$(\mathbf{i},\mathbf{j},\mathbf{k},\mathbf{l})$ in $\mathcal{D}$ with
$M(\mathbf{i},\mathbf{j},\mathbf{k},\mathbf{l})=(\chi_{1},\chi_{2},\varphi_{1},\varphi_{2})$.
Twenty-four of these tuples correspond to the setting where none of
$\mathbf{i},\mathbf{j},\mathbf{k},\mathbf{l}$ are equal, as there are twenty-
four permutations of the assignment of the labels
$\mathbf{i},\mathbf{j},\mathbf{k},\mathbf{l}$ to the four points. (This
corresponds to the first example). In this case the four points form a
rectangle, and we have
$|\vec{v}_{\mathbf{i}}\vec{v}_{\mathbf{j}}\vec{v}_{\mathbf{k}}\vec{v}_{\mathbf{l}}|\leq\frac{1}{2}((\vec{v}_{\chi_{1},\varphi_{1}}\vec{v}_{\chi_{2},\varphi_{2}})^{2}+(\vec{v}_{\chi_{1},\varphi_{2}}\vec{v}_{\chi_{2},\varphi_{1}})^{2})$.
Intuitively, in these cases, we assign the “weight” of the tuple to the
diagonals.
The remaining twelve tuples in
$M^{-1}(\chi_{1},\chi_{2},\varphi_{1},\varphi_{2})$ correspond to intervals.
(This corresponds to the second example.) In this case two of
$\mathbf{i},\mathbf{j},\mathbf{k},\mathbf{l}$ correspond to one endpoint of
the interval, and the other two labels correspond to the other endpoint. Hence
we have either
$|\vec{v}_{\mathbf{i}}\vec{v}_{\mathbf{j}}\vec{v}_{\mathbf{k}}\vec{v}_{\mathbf{l}}|=(\vec{v}_{\chi_{1},\varphi_{1}}\vec{v}_{\chi_{2},\varphi_{2}})^{2}$
or
$|\vec{v}_{\mathbf{i}}\vec{v}_{\mathbf{j}}\vec{v}_{\mathbf{k}}\vec{v}_{\mathbf{l}}|=(\vec{v}_{\chi_{1},\varphi_{2}}\vec{v}_{\chi_{2},\varphi_{1}})^{2}$,
and there are six tuples for each case.
Therefore for any $\chi_{1}<\chi_{2}\in[n]$ and
$\varphi_{1}<\varphi_{2}\in[n]$ we have:
$\sum_{\scriptsize{}\begin{subarray}{c}(\mathbf{i},\mathbf{j},\mathbf{k},\mathbf{l})\in\\\
M^{-1}(\chi_{1},\chi_{2},\varphi_{1},\varphi_{2})\end{subarray}}|v_{\mathbf{i}}v_{\mathbf{j}}v_{\mathbf{k}}v_{\mathbf{l}}|\leq
18((v_{\chi_{1},\varphi_{1}}v_{\chi_{2},\varphi_{2}})^{2}+(v_{\chi_{1},\varphi_{2}},v_{\chi_{2},\varphi_{1}})^{2}).$
The analysis is similar for the degenerate cases, where the constant 18 in the
bound above is now quite loose. When exactly one of $\chi_{1}=\chi_{2}$ or
$\varphi_{1}=\varphi_{2}$ holds, the size of
$M^{-1}(\chi_{1},\chi_{2},\varphi_{1},\varphi_{2})$ is $\binom{4}{2}=6$, and
the resulting intervals correspond to vertical or horizontal lines. When both
$\chi_{1}=\chi_{2}$ and $\varphi_{1}=\varphi_{2}$, then
$|M^{-1}(\chi_{1},\chi_{2},\varphi_{1},\varphi_{2})|=1$. In sum, we have
Following the same analysis as for the non-degenerate cases, we find
$\sum_{\mathbf{i},\mathbf{j},\mathbf{k},\mathbf{l}\in\mathcal{D}}|\vec{v}_{\mathbf{i}}\vec{v}_{\mathbf{j}}\vec{v}_{\mathbf{k}}\vec{v}_{\mathbf{l}}|=\sum_{\scriptsize{}\begin{subarray}{c}\chi_{1}\leq\chi_{2}\\\
\varphi_{1}\leq\varphi_{2}\end{subarray}}\sum_{\scriptsize{}\begin{subarray}{c}(\mathbf{i},\mathbf{j},\mathbf{k},\mathbf{l})\in\\\
M^{-1}(\chi_{1},\chi_{2},\varphi_{1},\varphi_{2})\end{subarray}}|\vec{v}_{\mathbf{i}}\vec{v}_{\mathbf{j}}\vec{v}_{\mathbf{k}}\vec{v}_{\mathbf{l}}|\\\
$ $\leq\sum_{\scriptsize{}\begin{subarray}{c}\chi_{1}<\chi_{2}\\\
\varphi_{1}<\varphi_{2}\end{subarray}}18((\vec{v}_{\chi_{1},\varphi_{1}}\vec{v}_{\chi_{2},\varphi_{2}})^{2}+(\vec{v}_{\chi_{1},\varphi_{2}}\vec{v}_{\chi_{2},\varphi_{1}})^{2})+\sum_{\scriptsize{}\begin{subarray}{c}\chi_{1}=\chi_{2}\\\
\varphi_{1}<\varphi_{2}\end{subarray}}6((\vec{v}_{\chi_{1},\varphi_{1}}\vec{v}_{\chi_{2},\varphi_{2}})^{2}+(\vec{v}_{\chi_{1},\varphi_{2}}\vec{v}_{\chi_{2},\varphi_{1}})^{2})$
$+\sum_{\scriptsize{}\begin{subarray}{c}\chi_{1}<\chi_{2}\\\
\varphi_{1}=\varphi_{2}\end{subarray}}6((\vec{v}_{\chi_{1},\varphi_{1}}\vec{v}_{\chi_{2},\varphi_{2}})^{2}+(\vec{v}_{\chi_{1},\varphi_{2}}\vec{v}_{\chi_{2},\varphi_{1}})^{2})+\sum_{\scriptsize{}\begin{subarray}{c}\chi_{1}=\chi_{2}\\\
\varphi_{1}=\varphi_{2}\end{subarray}}(\vec{v}_{\chi_{1},\varphi_{1}}\vec{v}_{\chi_{2},\varphi_{2}})^{2}\\\
$ $\leq 9\sum_{\scriptsize{}\begin{subarray}{c}\mathbf{i}\in[n]^{2}\\\
\mathbf{j}\in[n]^{2}\end{subarray}}(\vec{v}_{\mathbf{i}}\vec{v}_{\mathbf{j}})^{2}=9\mathrm{E}^{2}[Y].$
Finally, we have
$\sum_{\mathbf{i},\mathbf{j},\mathbf{k},\mathbf{l}\in[n]^{2}}\left|\mathrm{E}[H(\mathbf{i})H(\mathbf{j})H(\mathbf{k})H(\mathbf{l})]\right|\cdot|\vec{v}_{\mathbf{i}}\vec{v}_{\mathbf{j}}\vec{v}_{\mathbf{k}}\vec{v}_{\mathbf{l}}|\leq\sum_{\mathbf{i},\mathbf{j},\mathbf{k},\mathbf{l}\in\mathcal{D}}|\vec{v}_{\mathbf{i}}\vec{v}_{\mathbf{j}}\vec{v}_{\mathbf{k}}\vec{v}_{\mathbf{l}}|\leq
9\mathrm{E}^{2}[Y]$ and $\mathrm{Var}[Y]\leq 8\mathrm{E}[Y]^{2}$.
We emphasize the geometric interpretation of the above proof as follows. The
goal is to bound the variance by a constant times
$\mathrm{E}^{2}[Y]=\sum_{\scriptsize{}\begin{subarray}{c}\mathbf{i},\mathbf{j}\in[n]^{2}\end{subarray}}(\vec{v}_{\mathbf{i}}v_{\mathbf{j}})^{2}$,
where the index set is the set of all possible lines in plane $[n]^{2}$ (each
line appears twice). We first show that
$\mathrm{Var}[Y]\leq\sum_{\mathbf{i},\mathbf{j},\mathbf{k},\mathbf{l}\in\mathcal{D}}|\vec{v}_{\mathbf{i}}\vec{v}_{\mathbf{j}}\vec{v}_{\mathbf{k}}\vec{v}_{\mathbf{l}}|$,
where the 4-tuple index set corresponds to a set of rectangles in a natural
way. The main idea of [12] is to use inequalities of the form
$|\vec{v}_{\mathbf{i}}\vec{v}_{\mathbf{j}}\vec{v}_{\mathbf{k}}\vec{v}_{\mathbf{l}}|\leq\frac{1}{2}((\vec{v}_{\chi_{1},\varphi_{1}}\vec{v}_{\chi_{2},\varphi_{2}})^{2}+(\vec{v}_{\chi_{1},\varphi_{2}}\vec{v}_{\chi_{2},\varphi_{1}})^{2})$
to assign the “weight” of each $4$-tuple to the diagonals of the corresponding
rectangle. The above analysis shows that $18$ copies of all lines are
sufficient to accommodate all 4-tuples. While similar inequalities could also
assign the weight of a $4$-tuple to the vertical or horizontal edges of the
corresponding rectangle, using vertical or horizontal edges is problematic.
The reason is that there are $\Omega(n^{4})$ $4$-tuples but only $O(n^{3})$
vertical or horizontal edges, so some lines would receive $\Omega(n)$ weight,
requiring $\Omega(n)$ copies. This problem is already noted in [7].
Our bound here is $\mathrm{E}[Y^{2}]\leq 9\mathrm{E}^{2}[Y]$, while in [12]
the bound obtained is $\mathrm{E}[Y^{2}]\leq 3\mathrm{E}^{2}[Y]$. There
appears to have been an error in the derivation in [12]; some intuition comes
from the following example. We note that $|\mathcal{D}|$ is at least
$\binom{4}{2}^{2}\cdot\binom{n}{2}^{2}=9n^{4}-9n^{2}$. (This counts the number
of non-degenerate $4$-tuples.) Now if we set $v_{i}=1$ for all $1\leq i\leq
n^{2}$, we have
$\mathrm{E}[Y^{2}]\geq|\mathcal{D}|=9n^{4}-9n^{2}\sim\mathrm{9}\mathrm{E}^{2}(D)$,
which suggests $\mathrm{Var}[D]>3\mathrm{E}^{2}[D]$. Again, we emphasize this
discrepancy is of little importance to [12]; the point there is that the
variance is bounded by a constant factor times the square of the expectation.
It is here, where we are generalizing to $k\geq 3$, that the exact constant
factor is of some importance.
Given the bounds on the expectation and variance for the $D_{i,j}$, standard
techniques yield a bound on the performance of our algorithm.
###### Theorem 3.5.
For every $\epsilon>0$ and $\delta>0$, there exists a randomized algorithm
that computes, given a sequence
$(a^{1}_{1},a^{2}_{1}),\dots,(a^{1}_{m},a^{2}_{m})$, in one pass and using
$O(\epsilon^{-2}\log\frac{1}{\delta}(\log m+\log n))$ memory bits, a number
$\mathrm{Med}$ so that the probability $\mathrm{Med}$ deviates from
$\|v\|^{2}$ by more than $\epsilon$ is at most $\delta$.
###### Proof 3.6.
Recall the algorithm described in the beginning of Section 3: let
$s_{1}=72\epsilon^{-2}$ and $s_{2}=2\log\delta$. We first computes $s_{2}$
random variables $Y_{1},Y_{2},\dots,Y_{s_{2}}$ and outputs their median
$\mathrm{Med}$, where each $Y_{i}$ is the average of $s_{1}$ random variables
$Y_{ij}$: $1\leq j\leq s_{1}$ and $Y_{ij}$ are independent, identically
distributed random variables computed by Figure 1. By Chebyshev’s inequality,
we know that for any fixed $i$,
$\Pr\big{(}\big{|}Y_{i}-\|\vec{v}\|\big{|}\big{)}\geq\epsilon\|\vec{v}\|]\leq\frac{\mathrm{Var}(Y_{i})}{\epsilon^{2}\|\vec{v}\|^{2}}=\frac{(1/s_{1})\mathrm{Var}[Y]}{\epsilon^{2}\|\vec{v}\|^{2}}=\frac{(9\epsilon^{2}/72)\|\vec{v}\|^{2}}{\epsilon^{2}\|\vec{v}\|^{2}}=\frac{1}{8}.$
Finally, by standard Chernoff bound arguments (see for example Chapter 4 of
[14]), the probability that more than $s_{2}/2$ of the variables $Y_{i}$
deviate by more than $\epsilon\|\vec{v}\|$ from $\|\vec{v}\|$ is at most
$\delta$. In case this does not happen, the median $\mathrm{Med}$ supplies a
good estimate to the required quantity $\|\vec{v}\|$ as needed.
## 4\. The General Case $k\geq 3$
Now let us move to the general case where $k\geq 3$. Recall that ${\vec{v}}$
is a vector in $\mathrm{R}^{n^{k}}$ that maintains certain statistics of a
data stream, and we are interested in estimating its $\ell_{2}$ norm
$\|{\vec{v}}\|$. There is a natural generalization for Indyk and McGregor’s
method for $k=2$ to construct an estimator for $\|{\vec{v}}\|$: let
$h_{1},\dots,h_{k}:[n]\rightarrow\\{-1,1\\}$ be independent copies of 4-wise
independent hash functions (namely, $h_{i}(1),\dots,h_{i}(n)\in\\{-1,1\\}$ are
$4$-wise independent hash functions for each $i\in[k]$, and
$h_{1}(\cdot),\dots,h_{k}(\cdot)$ are mutually independent.). Let
$H({\mathbf{p}})=\prod_{i=1}^{k}h_{j}(p_{j})$. The estimator $Y$ is defined as
$Y\equiv\left(\sum_{{\mathbf{p}}\in[n]^{k}}\vec{v}_{{\mathbf{p}}}H({\mathbf{p}})\right)^{2}$.
Our goal is to show that $\mathrm{E}[Y]=\|\vec{v}\|^{2}$ and $\mathrm{Var}[Y]$
is reasonably small so that a streaming algorithm maintaining multiple
independent instances of estimator $Y$ will be able to output an approximately
correct estimation of $\|{\vec{v}}\|$ with high probability. Notice that when
$\|{\vec{v}}\|$ represents the $\ell_{2}$ distance between the joint
distribution and the tensors of the marginal distributions, the estimator can
be computed efficiently in a streaming model similarly to as in Figure 1. We
stress that our result is applicable to a broader class of $\ell_{2}$-norm
estimation problems, as long as the vector ${\vec{v}}$ to be estimated has a
corresponding efficiently computable estimator $Y$ in an appropriate streaming
model. Formally, we shall prove the following main lemma in the next
subsection.
###### Lemma 4.1.
Let ${\vec{v}}$ be a vector in $\mathrm{R}^{n^{k}}$, and
$h_{1},\dots,h_{k}:[n]\rightarrow\\{-1,1\\}$ be independent copies of 4-wise
independent hash functions. Define
$H({\mathbf{p}})=\prod_{i=1}^{k}h_{j}(p_{j})$, and
$Y\equiv\left(\sum_{{\mathbf{p}}\in[n]^{k}}\vec{v}_{{\mathbf{p}}}H({\mathbf{p}})\right)^{2}$.
We have ${\mathrm{E}}[Y]=||{\vec{v}}||$ and ${\mathrm{Var}}[Y]\leq
3^{k}{\mathrm{E}}[Y]^{2}$.
We remark that the bound on the variance in the above lemma is tight. One can
verify that when the vector ${\vec{v}}$ is a uniform vector (i.e., all entries
of ${\vec{v}}$ are the same), the variance of $Y$ is $\Omega(3^{k}E[Y]^{2})$.
With the above lemma, the following main theorem mentioned in the introduction
immediately follows by a standard argument presented in the proof of Theorem
3.5 in the previous section.
###### Theorem 4.2.
Let $\vec{v}$ be a vector in $\mathrm{R}^{[n]^{k}}$ that maintains an
arbitrary statistics in a data stream of size $m$, in which every item is from
$[n]^{k}$. Let $\epsilon,\delta\in(0,1)$ be real numbers. If there exists an
algorithm that maintains an instance of $Y$ using
$O(\mu(n,m,k,\epsilon,\delta))$ memory bits, then there exists an algorithm
$\Lambda$ such that:
1. (1)
With probability $\geq 1-\delta$ the algorithm $\Lambda$ outputs a value
between $[(1-\epsilon)\|{\vec{v}}\|^{2},(1+\epsilon)\|{\vec{v}}|^{2}]$ and
2. (2)
the space complexity of $\Lambda$ is
$O(3^{k}{1\over\epsilon^{2}}\log{1\over\delta}\mu(n,m,k,\epsilon,\delta))$.
As discussed above, an immediate corollary is the existence of a one-pass
space efficient streaming algorithm to detect the dependency of $k$ random
variables in $\ell_{2}$-norm:
###### Corollary 4.3.
For every $\epsilon>0$ and $\delta>0$, there exists a randomized algorithm
that computes, given a sequence $a_{1},\dots,a_{m}$ of $k$-tuples, in one pass
and using $O(3^{k}\epsilon^{-2}\log\frac{1}{\delta}(\log m+\log n))$ memory
bits, a number $Y$ so that the probability $Y$ deviates from the square of the
$\ell_{2}$ distance between product and joint distribution by more than a
factor of $(1+\epsilon)$ is at most $\delta$.
### 4.1. Analysis of the Sketch $Y$
This section is devoted to prove Lemma 4.1, where the main challenge is to
bound the variance of $Y$. The geometric approach of Indyk and McGregor [12]
presented in Section 3 for the case of $k=2$ can be extended to analyze the
general case. However, we remark that the generalization requires new ideas.
In particular, instead of performing “local analysis” that maps each rectangle
to its diagonals, a more complex “global analysis” is needed in higher
dimensions to achieve the desired bounds. The alternative proof we present
here utilizes similar ideas, but relies on a more combinatorial rather than
geometric approach.
For the expectation of $Y$, we have
$\displaystyle\mathrm{E}[Y]$ $\displaystyle=$
$\displaystyle\mathrm{E}\left[\sum_{{\mathbf{p}},{\mathbf{q}}\in[n]^{k}}{\vec{v}}_{{\mathbf{p}}}\cdot{\vec{v}}_{{\mathbf{q}}}\cdot
H({\mathbf{p}})\cdot H({\mathbf{q}})\right]$ $\displaystyle=$
$\displaystyle\sum_{{\mathbf{p}}\in[n]^{k}}\vec{v}_{{\mathbf{p}}}^{2}\cdot\mathrm{E}\left[H({\mathbf{p}})^{2}\right]+\sum_{{\mathbf{p}}\neq{\mathbf{q}}\in[n]^{k}}{\vec{v}}_{{\mathbf{p}}}\cdot{\vec{v}}_{{\mathbf{q}}}\cdot\mathrm{E}\left[H({\mathbf{p}})H({\mathbf{q}})\right]$
$\displaystyle=$
$\displaystyle\sum_{{\mathbf{p}}\in[n]^{k}}\vec{v}_{{\mathbf{p}}}^{2}=||{\vec{v}}||^{2},$
where the last equality follows by $H({\mathbf{p}})^{2}=1$, and
$\mathrm{E}\left[H({\mathbf{p}})H({\mathbf{q}})\right]=0$ for
${\mathbf{p}}\neq{\mathbf{q}}$.
Now, let us start to prove ${\mathrm{Var}}[Y]\leq 3^{k}\mathrm{E}[Y]^{2}$. By
definition, ${\mathrm{Var}}[Y]=\mathrm{E}[(Y-{\mathrm{E}}[Y])^{2}]$, so we
need to understand the following random variable:
$Err\equiv
Y-{\mathrm{E}}[Y]=\sum_{{\mathbf{p}}\neq{\mathbf{q}}\in[n]^{k}}H({\mathbf{p}})H({\mathbf{q}}){\vec{v}}_{{\mathbf{p}}}{\vec{v}}_{{\mathbf{q}}}.$
(3)
The random variable $Err$ is a sum of terms indexed by pairs
$({\mathbf{p}},{\mathbf{q}})\in[n]^{k}\times[n]^{k}$ with
${\mathbf{p}}\neq{\mathbf{q}}$. At a very high level, our analysis consists of
two steps. In the first step, we group the terms in $Err$ properly and
simplify the summation in each group. In the second step, we expand the square
of the sum in ${\mathrm{Var}}[Y]=\mathrm{E}[Err^{2}]$ according to the groups
and apply Cauchy-Schwartz inequality three times to bound the variance.
We shall now gradually introduce the necessary notation for grouping the terms
in $Err$ and simplifying the summation. We remind the reader that vectors over
the reals (e.g., ${\vec{v}}\in R^{n^{k}}$) are denoted by
${\vec{v}},\vec{w},\vec{r}$, and vectors over $[n]$ are denoted by
${\mathbf{p}},{\mathbf{q}},{\mathbf{a}},{\mathbf{b}},{\mathbf{c}},{\mathbf{d}}$
and referred as _index vectors_. We use $S\subseteq[k]$ to denote a subset of
$[k]$, and let ${\bar{S}}=[k]\backslash S$. We use
$\mathrm{Ham}({\mathbf{p}},{\mathbf{q}})$ to denote the _Hamming distance_ of
index vectors ${\mathbf{p}},{\mathbf{q}}\in[n]^{k}$, i.e., the number of
coordinates where ${\mathbf{p}}$ and ${\mathbf{q}}$ are different.
###### Definition 4.4.
_(Projection and inverse projection)_ Let ${\mathbf{c}}\in[n]^{k}$ be an index
vector and $S\subseteq[k]$ a subset. We define the _projection of
${\mathbf{c}}$ to $S$_, denoted by $\Phi_{S}({\mathbf{c}})\in[n]^{|S|}$, to be
the vector ${\mathbf{c}}$ restricted to the coordinates in $S$. Also, let
${\mathbf{a}}\in[n]^{|S|}$ and ${\mathbf{b}}\in[n]^{k-|S|}$ be index vectors.
We define the _inverse projection of ${\mathbf{a}}$ and ${\mathbf{b}}$ with
respect to $S$_, denoted by
$\Phi^{-1}_{S}({\mathbf{a}},{\mathbf{b}})\in[n]^{k}$, as the index vector
${\mathbf{c}}\in[n]^{k}$ such that $\Phi_{S}({\mathbf{c}})={\mathbf{a}}$ and
$\Phi_{{\bar{S}}}({\mathbf{c}})={\mathbf{b}}$.
We next define _pair groups_ and use the definition to group the terms in
$Err$.
###### Definition 4.5.
_(Pair Group)_ Let $S\subseteq[k]$ be a subset of size $|S|=t$. Let
${\mathbf{c}},{\mathbf{d}}\in[n]^{t}$ be a pair of index vectors with
$\mathrm{Ham}({\mathbf{c}},{\mathbf{d}})=t$ (i.e., all coordinates of
${\mathbf{c}}$ and ${\mathbf{d}}$ are distinct.). The _pair group_
$\sigma_{S}({\mathbf{c}},{\mathbf{d}})$ is the set of pairs
$({\mathbf{p}},{\mathbf{q}})\in[n]^{k}\times[n]^{k}$ such that (i) on
coordinate $S$, $\Phi_{S}({\mathbf{p}})={\mathbf{c}}$ and
$\Phi_{S}({\mathbf{q}})={\mathbf{d}}$, and (ii) on coordinate ${\bar{S}}$,
${\mathbf{p}}$ and ${\mathbf{q}}$ are the same, i.e.,
$\Phi_{{\bar{S}}}({\mathbf{p}})=\Phi_{{\bar{S}}}({\mathbf{q}})$. Namely,
$\sigma_{S}({\mathbf{c}},{\mathbf{d}})=\left\\{({\mathbf{p}},{\mathbf{q}})\in[n]^{k}\times[n]^{k}:\Big{(}{\mathbf{c}}=\Phi_{S}({\mathbf{p}})\Big{)}\wedge\Big{(}{\mathbf{d}}=\Phi_{S}({\mathbf{q}})\Big{)}\wedge\Big{(}\Phi_{{\bar{S}}}({\mathbf{p}})=\Phi_{{\bar{S}}}({\mathbf{q}})\Big{)}\right\\}.$
(4)
To give some intuition for the above definitions, we note that for every
${\mathbf{a}}\in[n]^{|{\bar{S}}|}$, there is a unique pair
$({\mathbf{p}},{\mathbf{q}})\in\sigma_{S}({\mathbf{c}},{\mathbf{d}})$ with
${\mathbf{a}}=\Phi_{{\bar{S}}}({\mathbf{p}})=\Phi_{{\bar{S}}}({\mathbf{q}})$,
and so $|\sigma_{S}({\mathbf{c}},{\mathbf{d}})|=n^{|{\bar{S}}|}$. On the other
hand, for every pair $({\mathbf{p}},{\mathbf{q}})\in[n]^{k}\times[n]^{k}$ with
${\mathbf{p}}\neq{\mathbf{q}}$, there is a unique non-emtpy $S\subseteq[k]$
such that ${\mathbf{p}}$ and ${\mathbf{q}}$ are distinct on exactly
coordinates in $S$. Therefore, $({\mathbf{p}},{\mathbf{q}})$ belongs to
exactly one pair group $\sigma_{S}({\mathbf{c}},{\mathbf{d}})$. It follows
that we can partition the summation in $Err$ according to the pair groups:
$Err=\sum_{\begin{subarray}{c}S\subseteq[k]\\\ S\neq\emptyset\end{subarray}}\
\ \ \ \ \sum_{\begin{subarray}{c}{\mathbf{c}},{\mathbf{d}}\in[n]^{|S|},\\\
\mathrm{Ham}({\mathbf{c}},{\mathbf{d}})=|S|\end{subarray}}\ \ \ \ \
\sum_{\begin{subarray}{c}({\mathbf{p}},{\mathbf{q}})\in\\\
\sigma_{S}({\mathbf{c}},{\mathbf{d}})\end{subarray}}H({\mathbf{p}})H({\mathbf{q}}){\vec{v}}_{{\mathbf{p}}}{\vec{v}}_{{\mathbf{q}}}.$
(5)
We next observe that for any pair
$({\mathbf{p}},{\mathbf{q}})\in\sigma_{S}({\mathbf{c}},{\mathbf{d}})$, since
${\mathbf{p}}$ and ${\mathbf{q}}$ agree on coordinates in ${\bar{S}}$, the
value of the product $H({\mathbf{p}})H({\mathbf{q}})$ depends only on $S$,
${\mathbf{c}}$ and ${\mathbf{d}}$. More precisely,
$H({\mathbf{p}})H({\mathbf{q}})=\prod_{i\in[k]}h_{i}(p_{i})h_{i}(q_{i})=\left(\prod_{i\in
S}h_{i}(p_{i})h_{i}(q_{i})\right)\cdot\left(\prod_{i\in{\bar{S}}}h_{i}(p_{i})^{2}\right)=\prod_{i\in
S}h_{i}(p_{i})h_{i}(q_{i}),$
which depends only on $S$, ${\mathbf{c}}$ and ${\mathbf{d}}$ since
$\Phi_{S}({\mathbf{p}})={\mathbf{c}}$ and
$\Phi_{S}({\mathbf{q}})={\mathbf{d}}$. This motivates the definition of
_projected hashing_.
###### Definition 4.6.
_(Projected hashing)_ Let $S=\\{s_{1},s_{2},\dots,s_{t}\\}$ be a subset of
$[k]$, where $s_{1}<s_{2}<\dots<s_{j}$. Let ${\mathbf{c}}\in[n]^{t}$. We
define the _projected hashing_ $H_{S}({\mathbf{c}})=\prod_{i\leq
t}h_{s_{i}}(c_{i})$.
We can now translate the random variable $Err$ as follows:
$Err=\sum_{\begin{subarray}{c}S\subseteq[k]\\\ S\neq\emptyset\end{subarray}}\
\ \ \ \ \sum_{\begin{subarray}{c}{\mathbf{c}},{\mathbf{d}}\in[n]^{|S|},\\\
\mathrm{Ham}({\mathbf{c}},{\mathbf{d}})=|S|\end{subarray}}\left(H_{S}({\mathbf{c}})H_{S}({\mathbf{d}})\sum_{\begin{subarray}{c}({\mathbf{p}},{\mathbf{q}})\in\\\
\sigma_{S}({\mathbf{c}},{\mathbf{d}})\end{subarray}}{\vec{v}}_{{\mathbf{p}}}{\vec{v}}_{{\mathbf{q}}}\right).$
(6)
Fix a pair group $\sigma_{S}({\mathbf{c}},{\mathbf{d}})$, we next consider the
sum
$\sum_{\begin{subarray}{c}({\mathbf{p}},{\mathbf{q}})\in\sigma_{S}({\mathbf{c}},{\mathbf{d}})\end{subarray}}{\vec{v}}_{{\mathbf{p}}}{\vec{v}}_{{\mathbf{q}}}$.
Recall that for every ${\mathbf{a}}\in[n]^{|{\bar{S}}|}$, there is a unique
pair $({\mathbf{p}},{\mathbf{q}})\in\sigma_{S}({\mathbf{c}},{\mathbf{d}})$
with
${\mathbf{a}}=\Phi_{{\bar{S}}}({\mathbf{p}})=\Phi_{{\bar{S}}}({\mathbf{q}})$.
The sum can be viewed as the inner product of two vectors of dimension
$n^{|{\bar{S}}|}$ with entries indexed by ${\mathbf{a}}\in[n]^{|{\bar{S}}|}$.
To formalize this observation, we introduce the definition of _hyper-
projection_ as follows.
###### Definition 4.7.
_(Hyper-projection)_ Let ${\vec{v}}\in R^{n^{k}}$, $S\subseteq[k]$, and
$\mathbf{{\mathbf{c}}}\in[n]^{|S|}$. The _hyper-projection_
$\Upsilon_{S,{\mathbf{c}}}({\vec{v}})$ of ${\vec{v}}$ (with respect to $S$ and
${\mathbf{c}}$) is a vector $\vec{w}=\Upsilon_{S,{\mathbf{c}}}({\vec{v}})$ in
$\mathrm{R}^{[n]^{k-|S|}}$ such that
$\vec{w}_{{\mathbf{d}}}={\vec{v}}_{\Phi^{-1}_{S}({\mathbf{c}},{\mathbf{d}})}$
for all ${\mathbf{d}}\in[n]^{k-|S|}$.
Using the above definition, we continue to rewrite the $Err$ as
$Err=\sum_{\begin{subarray}{c}S\subseteq[k]\\\ S\neq\emptyset\end{subarray}}\
\ \ \ \ \sum_{\begin{subarray}{c}{\mathbf{c}},{\mathbf{d}}\in[n]^{|S|},\\\
\mathrm{Ham}({\mathbf{c}},{\mathbf{d}})=|S|\end{subarray}}H_{S}({\mathbf{c}})H_{S}({\mathbf{d}})\cdot\langle\Upsilon_{S,{\mathbf{c}}}({\vec{v}}),\Upsilon_{S,{\mathbf{d}}}({\vec{v}})\rangle.$
(7)
Finally, we consider the product $H_{S}({\mathbf{c}})H_{S}({\mathbf{d}})$
again and introduce the following definition to further simplify the $Err$.
###### Definition 4.8.
_(Similarity and dominance)_ Let $t$ be a positive integer.
* •
Two pairs of index vectors
$({\mathbf{c}},{\mathbf{d}})\in[n]^{t}\times[n]^{t}$ and
$({\mathbf{a}},{\mathbf{b}})\in[n]^{t}\times[n]^{t}$ are _similar_ if for all
$i\in[t]$, the two sets $\\{c_{i},d_{i}\\}$ and $\\{a_{i},b_{i}\\}$ are equal.
We denote this as
$({\mathbf{a}},{\mathbf{b}})\sim({\mathbf{c}},{\mathbf{d}})$.
* •
Let ${\mathbf{c}}$ and ${\mathbf{d}}\in[n]^{t}$ be two index vectors. We say
${\mathbf{c}}$ _is dominated by_ ${\mathbf{d}}$ if $c_{i}<d_{i}$ for all
$i\in[t]$. We denote this as ${\mathbf{c}}\prec{\mathbf{d}}$. Note that
${\mathbf{c}}\prec{\mathbf{d}}\Rightarrow\mathrm{Ham}({\mathbf{c}},{\mathbf{d}})=t$.
Now, note that if
$({\mathbf{a}},{\mathbf{b}})\sim({\mathbf{c}},{\mathbf{d}})$, then
$H_{S}({\mathbf{a}})H_{S}({\mathbf{b}})=H_{S}({\mathbf{c}})H_{S}({\mathbf{d}})$
since the value of the product $H_{S}({\mathbf{c}})H_{S}({\mathbf{d}})$
depends on the values $\\{c_{i},d_{i}\\}$ only as a set. It is also not hard
to see that $\sim$ is an equivalence relation, and for every equivalent class
$[({\mathbf{a}},{\mathbf{b}})]$, there is a unique
$({\mathbf{c}},{\mathbf{d}})\in[({\mathbf{a}},{\mathbf{b}})]$ with
${\mathbf{c}}\prec{\mathbf{d}}$. Therefore, we can further rewrite the $Err$
as
$Err=\sum_{\begin{subarray}{c}S\subseteq[k]\\\ S\neq\emptyset\end{subarray}}\
\sum_{{\mathbf{c}}\prec{\mathbf{d}}\in[n]^{|S|}}H_{S}({\mathbf{c}})H_{S}({\mathbf{d}})\cdot\left(\sum_{({\mathbf{a}},{\mathbf{b}})\sim({\mathbf{c}},{\mathbf{d}})}\langle\Upsilon_{S,{\mathbf{a}}}({\vec{v}}),\Upsilon_{S,{\mathbf{b}}}({\vec{v}})\rangle\right).$
(8)
We are ready to bound the term $\mathrm{E}[Err^{2}]$ by expanding the square
of the sum according to Equation (8). We first show in Lemma 4.9 below that
all the cross terms in the following expansion vanish.
${\mathrm{Var}}[Y]=\sum_{\begin{subarray}{c}S,S^{\prime}\subseteq[k]\\\
S,S^{\prime}\neq\emptyset\end{subarray}}\
\sum_{\begin{subarray}{c}{\mathbf{c}}\prec{\mathbf{d}}\in[n]^{|S|}\\\
{\mathbf{c}}^{\prime}\prec{\mathbf{d}}^{\prime}\in[n]^{|S|^{\prime}}\end{subarray}}{\mathrm{E}}[H_{S}({\mathbf{c}})H_{S}({\mathbf{d}})H_{S^{\prime}}({\mathbf{c}}^{\prime})H_{S^{\prime}}({\mathbf{d}}^{\prime})]\cdot$
$\left[\left(\sum_{({\mathbf{a}},{\mathbf{b}})\sim({\mathbf{c}},{\mathbf{d}})}\langle\Upsilon_{S,{\mathbf{a}}}({\vec{v}}),\Upsilon_{S,{\mathbf{b}}}({\vec{v}})\rangle\right)\left(\sum_{({\mathbf{a}}^{\prime},{\mathbf{b}}^{\prime})\sim({\mathbf{c}}^{\prime},{\mathbf{d}}^{\prime})}\langle\Upsilon_{S^{\prime},{\mathbf{a}}^{\prime}}({\vec{v}}),\Upsilon_{S^{\prime},{\mathbf{b}}^{\prime}}({\vec{v}})\rangle\right)\right].$
(9)
###### Lemma 4.9.
Let $S$ and $S^{\prime}$ be subsets of $[k]$, and
${\mathbf{c}}\prec{\mathbf{d}}\in[n]^{|S|}$ and
${\mathbf{c}}^{\prime}\prec{\mathbf{d}}^{\prime}\in[n]^{|S^{\prime}|}$ index
vectors. We have
$E[H_{S}({\mathbf{c}})H_{S}({\mathbf{d}})H_{S^{\prime}}({\mathbf{c}}^{\prime})H_{S^{\prime}}({\mathbf{d}}^{\prime})]\in\\{0,1\\}$.
Furthermore, we have
${\mathrm{E}}[H_{S}({\mathbf{c}})H_{S}({\mathbf{d}})H_{S^{\prime}}({\mathbf{c}}^{\prime})H_{S^{\prime}}({\mathbf{d}}^{\prime})]=1$
iff
$(S=S^{\prime})\wedge({\mathbf{c}}={\mathbf{c}}^{\prime})\wedge({\mathbf{d}}={\mathbf{d}}^{\prime})$.
###### Proof 4.10.
Recall that $h_{1},\dots,h_{k}$ are independent copies of $4$-wise independent
uniform random variables over $\\{-1,1\\}$. Namely, for every $i\in[k]$,
$h_{i}(1),\dots,h_{i}(n)$ are $4$-wise independent, and
$h_{1}(\cdot),\dots,h_{k}(\cdot)$ are mutually independent. Observe that for
every $i\in[k]$, there are at most $4$ terms out of $h_{i}(1),\dots,h_{i}(n)$
appearing in the product
$H_{S}({\mathbf{c}})H_{S}({\mathbf{d}})H_{S^{\prime}}({\mathbf{c}}^{\prime})H_{S^{\prime}}({\mathbf{d}}^{\prime})$.
It follows that all distinct terms appearing in
$H_{S}({\mathbf{c}})H_{S}({\mathbf{d}})H_{S^{\prime}}({\mathbf{c}}^{\prime})H_{S^{\prime}}({\mathbf{d}}^{\prime})$
are mutually independent uniform random variable over $\\{-1,1\\}$. Therefore,
the expectation is either 0, if there is some $h_{i}(j)$ that appears an odd
number of times, or 1, if all $h_{i}(j)$ appear an even number of times. By
inspection, the latter case happens if and only if
$(S=S^{\prime})\wedge({\mathbf{c}}={\mathbf{c}}^{\prime})\wedge({\mathbf{d}}={\mathbf{d}}^{\prime})$.
By the above lemma, Equation (4.1) is simplified to
${\mathrm{Var}}[Y]=\sum_{\begin{subarray}{c}S\subseteq[k]\\\
S\neq\emptyset\end{subarray}}\
\sum_{{\mathbf{c}}\prec{\mathbf{d}}\in[n]^{|S|}}\left(\sum_{({\mathbf{a}},{\mathbf{b}})\sim({\mathbf{c}},{\mathbf{d}})}\langle\Upsilon_{S,{\mathbf{a}}}({\vec{v}}),\Upsilon_{S,{\mathbf{b}}}({\vec{v}})\rangle\right)^{2}.$
(10)
We next apply the Cauchy-Schwartz inequality three times to bound the above
formula. Consider a subset $S\subseteq[k]$ and a pair
${\mathbf{c}}\prec{\mathbf{d}}\in[n]^{|S|}$. Note that there are precisely
$2^{|S|}$ pairs $({\mathbf{a}},{\mathbf{b}})$ such that
$({\mathbf{a}},{\mathbf{b}})\sim({\mathbf{c}},{\mathbf{d}})$. Thus, by the
Cauchy-Schwartz inequality:
$\displaystyle\left(\sum_{\begin{subarray}{c}({\mathbf{a}},{\mathbf{b}})\in[n]^{|S|}\\\
({\mathbf{a}},{\mathbf{b}})\sim({\mathbf{c}},{\mathbf{d}})\end{subarray}}\langle\Upsilon_{S,{\mathbf{a}}}({\vec{v}}),\Upsilon_{S,{\mathbf{b}}}(\vec{v})\rangle\right)^{2}$
$\displaystyle\leq$ $\displaystyle
2^{|S|}\sum_{\begin{subarray}{c}({\mathbf{a}},{\mathbf{b}})\in[n]^{|S|}\\\
({\mathbf{a}},{\mathbf{b}})\sim({\mathbf{c}},{\mathbf{d}})\end{subarray}}(\langle\Upsilon_{S,{\mathbf{a}}},\Upsilon_{S,{\mathbf{b}}}\rangle)^{2}$
$\displaystyle\leq$ $\displaystyle
2^{|S|}\sum_{\begin{subarray}{c}({\mathbf{a}},{\mathbf{b}})\in[n]^{|S|}\\\
({\mathbf{a}},{\mathbf{b}})\sim({\mathbf{c}},{\mathbf{d}})\end{subarray}}\langle\Upsilon_{S,{\mathbf{a}}}({\vec{v}}),\Upsilon_{S,{\mathbf{a}}}({\vec{v}})\rangle\cdot\langle\Upsilon_{S,{\mathbf{b}}},\Upsilon_{S,{\mathbf{b}}}({\vec{v}})\rangle.$
Notice that in the second inequality, we applied Cauchy-Schwartz in a
component-wise manner. Next, for a subset $S\subseteq[k]$, we can apply the
Cauchy-Schwartz inequality a third time (from the third line to the fourth
line) as follows:
$\displaystyle\sum_{{\mathbf{c}}\prec{\mathbf{d}}\in[n]^{|S|}}\ \ \ \ \
\left(\sum_{\begin{subarray}{c}({\mathbf{a}},{\mathbf{b}})\in[n]^{|S|}\\\
({\mathbf{a}},{\mathbf{b}})\sim({\mathbf{c}},{\mathbf{d}})\end{subarray}}\langle\Upsilon_{S,{\mathbf{a}}}(\vec{v}),\Upsilon_{S,{\mathbf{b}}}(\vec{v})\rangle\right)^{2}$
$\displaystyle\leq$ $\displaystyle
2^{|S|}\sum_{{\mathbf{c}}\prec{\mathbf{d}}\in[n]^{|S|}}\ \ \
\sum_{\begin{subarray}{c}({\mathbf{a}},{\mathbf{b}})\in[n]^{|S|}\\\
({\mathbf{a}},{\mathbf{b}})\sim({\mathbf{c}},{\mathbf{d}})\end{subarray}}\langle\Upsilon_{S,{\mathbf{a}}}(\vec{v}),\Upsilon_{S,{\mathbf{a}}}(\vec{v})\rangle\cdot\langle\Upsilon_{S,{\mathbf{b}}}(\vec{v}),\Upsilon_{S,{\mathbf{b}}}(\vec{v})\rangle$
$\displaystyle=$ $\displaystyle
2^{|S|}\sum_{\begin{subarray}{c}{\mathbf{c}},{\mathbf{d}}\in[n]^{|S|}\\\
\mathrm{Ham}({\mathbf{c}},{\mathbf{d}})=|S|\end{subarray}}\langle\Upsilon_{S,{\mathbf{c}}}(\vec{v}),\Upsilon_{S,{\mathbf{c}}}(\vec{v})\rangle\cdot\langle\Upsilon_{S,{\mathbf{d}}}(\vec{v}),\Upsilon_{S,{\mathbf{d}}}(\vec{v})\rangle$
$\displaystyle\leq$ $\displaystyle
2^{|S|}\sum_{{\mathbf{c}},{\mathbf{d}}\in[n]^{|S|}}\langle\Upsilon_{S,{\mathbf{c}}}(\vec{v}),\Upsilon_{S,{\mathbf{c}}}(\vec{v})\rangle\cdot\langle\Upsilon_{S,{\mathbf{d}}}(\vec{v}),\Upsilon_{S,{\mathbf{d}}}(\vec{v})\rangle$
$\displaystyle=$ $\displaystyle
2^{|S|}\left(\sum_{{\mathbf{c}}\in[n]^{|S|}}\langle\Upsilon_{S,{\mathbf{c}}}(\vec{v}),\Upsilon_{S,{\mathbf{c}}}(\vec{v})\rangle\right)^{2}.$
Finally, we note that by definition, we have
$\sum_{{\mathbf{c}}\in[n]^{|S|}}\langle\Upsilon_{S,{\mathbf{c}}}(\vec{v}),\Upsilon_{S,{\mathbf{c}}}(\vec{v})\rangle=||{\vec{v}}||^{2}$,
which equals to $\mathrm{E}[Y]$. It follows that the variance in Equation (10)
can be bounded by
${\mathrm{Var}}[Y]\leq\sum_{S\subseteq[k],S\neq\emptyset}2^{|S|}\cdot\mathrm{E}[Y]^{2}=\mathrm{E}[Y]^{2}\sum_{i=1}^{k}{k\choose
i}2^{i}=(3^{k}-1)\mathrm{E}[Y]^{2},$
which finishes the proof of Lemma 4.1.
## 5\. Conclusion
There remain several open questions left in this space. Lower bounds,
particularly bounds that depend non-trivially on the dimension $k$, would be
useful. There may still be room for better algorithms for testing $k$-wise
independence in this manner using the $\ell_{2}$ norm. A natural
generalization would be to find a particularly efficient algorithm for testing
$k$-out-of-$s$-wise independence (other than handling each set of $k$ variable
separately). More generally, a question given in [12], to identify random
variables whose correlation exceeds some threshold according to some measure,
remains widely open.
## References
* [1] Noga Alon, Alexandr Andoni, Tali Kaufman, Kevin Matulef, Ronitt Rubinfeld, and Ning Xie. Testing k-wise and almost k-wise independence. In STOC ’07: Proceedings of the Thirty-ninth Annual ACM Symposium on Theory of Computing, pages 496–505, New York, NY, USA, 2007. ACM.
* [2] Noga Alon, László Babai, and Alon Itai. A fast and simple randomized parallel algorithm for the maximal independent set problem. J. Algorithms, 7(4):567–583, 1986.
* [3] Noga Alon, Oded Goldreich, and Yishay Mansour. Almost k-wise independence versus k-wise independence. Inf. Process. Lett., 88(3):107–110, 2003.
* [4] Noga Alon, Yossi Matias, and Mario Szegedy. The space complexity of approximating the frequency moments. J. Comput. Syst. Sci., 58(1):137–147, 1999.
* [5] T. Batu, L. Fortnow, E. Fischer, R. Kumar, R. Rubinfeld, and P. White. Testing random variables for independence and identity. In FOCS ’01: Proceedings of the 42nd IEEE symposium on Foundations of Computer Science, page 442, Washington, DC, USA, 2001. IEEE Computer Society.
* [6] Tugkan Batu, Ravi Kumar, and Ronitt Rubinfeld. Sublinear algorithms for testing monotone and unimodal distributions. In STOC ’04: Proceedings of the Thirty-sixth Annual ACM Symposium on Theory of Computing, pages 381–390, New York, NY, USA, 2004. ACM.
* [7] Vladimir Braverman and Rafail Ostrovsky. Measuring $k$-wise independence of streaming data. CoRR, abs/0806.4790v1, 2008.
* [8] Vladimir Braverman and Rafail Ostrovsky. Measuring independence of datasets. CoRR, abs/0903.0034, 2009.
* [9] Vladimir Braverman and Rafail Ostrovsky. AMS without $\bf{4}$-wise independence on product domains. CoRR, abs/0806.4790v3, 2009.
* [10] Kai-Min Chung, Zhenming Liu, and Michael Mitzenmacher. Testing $k$-wise independence over streaming data. Unpublished manuscript, available at http://www.eecs.harvard.edu/~michaelm/postscripts/sketchexttemp.pdf, 2009\.
* [11] Oded Goldreich and Dana Ron. On testing expansion in bounded-degree graphs. Technical report, Electronic Colloquium on Computational Complexity, 2000\.
* [12] Piotr Indyk and Andrew McGregor. Declaring independence via the sketching of sketches. In SODA ’08: Proceedings of the nineteenth annual ACM-SIAM symposium on Discrete algorithms, pages 737–745, Philadelphia, PA, USA, 2008\. Society for Industrial and Applied Mathematics.
* [13] E. L. Lehmann and Springer. Testing Statistical Hypotheses (Springer Texts in Statistics). Springer, January 1997.
* [14] Michael Mitzenmacher and Eli Upfal. Probability and Computing: Randomized Algorithms and Probabilistic Analysis. Cambridge University Press, New York, NY, USA, 2005.
* [15] Ronitt Rubinfeld and Rocco A. Servedio. Testing monotone high-dimensional distributions. In STOC ’05: Proceedings of the thirty-seventh annual ACM symposium on Theory of computing, pages 147–156, New York, NY, USA, 2005. ACM.
* [16] Mikkel Thorup and Yin Zhang. Tabulation based 4-universal hashing with applications to second moment estimation. In SODA ’04: Proceedings of the Fifteenth Annual ACM-SIAM Symposium on Discrete Algorithms, pages 615–624, Philadelphia, PA, USA, 2004\. Society for Industrial and Applied Mathematics.
| arxiv-papers | 2008-06-29T21:34:28 | 2024-09-04T02:48:56.439865 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Vladimir Braverman, Kai-Min Chung, Zhenming Liu, Michael Mitzenmacher,\n Rafail Ostrovsky",
"submitter": "Vladimir Braverman",
"url": "https://arxiv.org/abs/0806.4790"
} |
0806.4798 | # Coexistence of Supercondcutivity and Magnetism in LaFeAs(O0.94F0.06) Probed
by Muon Spin Relaxation
Soshi Takeshita1 E-mail address : soshi@post.kek.jp Ryosuke Kadono1,2
Masatoshi Hiraishi2
Masanori Miyazaki2 Akihiro Koda1,2 Yoichi Kamihara3 and Hideo Hosono3,4,5
1Institute of Materials Structure Science1Institute of Materials Structure
Science High Energy Accelerator Research Organization (KEK) High Energy
Accelerator Research Organization (KEK) 1-1 Oho 1-1 Oho Tsukuba Tsukuba
Ibaraki 305-0801 Ibaraki 305-0801 Japan
2Department of Materials Structure Science Japan
2Department of Materials Structure Science School of High Energy Accelerator
Science School of High Energy Accelerator Science The Graduate University for
Advanced Studies The Graduate University for Advanced Studies 1-1 Oho 1-1 Oho
Tsukuba Tsukuba Ibaraki 305-0801 Ibaraki 305-0801 Japan
3ERATO-SORST Japan
3ERATO-SORST JST JST in Frontier Research Center in Frontier Research Center
Tokyo Institute of Technology Tokyo Institute of Technology 4259 Nagatsuta
4259 Nagatsuta Midori-ku Midori-ku Yokohama Yokohama Kanagawa 226-8503
Kanagawa 226-8503 Japan
4Frontier Research Center Japan
4Frontier Research Center Tokyo Institute of Technology Tokyo Institute of
Technology 4259 Nagatsuta 4259 Nagatsuta Midori-ku Midori-ku Yokohama Yokohama
Kanagawa 226-8503 Kanagawa 226-8503 Japan
5Materials and Structures Laboratory Japan
5Materials and Structures Laboratory Tokyo Institute of Technology Tokyo
Institute of Technology 4259 Nagatsuta 4259 Nagatsuta Midori-ku Midori-ku
Yokohama Yokohama Kanagawa 226-8503 Kanagawa 226-8503 Japan
Japan
###### Abstract
The presence of macroscopic phase separation into superconducting and spin-
glass-like magnetic phases in LaFeAs(O1-xFx) is demonstrated by muon spin
rotation measurement in a sample near the phase boundary ($x=0.06$). Both
magnetism and superconductivity develop simultaneously below a common critical
temperature, $T_{\rm m}\simeq T_{\rm c}\simeq 18$ K. This remarkable
accordance strongly suggests that the electronic correlations leading to these
two competing ground states share a common origin.
oxypnictide superconductor, muon spin rotation, electronic correlation,
magnetism
The recent discovery of the oxypnictide superconductor LaFeAsO1-xFx (LFAO-F)
with a critical temperature ($T_{\rm c}$) of 26 K [1] and the successful
revelation of much increased $T_{\rm c}$ upon the substitution of La for other
rare-earth elements (such as Sm, leading to $\sim 43$ K [2]) and the
application of pressure for LFAO-F ($\sim 43$ K [3]) have triggered broad
interest in the mechanism yielding a relatively high $T_{\rm c}$ in this new
class of compounds. They have a layered structure like high-$T_{\rm c}$
cuprates, where the dopant and conducting layers are so separated that the
doped carriers (electrons introduced by the substitution of O2- with F- in the
La2O2 layers) move within the layers consisting of strongly bonded Fe and As
atoms. They exhibit another qualitative similarity to cuprates in that
superconductivity occurs upon carrier doping of pristine compounds that
exhibit magnetism. [4] Some preliminary results of the muon spin
rotation/relaxation ($\mu$SR) experiment on a variety of oxypnictide
superconductors showed that the superfluid density $n_{s}$ falls on the
empirical line on the $n_{s}$ vs $T_{\rm c}$ diagram observed for the
underdoped cuprates, [5, 6] from which possibility of the common mechanism of
superconductivity is argued between oxypnictides and cuprates.
However, in terms of the doping phase diagram, there are certain differences
between these two systems, e.g., (i) $T_{\rm c}$ ($>0$ for $0.4<x<0.12$) does
not vary much with $x$ [1] as in cuprates known as “bell-shaped” and (ii) the
magnetic (spin density wave, SDW) phase shares a boundary with the
superconducting phase near $x\simeq 0.04$.[1, 7] The insensitivity of $T_{\rm
c}$ to $x$ is reasonably understood from the conventional BCS theory where
condensation energy is predicted to be independent of carrier concentration.
The close relationship of magnetism and superconductivity suggests that a
detailed investigation of how these two phases coexist (and compete) near the
phase boundary will provide important clues to elucidating the paring
mechanism. Among various techniques, $\mu$SR has a great advantage in that it
can be applied in systems consisting of spatially inhomogeneous multiple
phases, providing information on respective phases according to their
fractional yield. Our $\mu$SR measurement in the LFAO-F sample with $x=0.06$
($T_{\rm c}\simeq 18$ K) reveals that these two phases indeed coexist in the
form of macroscopic phase separation, and more interestingly, that a spin
glass-like magnetic phase develops in conjunction with superconductivity in
the paramagnetic phase. This accordance strongly suggests a common origin of
the electronic correlation leading to these two competing phases.
Although the oxypnictide with rare-earth ($R$) substitution $R$FeAsO1-xFx
exhibits higher $T_{\rm c}$ than that of LFAO-F, strong random fields from
rare-earth ions preclude a detailed study of the ground state using sensitive
magnetic probes like $\mu$SR. Therefore, we chose the original LFAO-F system
for our $\mu$SR study. The target concentration of LaFeAsO1-xFx is set near
the phase boundary, $x=0.06$, for which a polycrystalline sample was
synthesized by solid state reaction. The detailed procedure for sample
preparation is described in an earlier report [1]. The sample was confirmed to
be mostly of single phase using X-ray diffraction analysis. Of two possible
impurity phases, namely, LaOF and FeAs, only the latter exhibits a magnetic
(helical) order with $T_{N}\simeq 77$ K.[8] As shown in Fig. 1, magnetic
susceptibility exhibits no trace of FeAs phase or local magnetic impurities
except below $\sim 50$ K where a small upturn is observed. The susceptibility
at a lower field [shown in Fig. 3(a)] provides evidence of bulk
superconductivity with $T_{\rm c}\sim 18$ K from the onset of diamagnetism.
Conventional $\mu$SR measurement was performed using the LAMPF spectrometer
installed on the M15 beamline of TRIUMF, Canada. During the measurement under
a zero field (ZF), residual magnetic field at the sample position was reduced
below $10^{-6}$ T with the initial muon spin direction parallel to the muon
beam direction [$\vec{P}_{\mu}(0)\parallel\hat{z}$]. For longitudinal field
(LF) measurement, a magnetic field was applied parallel to $\vec{P}_{\mu}(0)$.
Time-dependent muon polarization [$G_{z}(t)=\hat{z}\cdot\vec{P}_{\mu}(t)$] was
monitored by measuring decay-positron asymmetry along the $\hat{z}$-axis.
Transverse field (TF) condition was realized by rotating the initial muon
polarization so that $\vec{P}_{\mu}(0)\parallel\hat{x}$, where the asymmetry
was monitored along the $\hat{x}$-axis to obtain
$G_{x}(t)=\hat{x}\cdot\vec{P}_{\mu}(t)$. All the measurements under a magnetic
field were made by cooling the sample to the target temperature after the
field equilibrated.
Figure 1: (Color online) Magnetic susceptibility of LaFeAsO1-xFx with $x=0.06$
for the sample used for $\mu$SR measurement. Inset shows a reduced view of the
region below 35 K.
ZF-$\mu$SR is the most sensitive technique for examining magnetism in any
form, where the development of local magnetic moments leads to either the
spontaneous oscillation (for long-range order) or exponential damping
(inhomogeneous local magnetism) of $G_{z}(t)$. Figure 2 shows examples of
ZF-$\mu$SR time spectra collected at 2 and 30 K. The spectrum at 30 K
($>T_{\rm c}$) exhibits a Gaussian-like relaxation due to weak random local
fields from nuclear magnetic moments, indicating that the entire sample is in
the paramagnetic state. Meanwhile, the spectrum at 2 K is split into two
components, one that exhibits a steep relaxation and the other that remains to
show Gaussian-like relaxation. This indicates that there is a finite fraction
of implanted muons that sense hyperfine fields from local electronic moments.
The absence of oscillatory signal implies that the hyperfine field is highly
inhomogeneous, so that the local magnetism is characterized by strong
randomness or spin fluctuation. The fractional yield of the component showing
steep relaxation is as large as 25% (see below), which is hardly attributed to
impurity and therefore implies that the sample exhibits a macroscopic phase
separation into two phases.
Figure 2: (Color online) $\mu$SR time spectra observed in LaFeAs(O1-xFx) with
$x=0.06$ at 2 K under a longitudinal field (LF), a zero field (ZF), and a
transverse field (TF), and that under ZF at 30 K. The spectrum under TF is
plotted on a rotating reference frame of 6.78 MHz to extract the envelop
function.
The magnitude of the hyperfine field and that of spin fluctuation are
evaluated by observing the response of the $\mu$SR spectrum to a longitudinal
field (LF). It is shown in Fig. 2 that the relaxation in the paramagnetic
component is quenched by applying a weak magnetic field (LF=5 mT), which is
perfectly explained by the suppression of static nuclear dipolar fields
($<10^{0}$ mT). Meanwhile, the faster relaxation (seen for $0<t<1$ $\mu$s) due
to the magnetic phase is recovered only gradually over a field range of
$10^{1\sim 2}$ mT, and there still remains a slow relaxation even at the
highest field of 60 mT. This residual depolarization under LF is a clear sign
that local spins are slowly fluctuating, leading to the spin-lattice
relaxation of $\vec{P}_{\mu}(t)$. Such quasi-two-step relaxation is also
observed in dilute spin-glass systems,[9] which is understood as a
distribution of spin correlation time. A detailed analysis is made considering
that these two components coming from the magnetic phase (see below).
Under a transverse field, implanted muons experience an inhomogeneity of the
field [$B_{z}({\bf r})$] due to flux line lattice formation below $T_{\rm c}$
that leads to relaxation, in addition to those observed under a zero field.
The TF-$\mu$SR time spectrum in Fig. 2 (envelop part of the oscillation)
obtained under a field of 50 mT exhibits complete depolarization at 2 K,
indicating that the entire volume of the paramagnetic phase falls into the
superconducting state. The rapidly relaxing component observed under ZF is
also visible (although the coarse binning of the spectra for extracting the
envelop part makes it slightly obscure), indicating that the corresponding
part of the sample remains magnetic.
Considering the presence of the magnetic phase besides the paramagnetic
(=superconducting below $T_{\rm c}$) phase, we take special precaution to
analyze both TF and ZF/LF $\mu$SR spectra in a consistent manner. For the
determination of physical parameters describing the behavior of signals from
the magnetic phase, we first analyze ZF/LF spectra at 2 K using the
$\chi$-square minimization fit with the relaxation function
$G_{z}(t)=[w_{1}+\sum^{3}_{i=2}w_{i}\exp{(-\Lambda_{i}t)}]\cdot G_{\rm
KT}(\delta_{\rm N}:t),$ (1)
where $G_{\rm KT}(\delta_{\rm N}:t)$ is the Kubo-Toyabe relaxation function
for describing the Gaussian damping due to random local fields from nuclear
moments (with $\delta_{\rm N}$ being the linewidth) [10], $w_{1}$ is the
fractional yield for the paramagnetic phase, $w_{2}$ and $w_{3}$ are those for
the magnetic phase ($\sum w_{i}=1$) with $\Lambda_{2}$ and $\Lambda_{3}$ being
the corresponding relaxation rate described by the Redfield relation
$\Lambda_{i}=\frac{2\delta_{i}^{2}\nu_{i}}{\nu_{i}^{2}+\omega_{\mu}^{2}}\ \ \
(i=2,3),$ (2)
where $\omega_{\mu}=\gamma_{\mu}H_{\rm LF}$, $\gamma_{\mu}$ is the muon
gyromagnetic ratio ($=2\pi\times 135.53$ MHz/T), $H_{\rm LF}$ is the
longitudinal field, $\delta_{2}$ and $\delta_{3}$ are the means of the
hyperfine fields exerting on muons from local electronic moments, and
$\nu_{2}$ and $\nu_{3}$ are the fluctuation rates of the hyperfine field. The
solid curves in Fig. 2 show the result of analysis where all the spectra at
different fields (only ZF and LF=5mT are shown here) are fitted simultaneously
using eqs. (1) and (2) with common parameter values (except $\omega_{\mu}$
that is fixed to the respective value for each spectrum), which show excellent
agreement with all the spectra. The deduced parameters are as follows:
$w_{1}=0.754(9)$, $w_{2}=0.165(9)$, $w_{3}=0.081(4)$, $\delta_{2}=0.71(5)$
$\mu$s-1, $\delta_{3}=3.9(3)$ $\mu$s-1, $\nu_{2}=1.7(2)$ $\mu$s-1, and
$\nu_{3}=4(1)$ $\mu$s-1. Although the depolarization in the magnetic phase is
approximately represented by two components with different hyperfine couplings
($\delta_{i}$), the fluctuation rates ($\nu_{i}$) are close to each other (107
s-1 at 2 K), suggesting that the randomness is primarily due to the
distribution in the size of local moments (or in their distances to muons).
Since no impurity phase with a fraction as large as 25% is detected by X-ray
diffraction analysis, it is concluded that this magnetic phase is intrinsic.
In the analysis of temperature-dependent TF spectra, we used the relaxation
function
$\displaystyle G_{x}(t)$ $\displaystyle=$
$\displaystyle\exp(-\frac{1}{2}\delta^{2}_{\rm
N}t^{2})[w_{1}\exp(-\delta^{2}_{\rm s}t^{2})\cos{(2\pi f_{\rm s}t+\phi)}$ (3)
$\displaystyle+$ $\displaystyle(w_{2}+w_{3})\exp{(-\Lambda_{\rm
m}t)}\cos{(2\pi f_{\rm m}t+\phi)}],$
where $w_{i}$ and $\delta_{\rm N}$ are fixed to the values obtained by
analyzing ZF/LF-$\mu$SR spectra. The first component in the above equation
represents the contribution of flux line lattice formation in the
superconducting phase, where $\delta_{\rm s}$ corresponds to the linewidth
$\sigma_{\rm s}=\sqrt{2}\delta_{\rm s}=\gamma_{\mu}\langle(B({\bf
r})-B_{0})^{2}\rangle^{1/2}$ [with $B_{0}$ being the mean $B({\bf r})$], while
the second term represents the relaxation in the magnetic phase. Here, the
relaxation rate for the latter is represented by a single value $\Lambda_{\rm
m}$ (instead of $\Lambda_{2,3}$), as it turns out that the two components
observed under LF are hardly discernible in TF-$\mu$SR spectra. [This does not
affect the result of the analysis, because the amplitude is fixed to
$w_{2}+w_{3}$ so that $\Lambda_{\rm m}$ may represent a mean
$\simeq(w_{2}\Lambda_{2}+w_{3}\Lambda_{3})/(w_{2}+w_{3})$.] The fit analysis
using the above form indicates that all the spectra are perfectly reproduced
while the partial asymmetry is fixed to the value determined from ZF-$\mu$SR
spectra. This strengthens the presumption that the paramagnetic phase becomes
superconducting below $T_{\rm c}$. The result of analysis is summarized in
Fig. 3, together with the result of dc magnetization measured in the sample
from the same batch as that used for $\mu$SR.
It is interesting to note in Fig. 3(b) that, although the central frequency in
the superconducting phase ($f_{\rm s}$) does not show much change below
$T_{\rm c}\simeq 18$ K probably owing to a large magnetic penetration depth
(it is indeed large, see below), that in the magnetic phase ($f_{\rm m}$)
exhibits a clear shift in the negative direction below $T_{\rm m}\simeq T_{\rm
c}$. The magnitude of the shift is as large as $\sim 1$% and thus is readily
identified despite a relatively low external field of 50 mT. As shown in Fig.
3(c), the relaxation rate in the magnetic phase ($\Lambda_{\rm m}$) also
develops below $T_{\rm c}$ in accordance with the frequency shift,
demonstrating that a spin-glass-like magnetism sets in below $T_{\rm c}$.
Here, we note that the development of magnetic phase is already evident in the
ZF/LF-$\mu$SR spectra, and results are fully consistent with each other. The
onset of superconductivity below $T_{\rm c}$ is also confirmed by an increase
in $\delta_{\rm s}$, as observed in Fig. 3(c). This remarkable accordance of
onset temperature between magnetism and superconductivity strongly suggests
that there is an intrinsic relationship between the superconducting and
magnetic phases that leads to a common characteristic temperature.
Figure 3: (Color online) Temperature dependence of dc magnetic susceptibility
measured at 1mT (a), and that of physical parameters deduced by analyzing
TF-$\mu$SR spectra in superconducting ($f_{\rm s}$, $\delta_{\rm s}$) and
magnetic ($f_{\rm m}$, $\Lambda_{\rm m}$) phases (b-c), and of $\sigma_{\rm
s}$ ($=\sqrt{2}\delta_{\rm s}$) proportional to superfluid density (d). Curves
in (d) are fits by models described in the text.
The temperature dependence of $\sigma_{\rm s}$ in Fig. 3(d) is compared with
theoretical predictions for a variety of models with different order
parameters. The weak-coupling BCS model ($s$-wave, single gap) apparently
fails to reproduce the present data, as they exhibit a tendency to vary with
temperature over the region $T/T_{\rm c}<0.4$. Although a fit using a two-gap
model [11] shown by a dotted line seems to exhibit reasonable agreement with
data, the deduced gap parameters ($2\Delta_{i}/k_{\rm B}T_{\rm c}$) are
largely inconsistent with the prediction of the weak-coupling BCS model (see
Table 1). These observations suggest that the superconducting order parameter
is not described by a s-wave symmetry with a single gap. When a power law,
$\sigma_{\rm s}=\sigma_{0}[1-(T/T_{\rm c})^{\beta}]$, is used in fitting the
data, we obtain a curve shown by the broken line in Fig. 3(d) with an exponent
$\beta\simeq 2$. This is in good agreement with the case of $d$-wave symmetry
at the dirty limit.
Table 1: Parameters for defining the lines in Fig. 3(d). Two-gap | | Power law |
---|---|---|---
$T_{\rm c}$ (K) | 18.0(5) | $T_{\rm c}(K)$ | 18.9(4)
$\sigma(0)$ ($\mu{\rm s}^{-1}$) | 0.291(5) | $\sigma(0)$ ($\mu{\rm s}^{-1})$ | 0.291(4)
$w$ | 0.73(6) | $\beta$ | 1.7(1)
$2\Delta_{1}/k_{\rm B}T_{\rm c}$ | 2.6(3) | |
$2\Delta_{2}/k_{\rm B}T_{\rm c}$ | 1.1(3) | |
In the limit of extreme type II superconductors [i.e., $\lambda/\xi\gg 1$,
where $\lambda$ is the effective London penetration depth and
$\xi=\sqrt{\Phi_{0}/(2\pi H_{\rm c2})}$ is the Ginzburg-Landau coherence
length, $\Phi_{0}$ is the flux quantum, and $H_{\rm c2}$ is the upper critical
field], $\sigma_{\rm s}$ is determined by $\lambda$ using the relation [12]
$\sigma_{\rm s}/\gamma_{\mu}=2.74\times
10^{-2}(1-h)\left[1+3.9(1-h)^{2}\right]^{1/2}\Phi_{0}\lambda^{-2}$, where
$h=H_{\rm TF}/H_{\rm c2}$ and $H_{\rm TF}$ is the magnitude of external field.
From $\sigma_{\rm s}$ extrapolated to $T=0$ and taking $H_{\rm c2}\simeq 50$ T
(ref. [13]), we obtain $\lambda$=595(3) nm. Because of the large anisotropy
expected from the layered structure of this compound, $\lambda$ in the
polycrystalline sample would be predominantly determined by in-plane
penetration depth ($\lambda_{\rm ab}$). Using the formula
$\lambda=1.31\lambda_{\rm ab}$ for such a situation [14], we obtain
$\lambda_{\rm ab}$=454(2) nm. This value coincides with that expected from the
aforementioned empirical relation between $\lambda^{-2}_{\rm ab}$
superconductors [15, 6]. However, this may not be uniquely attributed to the
superfluid density because $\lambda$ depends not only $n_{\rm s}$ but also on
the effective mass, $\sigma_{\rm s}\propto\lambda^{-2}=n_{\rm
s}e^{2}/m^{*}c^{2}$.
Finally, we discuss the feature of the spin glass-like phase. Assuming that
the local moments are those of off-stoichiometric iron atoms with a moment
size close to that in the SDW phase ($\sim 0.36\mu_{B}$ [4]), the mean
distance between muon and iron moments in the relevant phase is estimated to
be $\sim 0.5$ nm from an average of $\delta_{i}$. Given the unit cell size
($a=0.403$ nm, $c=0.874$ nm [1]), this would mean that more than a quarter of
iron atoms in the magnetic phase (i.e., $\simeq 7$% of the entire sample)
should serve as a source of local moments. It is unlikely that such a
significant fraction of iron atoms remains as impurities in the present
sample.
It might also be noteworthy that there is an anomaly near $T_{\rm m2}\simeq
12$ K in the susceptibility [the onset of ZFC/FC hysteresis in Fig. 1 and a
steplike kink in Fig. 3(a)]. This seems to be in accordance with the onset of
a steeper increase in $\Lambda_{\rm m}$ below $T_{\rm m2}$, suggesting a
change in magnetic correlation.
$\mu$SR studies of LFAO-F have been made by a number of groups. According to
those preliminary studies, no clear sign of magnetism is observed in the
sample over relevant doping concentrations, except for a weak relaxation
observed far below $T_{\rm c}$ for $x=0.05$ and 0.075 in ZF-$\mu$SR spectra
and an unidentified additional relaxation observed in TF-$\mu$SR spectra for
$x=0.075$.[6, 5, 7] This led us to recall the sensitivity to chemical
stoichiometry in the emergence of the spin glass-like $A$-phase observed near
the boundary between the antiferromagnetic and superconducting phases in
CeCu2Si2.[16]. In addition to the $A$-phase, the present LFAO-F system
exhibits a closer similarity to this classical heavy-fermion superconductor
such as the phase diagram against pressure/doping.[17] Further study of the
dependence of fractional yield for the magnetic phase with varying $x$ (in
small steps near the phase boundary) would provide further insight into the
true nature of these phases and the mechanism of superconductivity itself that
is working behind the coexistence/competition.
In summary, it has been revealed by our $\mu$SR experiment that
superconducting and magnetic phases coexist in LaFeAs(O0.94F0.06) with
$x=0.06$. These two phases simultaneously develop just below $T_{\rm c}$,
strongly suggesting an intimate and intrinsic relationship between these two
phases. The result of TF-$\mu$SR measurement suggests that the
superconductivity of LaFeAs(O0.94F0.06) cannot be explained by the
conventional weak-BCS model (single gap, $s$-wave).
We would like to thank the TRIUMF staff for their technical support during the
$\mu$SR experiment. This work was partially supported by the KEK-MSL Inter-
University Program for Oversea Muon Facilities and by a Grant-in-Aid for
Creative Scientific Research on Priority Areas from the Ministry of Education,
Culture, Sports, Science and Technology, Japan.
## References
* [1] Y. Kamihara, T. Watanabe, M. Hirano, and H. Hosono: J. Am. Chem. Soc. 130 (2008) 3296.
* [2] X. H. Chen, T. Wu, R. H. Liu, H. Chen, and D. F. Fang : Nature 453 (2008) 761.
* [3] H. Takahashi, K. Igawa, K. Arii, Y. Kamihara, M. Hirano, and H. Hosono: Nature 453 (2008) 376.
* [4] C. de la Cruz, Q. Huang, J. W. Lynn, J. Li, W. Ratcliff II, J. L. Zarestky, H. A. Mook, G. F. Chen, J. L. Luo, N. L. Wang and P. Dai: Nature 453 (2008) 899\.
* [5] H. Luetkens, H. H. Klauss, R. Khasanov, A. Amato, R. Klingeler, I. Hellmann, N. Leps, A. Kondrat, C. Hess, A. Köhler, G. Behr, J. Werner, and B. Büchner: arXiv:0804.3115.
* [6] J. P. Carlo, Y. J. Uemura, T. Goko, G. J. MacDougall, J. A. Rodriguez, W. Yu, G. M. Luke, P. Dai, N. Shannon, S. Miyasaka, S. Suzuki, S. Tajima, G. F. Chen, W. Z. Hu, J. L. Luo, and N. L. Wang: arXiv:0805.2186.
* [7] H. Luetkens, H. H. Klauss, M. Kraken, F. J. Litterst, T. Dellmann, R. Klingeler, C. Hess, R. Khasanov, A. Amato, C. Baines, J. Hamann-Borrero, N. Leps, A. Kondrat, G. Behr, J. Werner and B. Büchner: arXiv:0806.3533.
* [8] K. Selte, A. Kjekshus and A. Andresen: Acta Chemica Scandinavica 26 (1972) 3101.
* [9] Y. J. Uemura, K. Nishiyama, T. Yamazaki and R. Nakai: Solid State Comm. 39 (1981) 461.
* [10] R. S. Hayano, Y. J. Uemura, J. Imazato, N. Nishida, T. Yamazaki, and R. Kubo: Phys. Rev. B 20 (1979) 850.
* [11] F. Bouquet, Y. Wang, R. A. Fisher, D. G. Hinks, J. D. Jorgensen, A. Junod, and N. E. Phillips: Europhys. Lett. 56 (2001)856.
* [12] E. H. Brandt: Phys. Rev. B 37 (1988) 2349.
* [13] F. Hunte, J. Jaroszynski, A. Gurevich, D. C. Larbalestier, R. Jin, A. S. Sefat, M. A. McGuire, B. C. Sales, D. K. Christen, and D. Mandrus: Nature 453 (2008) 903.
* [14] V. I. Fesenko, V. N. Gorbunov, and V. P. Smilga: Physica C 176 (1991) 551.
* [15] Y. J. Uemura, V. J. Emery, A. R. Moodenbaugh, M. Suenaga, D. C. Johnston, A. J. Jacobson, J. T. Lewandowski, J. H. Brewer, R. F. Kiefl, S. R. Kreitzman, G. M. Luke, T. Riseman, C. E. Stronach, W. J. Kossler, J. R. Kempton, X. H. Yu, D. Opie, and H. E. Schone: Phys. Rev. B 38 (1988) 909.
* [16] R. Feyerherm, A. Amato, C. Geibel, F. N. Gygax, P. Hellmann, R. H. Heffner, D. E. MacLaughlin, R. Müller-Reisener, G. J. Nieuwenhuys, A. Schenck, F. Steglich: Phys. Rev. B 56 (1997) 699.
* [17] P. Monthoux, D. Pines, and G. G. Lonzarich: Nature 450 (2007) 1177.
| arxiv-papers | 2008-06-30T03:14:09 | 2024-09-04T02:48:56.446081 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "S. Takeshita, R. Kadono, M. Hiraishi, M. Miyazaki, A. Koda, Y.\n Kamihara, H. Hosono",
"submitter": "Soshi Takeshita",
"url": "https://arxiv.org/abs/0806.4798"
} |
0806.4892 | # GD 552: a cataclysmic variable with a brown dwarf companion?
E. Unda-Sanzana1, T. R. Marsh2, B. T. Gänsicke2, P. F. L. Maxted3, L. Morales-
Rueda4, V. S. Dhillon5, T. D. Thoroughgood5, E. Tremou6, C. A. Watson5, R.
Hinojosa-Goñi1
1 Instituto de Astronomía, Universidad Católica del Norte, Antofagasta, Chile;
eundas@almagesto.org; rhg002@ucn.cl
2 Dept. of Physics, University of Warwick, Coventry, United Kingdom CV4 7AL;
t.r.marsh@warwick.ac.uk; boris.gaensicke@warwick.ac.uk
3 Dept. of Physics and Astronomy, Keele University, Keele, Staffordshire ST5
5BG, United Kingdom; pflm@astro.keele.ac.uk
4 Dept. of Astrophysics, University of Nijmegen, 6500 GL Nijmegen, The
Netherlands; lmr@astro.ru.nl
5 Dept. of Physics and Astronomy, University of Sheffield, Sheffield S3 7RH,
United Kingdom; vik.dhillon@sheffield.ac.uk, C.Watson@sheffield.ac.uk
6 Dept. of Physics, Sect. of Astrophysics, Astronomy & Mechanics, Univ. of
Thessaloniki, 541 24 Thessaloniki, Greece; etremo@physics.auth.gr
(Accepted . Received ; in original form )
###### Abstract
GD 552 is a high proper motion star with the strong, double-peaked emission
lines characteristic of the dwarf nova class of cataclysmic variable star, and
yet no outburst has been detected during the past 12 years of monitoring. We
present spectroscopy taken with the aim of detecting emission from the mass
donor in this system. We fail to do so at a level which allows us to rule out
the presence of a near-main-sequence star donor. Given GD 552’s orbital period
of $103$ minutes, this suggests that it is either a system that has evolved
through the $\sim 80$-minute orbital period minimum of cataclysmic variable
stars and now has a brown dwarf mass donor, or that has formed with a brown
dwarf donor in the first place. This model explains the low observed orbital
velocity of the white dwarf and GD 552’s low luminosity. It is also consistent
with the absence of outbursts from the system.
###### keywords:
binaries: close – binaries: spectroscopic – stars: individual: GD 552 –
accretion
††pagerange: GD 552: a cataclysmic variable with a brown dwarf
companion?–References††pubyear: 2007
## 1 Introduction
GD 552 is a blue, high proper motion star ($0.18^{\prime\prime}/$yr)
discovered by Giclas et al. (1970). It was first observed spectroscopically by
Greenstein & Giclas (1978), who found that it is a cataclysmic variable star
(CV) in which a white dwarf accretes matter from a low-mass companion (often
termed primary and secondary star, respectively). GD 552’s proper motion and
position close to the plane of the Galaxy (galactic latitude $4^{\circ}$),
combined with its blue colour all suggest that it is relatively close by.
Greenstein & Giclas (1978) suggest a distance of $\sim 70\,$pc which gives a
transverse motion which is reasonable for a member of the Galactic disk, and
combined with its magnitude $V=16.5$, suggests $M_{V}\sim 12.5$, the
equivalent of a $0.6\,{\rm M}_{\odot}$ white dwarf with a temperature of only
$9000\,$K. The probable low luminosity of GD 552, which is of central
importance to this paper, is backed up by an absence of any observed
outbursts, suggesting that it may very rarely or never have outbursts
(Cannizzo 1993).
The main spectroscopic study of GD 552 so far has been carried out by Hessman
& Hopp (1990) (hereafter HH1990) who determined the orbital period of GD 552
of $102.7\,$min. They observed an extreme Balmer decrement (H $\alpha$:H
$\beta$= 6.2:1.0), indicative of a cool, optically thin disc, another
indication of low luminosity (Williams & Shipman 1988). HH1990 measured the
white dwarf’s projected orbital velocity to be $K_{1}=17\pm
4\,\mathrm{km}\,{\mathrm{s}^{-1}}$. This is a very low value suggestive of a
low inclination system, since for edge-on systems with orbital periods similar
to GD 552 $K_{1}$ is typically $\sim 60-80\,\mathrm{km}\,{\mathrm{s}^{-1}}$.
However, the emission lines from the accretion disc suggest instead a
moderately inclined system as they display clearly separated double-peaks
(which come from the outer disc, Horne & Marsh 1986) with velocities of $\pm
450\,\mathrm{km}\,{\mathrm{s}^{-1}}$ which can be compared to typical peak
velocities of $\sim 600\,\mathrm{km}\,{\mathrm{s}^{-1}}$ for edge-on systems.
To solve this conundrum, HH1990 suggested that the white dwarf in GD 552 is
unusually massive – close to the Chandrasekhar limit in fact – allowing a low
orbital inclination ($\sim 20^{\circ}$) and therefore the small $K_{1}$ value
at the same time as large disc velocities. HH1990 were forced to their model
because they assumed that the companion to the white dwarf had to be a main-
sequence star. Since the companion fills its Roche lobe, Roche geometry and
the orbital period uniquely specify its density ($36.5\,{\rm g}\,{\rm
cm}^{-3}$), which, if it is a main-sequence star, also fixes its mass, which
turns out to be $M_{2}\sim 0.13\,{\rm M}_{\odot}$. This would cause a much
larger $K_{1}$ than observed, unless the inclination is low.
More recent theoretical work has revised HH1990’s estimate a little to
$M_{2}\sim 0.15\,{\rm M}_{\odot}$ from GD 552’s orbital period (Kolb & Baraffe
1999), but the problem is qualitatively unchanged. However, this work also
suggests a very different scenario. Cataclysmic variables at long orbital
periods are thought to evolve towards shorter periods until the companion
becomes degenerate at a period near 70 minutes. After this time, the mass
donor increases in size as it loses mass and the orbital period lengthens
(e.g. Howell, Nelson & Rappaport, 2001; Kolb 1993). Thus although if GD 552 is
approaching the period minimum its donor mass must be $M_{2}\sim 0.15\,{\rm
M}_{\odot}$, if it has already passed the minimum and is now evolving to
longer periods, it would be about a factor of four times less massive, and
there would be no need for HH1990’s massive white dwarf, face-on model.
Moreover, such a system would have an extremely low mass transfer rate,
consistent with GD 552’s lack of outbursts and probably low intrinsic
luminosity. The evolution just described is the standard explanation for the
observed minimum orbital periods of CVs, which however fails on two counts.
First, the observed minimum around 80 min is distinctly longer than the
theoretical value of 70 min (Kolb & Baraffe 1999; although Willems et al.
(2005) propose that the period minimum might be longer than expected because
of extra angular momentum loss from circumbinary disks), and, second, while we
expect most systems to have passed the period minimum there is just one single
well-established example of such a system known (SDSS 103533.03+055158.4,
Littlefair et al. 2006) although there are a few other good candidates
(Mennickent et al. 2001; Littlefair, Dhillon & Martin 2003; Patterson,
Thorstensen & Kemp 2005; Araujo-Betancor et al. 2005; Southworth et al. 2006).
Therefore it is of considerable interest to establish whether GD 552 is a pre-
or post-period-minimum system.
In this article we carry out a test to distinguish between the two models of
GD 552. If the pre-period-bounce model is correct, then, as we show in this
paper, we should be able to detect features from the M star in a low mass-
transfer rate system such as GD 552. If, conversely, the post-period-bounce
model is correct, the donor will be extremely faint and it should not be
detectable. In summary, our task is to set the strongest possible limit upon
the presence of a hypothetical M dwarf such that we can say that we would have
seen it had GD 552 been a pre-bounce system.
## 2 Observations and data reduction
We carried out observations of GD 552 in January and August 2001 on the island
of La Palma in the Canary Islands (see Table 1 for details). In January 2001,
the 2.5-m Isaac Newton Telescope (INT) was used in conjunction with the
Intermediate Dispersion Spectrograph (IDS) to acquire one dataset. Another
dataset obtained in January 2001 used the 4.2-m William Herschel Telescope
(WHT) in conjunction with the double-beamed high-resolution ISIS spectrograph.
In August 2001 only WHT/ISIS was used. The weather was good during both runs,
with no clouds and a typical seeing of 1 arcsecond.
An additional low-resolution spectrum was acquired on January 4, 2005, again
with WHT/ISIS (see Table 1 for details). This spectrum was acquired with a
vertical slit in order to obtain approximately correct relative fluxes. It was
calibrated using observations of Feige 34 (Oke 1990). The seeing on this night
was around $2.5"$ and variable and so the flux calibration is only
approximate. See Table 1 for details.
Figure 1: Average spectra for the high-resolution datasets. A correction for
telluric absorption has been made to the I-band spectra, although the very
heavy absorption at 7600 Å could not be successfully removed. Figure 2: Low
resolution spectra acquired in January 2005. The upper panel is a vertical
enlargement of the lower one. Note the broad absorption wings around the
emission lines.
We flux calibrated the earlier spectra using observations of the
spectrophotometric standard HD19445 which we also used to remove the effect of
telluric lines on the red data (Bessell 1999). Flat-fields and comparison arc
spectra were taken at regular intervals (every $\sim 60\,$min).
The spectra were optimally extracted (Marsh 1989), with flat fields
interpolated from the many ones taken during the night. The wavelength scales
for each science spectrum were obtained by interpolating the solutions of the
nearest two arc spectra. For the January 2001 datasets a comparison star was
included in the slit. In this case slit losses were corrected using the ratio
of the spectra of the comparison star to a spectrum taken with a wide slit
close to the zenith. For the August 2001 datasets no comparison star was
observed, so we simply normalised the continua of these spectra.
Six images were acquired with the WHT on 25 September 2006 using a Harris
I-band filter. One Landolt standard (Mark A1) with very similar airmass was
observed to flux calibrate the images. Observations from the nearby Carlsberg
Automatic Meridian
Circle111http://www.ast.cam.ac.uk/$\sim$dwe/SRF/camc_extinction.html on the
same night show that the night was mostly photometric with normal extinction.
Finally, 5.4h of filterless CCD photometry of GD552 was obtained on 29 July
2005 using a SI-502 $516\times 516$ pixel camera on the 1.2m telescope at
Kryoneri Observatory. The data were reduced using the pipeline described in
Gänsicke et al. (2004), where we used USNO-A2.0 1500-09346766 (located $\sim
1\arcmin$ north-east of GD552, $B=16.8$, $R=15.6$) as comparison star.
Ultraviolet STIS spectroscopy of GD552 was retrieved from the HST archive. The
system was observed with the far-ultraviolet (FUV) G140L grating and the
$52\arcsec\times 0.2\arcsec$ aperture on 24 October 2002 for a total of $\sim
4$ h. The observation was split into 15 individual exposures of 970 s each. An
additional three 1410 s near-ultraviolet spectra using the G230LB grating were
obtained on 31 August 2002, again using the $52\arcsec\times 0.2\arcsec$
aperture.
Table 1: Summary of the WHT and INT spectroscopic data used in the analysis. In this table MD stands for ’mean dispersion’. FWHM is the full width at half maximum of the unblended arc lines. T is the mean exposure time per frame and DT is the average dead time between exposures. N is the number of spectra collected per night per arm. Telescope/ | CCD/Grating | Date | Start - End | Orbits | $\lambda$ range | MD | FWHM | T/DT | N
---|---|---|---|---|---|---|---|---|---
Instrument | | | (UT) | covered | (Å) | (Å pixel-1) | (Å) | (s) |
INT/IDS | EEV10/R1200B | 12/13 Jan 2001 | 19:48-22:06 | 1.343 | 6318-6719 | 0.39 | 0.78 | 300/30 | 26
WHT/ISIS | EEV12/H2400B | 12/13 Jan 2001 | 19:52-21:13 | 0.788 | 4618-4985 | 0.21 | 0.42 | 300/12 | 32
WHT/ISIS | EEV12/R1200B | 13/14 Aug 2001 | 23:28-05:34 | 2.979 | 4301-4962 | 0.22 | 0.69 | 290/16 | 61
WHT/ISIS | TEK4/R316R | 13/14 Aug 2001 | 23:35-05:34 | 2.911 | 7306-8814 | 1.48 | 3.43 | 300/5 | 63
WHT/ISIS | EEV12/R1200B | 04/05 Jan 2005 | 19:50-19:58 | 0.081 | 3900-5390 | 0.88 | 1.8 | 500/0 | 1
WHT/ISIS | TEK4/R316R | 04/05 Jan 2005 | 19:50-19:58 | 0.081 | 6110-8930 | 1.65 | 3.2 | 500/0 | 1
## 3 Analysis
### 3.1 Average profiles and trailed spectra
Figure 1 shows average spectra for our data. Figure 2 shows the available low
resolution spectra. We observe double peaked profiles in all the detected
lines: He i 4471.68 Å (hereafter He i 4471), He ii 4685.750 Å (hereafter He
ii), He i 4921.93 Å (hereafter He i 4921), He i 6678.15 Å (hereafter He i
6678), O i 7773, O i 8446 and the Paschen and Balmer series. In the general
theory of CVs, the double peaked profiles are consistent with disc emission.
A second order polynomial was fitted to the continuum of each dataset and then
the data were divided by these fits. The normalised continua were subtracted,
and the datasets were binned in 20 (Jan) or 30 (Aug) orbital phase bins before
plotting Figure 3. These trailed spectra display double peaked emission
following a sinusoidal motion with orbital phase. We interpret this as the
rotating accretion disk, approximately tracking the emission of the white
dwarf it surrounds. The other remarkable feature in these spectra is a higher-
amplitude sinusoid also varying with orbital phase, but shifted with respect
to the assumed zero phase. It can be seen very clearly, for instance, in O i
7773 (Figure 3). This emission is produced in the stream/disc impact region,
which is usually termed the bright spot. Note that the phasing of this
sinusoidal is different for the January and August datasets. This is due to
the uncertainty in the zero phase of the ephemeris used to fold the data
(HH1990’s).
Using the average of the WHT spectra acquired in January 2005, we measured the
Balmer decrement defined as the ratio of line fluxes
$\hbox{$\hbox{H\,$\alpha$}$}:\hbox{$\hbox{H\,$\beta$}$}:\hbox{$\hbox{H\,$\gamma$}$}$.
We did this by integrating the flux from each profile after subtracting a low
order fit to the continua. We computed
$\hbox{$\hbox{H\,$\alpha$}$}:\hbox{$\hbox{H\,$\beta$}$}:\hbox{$\hbox{H\,$\gamma$}$}=2.0:1.0:0.56$.
Our values for the Balmer decrement differ noticeably from HH1990’s
($\hbox{$\hbox{H\,$\alpha$}$}:\hbox{$\hbox{H\,$\beta$}$}:\hbox{$\hbox{H\,$\gamma$}$}=6.2:1.0:0.26$),
being rather closer to Greenstein & Giclas (1978)
($\hbox{$\hbox{H\,$\alpha$}$}:\hbox{$\hbox{H\,$\beta$}$}:\hbox{$\hbox{H\,$\gamma$}$}=2.2:1.0:0.5$).
Qualitatively, though, our conclusion is the same: the Balmer decrements
suggest that the accretion disk is optically thin and cool (see, for example,
Williams & Shipman, 1988). This conclusion is strengthened by the detection of
O i 7773 which, according to Friend et al. (1988) is a good indicator of the
state of the disc, its emission indicating an optically thin accretion disc.
Figure 3: Phase binned data. The fluxes in this figure have been normalised by
the maximum value the flux reaches for each wavelength. We set the colour
scale to white for 0 flux and to black for 85% of this normalised flux, with
the exception of line He i 4921for which black corresponds to 40% of the
normalised flux.
### 3.2 The primary star
HH1990 measured a value of $17\pm 4\,\mathrm{km}\,{\mathrm{s}^{-1}}$ for
$K_{1}$. We tried to check and refine this value by using Schneider & Young
(1980)’s method. We convolved our data with a difference of Gaussians
equidistant from a candidate line centre, and define the velocity as the
displacement of the double Gaussians such that the convolution equals zero. By
increasing the separation of the Gaussians we gather information from the
wings of the profile, which is produced in regions of the disc closer to its
centre (Horne & Marsh 1986). We assume that the motion of these regions tracks
the motion of the accreting object, so the further into the wings the
measurement is made, the better the estimate of the motion of the white dwarf.
The process does not continue indefinitely because there is a maximum velocity
in the line wings, which corresponds to the Keplerian motion of the innermost
ring of gas just before settling onto the white dwarf. However, in practice,
before reaching this limit the reliability of the calculation is constrained
by the noise present at the continuum level.
The motion of each disc region is calculated by fitting the orbital solution:
$V=\gamma-K\sin\left(\frac{2\pi(t-t_{0})}{P}\right)$ (1)
to the measurements. $\gamma$ is the systemic velocity, $K$ is the velocity of
the source of the emission, $t_{0}$ is the time of superior conjunction of the
emission-line source, and P is the orbital period of the system. Further, we
avail ourselves of a diagnostic diagram (Shafter, Szkody & Thorstensen 1986)
to decide when the calculation is becoming dominated by noise as the
separation of Gaussians ($a$) increases. This should reveal itself as a sharp
rising of the statistics $\sigma_{K}/K$. At the same time, we also expect to
see convergence of the calculated parameters ($K$, $\gamma$) and a phasing
appropriate to the white dwarf.
In our diagrams (Fig. 4) we do not put a strong constraint on the value of
convergence for the phasing (bottom panels) because the ephemeris of this
system has a large uncertainty and thus its orientation is arbitrary for the
date of our observation. For $a$ larger than
$2500\,\mathrm{km}\,{\mathrm{s}^{-1}}$ the noise dominates both diagrams so
this is the upper limit for the abscissa in our plots. The H $\beta$ diagram
displays a reasonable degree of convergence when $a$ approaches
$2000\,\mathrm{km}\,{\mathrm{s}^{-1}}$. On reaching that value the noise takes
off. On the other hand, for large values of $a$ in H $\alpha$ we are picking
data contaminated by the wings of He i 6678 and thus we do not expect a clear
convergence in this diagram, although the noise seems to dominate once $a$
reaches $2200\,\mathrm{km}\,{\mathrm{s}^{-1}}$. Eventually, as it is not
obvious which value to favour in H $\alpha$ we take as $K_{1}$ the value of
$K$ for $a=1900\,\mathrm{km}\,{\mathrm{s}^{-1}}$ in H $\beta$, right before
the noise begins to dominate. This is $14.1\pm
2.7\,\mathrm{km}\,{\mathrm{s}^{-1}}$ consistent with HH1990. The uncertainties
were estimated by means of the bootstrap method, repeating the calculation for
10000 bootstrap samples (Diaconis & Efron 1983).
Figure 4: Diagnostic diagram for H $\alpha$ (left) and H $\beta$ (right),
using INT January 2001 and WHT August 2001 data respectively.
The high resolution B-band spectrum of Fig. 1 shows signs of broad absorption
wings flanking the Balmer lines. These are often seen in low luminosity dwarf
novae (e.g. Szkody et al. 2007) and originate in the photosphere of the white
dwarf. Low resolution spectra taken in January 2005 and September 2006 confirm
the presence of the white dwarf photospheric lines (Fig. 2). The white dwarf
is also revealed in HST/STIS ultraviolet spectra through the a very broad L
$\alpha$ feature extending to 1600A (Fig. 6). This feature, which results from
L $\alpha$ from neutral hydrogen plus quasi-molecular features at 1400 and
1600 Å from $H_{2}^{+}$ and $H_{2}$, is a hallmark of cool white dwarfs.
For the purpose of the spectral analysis, we have computed a grid of white
dwarf models covering effective temperatures in the range 8000 K–15 000 K and
surface gravities of $\log g=7.5$, 8.0, and 8.5 using the stellar atmosphere
and spectral synthesis codes TLUSTY/SYNSPEC (Hubeny & Lanz 1995). Metal
abundances were set to 0.1 times their solar values, as suggested by the
weakness of the Mg$\lambda\lambda$2796,2803 resonance doublet. The assumed
value for the metal abundance is within the typical range for CVs (Sion &
Szkody 2005). A first exploratory fit revealed that none of the white dwarf
model spectra could reproduce the STIS data alone, as the flux of the model
spectra rapidly drops to zero shortward of $\sim 1300$ Å, in contrast to the
observed spectrum that has a substantial non-zero flux down to 1150 Å.
Moreover, it turns out that the 15 individual far-ultraviolet spectra show
substantial flux variability, supporting the presence of a second (variable)
component. The ultraviolet flux variability shows a clear dependence on
wavelength, and we have fitted the continuum slope of the spectrum with a
linear slope, $F_{\lambda}=7.22\times 10^{-19}\lambda-5.9\times
10^{-16}\,\mathrm{erg\,cm^{-2}s^{-1}\AA^{-1}}$. Qualitatively, such a slope is
what could be expected from either an optically thin emission region (where
absorption is dominated by bound-free absorption) or an optically thick
relatively cool quasi-blackbody component (where the far-ultraviolet emission
is on the Wien part of the blackbody spectrum). We then explored two-component
fits to the STIS data consisting of this linear slope plus a white dwarf
model, with the contributions of the slope and the white dwarf being
normalised to match the observed continuum below 1300 Å and in the range
1580–1620 Å, respectively. The best fit for $\log g=7.5$ is shown in Fig. 5
(left panel). This simple model provides a surprisingly good match to the STIS
spectrum over the entire ultraviolet range. However, an extrapolation of this
simple linear slope over a broader wavelength range, i.e. into the optical, is
not warranted, as the exact nature of this emission component is not known.
The parameters determined from the fit are the white dwarf effective
temperature $T_{\mathrm{eff}}$, and the flux scale factor between the
synthetic and the observed spectrum. Assuming a white dwarf radius, the flux
scaling factor can be used to calculate the distance to the system. As the
white dwarf mass is unconstrained by this analysis, we ran the fit for $\log
g=7.5$, 8.0, and 8.5, corresponding to white dwarf masses and radii of (0.33
${\rm M}_{\odot}$, $1.15\times 10^{9}$ cm), (0.57 ${\rm M}_{\odot}$,
$8.65\times 10^{8}$ cm), and (0.9 ${\rm M}_{\odot}$, $6.15\times 10^{8}$ cm),
respectively (where we have assumed a Hamada-Salpeter (1961) Carbon-Oxygen
core mass-radius relation). The best fit parameters for $\log g=7.5$, 8.0, and
8.5 are (10 500 K, 125 pc), (10 900 K, 105 pc), and (11 300 K and 85 pc),
respectively. Because of the very strong temperature dependence of the 1600 Å
H${}_{2}^{+}$ the STIS spectrum constrains the white dwarf temperature within
a very narrow range.
Fig. 5 (right panel) shows that the white dwarf model obtained from the fit to
the STIS data extended into the optical range clearly underpredicts the
observed flux by $\sim 50$%. The presence of double-peaked Balmer emission
lines clearly reveals the presence of an accretion disc in the system that is
contributing light in the optical. We model the accretion disc by an
isobaric/isothermal hydrogen slab, which has three free parameters: the disc
temperature $T_{\mathrm{d}}$, the column density $\Sigma$, and the flux
scaling factor (see Gänsicke et al. 1997, 1999 for details). A disc
temperature of $T_{\mathrm{d}}=5800$ K and a column density of
$\Sigma=2.3\times 10^{-2}\mathrm{g\,cm^{-2}}$, scaled to the observed
H$\alpha$ emission line flux provides an adequate model of the optical
spectrum. The emission line flux ratios of the model are
$\hbox{$\hbox{H\,$\alpha$}$}:\hbox{$\hbox{H\,$\beta$}$}:\hbox{$\hbox{H\,$\gamma$}$}=2.2:1.0:0.56$,
fairly close to the observed values (Sect. 3.1). For a distance of $\sim 100$
pc, the area of hydrogen slab implied by the flux scaling factor is consistent
with the size of the Roche lobe of the white dwarf. The disc parameters found
here are very similar to those found in a number of other studies of quiescent
accretion discs in CVs (e.g. Williams 1980, Marsh 1987, Lin et al. 1988,
Rodriguez-Gil et al. 2005). We refrained from more detailed modelling of the
optical data as the spectrum was obtained under non-photometric conditions,
and the flux calibration is subject to some uncertainty. Nevertheless the
temperature derived from the UV data is consistent with our detection of the
white dwarf’s photosphere in our optical spectra.
The recent discovery of several non-radially pulsating white dwarfs in CVs
(e.g. van Zyl et al. 2004, Warner & Woudt 2004, Araujo-Betancor et al. 2005,
Patterson et al. 2005, Warner & Woudt 2005 and Gänsicke 2006) motivated the
short photometric time series observation of GD552 carried out at Kryoneri
observatory. Visual inspection of the light curve (Fig. 6) reveals substantial
variability on time scales of tens of minutes, plus a modulation over $\sim
4.5$ h. A Lomb-Scargle period analysis (Lomb 1976, Scargle 1982) shows no
substantial power at periods shorter than $\sim 15$ min, where non-radial
pulsations would be expected to show up. The conclusions from this admittedly
limited set of data is that the white dwarf in GD 552 is probably too cool to
drive ZZ Ceti pulsations in its envelope and the varibility seen in the light
curve is probably due to flickering in the accretion disc.
Figure 5: Left: The archival STIS ultraviolet spectrum of GD552, plotted in
gray. The solid line shows a two-component fit consisting of a linear slope
plus a white dwarf model spectrum ($T_{\mathrm{eff}}=10\,500$ K, $\log
g=7.5$), both components are plotted individually as dashed lines. Shown as
dotted line are the same two-component models, but for
$T_{\mathrm{eff}}=11\,000$ K (top curve) and 10 000 K (bottom curve). Right:
The STIS ultraviolet spectrum along with our WHT low-resolution spectrum,
plotted in gray. Plotted as solid black line is the sum of the white dwarf
model as shown in the left panel plus the spectrum of an isothermal/isobaric
hydrogen slab. Both model components are plotted individually as dotted lines.
Figure 6: Kryoneri CCD photometry of GD 552.
### 3.3 The secondary star
The presence of a secondary star (assumed to be an M dwarf since we are
testing for the presence of a star of mass $0.15\,{\rm M}_{\odot}$) is not
obvious either in the average or in the trailed spectra.
Our main aim was to detect photospheric absorption lines, especially the NaI
doublet near 8200 Å which is strong in late M dwarfs. To do so we compared our
GD 552 spectra with a set of M dwarf template spectra kindly provided by Kelle
Cruz (personal communication) following the classification by Kirkpatrick,
Henry & McCarthy (1991). The stars used are listed in Table 2.
The Na I doublet near 8200 Å and nearby molecular bands are the fingerprints
of M dwarfs. As we were not able to recognise any such features in the GD 552
data, we looked for a constraint upon its presence given that its features are
not detectable. First we normalised the spectra of GD 552 and of each template
in Table 2. We did this by dividing by a constant fit to the continuum in the
range 8100-8400 Å, excluding the $\sim 8200$ Å doublet. Then we subtracted
each normalised template from the GD 552 data in 5% steps. In Figure 7 we show
an example of this procedure. GD 552 is at the bottom of the plot. Each
profile above GD 552 increases the amount of template subtracted from the data
by 5%. Consistently, we found that for all the templates the presence of M
star features is obvious once we reach 10%.
Figure 7: From bottom to top: normalised template star (Gl 65A); normalised GD 552; and the four following curves are normalised GD 552 minus 5%, 10%, 15% and 20% of the normalised template respectively. When 10% of Gl 65A is subtracted from GD 552 (third curve, top to bottom), the presence of M star features is easily seen from the growing bump at 8200Å due to the Na I doublet. The vertical scale in the figure includes an offset of 0.5 between each plot and thus it is arbitrary. Table 2: List of M star templates used for estimating maximum contribution of an M star to the GD 552 profile. Spectral Type | Star
---|---
M0.5 | Gl 720A
M1 | Gl 229
M1.5 | Gl 205
M2.5 | Gl 250B
M3 | Gl 752A
M3.5 | Gl 273
M4 | Gl 213
M4.5 | Gl 83.1
M5.5 | Gl 65A
M6 | Gl 406
M6.5 | G 51-15
M7 | VB 8
M8 | VB 10
M9 | LHS 2065
### 3.4 System parameters
In section 3.3 we placed an upper limit of 10% upon the contribution of an M
star to the spectrum of GD 552 in the range 8100 to 8400Å. We now use this
restriction to estimate the mass of this hypothetical star. This will allow us
to compare directly to HH1990’s $M_{2}$ value, and thus to test their model.
We first need to convert our constraint into one upon the $I$-band magnitude
of the M star. We start from the relation between magnitudes and fluxes,
including a colour correction to make our data match the profile of the I
band:
$m_{I,10\%MS}=m_{I,GD~{}552}-2.5\log\left(\frac{\int^{\lambda_{f}}_{\lambda_{i}}\epsilon_{I}\lambda
f_{\lambda,10\%MS}d\lambda}{\int^{\lambda_{f}}_{\lambda_{i}}\epsilon_{I}\lambda
f_{\lambda,GD~{}552}d\lambda}\right)$ (2)
In this equation, $10\%MS$ stands for 10 percent of the template M star;
$\epsilon_{I}$ is the transmission for the I Band; $\lambda_{i}$ and
$\lambda_{f}$ are the limits of the observed range in wavelength.
$f_{\lambda,10\%MS}$ is 10% of the normalised flux density of the M star,
obtained by dividing by a constant fit to the 8100-8400 Å region (the Na I
doublet at 8200 Å excluded) and multiplied by 0.10. Similarly,
$f_{\lambda,GD~{}552}$ is the flux density of GD 552 calculated by dividing by
a constant fit to its continuum. A field star in front of GD 552 made
difficult to do this in previous years. By September 2006, however, the
interloper had moved enough to make this calculation straightforward (see
Figure 8). We measured $m_{I,GD552}=16.3$ from our data. Then we use equation
2 with the M star templates listed in Table 2, obtaining $m_{I,10\%MS}$ as a
(mild) function of the assumed spectral type (hereafter this will be referred
to simply as $m_{I}$). In Figure 9 we compare these values after conversion to
absolute magnitude $M_{I}$ for distances of $70$ and $125\,$pc with the
absolute magnitudes of young M dwarfs of the same spectral type from Leggett
(1992). In doing this we are assuming solar-like metallicities.
Figure 8: GD 552 (marked with an A) and a field star (marked with a B) in
September 2006 (image scale is $\sim 1^{\prime}$ by $1^{\prime}$). The field
star was in front of the system in previous years and made impossible to
estimate GD 552’s magnitude with any certainty. The high proper motion of GD
552 results in it and the field star now being resolved. Using our data we
measured $I=16.3$ for GD 552 and $I=17.6$ for the field star. Figure 9:
Comparison of the upper limit on $M_{I}$ derived from our I-band spectra for
the mass donor in GD 552 (solid line, open squares plus arrows, to indicate
the upper-limit nature of the point) and $M_{I}$ for young M dwarfs of the
same spectral type (dashed line, open circles, Leggett 1992). The vertical
line at M6.5 marks the limit set by Kirkpatrick & McCarthy (1994) for a
hydrogen burning star. The pannels assume different distances to the system.
The upper one uses the old Greenstein & Giclas (1978) value while the lower
panel uses the distance estimated in this paper.
The vertical line in Figure 9 marks the spectral type for a hydrogen burning
star with the lowest possible mass ($\sim 0.08\,{\rm M}_{\odot}$, Kirkpatrick
& McCarthy 1994). The difference between the top and bottom panels is the
distance assumed in each case. In the top panel we used the value $d\sim
70\,{\rm pc}$ from Greenstein & Giclas (1978); it is clear from it that the
absolute magnitudes allowed by our data are inconsistent with the assumption
that the mass donor in GD 552 is a main-sequence star. This implies that the
mass of the companion star $M_{2}<0.08M_{\sun}$, ruling out near main-sequence
models such as that of HH1990. Our conclusion remains unaltered even if GD 552
is at the maximum distance of 125 pc calculated for a $0.33{\rm M}_{\odot}$
white dwarf (Section 3.2) which is lower than any white dwarf mass measured in
any CV.
## 4 Discussion
Our model of GD 552 is radically different from that of HH1990. Where they
have a very high mass white dwarf in an almost face-on system, with an M dwarf
donor of mass $\sim 0.15\,{\rm M}_{\odot}$, our failure to detect the mass
donor requires that it is a brown dwarf with $M_{2}<0.08\,{\rm M}_{\odot}$.
This makes GD 552 an excellent candidate for a post-period-bounce system with
an age of about 7 Gyr (Politano et al. 1998). Our explanation of the low
$K_{1}$ is therefore simply a case of extreme mass ratio, and we do not
require a particularly low orbital inclination. If the secondary is low mass,
there is no need for the system to be face on to get
$K_{1}=17.4\,\mathrm{km}\,{\mathrm{s}^{-1}}$, and therefore neither is there
any need for the white dwarf to have a high mass to produce the double peaked
profiles. For example, the observed value of $K_{1}$ can be matched with a
white dwarf of mass $M_{1}=0.6{\rm M}_{\odot}$ in orbit with a companion of
mass $M_{2}=0.03{\rm M}_{\odot}$ with a moderate inclination of $i\sim
60^{\circ}$.
Patterson et al. (2005) suggested that GD 552 is a likely “period bouncer”
(post period minimum CV) because of its lack of outbursts and small $K_{1}$.
We have now given support to their suggestion. However, HH1990’s original
model _could_ have been correct and the search for the donor was a crucial
step in ruling out HH1990’s hypothesis. Mass donors are easily visible in
other systems with hotter white dwarfs and with orbital periods similar to GD
552 (e.g. HT Cas, $P_{orb}=106.1min$, M5-M6V secondary spectral type, Marsh
1990; Z Cha, $P_{orb}=107.3min$, M5.5V secondary spectral type, Wade & Horne
1988; compared to $P_{orb}=102.73min$ for GD 552). Our search for the mass
donor was thus a realistic and critical test for the suitability of HH1990’s
model.
There is a key point about our analysis that is worth emphasizing: we have
managed to derive particularly strong constraints because of GD 552’s
relatively long orbital period which maximises the difference between the pre-
and post-period-bounce systems. In particular it means that the mass donor in
the pre-bounce model is relatively easy to detect, so that failure to detect
it is a clear indication of the post-bounce alternative. We believe that this
makes GD 552 one of the most secure post-period-minimum CV known (RE J1255+266
may be even better if its relatively long period can be confirmed, see
Patterson et al. 2005). The method we employ is indirect, but necessarily so,
since if the donor is a brown dwarf it is likely to contribute far less light
than the upper limit we have derived.
It can be argued that there is also the possibility that GD552 is not a post-
bounce system at all, even if its components are a WD with a brown dwarf
donor. If GD552’s progenitor was a binary composed of a main sequence star and
a brown dwarf, instead of a double main sequence binary, it would have evolved
into the CV we see today without the need of going through the minimum period.
We would not be able to distinguish between these two systems at all. Politano
(2004) found that CVs that evolve directly to a short orbital period are
expected to be rare because the progenitor systems containing a solar-type
star and a brown dwarf are known to be rare (“the brown-dwarf desert”).
## 5 Conclusions
We have used I-band spectroscopy in an attempt to detect the mass donor of the
cataclysmic variable GD 552. Failure to detect the donor star puts strong
limits on its mass, ruling out a main sequence nature.
We have shown GD 552 data to be consistent with a model in which its
components are an ordinary white dwarf and a brown dwarf, at a moderate
inclination angle. Population synthesis calculations favour the idea that the
mass donor reached the stage of brown dwarf through evolution instead of being
born as such. This suggests that GD 552 is likely to be a post-period minimum
CV, making it the “period bouncer” with the longest securely-determined
orbital period known.
We confirm the value obtained by HH1990 for the radial velocity semiamplitude
of the white dwarf and determine for it a $T_{\rm eff}$ of $10900\pm 400$ K by
fitting the spectra with a grid of white dwarf models. The temperature
determined for the white dwarf in GD552 is consistent with the red edge of the
instability strip (11 000 $<T_{\rm eff}<$ 12 000 K; Bergeron et al. 1995).
From the 5.4 h stretch of photometry presented, we find no indication of the
presence of pulsations. The lightcurve shows significant varibility but not
with the periodicities and amplitudes characteristic of ZZ Cet stars.
## Acknowledgments
EU was supported by PPARC (UK) and Fundación Andes (Chile) under the program
Gemini PPARC-Andes throughout most of this research. EU and RH-G were also
supported by Universidad Católica del Norte’s Proyecto de Investigación en
Docencia 220401-10602015. TRM was supported under a PPARC SRF during some of
the period over which this work was undertaken. LM-R is supported by NWO VIDI
grant 639.042.201 to P. J. Groot. BTG was supported by a PPARC Advanced
Fellowship.
The authors acknowledge the data analysis facilities provided by the Starlink
Project which is run by the University of Southampton on behalf of PPARC. This
research has made use of NASA’s Astrophysics Data System Bibliographic
Services. We acknowledge with thanks the variable star observations from the
AAVSO International Database contributed by observers worldwide and used in
this research.
The INT and the WHT are operated on the island of La Palma by the Isaac Newton
Group in the Spanish Observatorio del Roque de los Muchachos of the Instituto
de Astrofisica de Canarias This research was also partly based on observations
made with the NASA/ESA Hubble Space Telescope, obtained from the data archive
at the Space Telescope Science Institute. STScI is operated by the Association
of Universities for Research in Astronomy, Inc. under NASA contract NAS
5-26555; these observations are associated wirh program 9406.
## References
* [1] Araujo-Betancor, S. et al., 2005, A&A 430, 629
* [2] Bessell, M., 1999, PASP 111, 1426
* [3] Bergeron P., Wesemael F., Lamontagne R., Fontaine G., Saffer R.A., Allard N.F., 1995, ApJ 449, 258
* [4] Cannizzo, J., 1993, ApJ 419, 318
* [5] Delfosse, X. et al., 2000, A&A 364, 217
* [6] Diaconis, P., Efron, B., 1983, Sci. Am., v. 248, n. 5, 96
* [7] Friend, M., Smith, R., Martin, J., Jones, D., 1988, MNRAS 233, 451
* [8] Gänsicke, B.T., Sion, E.M., Beuermann, K., Fabian, D., Cheng, F.H., Krautter, J., 1999, A&A 347, 178
* [9] Gänsicke, B.T., Araujo-Betancor, S., H.-J. Hagen, E.T. Harlaftis, S. Kitsionas, S. Dreizler, D. Engels, 2004, A&A 418, 270
* [10] Gänsicke, B.T., 2006, MNRAS, 365, 969
* [11] Gänsicke, B.T., Beuermann, K, Thomas, H.-C., 1997, MNRAS 289, 388
* [12] Gänsicke, B.T., Sion, E.M., Beuermann, K., Fabian, D., Cheng, F.H., Krautter, J., 1999, A&A 346, 151
* [13] Giclas, H., Burnham, R., Thomas, N., 1970, Lowell Obs. Bull. 153
* [14] Greenstein, J., Giclas, H., 1978, PASP 90, 460
* [15] Hamada, T., Salpeter, E., 1961, ApJ 134, 683
* [16] Hessman, F., Hopp, U., 1990, A&A 228, 387
* [17] Horne, K., Marsh, T., 1986, MNRAS 218, 761
* [18] Howell, S., Nelson, L., Rappaport, S., 2001, ApJ 550, 897
* [19] Hubeny, I, Lanz, T, 1995, ApJ 439, 875
* [20] Kirkpatrick, J., McCarthy, D., 1994, AJ 107, 333
* [21] Kirkpatrick, J., Henry, T., McCarthy, D., 1991, ApJS 77, 417
* [22] Kolb, U., 1993, A&A 271, 149
* [23] Kolb, U., Baraffe, I., 1999, MNRAS 309, 1034
* [24] Leggett, S., 1992, ApJS, 82, 351
* [25] Lin, D.N.C., Williams, R.E., Stover, R.J., 1988, ApJ 327, 234
* [26] Littlefair, S., Dhillon, V., Martin, E. , 2003, MNRAS 340, 264
* [27] Littlefair, S., Dhillon, V., Marsh, T., Gänsicke, B., Southworth, J., Watson, C., 2006, Science, 314, 1578
* [28] Lomb, N., 1976, Ap&SS 39, 447
* [29] Marsh, T.R., 1987, MNRAS 229, 779
* [30] Marsh, T.R., 1988, MNRAS 231, 1117
* [31] Marsh, T.R., 1989, PASP 101, 1032
* [32] Marsh, T.R., 1990, ApJ 357, 621
* [33] Marsh, T.R., Horne, K., 1988, MNRAS 235, 269
* [34] Mennickent, R. E., Diaz, M., Skidmore, W., Sterken, C., 2001, A&A 376, 448
* [35] Oke, J., 1990, AJ 99, 1621
* [36] Patterson, J., Thorstensen, J., Kemp, J., 2005, PASP 117, 427
* [37] Politano, M., 2004, ApJ 604, 817
* [38] Politano, M., Howell, S.B., Rappapport, S., 1998, in ”Wild stars in the Old West.”, Howell S., Kuulers E., Woodwards, C. (eds.), ASP Conference Series 137, p. 207
* [39] Putte, D., Smith, R., Hawkins, N., Martin, J., 2003, MNRAS 342, 151
* [40] Rodriguez-Gil, P., Gänsicke, B.T., Hagen, H.-J., Marsh, T.R., Harlaftis, E.T., Kitsionas, S., Engels, D., 2005, A&A 431, 269
* [41] Scargle, J., 1982, ApJ 263, 835
* [42] Schneider, D., Young, P., 1980, ApJ 238, 946
* [43] Shafter, A., Szkody, P., Thorstensen, J., 1986, MNRAS 308, 765
* [44] Sion, E., Szkody, P., 2005, in ”White Dwarfs: Cosmological and Galactic Probes”, Sion, E., Vennes, S., Shipman, H. (eds.), Astrophysics and Space Science Library 332, p. 217 ff.
* [45] Smak, J., 1981, Acta Astron. 31, 395
* [46] Smith, M., Dhillon, V., Marsh, T., 1998, MNRAS 296, 465
* [47] Southworth, J., Gänsicke, B., Marsh, T., de Martino, D., Hakala, P., Littlefair, S., Rodríguez-Gil, P., Szkody, P., 2006, MNRAS 373, 687
* [48] Szkody, P. et al., 2007, AJ 134, 185
* [49] van Zyl, L. et al., 2004, MNRAS 350, 307
* [50] Wade, B., Horne, K., 1988, ApJ 324, 411
* [51] Warner, B., Woudt, P.A., 2004, MNRAS 348, 599
* [52] Warner, B., Woudt, P.A., 2005, PASP 334, 453
* [53] Willems, B., Kolb, U., Sandquist, E., Taam, R., Dubus, G., 2005, ApJ 635, 1263
* [54] Williams, G., Shipman, H., 1988, ApJ 326, 738
* [55] Williams, R.E., 1980, ApJ 235, 939
* [56] Wilson, R., 1953, Carnegie Inst. Washington D.C. Publ. 601
| arxiv-papers | 2008-06-30T13:38:42 | 2024-09-04T02:48:56.451530 | {
"license": "Public Domain",
"authors": "E. Unda-Sanzana, T. R. Marsh, B. T. Gansicke, P. F. L. Maxted, L.\n Morales-Rueda, V. S. Dhillon, T. D. Thoroughgood, E. Tremou, C. A. Watson, R.\n Hinojosa-Goni",
"submitter": "Eduardo Unda-Sanzana",
"url": "https://arxiv.org/abs/0806.4892"
} |
0806.4929 | Quadratic-Argument Approach to Nonlinear
Schrödinger Equation and Coupled Ones1112000 Mathematical Subject
Classification. Primary 35C05, 35Q55; Secondary 37K10.
Xiaoping Xu
Institute of Mathematics, Academy of Mathematics & System Sciences
Chinese Academy of Sciences, Beijing 100190, P.R. China 222Research supported
by China NSF 10431040
Abstract
The two-dimensional cubic nonlinear Schrödinger equation is used to describe
the propagation of an intense laser beam through a medium with Kerr
nonlinearity. The coupled two-dimensional cubic nonlinear Schrödinger
equations are used to describe interaction of electromagnetic waves with
different polarizations in nonlinear optics. In this paper, we solve the above
equations by imposing a quadratic condition on the related argument functions
and using their symmetry transformations. More complete families of exact
solutions of such type are obtained. Many known interesting solutions, such
soliton ones, turn out to be special cases of our solutions.
## 1 Introduction
The two-dimensional cubic nonlinear Schrödinger equation:
$i\psi_{t}+c(\psi_{xx}+\psi_{yy})+a|\psi|^{2}\psi=0$ $None$
is used to describe the propagation of an intense laser beam through a medium
with Kerr nonlinearity, where $t$ is the distance in the direction of
propagation, $x$ and $y$ are the transverse spacial coordinates, $\psi$ is a
complex valued function in $t,x,y$ standing for electric field amplitude, and
$a,c$ are nonzero real constants. We refer the introduction of [SEG] for more
systematic exposition of the equation. Akhnediev, Eleonskii and Kulagin [AEK]
found certain exact solutions of (1.1) whose real and imaginary parts are
linearly dependent over the functions of $t$. Moreover, Gagnon and Winternitz
[GW] found exact solutions of the cubic and quintic nonlinear Schrödinger
equation for a cylindrical geometry. Mihalache and Panoin [MN] used the method
in [AEK] to obtain new solutions which describe the propagation of dark
envelope soliton light pulses in optical fibers in the normal group velocity
dispersion regime. Furthermore, Saied, EI-Rahman and Ghonamy [SEG] used
various similarity variables to reduce the above equation to certain ordinary
differential equations and obtain some exact solutions. However, many of their
solutions are equivalent to each other under the action of the known symmetry
transformations of the above equation.
The coupled two-dimensional cubic nonlinear Schrödinger equations
$i\psi_{t}+c_{1}(\psi_{xx}+\psi_{yy})+(a_{1}|\psi|^{2}+b_{1}|\varphi|^{2})\psi=0,$
$None$
$i\varphi_{t}+c_{2}(\varphi_{xx}+\varphi_{yy})+(a_{2}|\psi|^{2}+b_{2}|\varphi|^{2})\varphi=0$
$None$
are used to describe interaction of electromagnetic waves with different
polarizations in nonlinear optics, where $a_{1},a_{2},b_{1},b_{2},c_{1}$ and
$c_{2}$ are real constants. Radhakrishnan and Lakshmanan [RL1] used Painlevé
analysis to find a Hirota bilinearization of the above system of partial
differential equations and obtained bright and dark multiple soliton soutions.
They also generalized their results to the coupled nonlinear Schrödinger
equations with higher-order effects in [RL2]. Grébert and Guillot [GG]
construcetd periodic solutions of coupled one-dimensional nonlinear
Schrödinger equations with periodic boundary conditions in some resonance
situations. Moreover, Hioe and Salter [HS] found a connections between Lamé
functions and solutions of the above coupled equations.
In terms of real-valued functions, the above equations form systems of
nonlinear partial differential equations. Such systems can not be solved
exactly without pre-assumptions. We observe that the argument functions of
many known solutions for these equations are quadratic in the spacial
variables $x$ and $y$, in particular, those in [SEG]. Moreover, some of these
solutions are actually equivalent to each other under the Lie point symmetries
of their corresponding equations. These facts motivate us to solve the above
equations in this paper by imposing the quadratic condition on the related
argument functions and using their symmetry transformations. More complete
families of explicit exact solutions of this type with multiple parameter
functions are obtained. Many known interesting solutions, such soliton ones,
turn out to be special cases of our solutions. Various singular solutions and
periodic solutions that we obtain may reflect some important physical
phenomena in practical models. Our solutions can also be used to solve some
boundary-value problems. Below we give more details.
For convenience, we always assume that all the involved partial derivatives of
related functions always exist and we can change orders of taking partial
derivatives. We also use prime ′ to denote the derivative of any one-variable
function.
It is known that the equation (1.1) is invariant under the following known
symmetric transformations:
$T_{1}(\psi)=de^{d_{3}i}\psi(d^{2}t+d_{2},d(x\cos d_{1}+y\sin d_{1}),d(-x\sin
d_{1}+y\cos d_{1})),$ $None$
$T_{2}(\psi)=e^{[2(d_{1}x+d_{3}y)-(d_{1}^{2}+d_{3}^{2})t]i/4c}\psi(t,x-d_{1}t+d_{2},y-d_{3}t+d_{4}),$
$None$
where $d,d_{1},d_{2},d_{3},d_{4}\in\mathbb{R}$ with $d\neq 0$. In other words,
the above transformations transform one solution of (1.1) into another
solution. Our solutions contain all the solutions in [SEG] up to the above
transformations. In particular, our solutions with elliptic functions were not
given in [SEG]. Our approach is quite elementary and accessible to large
audiences such as physicists and engineers. For the reader’s convenience, we
list in this paper all the solutions of the equation (1.1) found by our method
although some of them are known and obvious. This may help non-mathematicians
to apply the solutions of the Schrödinger equation to their fields. In fact,
applying the transformations in (1.4) and (1.5) to any of our solutions will
yield more sophisticated one.
Similarly, we have the following known symmetric transformations of the
coupled equations (1.2) and (1.3):
$T_{1}(\psi)=de^{d_{3}i}\psi(d^{2}t+d_{2},d(x\cos d_{1}+y\sin d_{1}),d(-x\sin
d_{1}+y\cos d_{1})),$ $None$
$T_{1}(\varphi)=de^{d_{4}i}\varphi(d^{2}t+d_{2},d(x\cos d_{1}+y\sin
d_{1}),d(-x\sin d_{1}+y\cos d_{1}));$ $None$
$T_{2}(\psi)=e^{[2(d_{1}x+d_{3}y)-(d_{1}^{2}+d_{3}^{2})t]i/4c_{1}}\psi(t,x-d_{1}t+d_{2},y-d_{3}t+d_{4}),$
$None$
$T_{2}(\varphi)=e^{[2(d_{1}x+d_{3}y)+(d_{1}^{2}+d_{3}^{2})t]i/4c_{2}}\varphi(t,x-d_{1}t+d_{2},y-d_{3}t+d_{4});$
$None$
where $d,d_{1},d_{2},d_{3},d_{4}\in\mathbb{R}$ with $d\neq 0$. In addition to
the above symmetries, we also solve the coupled equations modulo the following
symmetry:
$(\psi,a_{1},b_{1},c_{1})\leftrightarrow(\varphi,a_{2},b_{2},c_{2}).$ $None$
Again for the reader’s convenience, we list in this paper all the solutions of
the coupled equations (1.2) and (1.3) found by our method although some of
them are known and obvious. For convenience, we always assume that all the
involved partial derivatives of related functions always exist and we can
change orders of taking partial derivatives. We also use prime ′ to denote the
derivative of any one-variable function.
In Section 2, we solve the Schrödinger equation (1.1). In Section 3, we use
the results in Section 2 to solve the coupled Schrödinger equations (1.2) and
(1.3).
## 2 Exact Solutions of the Schrödinger Equation
In this section, we will present our quadratic-argument approach to the two-
dimensional cubic nonlinear Schrödinger equation (1.1) and find more exact
solutions than [SEG] in the modulo sense.
Write
$\psi=\xi(t,x,y)e^{i\phi(t,x,y)},$ $None$
where $\xi$ and $\phi$ are real functions in $t,x,y$. Note
$\psi_{t}=(\xi_{t}+i\xi\phi_{t})e^{i\phi},\qquad\psi_{x}=(\xi_{x}+i\xi\phi_{x})e^{i\phi},\qquad\psi_{y}=(\xi_{y}+i\xi\phi_{y})e^{i\phi},$
$None$
$\psi_{xx}=(\xi_{xx}-\xi\phi_{x}^{2}+i(2\xi_{x}\phi_{x}+\xi\phi_{xx}))e^{i\phi},\;\;\psi_{yy}=(\xi_{yy}-\xi\phi_{y}^{2}+i(2\xi_{y}\phi_{y}+\xi\phi_{yy}))e^{i\phi}.$
$None$
So the equation (1.1) becomes
$\displaystyle
i\xi_{t}-\phi_{t}\xi+a\xi^{3}+c[\xi_{xx}+\xi_{yy}-\xi(\phi_{x}^{2}+\phi_{y}^{2})$
$\displaystyle+i(2\xi_{x}\phi_{x}+2\xi_{y}\phi_{y}+\xi(\phi_{xx}+\phi_{yy}))]=0,\hskip
153.6447pt(2.4)$
equivalently,
$\xi_{t}+c(2\xi_{x}\phi_{x}+2\xi_{y}\phi_{y}+\xi(\phi_{xx}+\phi_{yy}))=0,$
$None$
$-\xi[\phi_{t}+c(\phi_{x}^{2}+\phi_{y}^{2})]+c(\xi_{xx}+\xi_{yy})+a\xi^{3}=0.$
$None$
Note that it is very difficult to solve the above system without pre-
assumptions. We observe that the functions $\phi$ in all the solutions of
[SEG] are quadratic in $x$ and $y$. From the algebraic characteristics of the
above system of partial differential equations, it is most affective to assume
that $\phi$ is quadratic in $x$ and $y$. After sorting case by case, we only
have the following four cases that lead us to exact solutions, modulo the
transformations in (1.4) and (1.5).
Case 1. $\phi=\beta(t)$ is a function of $t$.
According to (2.5), $\xi_{t}=0$. Moreover, (2.6) becomes
$-\beta^{\prime}\xi+c(\xi_{xx}+\xi_{yy})+a\xi^{3}=0.$ $None$
So we take
$\beta=bt+d,\qquad b,d\in\mathbb{R}.$ $None$
If $b=0$ and $ac<0$, modulo the transformations (1.4), we take $d=0$ and the
following solutions:
$\xi=\frac{1}{x}\sqrt{-\frac{2c}{a}}\qquad\mbox{or}\qquad\sqrt{-\frac{c}{a(x^{2}+y^{2})}}.$
$None$
Next we assume $b\neq 0$. Modulo the transformation (1.4), we can take $d=0$.
Note that
${(\tan s)^{\prime}}^{\prime}=2(\tan^{3}s+\tan s),\qquad{(\sec
s)^{\prime}}^{\prime}=2\sec^{3}s-\sec s,$ $None$ ${(\coth
s)^{\prime}}^{\prime}=2(\coth^{3}s-\coth
s),\qquad{(\mbox{csch}\>s)^{\prime}}^{\prime}=2\mbox{csch}\>^{3}s+\mbox{csch}\>s.$
$None$
Denote Jacobi elliptic functions
$\mbox{sn}\>s=\mbox{sn}\>(s|m),\qquad\mbox{cn}\>s=\mbox{cn}\>(s|m),\qquad\mbox{dn}\>s=\mbox{dn}\>(s|m),$
$None$
where $m$ is the elliptic modulus (e.g., cf. [WG]). Then
${(\mbox{sn}\>s)^{\prime}}^{\prime}=2m^{2}\mbox{sn}\>^{3}s-(1+m^{2})\mbox{sn}\>s,$
$None$
${(\mbox{cn}\>s)^{\prime}}^{\prime}=-2m^{2}\mbox{cn}\>^{2}s+(2m^{2}-1)\mbox{cn}\>s,$
$None$
${(\mbox{dn}\>s)^{\prime}}^{\prime}=-2\mbox{dn}\>^{3}s+(2-m^{2})\mbox{dn}\>s.$
$None$
Moreover,
$\lim_{m\rightarrow 1}\mbox{sn}\>s=\tanh s,\qquad\lim_{m\rightarrow
1}\mbox{cn}\>s=\lim_{m\rightarrow 1}\mbox{dn}\>s=\mbox{sech}\>s.$ $None$
Consider solutions modulo the transformation (1.4). If $ac<0$, we have the
following solutions:
$\xi=\sqrt{-\frac{2c}{a}}\>\tan x,\qquad b=2c;$ $None$
$\xi=\sqrt{-\frac{2c}{a}}\>\sec x,\qquad b=-c;$ $None$
$\xi=\sqrt{-\frac{2c}{a}}\>\coth x,\qquad b=-2c;$ $None$
$\xi=\sqrt{-\frac{2c}{a}}\>\mbox{csch}\>x,\qquad b=c;$ $None$
$\xi=m\sqrt{-\frac{2c}{a}}\>\mbox{sn}\>x,\qquad b=-(1+m^{2})c.$ $None$
When $ac>0$, we get the following solutions:
$\xi=m\sqrt{\frac{2c}{a}}\>\mbox{cn}\>x,\qquad b=(2m^{2}-1)c,$ $None$
$\xi=\sqrt{\frac{2c}{a}}\>\mbox{dn}\>x,\qquad b=(2-m^{2})c.$ $None$
Theorem 2.1.Let $m\in\mathbb{R}$. The following function are solutions $\psi$
of the two-dimensional cubic nonlinear cubic nonlinear Schrödinger equation
(1.1): if $ac<0$,
$\frac{1}{x}\sqrt{-\frac{2c}{a}},\qquad\sqrt{-\frac{c}{a(x^{2}+y^{2})}},\qquad
e^{2cti}\sqrt{-\frac{2c}{a}}\>\tan x,\qquad e^{-cti}\sqrt{-\frac{2c}{a}}\sec
x,$ $None$ $e^{-2cti}\sqrt{-\frac{2c}{a}}\>\coth x,\qquad
e^{cti}\sqrt{-\frac{2c}{a}}\>\mbox{csch}\>x,\qquad
me^{-(1+m^{2})cti}\sqrt{-\frac{2c}{a}}\mbox{sn}\>x;$ $None$
when $ac>0$,
$me^{(2m^{2}-1)cti}\sqrt{\frac{2c}{a}}\>\mbox{cn}\>x,\qquad
e^{(2-m^{2})cti}\sqrt{\frac{2c}{a}}\>\mbox{dn}\>x.$ $None$
Remark 2.2. Although the above solution are simple, we can obtain more
sophisticated ones by applying the transformations (1.4) and (1.5) to them.
For instance, applying the transformation (1.4) to the first solution in
(2.24), we get a solution:
$\psi=\frac{e^{d_{2}i}}{x\cos d_{1}+y\sin d_{1}}\sqrt{-\frac{2c}{a}},\qquad
d_{1},d_{2}\in\mathbb{R}.$ $None$
Applying the transformation (1.5) to the above solution, we obtain another
solution:
$\psi=\frac{e^{[2(d_{3}x+d_{4}y)+(d_{3}^{2}+d_{4}^{2})t+d_{2}]i/4c}}{(x-d_{3}t)\cos
d_{1}+(y-d_{4}t)\sin d_{1}+d_{5}}\sqrt{-\frac{2c}{a}},\qquad
d_{1},d_{2},d_{3},d_{4},d_{5}\in\mathbb{R}.$ $None$
Case 2. $\phi=x^{2}/4ct+\beta$ for some function $\beta$ of $t$.
In this case, (2.5) becomes
$\xi_{t}+\frac{x}{t}\xi_{x}+\frac{1}{2t}\xi=0.$ $None$
Thus
$\xi=\frac{1}{\sqrt{t}}\zeta(u,y),\qquad u=\frac{x}{t},$ $None$
for some two-variable function $\zeta$. Now (2.6) becomes (2.7). Note
$\xi_{xx}=t^{-5t/2}\zeta_{uu},\qquad\xi_{yy}=t^{-1/2}\zeta_{yy},\qquad\xi^{3}=t^{-3/2}\zeta^{3}.$
$None$
So (2.7) becomes
$-\frac{\beta^{\prime}}{\sqrt{t}}\zeta+c(t^{-5t/2}\zeta_{uu}+t^{-1/2}\zeta_{yy})+at^{-3/2}\zeta^{3}=0,$
$None$
whose coefficients of $t^{-3/2}$ force us to take
$\xi=\frac{b}{\sqrt{t}},\qquad b\in\mathbb{R}.$ $None$
Now (2.7) becomes
$-\beta^{\prime}+\frac{ab^{2}}{t}=0\Longrightarrow\beta=ab^{2}\ln t$ $None$
modulo the transformation in (1.5).
Case 3. $\phi=x^{2}/4ct+y^{2}/4c(t-d)+\beta$ for some function $\beta$ of $t$
with $0\neq d\in\mathbb{R}$.
In this case, (2.5) becomes
$\xi_{t}+\frac{x}{t}\xi_{x}+\frac{y}{t-d}\xi_{y}+\left(\frac{1}{2t}+\frac{1}{2(t-d)}\right)\xi=0.$
$None$
So we have:
$\xi=\frac{1}{\sqrt{t(t-d)}}\zeta(u,v),\qquad
u=\frac{x}{t},\;v=\frac{y}{t-d},$ $None$
for some two-variable function $\zeta$. Again (2.6) becomes (2.7). Note
$\xi_{xx}=t^{-5/2}(t-c)^{-1/2}\zeta_{uu},\qquad\xi_{yy}=t^{-1/2}(t-c)^{-5/2}\zeta_{vv},\qquad\xi^{3}=t^{-3/2}(t-c)^{-3/2}\zeta^{3}.$
$None$
So (2.7) becomes
$-\frac{\beta^{\prime}}{\sqrt{t(t-d)}}\zeta+c(t^{-5/2}(t-c)^{-1/2}\zeta_{uu}+t^{-1/2}(t-c)^{-5/2}\zeta_{vv})+at^{-3/2}(t-c)^{-3/2}\zeta^{3}=0,$
$None$
whose coefficients of $t^{-3/2}(t-c)^{-3/2}$ force us to take
$\xi=\frac{b}{\sqrt{t(t-d)}},\qquad b\in\mathbb{R}.$ $None$
Now (2.7) becomes
$-\beta^{\prime}+\frac{ab^{2}}{t(t-d)}=0\Longrightarrow\beta=\frac{ab^{2}}{d}\ln\frac{t-d}{t}$
$None$
modulo the transformation in (1.5).
Theorem 2.3. Let $b,d\in\mathbb{R}$ with $d\neq 0$. The following function are
solutions $\psi$ of the two-dimensional cubic nonlinear cubic nonlinear
Schrödinger equation:
$bt^{ab^{2}i-1/2}e^{x^{2}i/4ct},\qquad
bt^{-ab^{2}i/d-1/2}(t-d)^{ab^{2}i/d-1/2}e^{x^{2}i/4ct+y^{2}i/4c(t-d)}.$ $None$
Remark 2.4. Applying (1.4) to the above first solution, we get another
solution
$\psi=dd_{1}(d_{1}^{2}t+d_{4})^{ad^{2}i-1/2}\exp\left(\frac{d_{1}^{2}(x\cos
d_{2}+y\sin d_{2})^{2}}{4c(d_{1}^{2}t+d_{4})}+d_{3}\right)i,$ $None$
for $d_{1},d_{2},d_{3},d_{4}\in\mathbb{R}$. Moreover, we obtain a more
sophisticated solution:
$\displaystyle\hskip 56.9055pt\psi$ $\displaystyle=$ $\displaystyle
dd_{1}(d_{1}^{2}t+d_{4})^{ad^{2}i-1/2}\exp\frac{d_{1}^{2}((x-d_{5}t)\cos
d_{2}+(y-d_{6}t)\sin d_{2}+d_{7})^{2}i}{4c(d_{1}^{2}t+d_{4})}$
$\displaystyle\times\exp\left(\frac{2(d_{5}x+d_{6}y)+(d_{5}^{2}+d_{6}^{2})t}{4c}+d_{3}\right)i\hskip
119.50148pt(2.43)$
by applying the transformation (1.5) to (2.42), where $b_{r}\in\mathbb{R}$.
Case 4. $\phi=(x^{2}+y^{2})/4ct+\beta$ for some function $\beta$ of $t$.
Under our assumption, (2.5) becomes
$\xi_{t}+\frac{x}{t}\xi_{x}+\frac{y}{t}\xi_{y}+\frac{1}{t}\xi=0.$ $None$
Thus we have:
$\xi=\frac{1}{t}\zeta(u,v),\qquad u=\frac{x}{t},\;v=\frac{y}{t},$ $None$
for some two-variable function $\zeta$. Moreover, (2.6) becomes
$-\beta^{\prime}\zeta+\frac{c}{t^{2}}(\zeta_{uu}+\zeta_{vv})+\frac{a}{t^{2}}\zeta^{3}=0.$
$None$
An obvious solution is
$\zeta=d,\qquad\beta=-\frac{ad^{2}}{t},\qquad d\in\mathbb{R}.$ $None$
If $ac<0$, we have the simple following solutions with $\beta=0$:
$\zeta=\frac{1}{\ell_{1}u+\ell_{2}v+\ell_{3}}\sqrt{-\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{a}}\qquad\mbox{or}\qquad\sqrt{-\frac{c}{a((u-\ell_{1})^{2}+(v-\ell_{2})^{2})}}$
$None$
for $\ell_{1},\ell_{2},\ell_{3}\in\mathbb{R}$.
Next we assume
$\beta^{\prime}=\frac{b}{t^{2}}\Longrightarrow\beta=-\frac{b}{t}$ $None$
modulo the transformation in (1.2), where $b$ is a real constant to be
determined. Suppose
$\zeta=\Im(\varpi),\qquad\varpi=\ell_{1}u+\ell_{2}v+\ell_{3}$ $None$
for $\ell_{1},\ell_{2},\ell_{3}\in\mathbb{R}$ such that
$(\ell_{1},\ell_{2})\neq(0,0)$. Then (2.46) is equivalent to:
$-b\Im+c(\ell_{1}^{2}+\ell_{2}^{2}){\Im^{\prime}}^{\prime}+a\Im^{3}=0.$ $None$
According to (2.10), (2.11) and (2.13)-(2.15), we have the following
solutions: If $ac<0$, we have the following solutions:
$\Im=\sqrt{-\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{a}}\>\tan\varpi,\qquad
b=2c(\ell_{1}^{2}+\ell_{2}^{2});$ $None$
$\Im=\sqrt{-\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{a}}\>\sec\varpi,\qquad
b=-c(\ell_{1}^{2}+\ell_{2}^{2});$ $None$
$\Im=\sqrt{-\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{a}}\>\coth\varpi,\qquad
b=-2c(\ell_{1}^{2}+\ell_{2}^{2});$ $None$
$\Im=\sqrt{-\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{a}}\>\mbox{csch}\>\varpi,\qquad
b=c(\ell_{1}^{2}+\ell_{2}^{2});$ $None$
$\Im=m\sqrt{-\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{a}}\>\mbox{sn}\>\varpi,\qquad
b=-(1+m^{2})c(\ell_{1}^{2}+\ell_{2}^{2}).$ $None$
When $ac>0$, we get the following solutions:
$\Im=m\sqrt{\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{a}}\>\mbox{cn}\>\varpi,\qquad
b=(2m^{2}-1)c(\ell_{1}^{2}+\ell_{2}^{2}),$ $None$
$\Im=\sqrt{\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{a}}\>\mbox{dn}\>\varpi,\qquad
b=(2-m^{2})c(\ell_{1}^{2}+\ell_{2}^{2}).$ $None$
Theorem 2.5.Let $\ell_{1},\ell_{2},\ell_{3},\ell_{4},\ell_{5},m\in\mathbb{R}$
such that $(\ell_{1},\ell_{2})\neq(0,0)$. The following functions are
solutions $\psi$ of the two-dimensional cubic nonlinear cubic nonlinear
Schrödinger equation: if $ac<0$,
$\frac{1}{\ell_{1}x+\ell_{2}y+\ell_{3}t}\sqrt{-\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{a}}\;e^{(x^{2}+y^{2})i/4ct},$
$None$
$\sqrt{-\frac{c}{a((x-\ell_{4}t)^{2}+(y-\ell_{5}t)^{2})}}\;e^{(x^{2}+y^{2})i/4ct},$
$None$
$\sqrt{-\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{a}}\>\frac{1}{t}\tan\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}\;\exp\left(\frac{x^{2}+y^{2}}{4ct}-\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i,$
$None$
$\sqrt{-\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{a}}\>\frac{1}{t}\sec\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}\;\exp\left(\frac{x^{2}+y^{2}}{4ct}+\frac{c(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i,$
$None$
$\sqrt{-\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{a}}\>\frac{1}{t}\coth\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}\;\exp\left(\frac{x^{2}+y^{2}}{4ct}+\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i,$
$None$
$\sqrt{-\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{a}}\>\frac{1}{t}\mbox{csch}\>\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}\;\exp\left(\frac{x^{2}+y^{2}}{4ct}-\frac{c(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i,$
$None$
$m\sqrt{-\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{a}}\>\frac{1}{t}\mbox{sn}\>\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}\;\exp\left(\frac{x^{2}+y^{2}}{4ct}+\frac{c(1+m^{2})(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i;$
$None$
when $ac>0$,
$m\sqrt{\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{a}}\>\frac{1}{t}\mbox{cn}\>\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}\;\exp\left(\frac{x^{2}+y^{2}}{4ct}-\frac{c(2m^{2}-1)(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i,$
$None$
$\sqrt{\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{a}}\>\frac{1}{t}\mbox{dn}\>\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}\;\exp\left(\frac{x^{2}+y^{2}}{4ct}-\frac{c(2-m^{2})(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i.$
$None$
Remark 2.6. Applying the transformation (1.5) to the solution (2.59), we get
another solution:
$\psi=\frac{e^{[2(d_{1}x+d_{3}y)+(d_{1}^{2}+d_{3}^{2})t]i/4c}\sqrt{-\frac{2c(\ell_{1}^{2}+\ell_{2}^{2})}{a}}}{\ell_{1}(x-d_{1}t+d_{2})+\ell_{2}(y-d_{3}t+d_{4})+\ell_{3}t}\;e^{((x-d_{1}t+d_{2})^{2}+(y-d_{3}t+d_{4})^{2})i/4ct},$
$None$
where $d_{1},d_{2},d_{3},d_{4}\in\mathbb{R}$.
## 3 Exact Solutions of the Coupled Equations
In this section, we will use our results in previous section to find exact
solutions of the coupled two-dimensional cubic nonlinear Schrödinger equations
(1.2) and (1.3).
Write
$\psi=\xi(t,x,y)e^{i\phi(t,x,y)},\qquad\varphi=\eta(t,x,y)e^{i\mu(t,x,y)}$
$None$
where $\xi,\phi,\eta$ and $\mu$ are real functions in $t,x,y$. As the
arguments in (2.1)-(2.6), the system (1.2) and (1.3) is equivalent to the
following system for real functions:
$\xi_{t}+c_{1}(2\xi_{x}\phi_{x}+2\xi_{y}\phi_{y}+\xi(\phi_{xx}+\phi_{yy}))=0,$
$None$
$-\xi[\phi_{t}+c_{1}(\phi_{x}^{2}+\phi_{y}^{2})]+c_{1}(\xi_{xx}+\xi_{yy})+(a_{1}\xi^{2}+b_{1}\eta^{2})\xi=0,$
$None$
$\eta_{t}+c_{2}(2\eta_{x}\mu_{x}+2\eta_{y}\mu_{y}+\eta(\mu_{xx}+\mu_{yy}))=0,$
$None$
$-\eta[\mu_{t}+c_{2}(\mu_{x}^{2}+\mu_{y}^{2})]+c_{2}(\eta_{xx}+\eta_{yy})+(a_{2}\xi^{2}+b_{2}\eta^{2})\eta=0.$
$None$
Based on our experience in last section, we will solve the above system
according to the following cases. For the convenience, we always assume the
conditions on the constants involved in an expression such that it make sense.
For instance, when we use $\sqrt{d_{1}-d_{2}}$, we naturally assume $d_{1}\geq
d_{2}$.
Case 1. $(\phi,\mu)=(0,0)$ and $a_{1}b_{2}-a_{2}b_{1}\neq 0$.
In this case, $\xi_{t}=\eta_{t}=0$ by (3.2) and (3.4). Moreover, (3.3) and
(3.5) become
$c_{1}(\xi_{xx}+\xi_{yy})+(a_{1}\xi^{2}+b_{1}\eta^{2})\xi=0,\qquad
c_{2}(\eta_{xx}+\eta_{yy})+(a_{2}\xi^{2}+b_{2}\eta^{2})\eta=0.$ $None$
Assume
$\xi=\frac{\iota_{1}}{x},\qquad\eta=\frac{\iota_{2}}{x}.$ $None$
Then (3.6) is equivalent to:
$a_{1}\iota_{1}^{2}+b_{1}\iota_{2}^{2}+2c_{1}=0,\qquad
a_{2}\iota^{2}+b_{2}\iota^{2}+2c_{2}=0.$ $None$
Solving the above linear algebraic equations for $\iota_{1}^{2}$ and
$\iota_{2}^{2}$, we have:
$\iota_{1}^{2}=\frac{2(b_{1}c_{2}-b_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}},\qquad\iota_{2}^{2}=\frac{2(a_{2}c_{1}-a_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}.$
$None$
Thus we have the following solution
$\xi=\frac{\epsilon_{1}}{x}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}}},\qquad\eta=\frac{\epsilon_{2}}{x}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}}$
$None$
for $\epsilon_{1},\epsilon_{2}\in\\{1,-1\\}$. Similarly, we have the solution:
$\xi=\epsilon_{1}\sqrt{\frac{b_{1}c_{2}-b_{2}c_{1}}{(a_{1}b_{2}-a_{2}b_{1})(x^{2}+y^{2})}},\qquad\eta=\epsilon_{2}\sqrt{\frac{a_{2}c_{1}-a_{1}c_{2}}{(a_{1}b_{2}-a_{2}b_{1})(x^{2}+y^{2})}}.$
$None$
Case 2. $(\phi,\mu)=(k_{1}t,k_{2}t)$ with $k_{1},k_{2}\in\mathbb{R}$.
Again we have $\xi_{t}=\eta_{t}=0$ by (3.2) and (3.4). Moreover, (3.3) and
(3.5) become
$-k_{1}\xi+c_{1}(\xi_{xx}+\xi_{yy})+(a_{1}\xi^{2}+b_{1}\eta^{2})\xi=0,\;\;-k_{2}\eta+c_{2}(\eta_{xx}+\eta_{yy})+(a_{2}\xi^{2}+b_{2}\eta^{2})\eta=0.$
$None$
First we assume $a_{1}b_{2}-a_{2}b_{1}\neq 0$ and
$\xi=\iota_{1}\Im(x),\qquad\eta=\iota_{2}\Im(x).$ $None$
Then (3.12) becomes
$-k_{1}\Im+c_{1}{\Im^{\prime}}^{\prime}+(a_{1}\iota_{1}^{2}+b_{1}\iota_{2}^{2})\Im^{3}=0,\qquad-
k_{2}\Im+c_{2}{\Im^{\prime}}^{\prime}+(a_{2}\iota^{2}+b_{2}\iota^{2})\Im^{3}=0.$
$None$
According to (2.10) and (2.11), when $\Im=\tan x,\;\sec x,\;\coth x$ and
$\mbox{csch}\>x$, we always have
$a_{1}\iota_{1}^{2}+b_{1}\iota_{2}^{2}+2c_{1}=0,\qquad
a_{2}\iota^{2}+b_{2}\iota^{2}+2c_{2}=0.$ $None$
Thus for $\epsilon_{1},\epsilon_{2}\in\\{1,-1\\}$, we have the following
solutions:
$\xi=\epsilon_{1}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}}}\>\tan
x,\;\;\eta=\epsilon_{2}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}}\>\tan
x,\;\;(k_{1},k_{2})=2(c_{1},c_{2});$ $None$
$\xi=\epsilon_{1}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}}}\>\sec
x,\;\;\eta=\epsilon_{2}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}}\>\sec
x,\;\;(k_{1},k_{2})=-(c_{1},c_{2});$ $None$
$\xi=\epsilon_{1}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}}}\>\coth
x,\;\;\eta=\epsilon_{2}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}}\>\coth
x$ $None$
and $(k_{1},k_{2})=-2(c_{1},c_{2});$
$\xi=\epsilon_{1}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}}}\>\mbox{csch}\>x,\;\;\eta=\epsilon_{2}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}}\>\mbox{csch}\>x,\;\;(k_{1},k_{2})=(c_{1},c_{2}).$
$None$
Similarly, (2.13)-(2.15) give us the following solutions:
$\xi=m\epsilon_{1}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}}}\;\mbox{sn}\>x,\qquad\eta=m\epsilon_{2}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;\mbox{sn}\>x$
$None$
and $(k_{1},k_{2})=-(1+m^{2})(c_{1},c_{2});$
$\xi=m\epsilon_{1}\sqrt{\frac{2(b_{2}c_{1}-b_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;\mbox{cn}\>x\qquad\eta=m\epsilon_{2}\sqrt{\frac{2(a_{1}c_{2}-a_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}}}\;\mbox{cn}\>x$
$None$
and $(k_{1},k_{2})=(2m^{2}-1)(c_{1},c_{2});$
$\xi=\epsilon_{1}\sqrt{\frac{2(b_{2}c_{1}-b_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;\mbox{dn}\>x,\qquad\eta=\epsilon_{2}\sqrt{\frac{2(a_{1}c_{2}-a_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}}}\;\mbox{dn}\>x$
$None$
and $(k_{1},k_{2})=(2-m^{2})(c_{1},c_{2}).$
If $(a_{1},b_{1})=a_{1}(1,d^{2})$ and $(a_{2},b_{2})=a_{2}(1,d^{2})$ with
$d\in\mathbb{R}$, we have the following solution of (3.12):
$\xi=d\ell\sin x,\qquad\eta=\ell\cos
x,\qquad(k_{1},k_{2})=(a_{1}(d\ell)^{2}-c_{1},a_{2}(d\ell)^{2}-c_{2})$ $None$
for $\ell\in\mathbb{R}$. When $(a_{1},b_{1})=a_{1}(1,-d^{2})$ and
$(a_{2},b_{2})=a_{2}(1,-d^{2})$ with $d\in\mathbb{R}$, we get the solution:
$\xi=d\ell\cosh x,\qquad\eta=\ell\sinh
x,\qquad(k_{1},k_{2})=(a_{1}(d\ell)^{2}+c_{1},a_{2}(d\ell)^{2}+c_{2}).$ $None$
In summary, we have the following theorem.
Theorem 3.1. Let $d,\ell,m\in\mathbb{R}$ and let
$\epsilon_{1},\epsilon_{2}\in\\{1,-1\\}$. If $a_{1}b_{2}-a_{2}b_{1}\neq 0$, we
have the following solutions of the coupled two-dimensional cubic nonlinear
Schrödinger equations (1.2) and (1.3):
$\psi=\frac{\epsilon_{1}}{x}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}}},\qquad\varphi=\frac{\epsilon_{2}}{x}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}};$
$None$
$\psi=\epsilon_{1}\sqrt{\frac{b_{1}c_{2}-b_{2}c_{1}}{(a_{1}b_{2}-a_{2}b_{1})(x^{2}+y^{2})}},\qquad\varphi=\epsilon_{2}\sqrt{\frac{a_{2}c_{1}-a_{1}c_{2}}{(a_{1}b_{2}-a_{2}b_{1})(x^{2}+y^{2})}};$
$None$
$\psi=\epsilon_{1}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}}}\>e^{2c_{1}ti}\tan
x,\;\;\varphi=\epsilon_{2}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}}\>e^{2c_{2}ti}\tan
x;$ $None$
$\psi=\epsilon_{1}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}}}\>e^{-c_{1}ti}\sec
x,\;\;\varphi=\epsilon_{2}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}}\>e^{-c_{2}ti}\sec
x;$ $None$
$\psi=\epsilon_{1}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}}}\>e^{-2c_{1}ti}\coth
x,\;\;\varphi=\epsilon_{2}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}}\>e^{-2c_{2}ti}\coth
x;$ $None$
$\psi=\epsilon_{1}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}}}\>e^{c_{1}ti}\mbox{csch}\>x,\qquad\varphi=\epsilon_{2}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}}\>e^{c_{2}ti}\mbox{csch}\>x;$
$None$
$\psi=m\epsilon_{1}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}}}\;e^{-(1+m^{2})c_{1}ti}\mbox{sn}\>x,$
$None$
$\varphi=m\epsilon_{2}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;e^{-(1+m^{2})c_{2}ti}\mbox{sn}\>x;$
$None$
$\psi=m\epsilon_{1}\sqrt{\frac{2(b_{2}c_{1}-b_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;e^{(2m^{2}-1)c_{1}ti}\mbox{cn}\>x,$
$None$
$\varphi=m\epsilon_{2}\sqrt{\frac{2(a_{1}c_{2}-a_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}}}\;e^{(2m^{2}-1)c_{2}ti}\mbox{cn}\>x;$
$None$
$\psi=\epsilon_{1}\sqrt{\frac{2(b_{2}c_{1}-b_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;e^{(2-m^{2})c_{1}ti}\mbox{dn}\>x,\;\;\varphi=\epsilon_{2}\sqrt{\frac{2(a_{1}c_{2}-a_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}}}\;e^{(2-m^{2})c_{1}ti}\mbox{dn}\>x.$
$None$
If $(a_{1},b_{1})=a_{1}(1,d^{2})$ and $(a_{2},b_{2})=a_{2}(1,d^{2})$,
$\psi=d\ell e^{(a_{1}(d\ell)^{2}-c_{1})ti}\sin x,\qquad\varphi=\ell
e^{(a_{2}(d\ell)^{2}-c_{2})ti}\cos x.$ $None$
When $(a_{1},b_{1})=a_{1}(1,-d^{2})$ and $(a_{2},b_{2})=a_{2}(1,-d^{2})$,
$\psi=d\ell e^{(a_{1}(d\ell)^{2}+c_{1})ti}\cosh x,\qquad\eta=\ell
e^{(a_{2}(d\ell)^{2}+c_{2})ti}\sinh x.$ $None$
Remark 3.2. Applying the symmetric transformations (1.6)-(1.9) to the above
solutions, we can get more sophisticated ones. For instance, applying $T_{1}$
in (1.6) and (1.7) to (3.25) and (3.37), we get
$\psi=\frac{\epsilon_{1}e^{d_{2}i}}{x\cos d_{1}+y\sin
d_{1}}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})}{a_{1}b_{2}-a_{2}b_{1}}},\;\;\varphi=\frac{\epsilon_{2}e^{d_{3}i}}{x\cos
d_{1}+y\sin
d_{1}}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})}{a_{1}b_{2}-a_{2}b_{1}}};$ $None$
and
$\psi=dd_{2}\ell e^{[(a_{1}(d\ell)^{2}+c_{1})d_{2}^{2}t+d_{3}]i}\cosh
d_{2}(x\cos d_{1}+y\sin d_{1}),$ $None$ $\eta=d_{2}\ell
e^{[(a_{2}(d\ell)^{2}+c_{2})d_{2}^{2}t+d_{4}]i}\sinh d_{2}(x\cos d_{1}+y\sin
d_{1}).$ $None$
Applying $T_{2}$ in (1.8) and (1.9) to (3.26), we obtain:
$\psi=\epsilon_{1}e^{[2(d_{1}x+d_{3}y)+(d_{1}^{2}+d_{3}^{2})t]i/4c_{1}}\sqrt{\frac{b_{1}c_{2}-b_{2}c_{1}}{(a_{1}b_{2}-a_{2}b_{1})((x-d_{1}t+d_{2})^{2}+(y-d_{2}t+d_{4})^{2})}},$
$None$
$\varphi=\epsilon_{2}e^{[2(d_{1}x+d_{3}y)+(d_{1}^{2}+d_{3}^{2})t]i/4c_{2}}\sqrt{\frac{a_{2}c_{1}-a_{1}c_{2}}{(a_{1}b_{2}-a_{2}b_{1})((x-d_{1}t+d_{2})^{2}+(y-d_{2}t+d_{4})^{2})}}.$
$None$
Case 3. $\phi=x^{2}/4c_{1}t+\beta_{1}$ and
$\mu=(x-d)^{2}/4c_{2}(t-\ell)+\beta_{2}$ or
$\mu=y^{2}/4c_{2}(t-\ell)+\beta_{2}$ for some functions $\beta_{1}$ and
$\beta_{2}$ of $t$ and real constants $d$ and $\ell$.
First we assume $\mu=(x-d)^{2}/4c_{2}(t-\ell)+\beta_{2}$. Then (3.2) and (3.4)
become
$\xi_{t}+\frac{x}{t}\xi_{x}+\frac{1}{2t}\xi=0,\qquad\eta_{t}+\frac{x-d}{t-\ell}\eta_{x}+\frac{1}{2(t-\ell)}\eta=0.$
$None$
Thus
$\xi=\frac{1}{\sqrt{t}}\hat{\xi}(t^{-1}x,y),\qquad\eta=\frac{1}{\sqrt{t-\ell}}\hat{\eta}((t-\ell)^{-1}(x-d),y)$
$None$
for some two-variable functions $\hat{\xi}$ and $\hat{\eta}$. On the other
hand, (3.3) and (3.5) become:
$-\beta_{1}^{\prime}\xi+c_{1}(\xi_{xx}+\xi_{yy})+(a_{1}\xi^{2}+b_{1}\eta^{2})\xi=0,$
$None$
$-\beta_{2}^{\prime}\eta+c_{2}(\eta_{xx}+\eta_{yy})+(a_{2}\xi^{2}+b_{2}\eta^{2})\eta=0.$
$None$
As (2.31)-(2.33), the above two equations force us to take
$\xi=\frac{k_{1}}{\sqrt{t}},\qquad\eta=\frac{k_{2}}{\sqrt{t-\ell}}.$ $None$
So (3.45) and (3.46) are implied by the equations:
$\beta_{1}^{\prime}=\frac{a_{1}k_{1}^{2}}{t}+\frac{b_{1}k_{2}^{2}}{t-\ell},\qquad\beta_{2}^{\prime}=\frac{a_{2}k_{1}^{2}}{t}+\frac{b_{2}k_{2}^{2}}{t-\ell}.$
$None$
Modulo the transformation (1.6) and (1.7), we take
$\beta_{1}=a_{1}k_{1}^{2}\ln
t+b_{1}k_{2}^{2}\ln(t-\ell),\qquad\beta_{2}=a_{2}k_{1}^{2}\ln
t+b_{2}k_{2}^{2}\ln(t-\ell).$ $None$
Exact same approach holds for $\mu=y^{2}/4c_{2}(t-\ell)+\beta_{2}$.
Theorem 3.3. Let $d,\ell,k_{1},k_{2}\in\mathbb{R}$. We have the following
solutions of the coupled two-dimensional cubic nonlinear Schrödinger equations
(1.2) and (1.3):
$\psi=k_{1}t^{a_{1}k_{1}^{2}i-1/2}(t-\ell)^{b_{1}k_{2}^{2}i}e^{x^{2}i/2c_{1}t},\qquad\varphi=k_{2}t^{a_{2}k_{1}^{2}i}(t-\ell)^{b_{2}k_{2}^{2}i-1/2}e^{(x-d)^{2}i/2c_{2}(t-\ell)};$
$None$
$\psi=k_{1}t^{a_{1}k_{1}^{2}i-1/2}(t-\ell)^{b_{1}k_{2}^{2}i}e^{x^{2}i/2c_{1}t},\qquad\varphi=k_{2}t^{a_{2}k_{1}^{2}i}(t-\ell)^{b_{2}k_{2}^{2}i-1/2}e^{y^{2}i/2c_{2}(t-\ell)}.$
$None$
Case 4. $\phi=x^{2}/4c_{1}t+\beta_{1}$ and
$\mu=(x-d)^{2}/4c_{2}(t-d_{1})+y^{2}/4c_{2}(t-d_{2})+\beta_{2}$ for some
functions $\beta_{1}$ and $\beta_{2}$ of $t$ and real constants $d,d_{1}$ and
$d_{2}$.
In this case, (3.2) and (3.4) become
$\xi_{t}+\frac{x}{t}\xi_{x}+\frac{1}{2t}\xi=0,\;\;\eta_{t}+\frac{x-d}{t-d_{1}}\eta_{x}+\frac{y}{t-d_{2}}\eta_{y}+\left(\frac{1}{2(t-d_{1})}+\frac{1}{2(t-d_{2})}\right)\xi=0.$
$None$
Thus
$\xi=\frac{1}{\sqrt{t}}\hat{\xi}(t^{-1}x,y),\qquad\eta=\frac{1}{\sqrt{(t-d_{1})(t-d_{2})}}\hat{\eta}((t-d_{1})^{-1}(x-d),(t-d_{2})^{-1}y)$
$None$
for some two-variable functions $\hat{\xi}$ and $\hat{\eta}$. Again (3.3) and
(3.5) become (3.45) and (3.46), respectively. Moreover, they force us to take
$\xi=\frac{k_{1}}{\sqrt{t}},\qquad\eta=\frac{k_{2}}{\sqrt{(t-d_{1})(t-d_{2})}}.$
$None$
So (3.3) and (3.5) are implied by the equations:
$\beta_{1}^{\prime}=\frac{a_{1}k_{1}^{2}}{t}+\frac{b_{1}k_{2}^{2}}{(t-d_{1})(t-d_{2})},\qquad\beta_{2}^{\prime}=\frac{a_{2}k_{1}^{2}}{t}+\frac{b_{2}k_{2}^{2}}{(t-d_{1})(t-d_{2})}.$
$None$
Modulo the transformation (1.6) and (1.7), we get
$\beta_{1}=a_{1}k_{1}^{2}\ln
t+\frac{b_{1}k_{2}^{2}}{d_{2}-d_{1}}\ln\frac{t-d_{1}}{t-d_{2}},\qquad\beta_{2}=a_{2}k_{1}^{2}\ln
t+\frac{b_{2}k_{2}^{2}}{d_{2}-d_{1}}\ln\frac{t-d_{1}}{t-d_{2}}$ $None$
if $d_{1}\neq d_{2}$, and
$\beta_{1}=a_{1}k_{1}^{2}\ln
t-\frac{b_{1}k_{2}^{2}}{t-d_{1}},,\qquad\beta_{2}=a_{2}k_{1}^{2}\ln
t-\frac{b_{2}k_{2}^{2}}{t-d_{1}}$ $None$
when $d_{1}=d_{2}$.
Theorem 3.4. Let $d_{1},d_{2},k_{1},k_{2}\in\mathbb{R}$ such that $d_{1}\neq
d_{2}$. We have the following solutions of the coupled two-dimensional cubic
nonlinear Schrödinger equations (1.2) and (1.3):
$\psi=k_{1}t^{a_{1}k_{1}^{2}i-1/2}(t-d_{1})^{b_{1}k_{2}^{2}(\ell_{2}-d_{1})^{-1}i}(t-d_{2})^{-b_{1}k_{2}^{2}(\ell_{2}-d_{1})^{-1}i}e^{x^{2}i/4c_{1}t},$
$None$ $\displaystyle\hskip 56.9055pt\varphi$ $\displaystyle=$ $\displaystyle
k_{2}t^{a_{2}k_{1}^{2}i}(t-d_{1})^{b_{1}k_{2}^{2}(\ell_{2}-d_{1})^{-1}i-1/2}(t-d_{2})^{-b_{1}k_{2}^{2}(\ell_{2}-d_{1})^{-1}i-1/2}$
$\displaystyle\times\exp\left(\frac{(x-d)^{2}i}{4c_{2}(t-d_{1})}+\frac{y^{2}i}{4c_{2}(t-d_{1})}\right);\hskip
147.95424pt(3.59)$
$\psi=k_{1}t^{a_{1}k_{1}^{2}i-1/2}\exp\left(\frac{x^{2}i}{4c_{1}t}-\frac{b_{1}k_{2}^{2}i}{t-d_{1}}\right),$
$None$
$\varphi=\frac{k_{2}t^{a_{2}k_{1}^{2}i}}{t-d_{1}}\exp\frac{((x-d)^{2}+y^{2}-4c_{2}b_{2}k_{2}^{2})i}{4c_{2}(t-d_{1})}.$
$None$
Case 5. For $\ell_{1},\ell_{2},\ell,d_{1},d_{2}\in\mathbb{R}$ and functions
$\beta_{1},\beta_{2}$ of $t$,
$\phi=\frac{x^{2}}{4c_{1}t}+\frac{y^{2}}{4c_{1}(t-\ell)}+\beta_{1},\qquad\mu=\frac{(x-d_{1})^{2}}{4c_{2}(t-\ell_{1})}+\frac{(y-d_{2})^{2}}{4c_{1}(t-\ell_{2})}+\beta_{2}.$
$None$
As in the above case, we get
$\xi=\frac{k_{1}}{\sqrt{t(t-\ell)}},\qquad\eta=\frac{k_{2}}{\sqrt{(t-\ell_{1})(t-\ell_{2})}}.$
$None$
So (3.3) and (3.5) are implied by the equations:
$\beta_{1}^{\prime}=\frac{a_{1}k_{1}^{2}}{t(t-\ell)}+\frac{b_{1}k_{2}^{2}}{(t-\ell_{1})(t-\ell_{2})},\qquad\beta_{2}^{\prime}=\frac{a_{2}k_{1}^{2}}{t}+\frac{b_{2}k_{2}^{2}}{(t-\ell_{1})(t-\ell_{2})}.$
$None$
Modulo the transformation (1.6) and (1.7), we have
$\beta_{1}=\frac{a_{1}k_{1}^{2}}{\ell}\ln\frac{t-\ell}{t}+\frac{b_{1}k_{2}^{2}}{\ell_{2}-\ell_{1}}\ln\frac{t-\ell_{1}}{t-\ell_{2}},\qquad\beta_{2}=\frac{a_{2}k_{1}^{2}}{\ell}\ln\frac{t-\ell}{t}+\frac{b_{2}k_{2}^{2}}{\ell_{2}-\ell_{1}}\ln\frac{t-\ell_{1}}{t-\ell_{2}}$
$None$
if $\ell\neq 0$ and $\ell_{1}\neq\ell_{2}$;
$\beta_{1}=-\frac{a_{1}k_{1}^{2}}{t}+\frac{b_{1}k_{2}^{2}}{\ell_{2}-\ell_{1}}\ln\frac{t-\ell_{1}}{t-\ell_{2}},\qquad\beta_{2}=-\frac{a_{2}k_{1}^{2}}{t}+\frac{b_{2}k_{2}^{2}}{\ell_{2}-\ell_{1}}\ln\frac{t-\ell_{1}}{t-\ell_{2}}$
$None$
when $\ell=0$ and $\ell_{1}\neq\ell_{2}$;
$\beta_{1}=\frac{a_{1}k_{1}^{2}}{t}-\frac{b_{1}k_{2}^{2}}{t-\ell_{1}},,\qquad\beta_{2}=\frac{a_{2}k_{1}^{2}}{t}t-\frac{b_{2}k_{2}^{2}}{t-\ell_{1}}$
$None$
if $\ell=0$ and $\ell_{1}=\ell_{2}$. Therefore, we obtain:
Theorem 3.5. Let $\ell_{1},\ell_{2},\ell,d_{1},d_{2},k_{1},k_{2}\in\mathbb{R}$
such that $\ell\neq 0$ and $\ell_{1}\neq\ell_{2}$. We have the following
solutions of the coupled two-dimensional cubic nonlinear Schrödinger equations
(1.2) and (1.3):
$\psi=\frac{k_{1}}{t}\exp\left(\frac{(x^{2}+y^{2}-4c_{1}a_{1}k_{1}^{2})i}{4c_{1}t}-\frac{b_{1}k_{2}^{2}i}{t-\ell_{1}}\right),$
$None$
$\varphi=\frac{k_{2}}{t-\ell_{1}}\exp\left(\frac{((x-d_{1})^{2}+(y-d_{2})^{2}-4c_{2}b_{2}k_{2}^{2})i}{4c_{2}(t-\ell_{1})}-\frac{a_{2}k_{1}^{2}i}{t}\right);$
$None$
$\psi=\frac{k_{1}(t-\ell_{1})^{b_{1}k_{2}^{2}(\ell_{2}-\ell_{1})^{-1}i}(t-\ell_{2})^{-b_{1}k_{2}^{2}(\ell_{2}-\ell_{1})^{-1}i}}{t}\exp\frac{(x^{2}+y^{2}-4c_{1}a_{1}k_{1}^{2})i}{4c_{1}t},$
$None$ $\displaystyle\hskip 56.9055pt\varphi$ $\displaystyle=$ $\displaystyle
k_{2}(t-\ell_{1})^{b_{2}k_{2}^{2}(\ell_{2}-\ell_{1})^{-1}i-1/2}(t-\ell_{2})^{-b_{2}k_{2}^{2}(\ell_{2}-\ell_{1})^{-1}i-1/2}$
$\displaystyle\times\exp\left(\frac{(x-d_{1})^{2}i}{4c_{2}(t-\ell_{1})}+\frac{(y-d_{2})^{2}i}{4c_{2}(t-\ell_{2})}-\frac{a_{2}k_{1}^{2}i}{t}\right);\hskip
105.2751pt(3.71)$ $\displaystyle\hskip 54.06006pt\psi$ $\displaystyle=$
$\displaystyle
k_{1}t^{-a_{1}k_{1}^{2}\ell^{-1}i-1/2}(t-\ell)^{a_{1}k_{1}^{2}\ell^{-1}i-1/2}(t-\ell_{1})^{b_{1}k_{2}^{2}(\ell_{2}-\ell_{1})^{-1}i}$
$\displaystyle\times(t-\ell_{2})^{-b_{1}k_{2}^{2}(\ell_{2}-\ell_{1})^{-1}i}\exp\left(\frac{x^{2}i}{4c_{1}t}+\frac{y^{2}i}{4c_{1}(t-\ell)}\right),\hskip
85.35826pt(3.72)$ $\displaystyle\hskip 28.45274pt\varphi$ $\displaystyle=$
$\displaystyle
k_{2}t^{-a_{2}k_{1}^{2}\ell^{-1}i}(t-\ell)^{a_{2}k_{1}^{2}\ell^{-1}i}(t-\ell_{1})^{b_{2}k_{2}^{2}(\ell_{2}-\ell_{1})^{-1}i-1/2}$
$\displaystyle\times(t-\ell_{2})^{-b_{2}k_{2}^{2}(\ell_{2}-\ell_{1})^{-1}i-1/2}\exp\left(\frac{(x-d_{1})^{2}i}{4c_{2}(t-\ell_{1})}+\frac{(y-d_{2})^{2}i}{4c_{2}(t-\ell_{2})}\right).\hskip
56.9055pt(3.73)$
Case 6. For two functions $\beta_{1},\beta_{2}$ of $t$,
$\phi=\frac{x^{2}+y^{2}}{4c_{1}t}+\beta_{1},\qquad\mu=\frac{x^{2}+y^{2}}{4c_{2}t}+\beta_{2}.$
$None$
As in Case 4, (3.2) and (3.4) imply
$\xi=\frac{1}{t}\hat{\xi}(u,v),\qquad\eta=\frac{1}{t}\hat{\eta}(u,v),\qquad
u=\frac{x}{t},\;v=\frac{y}{t}.$ $None$
Moreover, (3.3) and (3.5) become
$-\beta_{1}^{\prime}\hat{\xi}+\frac{c_{1}}{t^{2}}(\hat{\xi}_{uu}+\hat{\xi}_{vv})+\frac{1}{t^{2}}(a_{1}\hat{\xi}^{2}+b_{1}\hat{\eta}^{2})\hat{\xi}=0,$
$None$
$-\beta_{2}^{\prime}\hat{\eta}+\frac{c_{2}}{t^{2}}(\hat{\eta}_{uu}+\hat{\eta}_{vv})+\frac{1}{t^{2}}(a_{2}\hat{\xi}^{2}+b_{2}\hat{\eta}^{2})\hat{\eta}=0.$
$None$
To solve the above system, we assume
$\beta_{1}=-\frac{k_{1}}{t},\qquad\beta_{2}=-\frac{k_{2}}{t},\qquad
k_{1},k_{2}\in\mathbb{R}.$ $None$
Then (3.77) and (3.78) are equivalent to:
$-k_{1}\hat{\xi}+c_{1}(\hat{\xi}_{uu}+\hat{\xi}_{vv})+(a_{1}\hat{\xi}^{2}+b_{1}\hat{\eta}^{2})\hat{\xi}=0,$
$None$
$-k_{2}\hat{\eta}+c_{2}(\hat{\eta}_{uu}+\hat{\eta}_{vv})+(a_{2}\hat{\xi}^{2}+b_{2}\hat{\eta}^{2})\hat{\eta}=0.$
$None$
For $\ell_{1},\ell_{2},\ell_{3}\in\mathbb{R}$, we set
$\varpi=\ell_{1}u+\ell_{2}v+\ell_{3}.$ $None$
If $(a_{1},b_{1})=a_{1}(1,d^{2})$ and $(a_{2},b_{2})=a_{2}(1,d^{2})$ with
$d\in\mathbb{R}$, we have the following solution:
$\hat{\xi}=d\ell\sin\varpi,\;\;\hat{\eta}=\ell\cos\varpi,\;\;(k_{1},k_{2})=(a_{1}(d\ell)^{2}-c_{1}(\ell_{1}^{2}+\ell_{2}^{2}),a_{2}(d\ell)^{2}-c_{2}(\ell_{1}^{2}+\ell_{2}^{2}))$
$None$
for $\ell\in\mathbb{R}$. When $(a_{1},b_{1})=a_{1}(1,-d^{2})$ and
$(a_{2},b_{2})=a_{2}(1,-d^{2})$ with $d\in\mathbb{R}$, we get the solution:
$\hat{\xi}=d\ell\cosh\varpi,\;\;\hat{\eta}=\ell\sinh\varpi,\;\;(k_{1},k_{2})=(c_{1}(\ell_{1}^{2}+\ell_{2}^{2})+a_{1}(d\ell)^{2},c_{2}(\ell_{1}^{2}+\ell_{2}^{2})+a_{2}(d\ell)^{2})$
$None$
for $\ell\in\mathbb{R}$.
Theorem 3.6. For $d,\ell,\ell_{1},\ell_{2},\ell_{3}\in\mathbb{R}$, we have the
following solutions of the coupled two-dimensional cubic nonlinear Schrödinger
equations (1.2) and (1.3):
$\psi=\frac{d\ell\sin\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}}{t}\exp\left(\frac{x^{2}+y^{2}}{4c_{1}t}+\frac{c_{1}(\ell_{1}^{2}+\ell_{2}^{2})-a_{1}(d\ell)^{2}}{t}\right)i,$
$None$
$\varphi=\frac{\ell\cos\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}}{t}\exp\left(\frac{x^{2}+y^{2}}{4c_{2}t}+\frac{c_{2}(\ell_{1}^{2}+\ell_{2}^{2})-a_{2}(d\ell)^{2}}{t}\right)i$
$None$
if $(a_{1},b_{1})=a_{1}(1,d^{2})$ and $(a_{2},b_{2})=a_{2}(1,d^{2})$;
$\psi=\frac{d\ell\cosh\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}}{t}\exp\left(\frac{x^{2}+y^{2}}{4c_{1}t}-\frac{c_{1}(\ell_{1}^{2}+\ell_{2}^{2})+a_{1}(d\ell)^{2}}{t}\right)i,$
$None$
$\varphi=\frac{\ell\sinh\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}}{t}\exp\left(\frac{x^{2}+y^{2}}{4c_{2}t}-\frac{c_{2}(\ell_{1}^{2}+\ell_{2}^{2})+a_{2}(d\ell)^{2}}{t}\right)i$
$None$
when $(a_{1},b_{1})=a_{1}(1,-d^{2})$ and $(a_{2},b_{2})=a_{2}(1,-d^{2})$.
Finally, we assume $a_{1}b_{2}-a_{2}b_{1}\neq 0$. Recall the notion in (3.81).
Taking $k_{1}=k_{2}=0$, the we have the following solutions of the system of
(3.79) and (3.80):
$\hat{\xi}=\frac{1}{\varpi}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}},\qquad\hat{\eta}=\frac{1}{\varpi}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}}$
$None$
and
$\hat{\xi}=\sqrt{\frac{b_{1}c_{2}-b_{2}c_{1}}{(a_{1}b_{2}-a_{2}b_{1})((u-d_{1})^{2}+(v-d_{2})^{2})}},$
$None$
$\hat{\eta}=\sqrt{\frac{a_{2}c_{1}-a_{1}c_{2}}{(a_{1}b_{2}-a_{2}b_{1})((u-d_{1})^{2}+(v-d_{2})^{2})}}$
$None$
for $\ell_{1},\ell_{2}\in\mathbb{R}$. In general, we assume
$\hat{\xi}=\iota_{1}\Im(\varpi),\qquad\hat{\eta}=\iota_{2}\Im(\varpi)$ $None$
for some one-variable function $\Im$ and $k_{1},k_{2}\in\mathbb{R}$. Then
(3.79) and (3.80) implied by
$-k_{1}\Im+c_{1}(\ell_{1}^{2}+\ell_{2}^{2}){\Im^{\prime}}^{\prime}+(a_{1}\iota_{1}^{2}+b_{1}\iota_{2}^{2})\Im^{3}=0,$
$None$
$-k_{2}\Im+c_{2}(\ell_{1}^{2}+\ell_{2}^{2}){\Im^{\prime}}^{\prime}+(a_{2}\iota_{1}^{2}+b_{2}\iota_{2}^{2})\Im^{3}=0.$
$None$
Recall $\epsilon_{1},\epsilon_{2}\in\\{1,-1\\}$. Again by (2.10)-(2.15), Case
1 and Case 2, we have:
$\iota_{1}=\epsilon_{1}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}},\;\;\iota_{2}=\epsilon_{2}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}}$
$None$
with $(\Im,k_{1},k_{2})$ as follows:
$(\tan\varpi,2c_{1}(\ell_{1}^{2}+\ell_{2}^{2}),2c_{2}(\ell_{1}^{2}+\ell_{2}^{2})),\;\;(\sec\varpi,-c_{1}(\ell_{1}^{2}+\ell_{2}^{2}),-c_{2}(\ell_{1}^{2}+\ell_{2}^{2})),$
$None$
$(\coth\varpi,-2c_{1}(\ell_{1}^{2}+\ell_{2}^{2}),-2c_{2}(\ell_{1}^{2}+\ell_{2}^{2})),\;\;(\mbox{csch}\>\varpi,c_{1}(\ell_{1}^{2}+\ell_{2}^{2}),c_{2}(\ell_{1}^{2}+\ell_{2}^{2}));$
$None$
$\iota_{1}=m\epsilon_{1}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}},\;\;\iota_{2}=m\epsilon_{2}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}},,$
$None$
$\Im=\mbox{sn}\>\varpi$ and
$(k_{1},k_{2})=-(1+m^{2})(\ell_{1}^{2}+\ell_{2}^{2})(c_{1},c_{2});$
$\iota_{1}=m\epsilon_{1}\sqrt{\frac{2(b_{2}c_{1}-b_{1}c_{2})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}},\;\;\iota_{2}=m\epsilon_{2}\sqrt{\frac{2(a_{1}c_{2}-a_{2}c_{1})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}},$
$None$
$\Im=\mbox{cn}\>\varpi$ and
$(k_{1},k_{2})=(2m^{2}-1)(\ell_{1}^{2}+\ell_{2}^{2})(c_{1},c_{2});$
$\iota_{1}=\epsilon_{1}\sqrt{\frac{2(b_{2}c_{1}-b_{1}c_{2})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}},\;\;\iota_{2}=\epsilon_{2}\sqrt{\frac{2(a_{1}c_{2}-a_{2}c_{1})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}},$
$None$
$\Im=\mbox{dn}\>\varpi$ and
$(k_{1},k_{2})=(2-m^{2})(\ell_{1}^{2}+\ell_{2}^{2})(c_{1},c_{2}).$
Theorem 3.7. Let $d_{1},d_{2},\ell_{1},\ell_{2},\ell_{3}\in\mathbb{R}$ and let
$\epsilon_{1},\epsilon_{2}\in\\{1,-1\\}$. We have the following solutions of
the coupled two-dimensional cubic nonlinear Schrödinger equations (1.2) and
(1.3):
$\psi=\frac{\epsilon_{1}e^{(x^{2}+y^{2})i/4c_{1}t}}{\ell_{1}x+\ell_{2}y+\ell_{3}t}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}},$
$None$
$\varphi=\frac{\epsilon_{2}e^{(x^{2}+y^{2})i/4c_{2}t}}{\ell_{1}x+\ell_{2}y+\ell_{3}t}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}};$
$None$
$\psi=\epsilon_{1}e^{(x^{2}+y^{2})i/4c_{1}t}\sqrt{\frac{b_{1}c_{2}-b_{2}c_{1}}{(a_{1}b_{2}-a_{2}b_{1})((x-d_{1}t)^{2}+(y-d_{2}t)^{2})}},$
$None$
$\varphi=\epsilon_{2}e^{(x^{2}+y^{2})i/4c_{2}t}\sqrt{\frac{a_{2}c_{1}-a_{1}c_{2}}{(a_{1}b_{2}-a_{2}b_{1})((x-d_{1}t)^{2}+(y-d_{2}t)^{2})}};$
$None$ $\displaystyle\hskip 56.9055pt\psi$ $\displaystyle=$
$\displaystyle\epsilon_{1}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;\frac{1}{t}\tan\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}$
$\displaystyle\times\exp\left(\frac{x^{2}+y^{2}}{4c_{1}t}-\frac{2c_{1}(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i,\hskip
150.79968pt(3.104)$ $\displaystyle\hskip 56.9055pt\varphi$ $\displaystyle=$
$\displaystyle\epsilon_{2}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;\frac{1}{t}\tan\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}$
$\displaystyle\times\exp\left(\frac{x^{2}+y^{2}}{4c_{2}t}-\frac{2c_{2}(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i;\hskip
150.79968pt(3.105)$ $\displaystyle\hskip 56.9055pt\psi$ $\displaystyle=$
$\displaystyle\epsilon_{1}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;\frac{1}{t}\sec\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}$
$\displaystyle\times\exp\left(\frac{x^{2}+y^{2}}{4c_{1}t}+\frac{c_{1}(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i,\hskip
156.49014pt(3.106)$ $\displaystyle\hskip 56.9055pt\varphi$ $\displaystyle=$
$\displaystyle\epsilon_{2}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;\frac{1}{t}\sec\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}$
$\displaystyle\times\exp\left(\frac{x^{2}+y^{2}}{4c_{2}t}+\frac{c_{2}(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i;\hskip
156.49014pt(3.107)$ $\displaystyle\hskip 56.9055pt\psi$ $\displaystyle=$
$\displaystyle\epsilon_{1}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;\frac{1}{t}\coth\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}$
$\displaystyle\times\exp\left(\frac{x^{2}+y^{2}}{4c_{1}t}+\frac{2c_{1}(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i,\hskip
150.79968pt(3.108)$ $\displaystyle\hskip 56.9055pt\varphi$ $\displaystyle=$
$\displaystyle\epsilon_{2}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;\frac{1}{t}\coth\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}$
$\displaystyle\times\exp\left(\frac{x^{2}+y^{2}}{4c_{2}t}+\frac{2c_{2}(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i;\hskip
150.79968pt(3.109)$ $\displaystyle\hskip 56.9055pt\psi$ $\displaystyle=$
$\displaystyle\epsilon_{1}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;\frac{1}{t}\mbox{csch}\>\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}$
$\displaystyle\times\exp\left(\frac{x^{2}+y^{2}}{4c_{1}t}-\frac{c_{1}(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i,\hskip
156.49014pt(3.110)$ $\displaystyle\hskip 56.9055pt\varphi$ $\displaystyle=$
$\displaystyle\epsilon_{2}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;\frac{1}{t}\mbox{csch}\>\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}$
$\displaystyle\times\exp\left(\frac{x^{2}+y^{2}}{4c_{2}t}-\frac{c_{2}(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i;\hskip
156.49014pt(3.111)$ $\displaystyle\hskip 56.9055pt\psi$ $\displaystyle=$
$\displaystyle
m\epsilon_{1}\sqrt{\frac{2(b_{1}c_{2}-b_{2}c_{1})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;\frac{1}{t}\mbox{sn}\>\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}$
$\displaystyle\times\exp\left(\frac{x^{2}+y^{2}}{4c_{1}t}+\frac{c_{1}(1+m^{2})(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i,\hskip
110.96556pt(3.112)$ $\displaystyle\hskip 56.9055pt\varphi$ $\displaystyle=$
$\displaystyle
m\epsilon_{2}\sqrt{\frac{2(a_{2}c_{1}-a_{1}c_{2})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;\frac{1}{t}\mbox{sn}\>\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}$
$\displaystyle\times\exp\left(\frac{x^{2}+y^{2}}{4c_{2}t}+\frac{c_{2}(1+m^{2})(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i;\hskip
110.96556pt(3.113)$ $\displaystyle\hskip 56.9055pt\psi$ $\displaystyle=$
$\displaystyle
m\epsilon_{1}\sqrt{\frac{2(b_{2}c_{1}-b_{1}c_{2})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;\frac{1}{t}\mbox{cn}\>\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}$
$\displaystyle\times\exp\left(\frac{x^{2}+y^{2}}{4c_{1}t}-\frac{c_{1}(2m^{2}-1)(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i,\hskip
105.2751pt(3.114)$ $\displaystyle\hskip 56.9055pt\varphi$ $\displaystyle=$
$\displaystyle
m\epsilon_{2}\sqrt{\frac{2(a_{1}c_{2}-a_{2}c_{1})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;\frac{1}{t}\mbox{cn}\>\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}$
$\displaystyle\times\exp\left(\frac{x^{2}+y^{2}}{4c_{2}t}-\frac{c_{2}(2m^{2}-1)(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i;\hskip
102.43008pt(3.115)$ $\displaystyle\hskip 62.59596pt\psi$ $\displaystyle=$
$\displaystyle\epsilon_{1}\sqrt{\frac{2(b_{2}c_{1}-b_{1}c_{2})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;\frac{1}{t}\mbox{dn}\>\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}$
$\displaystyle\times\exp\left(\frac{x^{2}+y^{2}}{4c_{1}t}+\frac{c_{1}(m^{2}-2)(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i,\hskip
105.2751pt(3.116)$ $\displaystyle\hskip 62.59596pt\varphi$ $\displaystyle=$
$\displaystyle\epsilon_{2}\sqrt{\frac{2(a_{1}c_{2}-a_{2}c_{1})(\ell_{1}^{2}+\ell_{2}^{2})}{a_{1}b_{2}-a_{2}b_{1}}}\;\frac{1}{t}\mbox{dn}\>\frac{\ell_{1}x+\ell_{2}y+\ell_{3}t}{t}$
$\displaystyle\times\exp\left(\frac{x^{2}+y^{2}}{4c_{2}t}+\frac{c_{2}(m^{2}-2)(\ell_{1}^{2}+\ell_{2}^{2})}{t}\right)i.\hskip
108.12054pt(3.117)$
References
[AEK]
N. Akhmediev, V. Eleonskii and N. Kulagin, First-order exact solutions of the
nonlinear Schrödinger equation, Teoret. Mat. Fiz. 72 (1987), 183-196.
[GG]
B. Grébert and J. Guillot, Periodic solutions of coupled nonlinear Schrödinger
equations in nonlinear optics: the resonant case, Appl. Math. Lett. 9 (1996),
65-68.
[GW]
L. Gagnon and P. Winternitz, Exact solutions of the cubic and quintic
nonlinear Schrödinger equation for a cylindrical geometry, Phys. Rev. A 22
(1989), 296
[HS]
F. Hioe and T. Salter, Special set and solutions of coupled nonlinear
Schrödinger equations, J. Phys. A: Math. Gen. 35 (2002), no. 42, 8913-8928.
[I]
N. H. Ibragimov, Lie Group Analysis of Differential Equations, Volume 2, CRC
Handbook, CRC Press, 1995.
[MP]
D. Mihalache and N. Panoin, Exact solutions of nonlinear Schrödinger equation
for positive group velocity dispersion, J. Math. Phys. 33 (1992), no. 6,
2323-2328.
[RL1]
R. Radhakrishnan and M. Lakshmanan, Bright and dark soliton solutions to
coupled nonlinear Schrödinger equations, J. Phys. A: Math. Gen. 28 (1995), no.
9, 2683-2692.
[RL2]
R. Radhakrishnan and M. Lakshmanan, Exact soliton solutions to coupled
nonlinear Schrödinger equations with higher-order effects, Phys. Rev. E(3) 54
(1995), no. 3, 2949-2955.
[SEG]
E. Saied, R. EI-Rahman and M. Ghonamy, On the exact solution of
(2+1)-dimensional cubic nonlinear Schrödinger (NLS) equation”, J. Phy. A:
Math. Gen. 36 (2003), 6751-6770.
[WG]
Z. Wang and D. Guo, Special functions, World Scientific, Singapore, 1998.
| arxiv-papers | 2008-06-30T15:48:57 | 2024-09-04T02:48:56.458209 | {
"license": "Public Domain",
"authors": "Xiaoping Xu",
"submitter": "Xiaoping Xu",
"url": "https://arxiv.org/abs/0806.4929"
} |
0807.0023 | # Automatic Metadata Generation using Associative Networks111Rodriguez M.A.,
Bollen, J., Van de Sompel, H., “Automatic Metadata Generation using
Associative Networks”, ACM Transactions on Information Systems, volume 27,
number 2, pages 1-20, doi:10.1145/1462198.1462199, ISSN: 1046-8188, ACM Press,
February 2009.
Marko A. Rodriguez marko@lanl.gov Digital Library Research and Prototyping
Team
Los Alamos National Laboratory
Los Alamos, New Mexico 87545 Johan Bollen jbollen@lanl.gov Digital Library
Research and Prototyping Team
Los Alamos National Laboratory
Los Alamos, New Mexico 87545 Herbert Van de Sompel herbertv@lanl.gov Digital
Library Research and Prototyping Team
Los Alamos National Laboratory
Los Alamos, New Mexico 87545
###### Abstract
In spite of its tremendous value, metadata is generally sparse and incomplete,
thereby hampering the effectiveness of digital information services. Many of
the existing mechanisms for the automated creation of metadata rely primarily
on content analysis which can be costly and inefficient. The automatic
metadata generation system proposed in this article leverages resource
relationships generated from existing metadata as a medium for propagation
from metadata-rich to metadata-poor resources. Because of its independence
from content analysis, it can be applied to a wide variety of resource media
types and is shown to be computationally inexpensive. The proposed method
operates through two distinct phases. Occurrence and co-occurrence algorithms
first generate an associative network of repository resources leveraging
existing repository metadata. Second, using the associative network as a
substrate, metadata associated with metadata-rich resources is propagated to
metadata-poor resources by means of a discrete-form spreading activation
algorithm. This article discusses the general framework for building
associative networks, an algorithm for disseminating metadata through such
networks, and the results of an experiment and validation of the proposed
method using a standard bibliographic dataset.
††preprint: LA-UR-06-3445
## I Introduction
Resource metadata plays a pivotal role in the functionality and
interoperability of digital information repositories. However, in spite of its
value, high quality metadata is difficult to come by duval:meta2002 .
dc:ward2003 demonstrates that although as many as 15 possible metadata
properties can theoretically be included in the widely used Dublin Core
standard222The Dublin Core 1.1 specification is available at:
http://dublincore.org/documents/1999/07/02/dces/, few are frequently used in
collections whose metadata are generally created by the author’s
themselves333The most frequently used are creator, identifier, title, date,
and type.. The problem of poor and incomplete metadata is expected to worsen
as repositories are applied to materials collected beyond the traditional,
centralized methods of publication and start to obtain data from web pages,
blogs, personal multimedia collections, and collaborative tagging
environments.
Metadata is a costly resource to create, maintain, and/or recover manually.
There has therefore been significant research on automated metadata generation
(e.g. by extracting metadata from the content of resources). Natural language
processing meta:yang2005 and document image analysis techniques
meta:guirida2000 ; sebastiani02machine ; meta:han2003 ; meta:mao2004 may
extract keywords, subject categories, author, and citations (e.g.
CiteSeer444CiteSeer available at: http://citeseer.ist.psu.edu/) from
manuscripts. Furthermore, in meta:greenburg2004 , two metadata generators are
demonstrated that successfully harvest and extract metadata from existing
resource source and content. Such content-based techniques are much less
efficient for multimedia resources, e.g. video, music, images, and datasets.
Reliable content analysis for such data is still an active research area and
existing methods generally yield little content-related metadata. In addition,
content-based approaches can be prohibitively expensive in computational terms
kuwano:meta2004 .
For the reasons outlined above, methods for the generation of metadata that do
not rely on resource content have generated considerable interest. The recent
growth in applications of “folksonomies” (i.e. community-based “tagging”
tag:mathes2004 ; tagging:hub2006 ), has been, to some extent, inspired by the
shortcomings of existing metadata generation methods. Unfortunately, human
tagging only works well in situations where the number of participants greatly
exceeds the number of resources to be tagged and where there is no requirement
for controlled vocabularies or standardized metadata formats.
In this article, we propose a system for automated metadata generation that
starts from a common scenario: a heterogeneous repository contains resources
for which varying degrees of metadata are available. Some resources have been
imbued with rich, vetted metadata, whereas others have not. However, if it can
be assumed that resources that are “similar” (e.g. similar in publication
venue, authorship, date, citations, etc.) are more likely to have shared
metadata, then the problem of metadata generation can be reformulated as one
of extrapolating metadata from metadata-rich to related, but metadata-poor
resources. This article’s experiment focuses on identifying which aspects of
metadata similarity are best used to extrapolate resource metadata in a
bibliographic dataset.
As a case in point, levera:naaman2005 describes a method to support the
annotation of personal photograph collections. Once a user has annotated a
photograph its metadata is automatically transferred to photographs taken at
similar times and locations. For example, a user photographs a group of
friends at 3:45PM. Another photograph is made at 3:47PM. Since the second
photograph was taken only two minutes after the first, it is likely that it
depicts a similar scene. The system therefore transfers metadata from
photograph 1 to photograph 2. Similarly, metaprop:prime2005 proposes a method
of web page metadata propagation using co-citation networks. The general idea
is that if two web pages cite other web pages in common, then the probability
that they share similar metadata is higher. The user can later correct and
augment any transferred metadata.
The mentioned systems are strongly related to collaborative filtering
collab:herlocker2006 . Collaborative filtering systems are commonly employed
in online retail systems to recommend items of interest to individual users.
Using the principle that similar users are more likely to appreciate similar
items, users are recommended items that are missing from their profiles but
occur in the profiles of similar users. The collaborative filtering process
can thus be regarded as an instance of metadata propagation. If users are
considered resources and their profiles are considered “resource metadata”, it
can be said that collaborative filtering systems “recommend” metadata from one
resource to another based on resource similarity.
A generalization of the above metadata propagation systems can be made in
terms of the following elements:
1. 1.
A mechanism to generate resource relations, i.e. assess their similarity.
2. 2.
The determination of a metadata-rich subset of the repository’s collection
that can serve as a reference set.
3. 3.
A means of propagating metadata from the metadata-rich reference set to a
metadata-poor subset of the collection using the established resource
relations as a substrate.
Such systems for the generation of metadata can be said to operate on a “Robin
Hood” principle; they take from metadata-rich resources and give to metadata-
poor resources, with the exception that metadata is not a zero-sum resource.
This mode of operation has a number of desirable properties. First, it reduces
the need for the costly generation of metadata; metadata is automatically
extrapolated from an existing metadata-rich reference collection to a
metadata-poor subset. Second, resource relations can be defined independent of
content and metadata extrapolation can thus be implemented for wide range of
heterogeneous resources, e.g. audio, video, and images.
This article outlines a proposal for a metadata propagation system designed
for scholarly repositories that takes advantage of the multiple means by which
two resources can be related (e.g. co-citation, citation, co-author, co-
keyword, etc.). Figure 1 presents the outline of the proposed system’s
components and processing stages. First, resource metadata is extracted from
the collection of a repository. Second, an associative multi-relational
network (i.e. a directed labeled graph) of resource relations is derived from
a subset of the available metadata. Third, a metadata-rich subset of the
collection is selected to serve as a reference data set. Fourth, and finally,
metadata is propagated (i.e. extrapolated) from the metadata-rich reference
set to all other metadata-poor resources over the associative network of
resources after which the repository is updated. Human validation can vet the
results of the metadata extrapolation before insertion into the repository
occurs.
Figure 1: System outline
It is important to emphasize that this system requires the existence of some
preliminary metadata both for the construction of resource relations and for
metadata propagation. Furthermore, the quality or accuracy of the preliminary
metadata is important in ensuring successful results (i.e. to avoid a “garbage
in, garbage out” scenario). However, the metadata being propagated can be
different from the metadata used to generate resource relations. For instance,
in the manuscript domain, the propagation of keyword metadata may be most
efficient along resource relations derived from citation metadata. Therefore,
two aspects affect the efficiency of metadata propagation: the type of
resource relations and the algorithm used to propagate metadata. It is
important to note that no new metadata values are created in model proposed in
this article. While it is important for resources to maintain metadata, this
method only propagates pre-existing metadata values and thus, does not
increase the discriminatory aspects that metadata should and generally
provides. While like resources should have similar metadata, variations should
also exist to make sure that a resource’s metadata accentuates the unique
characteristics of the resource.
This paper will first discuss two algorithms to define sets of resource
relations and represent these relations in terms of associative networks. It
will then formally define a metadata propagation algorithm which can operate
on the basis of the generated resource relations. Finally, the proposed
metadata generation system is validated using a modified version of the KDD
Cup 2003 High-Energy Physics bibliographic dataset (hep-th 2003)555hep-th 2003
available at: http://www.cs.cornell.edu/projects/kddcup/datasets.html. While
it is theoretically possible for this method to work on other resource types
(e.g. video, audio, etc.) as it doesn’t require an analysis of the content of
the resources, only their metadata; it is only speculated that the results of
such a method would be viable in these other, non-tested, domains.
## II Constructing an Associative Network of Repository Resources
An associative network is a network that connects resources according to some
measure of similarity. An associative network is represented by the data
structure $G=(N,E,W)$ where $N$ is the set of resources, $E\subseteq N\times
N$ the set of directed relationships between resources, and $W$ is the set of
weight values for all edges such that $|W|=|E|$. Any edge $e_{i,j,\mu}$ with
corresponding weight $w_{i,j,\mu}$ expresses that there exists a directed
weighted relationship constructed using properties of type $\mu$ from resource
$n_{i}$ to resource $n_{j}$. The explicit representation of $\mu$ is necessary
because an associative network can be constructed according to different
properties (i.e. authorship, citations, keywords, etc.). As will be
demonstrated, certain network $\mu$ relationships are better (in terms of
precision and recall) at propagating certain property types than others.
The remainder of this section will describe two associative network
construction algorithms. One is based on occurrence metadata where a resource
is considered similar to another if there is a direct reference from one
resource to the other (e.g. a direct citation). The other algorithm is based
on co-occurrence metadata and thus, considers two resources to be similar if
they share similar metadata. That is, two resources are deemed similar if the
same metadata values occur in both their properties (i.e. same authors, same
keywords, same publication venue, etc.). Depending on how the repository
represents its metdata some property types will be direct reference properties
and others will have to be infered through indirect, co-occurence algorithms.
### II.1 Occurrence Associative Networks
An associative network can be constructed if direct references connect one
resource to another. The World Wide Web, for instance, is an associative
network based on occurrence data because a web-page makes a direct reference
to another web-page via a hyper-link (i.e. the href HTML tag). For manuscript
resources, occurrence information usually exists in citations. For instance,
if resource $n_{i}$ references (i.e. cites) resource $n_{j}$ then their exists
an edge $e_{i,j,\mathrm{cite}}$. One potential algorithm for determining the
edge weight is to first determine how many other citations resource $n_{i}$
currently maintains. That is, if resource $n_{i}$ also cites $50$ other
resources then resource $n_{i}$ is $\frac{1}{50}$ as similar to $n_{j}$,
$w_{i,j,\mathrm{cite}}=\frac{1}{50}$. Similarly, if resource $n_{i}$ only
cites resource $n_{j}$ then the strength of tie to resource $n_{j}$ is
greater, $w_{i,j,\mathrm{cite}}=1.0$. The general equation is defined by Eq.
II.1 where the function $\mathrm{meta}(n_{i},\mathrm{cite})$ returns the set
of all citations for resource $n_{i}$. This equation only holds if resource
$n_{j}\in\mathrm{meta}(n_{i},\mathrm{cite})$. Eq. II.1 makes use of the $\mu$
notation in order to generalize the equation for use with any direct reference
property types.
$w_{i,j,\mathrm{\mu}}=\frac{1}{\left|\mathrm{meta}(n_{i},\mathrm{\mu})\right|}\;:\;n_{j}\in\mathrm{meta}(n_{i},\mathrm{\mu})$
The running time of the algorithm to construct an associative network based on
direct, occurrence property types is $O(N)$ since each resource must be
checked once and only once for direct reference to other resources.
### II.2 Co-occurrence Associative Networks
Co-occurrence networks are created when resources share the same metadata
property values. For instance, if two resources share the same keyword,
author, or citation values then there exists some degree of similarity. For a
co-occurrence network the edge weight for any two resources,
$w_{i,j,\mathrm{co}\mu}$ and $w_{j,i,\mathrm{co}\mu}$, is a function of the
amount of metadata properties of type $\mu$ that $n_{i}$ and $n_{j}$ share in
common. A specific example of this could be a co-keyword associative network
created when two resources have similar keywords. For example, suppose the
resource nodes $n_{i}$ and $n_{j}$ have the following list of keyword
properties presented in Table 1.
resource | keyword-1 | keyword-2 | keyword-3
---|---|---|---
$n_{i}$ | repository | metadata | particle
$n_{j}$ | images | repository | metadata
Table 1: Keyword metadata for resources $n_{i}$ and $n_{j}$
In Table 1, resource $n_{i}$ and $n_{j}$ share two keywords in common, namely
repository and metadata. The edge weight between these two resources is a
function of the amount of keywords they share in common, Eq. II.2, and the
size of the keyword count of both resources. Therefore, according to Eq. II.2,
the edges connecting resource $n_{i}$ to $n_{j}$ and $n_{j}$ to $n_{i}$ have a
weight of $w_{n_{i},n_{j},\mathrm{cokey}}=w_{n_{j},n_{i},\mathrm{cokey}}=0.5$.
$\mathrm{co}(n_{i},n_{j},\mu)=\mathrm{meta}(n_{i},\mu)\cap\mathrm{meta}(n_{j},\mu)$
so that
$w_{i,j,\mathrm{co}\mu}=w_{j,i,\mathrm{co}\mu}=\frac{|\mathrm{co}(n_{i},n_{j},\mu)|}{[|\mathrm{meta}(n_{i},\mu)|+|\mathrm{meta}(n_{j},\mu)|]-|\mathrm{co}(n_{i},n_{j},\mu)|}$
Notice that the co-occurrence algorithm in Eq. II.2 returns a $\mathrm{co}\mu$
representation. This means for keyword properties, the returned weight is a
co-keyword similarity weight. Similarly, for authorship metadata, the returned
weight is a co-authorship weight. The running time of the algorithm to
construct a co-occurrence network is $O(\frac{N^{2}-N}{2})$ since each
resource’s $\mu$-properties must be checked against every other resource’s
$\mu$-properties ($N^{2}$), except itself ($-N$), once and only once
($\frac{1}{2}$).
## III Metadata Propagation Algorithm
Reconstructing the metadata for a metadata-poor collection of resources is
dependent not only on the associative network data structure, but also upon
the use of a metadata propagation algorithm. The algorithm chosen is a
derivative of the particle-swarm algorithm socialgrammar:rodriguez2007 .
Particle-swarm algorithms are a discrete form of the spreading activation
algorithms spread:collins1975 ; inform:cohen1987 ; applic:crestani1997 ;
search:crestani2000 ; applyi:huang2004 . Because particles are indivisible
entities, it is easy to represent metadata properties as being encapsulated
inside a particle. These metadata particles are then propagated over the edges
of the associative network. Upon reaching a resource node that is missing a
particular property type, the particle recommends its property value to the
visited resource. This section will formally describe the metadata propagation
algorithm before discussing the results of an experiement using a
bibliographic dataset.
Every resource node in an associative network is supplied with a single
particle, $p_{i}\in P$, such that $|P|=|N|$. The particle $p_{i}$ encapsulates
all the metadata properties of a particular resource $n_{i}$. Therefore,
$\mathrm{meta}(n_{i},\mu)\equiv\mathrm{meta}(p_{i},\mu)$ for all $\mu$.
Particle $p_{i}$ has a reference to its current node $c_{i}\in N$ such that at
$t=0$, $c_{i}=n_{i}$. The particle $p_{i}$ begins its journey ($t=0$) at its
home node, $n_{i}$, and traverses an outgoing edge of $n_{i}$. Particle edge
traversal is a stochastic process that requires the outgoing edge weights of
each node to form a probability distribution. Therefore, the set of outgoing
edge weights of relation type $\mu$ for $n_{i}$, $\mathrm{out}(n_{i},\mu)$,
must be normalized as represented in Eq. III and Eq. III. Unlike Eq. II.2, for
co-occurrence edges, these equations do not guarantee that
$w_{i,j,\mu}=w_{j,i,\mu}$.
$w_{i,j,\mu}=\frac{w_{i,j,\mu}}{\sum_{\forall k\in e_{i,k,\mu}}w_{i,k,\mu}}$
such that
$\sum_{\forall j\in e_{i,j,\mu}}w_{i,j,\mu}=1.0$
The function $\theta(\mathrm{out}(n_{i},\mu))$ is defined such that it takes a
set of outgoing edges of relation type $\mu$ of node $n_{i}$ and returns a
single node $n_{j}$ based upon the outgoing edge weight probability
distribution, where $e_{i,j,\mu}\in\mathrm{out}(n_{i},\mu)$. This is how a
particle traverses an associative network.
The particle $p_{i}$ also has an associated energy value
$\epsilon_{i}\in[0,1]$. Each time an edge is traversed, the particle $p_{i}$
decays its energy content, $\epsilon_{i}$, according to a global decay value,
$\delta\in[0,1]$. Particle energy decay over discrete time $t$ is represented
in Eq. III. The rational for decay is based on the intuition that the metadata
property values of a particular particle become less relevant the further the
particle travels away from its source node ($c_{i}$ at $t=0$). Therefore, the
further a particle travels in the network, the more that particle’s energy
value (or recommendation influence), $\epsilon$, is decayed.
$\epsilon_{i}(t+1)=(1-\delta)\epsilon_{i}(t)$
The energy value of a particle defines how much recommendation influence a
particle’s metadata property values has on a visited metadata-poor node. Each
time a particle traverses a node with missing metadata properties, it not only
recommends its metadata property values to that node, but also increments the
appropriate property value with its current energy value $\epsilon_{i}$. In
Figure 2, at $t=0$, before the propagation algorithm has been executed,
resource $n_{3}$ has no keyword values. Therefore, when particle $p_{1}$
reaches $n_{3}$ at $t=1$, particle $p_{1}$ recommends its keyword property
values (keyword={swarm, algorithms}) to node $n_{3}$ with an influence of
$\epsilon_{1}=0.85$. At $t=2$, particle $p_{2}$, with $\epsilon_{2}=0.723$,
recommends its keyword property (keyword={swarm}) to node $n_{3}$. Notice that
the recommendation of the keyword property value ‘swarm’ is reinforced each
time that property value is presented to $n_{3}$.
Figure 2: Particles recommending metadata information to a metadata-poor node
The function of a single particle, $p_{i}$, at a particular node, $n_{j}$, is
represented in pseudo-code in Algorithm 1 where $\mathrm{rec}(n_{j},\mu)$
returns the set of previous property values to $n_{j}$ for a property of type
$\mu$.
Input: $\mathrm{recommendMeta}(n_{j},p_{i})$
# $p_{i}$ updates the metadata of $n_{j}$ for all property types;AlgoLine0.1
foreach _$(\mu\mathrm{-property})$_ do # first ensure that $n_{j}$ is
metadata-poor at the particular $\mu$-property; AlgoLine0.2
if _$(|\mathrm{meta}(n_{j},\mu)|==0)$_ then # update the metadata-poor node’s
$\mu$-property with the $\mu$ property value of $p_{i}$; AlgoLine0.3
foreach _$(x\in\mathrm{meta}(p_{i},\mu))$_ do $\mathrm{found}=false$;
AlgoLine0.4
# if property value already exists, increment its energy value with
$e_{i}$;AlgoLine0.5
foreach _$(y\in\mathrm{rec}(n_{j},\mu))$_ do if _$(x==\mathrm{value}(y))$_
then $\mathrm{energy}(y)=\mathrm{energy}(y)+\epsilon_{i}$; AlgoLine0.6
$\mathrm{found}=true$;AlgoLine0.7
end end # if no recommended value exists, add to $n_{j}$’s
recommendations;AlgoLine0.8
if _$(!\mathrm{found})$_ then $\mathrm{addRec}(x,\epsilon_{i})=x$; AlgoLine0.9
end end end end
Algorithm 1 Particle $p_{i}$ recommending metadata properties values to
$n_{j}$
9
9
9
9
9
9
9
9
9
If Algorithm 1 is called $\mathrm{recommendMeta}(n_{j},p_{i})$ then the full
particle propagation algorithm can be described by the pseudo-code in
Algorithm 2. The process of moving metadata particles through the associative
network and recommending metadata-poor nodes metadata property values
continues until some desired $t$ is reached or all particle energy in the
network has decayed to $0.0$, $\sum_{\forall i}\epsilon_{i}\cong 0.0$.
Input: $\mathrm{propagate}(\mu)$
# $\delta$ is a global energy decay value ;AlgoLine0.1
$\delta=0.15$ ;AlgoLine0.2
# create a particle for each node ;AlgoLine0.3
foreach _$(n_{i}\in N)$_ do
$\mathrm{meta}(p_{i},\mu)=\mathrm{meta}(n_{i},\mu)\;:\;\forall\mu$ ;
AlgoLine0.4
$\epsilon_{i}=1.0$ ;AlgoLine0.5
$c_{i}=n_{i}$ ;AlgoLine0.6
end # propagate metadata particles throughout $\mu$ network ;AlgoLine0.7
$t=0$ ;AlgoLine0.8
while _$(\sum_{\forall p_{i}\in P}\epsilon_{i}
>0.0\;\;\&\&\;\;t<\mathrm{maxSteps})$_ do foreach _$(p_{i}\in P)$_ do # if
$c_{i}$ has no outgoing edges, freeze the particle ; AlgoLine0.9
if _$(|\mathrm{out}(c_{i},\mu)| >0)$_ then
$c_{i}=\theta(\mathrm{out}(c_{i},\mu))$ ; AlgoLine0.10
$\epsilon_{i}=\epsilon_{i}*(1-\delta)$ ;AlgoLine0.11
# do not recommend metadata to the particle’s home node ;AlgoLine0.12
if _$(c_{i}\;\mathrm{!=}\;n_{i})$_ then $\mathrm{recommendMeta}(c_{i},p_{i})$
; AlgoLine0.13
end end end $t=t+1$ ;AlgoLine0.14
end
Algorithm 2 Propagating metadata particles through an associative network of
type $\mu$
14
14
14
14
14
14
14
14
14
14
14
14
14
14
## IV An Experiment using the 2003 HEP-TH Bibliographic Dataset
This section will present the results of the proposed metadata generation
system when attempting to reconstruct an artificially atrophied bibliographic
dataset. By artificially reducing the amount of metadata in the full
bibliographic dataset, it is possible to simulate a metadata-poor environment
and at the same time still be able to validate the results of the metadata
propagation algorithm. The section is outlined as follows. First, the dataset
used for this experiment is described. Second, a short review of the
validation metrics (precision, recall, and F-score) is presented. Third, the
various system parameters are discussed. Finally, the results of the
experiment are presented as a validation of the systems use for manuscript-
based digital library repositories. Further research into other domains
besides manuscripts will demonstrate the validity of this method for other
resource types.
The dataset used to validate the proposed system is a modified version of the
hep-th 2003 bibliographic dataset for high energy physics and theory
mcgovern03exploiting .666The details of how the hep-th dataset was created can
be found in the disseminated README (http://kdl.cs.umass.edu/data/hepth/hepth-
README.txt) of the data set and a few specifics are quoted here: “Object and
link properties such as title, authors, journal (if published), and various
dates were extracted from the abstract files with additional date information
coming from the slacdates citation tarball. Because institutions were often
not presented in a consistent format, the email domain of the submitter (if
available) was used as a surrogate for institution. Because many authors had
no associated email address, domain information is not available for all
authors. Consolidation was performed on journal names, domains, and author
names. A nominal amount of hand-cleaning to correct spelling or formatting
problems was also performed.” A modified version of the hep-th dataset, as
used in unsupervised:lin2004 , is represented as a semantic network containing
manuscripts (29,014), authors (12,755), journals (267), organizations (963),
keywords (40), and publication date in year/season pairs (60). These nodes are
then connected according to the following semantics:
* •
writes($a$,$m$): author $a$ wrote manuscript $m$
* •
date_published($m$,$d$): manuscript $m$ was published on date $d$
* •
organization_of($a$,$o$): author $a$ works for organization $o$
* •
published_in($m$,$j$): manuscript $m$ was published in journal $j$
* •
cites($m_{x}$,$m_{y}$): manuscript $m_{x}$ cites manuscript $m_{y}$
* •
keyword_of($m$,$k$): manuscript $m$ has keyword $k$
For the purposes of this experiment, the semantic network from
unsupervised:lin2004 was transformed into a list of manuscripts and their
associated metadata property name/value pairs. These manuscript properites
include: authors, date of publication, citations, keywords, publishing
journal, and organizations. Of the 29,014 manuscript nodes, different
occurrence and co-occurrence algorithms were used to construct the following
associative networks:
1. 1.
citation: manuscript $m_{i}$ maintains an edge to manuscript $m_{j}$ if
$m_{i}$ cites $m_{j}$ (27,240 edges)
2. 2.
co-author: manuscripts maintain an edge if they share authors (724,406 edges)
3. 3.
co-citation: manuscripts maintain an edge if they share citations (23,089,616
edges)
4. 4.
co-keyword: manuscripts maintain an edge if they share keywords (12,418,172
edges)
5. 5.
co-organization: manuscript maintain an edge if they share organizations
(33,947,083 edges)
Though not explored empirically, it is worth noting that link prediction
algorithms can be employed to resolve issues relating to edge sparsity in the
network. In particular, the methods proposed in linkpr:liben2003 and
linkpr:hunag2005 are such algorithms.
### IV.1 A Review of Precision, Recall, and F-Score
The results of the metadata generation experiment are evaluated according to
the F-score measure so therefore, it is important to provide a quick review of
precision, recall, and F-score within the framework of the notation presented
thus far. For a particular property $\mu$, precision is defined as the amount
of property values of type $\mu$ received that were relevant relative to the
total number of property values retrieved overall. This is represented in Eq.
IV.1 where the function $\mathrm{rec}(n_{i},\mu)$ returns the set of
recommended property values for resource $n_{i}$ of type $\mu$, while
$\mathrm{meta}(n_{i},\mu)$ returns the set properties values of type $\mu$
previously existing for resource $n_{i}$. Since the validation is against an
artificially atrophied resource set, the recommended property values are
checked against the previously existing property values (prior to artificial
atrophy).
$Pr(\mu)=\frac{|\mathrm{meta}(n_{i},\mu)\cap\mathrm{rec}(n_{i},\mu)|}{|\mathrm{rec}(n_{i},\mu)|}$
Recall, Eq. IV.1, on the other hand, is defined as the proportion of relevant
property values received to the total amount of relevant property values
possible. For example, if resource $n_{i}$ previously (before artificial
atrophy) had the property value keyword=$\\{$swarm$\\}$ and is recommended the
property value keyword=$\\{$swarm$\\}$, then there is a 100% recall. On the
other hand, if resource $n_{i}$ previously had the property values
keyword=$\\{$swarm, network$\\}$ and is recommended the property value
keyword=$\\{$swarm$\\}$, then there is a 50% recall, whereas its precision is
100% in both cases.
$Re(\mu)=\frac{|\mathrm{meta}(n_{i},\mu)\cap\mathrm{rec}(n_{i},\mu)|}{|\mathrm{meta}(n_{i},\mu)|}$
Precision and recall tend to be inversely related, $Pr\approx\frac{1}{Re}$.
This inverse relationship is understood best when examining the extreme cases.
If every possible property value was provided to a resource
($|\mathrm{rec}(n_{i},\mu)|\rightarrow\infty$), and that resource originally
only had one property value ($|\mathrm{meta}(n_{i},\mu)|=1$) then the recall
would be 100% while the precision would be near 0%. At the opposite extreme,
if a resource previously had every possible property value in its original
metadata ($|\mathrm{meta}(n_{i},\mu)|\rightarrow\infty$) and was recommend
only one property value ($|\mathrm{rec}(n_{i},\mu)|=1$), then the precision
would be 100%, but the recall would be near 0%. While, in some systems,
precision and recall can be inversely related, it is the goal of information
retrieval systems that are validated according to this criterion to achieve
both high precision and recall values.
Finally, F-score, Eq. IV.1, can be used to combine precision and recall into a
single measure. Note that different associative networks will perform
differently for different property types. For instance, co-citation networks
will, intuitively, preform better at propagating keyword values than co-
organization networks. Therefore, the F-score measure will be represented as
$F(\mu_{x},\mu_{y})$ in order to express the F-score of a network created from
metadata properties of type $\mu_{y}$ propagating property values of type
$\mu_{x}$. Precision and recall can be represented in a similar fashion though
the results of the experiment to follow are expressed according to the F-score
measure only.
$F(\mu_{x},\mu_{y})=\frac{2\cdot Pr(\mu_{x})\cdot
Re(\mu_{x})}{Pr(\mu_{x})+Re(\mu_{x})}$
### IV.2 Experiment Parameters
The experiment was set up to determine various F-scores, $F(\mu_{x},\mu_{y})$,
where
$\mu_{x}\in\\{\mathrm{auth},\mathrm{cite},\mathrm{date},\mathrm{jour},\mathrm{key},\mathrm{org}\\}$
and
$\mu_{y}\in\\{\mathrm{cite},\mathrm{coauth},\mathrm{cocite},\mathrm{cokey},\mathrm{coorg}\\}$.
This means that for every type of associative network generated, an F-score
for each metadata property type was determined. Since the hep-th 2003
bibliographic dataset is a metadata-rich dataset, it was necessary to destroy
a percentage of the metadata to test whether or not the metadata generation
algorithm could reconstruct the property values for the selected metadata-poor
resources. Therefore, the tunable parameter, density, $\partial\in[0.01,0.9]$,
was created. The density of the network metadata ranges from 1% of the network
resources containing metadata to 99% of the resources. Given the percentage
parameter, resources were randomly selected for atrophy before the metadata
propagation algorithm was run.
With the potential for 99% of the network containing metadata, the propagation
of metadata to the lacking 1% would be overwhelming (a high recall with a low
precision). In order to allow nodes to regulate the amount of metadata
property values they accept, a second parameter exists. The percentile
parameter, $\rho\in[0,1]$, determines the energy threshold for property value
recommendations. Since each $\mathrm{rec}(n_{i},\mu)$ entry has an associated
energy value (recommendation influence), a range from 0th percentile, meaning
all provided property values are accepted to 100th percentile, meaning only
the top energy property value is accepted, is used. The pseudo-code for the
experimental set-up is presented in Algorithm 3. In Algorithm 3,
$\mathrm{killMeta}()$, $\mathrm{acceptMeta}()$, and $\mathrm{calculateF}()$ do
not have accompanying pseudo-code.
Input: $\mathrm{experiment}()$
# run the metadata propagation algorithm for each associative network type
;AlgoLine0.1
foreach _$(\mu_{y}\in[\mathrm{coauth,cocite,cokey,cite}])$_ do
$\mathrm{loadNetwork}(\mu_{y})$ ; AlgoLine0.2
foreach _$(\mu_{x}\in[\mathrm{auth,cite,date,jour,key,org}])$_ do # atrophy a
randomly selected percentage of the network ; AlgoLine0.3
for _$(\partial=0.01,\;\partial <1.0,\;\partial=\partial+0.2)$_ do
$\mathrm{killMeta}(1-\partial)$ ; AlgoLine0.4
$\mathrm{propagateMeta}(\mu_{x})$ ;AlgoLine0.5
# allow metadata-poor resources to accept only a certain percentage of their
recommended property values ;AlgoLine0.6
for _$(\rho=0.0,\;\rho <=1.0,\;\rho=\rho+0.1)$_ do $\mathrm{acceptMeta}(\rho)$
; AlgoLine0.7
$\mathrm{calcuateF}(\mu_{x},\mu_{y})$ ;AlgoLine0.8
end end end end
Algorithm 3 Determining the F-score for the various experimental parameters
8
8
8
8
8
8
8
8
The general expected trend is that as the density of the network increases,
the recall increases and the precision decreases. With more property values
being propagated, any metadata-poor record, on average, will receive more
recommendations than are needed. For instance, a manuscript only has one
publishing journal, therefore a recommendation of $100$ journals is going to
yield a very low precision ($0.01$). To balance this effect, a percentile
increase will tend to increase the precision of the algorithm at the expense
of recall. When only the highest energy recommendations are accepted, the
probability of rejecting a useful recommendation increases. In the case of
journal propagation, if only the 100th percentile recommendation is accepted,
then only the highest energy recommendation is accepted. If this journal
recommendation is the correct publishing venue, then there is 100% recall and
precision. If not, then there is 0% recall and precision. Depending on the
amount of values needed to fill a particular property, different $\rho$ values
will be most suitable than others.
### IV.3 The Results
This section presents the results of the experiment outlined previously in
Algorithm 3. For every associative network type and for every metadata type, a
F-score matrix was determined for every combination of $\partial$ (density)
and $\rho$ (percentile). These F-score values were calculated as the average
over $20$ different runs of the experiment. Tables 2 and 3 provide the max and
mean F-scores for each network/metadata pair over the entire $\partial$/$\rho$
set. Note that the bold faced values are those $\mu_{x}$/$\mu_{y}$ pairs for
which a landscape plot is provided. The italicized values are experimental
anomalies since the same metadata that was used to generate the associative
network is also the same metadata being propagated. For all other
combinations, metadata of a particular $\mu$ type exists to create an
associative network and metadata properties of a different $\mu$ type is being
propagated over those edges. For instance, a co-authorship network is used to
propagate citation property values.
network/metadata | author | citation | date | journal | keyword | organization
---|---|---|---|---|---|---
citation | 0.1829 | 0.1757 | 0.0606 | 0.2438 | 0.3913 | 0.2782
co-author | 0.6218 | 0.1300 | 0.0717 | 0.2630 | 0.2795 | 0.6457
co-citation | 0.0770 | 0.1821 | 0.0780 | 0.2081 | 0.2213 | 0.1350
co-keyword | 0.0073 | 0.0248 | 0.0472 | 0.1904 | 0.8689 | 0.0420
co-organization | 0.0709 | 0.0236 | 0.0508 | 0.1918 | 0.1180 | 0.5000
Table 2: Max F-scores network/metadata | author | citation | date | journal | keyword | organization
---|---|---|---|---|---|---
citation | 0.1367 | 0.1327 | 0.0441 | 0.2133 | 0.3246 | 0.2218
co-author | 0.2848 | 0.0780 | 0.0548 | 0.2004 | 0.1958 | 0.3935
co-citation | 0.0338 | 0.0697 | 0.0539 | 0.1554 | 0.1509 | 0.0768
co-keyword | 0.0032 | 0.0160 | 0.0385 | 0.1468 | 0.3240 | 0.0330
co-organization | 0.0312 | 0.0145 | 0.0392 | 0.1410 | 0.0909 | 0.1554
Table 3: Mean F-scores
The following landscape plots expose the relationship between $\partial$ and
$\rho$. A short explanation of the intuition behind each plot is also
provided.
Figure 3: Co-authorship network propagating a. citation
$F(\mathrm{cite},\mathrm{coauth})$ and b. journal
$F(\mathrm{jour},\mathrm{coauth})$ metadata
Intuitively, it makes sense that a co-authorship network would perform well
when propagating citation, journal, keyword, and organization property values
which are represented in Figure 3a, Figure 3b, Figure 4a, Figure 4b
respectively. The performance is a result of the fact that collaborating
authors tend to cite themselves, publish in similar journals, write about
similar topics, and are within similar organizations. Notice the effect that
percentile ($\rho$) has on Figure 3a as opposed to Figure 3b. Since there tend
to exist many citation property values (manuscripts cite many manuscripts),
lower percentile values ($\rho\approx 0$) ensures that there is a high recall.
When $\rho=1.0$, only the top citation is accepted and therefore the F-score
drops (very poor recall). On the other other hand, in Figure 4, when
$\rho=0.0$, there are many journal recommendations. This is not desirable
since a journal property only has one value (a manuscript is published in only
one venue). Therefore, at $\rho=1.0$, only one journal value is accepted into
the resource’s journal property. In situations where few property values are
expected, the F-score is best with a high $\rho$.
Figure 4: Co-authorship network propagating a. keyword
$F(\mathrm{key},\mathrm{coauth})$ and b. organization
$F(\mathrm{org},\mathrm{coauth})$ properties
A co-citation network, Figure 5, performs best with journal and keyword
properties. This means that manuscripts are likely to cite other manuscripts
with similar journal venues and since citation tends to be within the same
subject domain, the probability of similar keyword metadata increases. Again,
notice the effect of $\rho$ on journal metadata propagation. The shape of the
Figure 5a graph nearly mimics the shape of Figure 3b. Likewise, for Figure 5b
and Figure 4a. Again, the expected property value number is a major factor in
determining the system’s $\rho$ parameter.
Figure 5: Co-citation network propagating a. journal
$F(\mathrm{jour},\mathrm{cocite})$ and b. keyword
$F(\mathrm{key},\mathrm{cocite})$ properties
A citation network, like a co-citation network performs well for author,
journal, keyword, and organizational properties Figure 6 and Figure 8. It is
interesting to note how much better a citation network works for $\rho\approx
0.0$. Since a citation network isn’t symmetric, there is a chance that a
particle will reach a dead end. When a particle reaches a dead end, it no
longer recommends property values. Furthermore, citations are in a hierarchy
with more recent publications being at the top of the hierarchy (manuscripts
can not cite forward in time). Particles therefore trickle down the hierarchy
via a single, non-recurrent path from top to bottom. This “plinko ball” effect
is represented in Figure 7. The lack of recurrence in citation networks tends
to produce a high precision with a lower recall. High precision and low recall
is exactly what a low $\rho$ produces. Therefore, since the topology of the
citation network yields the same effect, the effect of $\rho$ as
$\rho\rightarrow 0.0$ isn’t as pronounced.
Figure 6: Citation network propagating a. author
$F(\mathrm{auth},\mathrm{cite})$ and b. journal
$F(\mathrm{jour},\mathrm{cite})$ properties Figure 7: Citation networks are
non-recurrent networks
As can be noticed from Table 2, Table 3, and Figure 8a, the keyword property
performs best in a citation network. A direct reference from one document to
another is a validation of the similarity between documents with respect to
subject domain. Therefore, the tendency for citing documents to contains
similar keyword values is high. For instance, refer to the citations of this
article (references in this manuscript’s bibliography). Every cited manuscript
is either about automatic metadata generation, bibliographic networks, or
network analysis.
Figure 8: Citation network propagating a. keyword
$F(\mathrm{key},\mathrm{cite})$ and b. organization
$F(\mathrm{org},\mathrm{cite})$ properties
A co-keyword network does not perform well for most property types except the
journal property, Figure 9a. This makes sense since manuscripts on similar
topics are likely to be published in similar journals.
Figure 9: Co-keyword and co-organization networks propagating a. journal
$F(\mathrm{jour},\mathrm{cokey})$ and b. $F(\mathrm{jour},\mathrm{coorg})$
properties, respectively
## V Future Work
This paper has provided a preliminary exploration of metadata generation in
terms of metadata property propagation within an associative network of
repository resources. Further research in this area may prove useful for other
network types such as those generated from other metadata properties. For
instance, it may be of interest to study the effect of this algorithm on usage
networks altern:bollen2005 . Usage metadata, unlike citation and journal
metadata, is applicable to every accessible resource. It would be interesting
to see what co-usage means for a particular genera of resources by determining
which metadata properties these networks are best at propagating.
A variety of propagation algorithms may also be explored. It is assumed that a
particle will take only edges of a particular $\mu$ type for the duration of
their life-span. Different path types might be an important aspect of
increasing the precision and recall performance of this method. For instance,
keyword metadata that first propagates over co-authorship edges and then over
co-citation edges might provide better results. Methods to implement such
propagation algorithms have been presented in socialgrammar:rodriguez2007 ;
grammar:rodriguez2007 . Also, different edge types can be merged such that all
co-keyword and co-authorship edges are collapsed to form a single edge.
What has been presented in this study is the results of this algorithm without
the intervention of any human components (besides the initial creation of
metadata through the hep-th dataset creation process). Future work that
studies this method with the inclusion of humans that help to validate and
“clean” the recommended metadata would be telling of how much this method is
able to speed up the process of generating accurate and reliable metadata for
metadata-poor resources. Such an analysis is left to future research.
Finally, multiplicative effects due to particle interaction may effect the
results of the algorithm. For instance, if two particles, $p_{i}$ and $p_{j}$,
meet at a particular node, $n_{k}$, and $p_{i}$ and $p_{j}$ have similar
metadata then the footprint they leave at $n_{k}$ should be more noticeable.
Because two different metadata sources are supplying the same property values,
there is an increased probability of that recommended metadata value being
correct. Currently, only a summation is being provided. It may be interesting
to multiply this summation by the number of unique particles that provided
energy for a particular recommended metadata value. The variations of this
preliminary framework will be explored in future work.
## VI Conclusion
Automatic metadata generation is becoming an increasingly important field of
research as digital library repositories become more prevalent and move into
the arena of less strongly controlled, decentralized collections (e.g. arXiv
and CiteSeer). The creation and maintenance of high-quality, detailed metadata
is hampered on numerous levels. Manual metadata creation methods are costly.
Recent efforts to leverage the collective power of social tagging (i.e.
“folksonomies”) may address some of the shortcomings of the manual creation of
metadata and result in viable models for online resources that do not require
strongly controlled vocabularies and metadata ontologies. However, it is
doubtful that “folksonomies” can be generalized to situations that require
vetted, well-standardized metadata. The automated creation of metadata on the
basis of content-analysis is a promising alternative to the manual creation of
metadata. It is conceivably more efficient in situations where textual data is
available and allows for more formal control of the type and nature of
metadata that is extracted. However, it can be unreliable for non-text
resources, yield low-quality metadata and can be computationally expensive.
This article proposed another possible component of the metadata generation
toolkit which may complement and support the above mentioned approaches.
Instead of creating new metadata, metadata is propagated from a metadata-rich
subset of the collection to similar, but metadata-poor subsets. The substrate
for this extrapolation is an associative network of resource relations created
from other available metadata. Metadata propagation may provide a
computationally feasible means of generating large amounts of metadata for
heterogeneous resources which can later be fine-tuned by manual intervention
or cross-validation with content-based methods. The article finally provided
experimental results using the High-Energy Physics bibliographic data set
(hep-th 2003). Human intervention may play an important role in fine-tuning
the metadata propagation algorithm. The results of this experiment are
promising and there still exists a range of potential modifications to this
basic framework that may lead to even better results.
## VII Acknowledgements
This research was financially supported by the Research Library at the Los
Alamos National Laboratory. The modified hep-th 2003 bibliographic dataset was
generously provided by Shou-de Lin and Jennifer H. Watkins provided editorial
assistance. Finally, the hep-th 2003 database is based on data from the arXiv
archive and the Stanford Linear Accelerator Center SPIRES-HEP database
provided for the 2003 KDD Cup competition with additional preparation
performed by the Knowledge Discovery Laboratory, University of Massachusetts
Amherst.
## References
* (1) Johan Bollen, Herbert Van de Sompel, Joan Smith, and Rick Luce. Toward alternative metrics of journal impact: a comparison of download and citation data. Information Processing and Management, 41(6):1419–1440, 2005.
* (2) Paul R. Cohen and Rick Kjeldsen. Information retrieval by constrained spreading activation in semantic networks. Information Processing and Management, 23(4):255–268, 1987.
* (3) A.M. Collins and E.F. Loftus. A spreading activation theory of semantic processing. Psychological Review, 82:407–428, 1975.
* (4) Fabio Crestani. Application of spreading activation techniques in information retrieval. Artificial Intelligence Review, 11(6):453–582, 1997.
* (5) Fabio Crestani and Puay Leng Lee. Searching the web by constrained spreading activation. Information Processing and Management, 36(4):585–605, 2000.
* (6) Erik Duval, Wayne Hodgins, Stuart Sutton, and Stuart L. Weibel. Metadata principles and practices. D-Lib Magazine, 8(4), 2002.
* (7) G. Giurida and J. Shek, E. Yang. Knowledge-based metadata extraction from postscript file. In Proceedings of the International Conference on Digital Libraries, pages 77–84. ACM, May 2000.
* (8) Scott A. Golder and Bernardo A. Huberman. Usage patterns of collaborative tagging systems. Journal of Information Science, 32(2):198–208, 2006.
* (9) Jane Greenburg. Metadata extraction and harvesting: A comparison of two automatic metadata generation applications. Journal of Internet Cataloging, 6(4):59–82, 2004.
* (10) Hui Han, C. Lee Giles, Eren Manavoglu, and Hongyuan Zha. Automatic document metadata extraction using support vector machines. In Proceedings of the Joint Conference on Digital Libraries (JCDL03), Huston, TX, May 2003. ACM.
* (11) Johnathan L. Herlocker, Joseph A. Konstan, Loren G. Terveen, and John T. Riedl. Evaluating collaborative filtering recommender systems. ACM Transactions on Information Systems, 22(1):5–53, 2004.
* (12) Zan Huang, Hsinchun Chen, and Daniel Zeng. Applying associative retrieval techniques to alleviate the sparsity problem in collaborative filtering. ACM Trans. Inf. Syst., 22(1):116–142, 2004.
* (13) Zan Huang, Xin Li, and Hsinchun Chen. Link prediction approach to collaborative filtering. In Proceedings of the Joint Conference on Digital Libraries (JCDL05), Denver, CO, June 2005. ACM.
* (14) H. Kuwano, Y. Matsuo, and K. Kawazoe. Reducing the cost of metadata generation by using video/audio indexing and natural language processing techniques. NTT Technical Review, pages 68–74, 2004.
* (15) David Liben-Nowell and Jon Kleinberg. The link prediction problem for social networks. In Twelfth International Conference on Information and Knowledge Management, pages 556–559. ACM, November 2003.
* (16) Shou-de Lin and Hans Chalupsky. Issues of verification for unsupervised discovery systems. In KDD04 Workshop on Link Discovery, Seattle, WA, 2004\.
* (17) Song Mao, Jong Woo Kim, and George R. Thoma. A dynamic feature generation system for automated metadata extraction in preservation of digital materials. In First International Workshop on Document Image Analysis for Libraries DIAL’04. IEEE, January 2004.
* (18) Adam Mathes. Folksonomies - cooperative classification and communication through shared metadata. Computer Mediated Communication - LIS590CMC (graduate course), December 2004.
* (19) A. McGovern, L. Friedland, M. Hay, B. Gallagher, A. Fast, J. Neville, and D. Jensen. Exploiting relational structure to understand publication patterns in high-energy physics. ACM SIGKDD Explorations Newsletter, 5(2):165–172, December 2003\.
* (20) Mor Naaman, Ron B. Yeh, Hector Garcia-Molina, and Andreas Paepcke. Leveraging context to resolve identity in photo albums. In Proceedings of the 5th Joint Conference on Digital Libraries (JCDL 05), Denver, CO, June 2005.
* (21) Camille Prime-Claverie, Michel Beigbeder, and Thierry Lafouge. Metadata propagation in the web using co-citations. In Proceedings of the 2005 IEEE/ACM International Conference on Web Intelligence, Compiegne University of Technology, France, September 2005. IEEE Computer Society.
* (22) Marko A. Rodriguez. Social decision making with multi-relational networks and grammar-based particle swarms. In Proceedings of the Hawaii International Conference on Systems Science, pages 39–49, Waikoloa, Hawaii, January 2007. IEEE Computer Society.
* (23) Marko A. Rodriguez. Grammar-based random walkers in semantic networks. Knowledge-Based Systems, 21(7):727–739, 2008.
* (24) Fabrizio Sebastiani. Machine learning in automated text categorization. ACM Computing Surveys, 34(1):1–47, 2002.
* (25) Jewel Ward. A quantitative analysis of unqualified dublin core metadata element set usage within data providers registered with the open archives initiative. In JCDL ’03: Proceedings of the 3rd ACM/IEEE-CS joint conference on Digital libraries, pages 315–317, Washington, DC, USA, 2003. IEEE Computer Society.
* (26) Hsin-Chang Yang and Chung-Hong Lee. Automatic metadata generation for web pages using a text mining approach. In International Workshop on Challenges in Web Information Retrieval and Integration, pages 186–194. IEEE, April 2005\.
| arxiv-papers | 2008-06-30T21:23:28 | 2024-09-04T02:48:56.467502 | {
"license": "Public Domain",
"authors": "Marko A. Rodriguez, Johan Bollen, Herbert Van de Sompel",
"submitter": "Marko A. Rodriguez",
"url": "https://arxiv.org/abs/0807.0023"
} |
0807.0040 | # Bulge $n$ and $B/T$ in High Mass Galaxies: Constraints on the Origin of
Bulges in Hierarchical Models
Tim Weinzirl11affiliation: Department of Astronomy, University of Texas at
Austin, Austin, TX , Shardha Jogee11affiliation: Department of Astronomy,
University of Texas at Austin, Austin, TX , Sadegh Khochfar22affiliation:
Sub-Department of Astrophysics, University of Oxford, Denys Wilkinson Bldg.,
Keble Road, OX1 3RH, Oxford, UK 33affiliation: Max Planck Institut für
extraterrestrische Physik, P.O Box 1312, D-85478 Garching, Germany , Andreas
Burkert44affiliation: Universitäts-Sternwarte München, Scheinerstr. 1, 81679
München, Germany , John Kormendy11affiliation: Department of Astronomy,
University of Texas at Austin, Austin, TX
###### Abstract
We use the bulge Sérsic index $n$ and bulge-to-total ratio ($B/T$) to explore
the fundamental question of how bulges form. We perform $2D$ bulge-disk-bar
decomposition on $H$-band images of 143 bright, high mass ($M_{\star}\geq
1.0\times 10^{10}M_{\odot}$) low-to-moderately inclined ($i<70^{\circ}$)
spirals. Our results are: (1) Our $H$-band bar fraction ($\sim 58\%$) is
consistent with that from ellipse fits. (2) 70% of the stellar mass is in
disks, 10% in bars, and 20% in bulges. (3) A large fraction ($\sim$ 69%) of
bright spirals have $B/T\leq$ 0.2, and $\sim$ 76% have low $n\leq 2$ bulges.
These bulges exist in barred and unbarred galaxies across a wide range of
Hubble types. (4) About 65% (68%) of bright spirals with $n\leq 2$ ($B/T\leq
0.2$) bulges host bars, suggesting a possible link between bars and bulges.
(5) We compare the results with predictions from a set of $\Lambda$CDM models.
In the models, a high mass spiral can have a bulge with a present-day low
$B/T\leq$ 0.2 only if it did not undergo a major merger since $z\leq 2$. The
predicted fraction ($\sim$ 1.6%) of high mass spirals, which have undergone a
major merger since $z\leq 4$ and host a bulge with a present-day low $B/T\leq
0.2$, is a factor of over thirty smaller than the observed fraction ($\sim
66\%$) of high mass spirals with $B/T\leq 0.2$. Thus, contrary to common
perception, bulges built via major mergers since $z\leq 4$ seriously fail to
account for the bulges present in $\sim 66\%$ of high mass spirals. Most of
these present-day low $B/T\leq 0.2$ bulges are likely to have been built by a
combination of minor mergers and/or secular processes since $z\leq 4$.
galaxies: bulges — galaxies: evolution — galaxies: formation — galaxies:
fundamental parameters — galaxies: interactions — galaxies: structure
## 1 Introduction
The formation of galaxies is a classic problem in astrophysics. Contemporary
galaxy formation models combine the well-established $\Lambda$ Cold Dark
Matter ($\Lambda$CDM) cosmology, which describes behavior of dark matter on
very large scales, with baryonic physics to model galaxy formation. In the
early Universe, pockets of dark matter decoupled from the Hubble flow,
collapsed into virialized halos, and then clustered hierarchically into larger
structures. Meanwhile, gas aggregated in the interiors of the halos to form
rotating disks, which are the building blocks of galaxies (Steinmetz & Navarro
2002; Cole et al. 2000). Such disks are typically destroyed during major
mergers of galaxies with mass ratio $M_{1}/M_{2}>1/4$ (e.g. Steinmetz &
Navarro 2002; Naab & Burkert 2003; Burkert & Naab 2004; Mihos & Hernquist
1996). When the mass ratio is close to unity, the remnant is a spheroid with
properties close to that of a classical bulge, namely a steep de Vaucouleurs
$r^{1/4}$ surface brightness profile and a high ratio of ordered-to-random
motion ($v/\sigma$). We shall return to this point in $\S$ 5. Within this
hierarchical framework, the disk of spiral galaxies forms when gas of higher
specific angular momentum subsequently accretes around the bulge (Steinmetz &
Navarro 2002; Burkert & Naab 2004).
$\Lambda$CDM-based simulations of galaxy formation face several challenges.
One issue is the angular momentum problem; simulated galaxy disks have smaller
scalelengths and, therefore, less specific angular momentum than their
counterparts in nature (Navarro & Steinmetz 2000; Burkert & D’Onghia 2004;
D’Onghia et al. 2006). A second issue is the problem of bulgeless or low
bulge-to-total mass ratio ($B/T$) spirals. Within the $\Lambda$CDM paradigm,
galaxies that had a past major merger at a time when its mass was a fairly
large fraction of its present-day mass are expected to have a significant
bulge with large $B/T$ and high Sérsic index. Depending on the merger history
and hence the fraction of spiral galaxies that fulfill this criterion (see
$\S$ 5.8) we can end up with a small or large fraction of present-day galaxies
with low $B/T$.
There is rising evidence that low $B/T$ and bulgeless galaxies are quite
common in the local Universe, especially in low mass or late-type galaxies.
Late-type Sd galaxies often harbor no bulge (Böker et al. 2002). Kautsch et
al. (2006) and Barazza, Jogee & Marinova (2007, 2008) also find from the
analysis of several thousand late-type SDSS galaxies that 15-20% of such disk
galaxies out to $z\sim 0.03$ appear bulgeless. Of the 19 local galaxies ($D<8$
Mpc) with circular velocity $V_{c}>150$ km s-1, 11 (58%) have pseudobulges
instead of merger-built classical bulges (Kormendy & Fisher 2008).
Theoretical work by Koda et al. (2007) conclude the survival of disk-dominated
systems in a $\Lambda$CDM universe is compatible with observational
constraints provided classical bulges form only in mergers where
$M_{1}/M_{2}>0.3$ and the primary halo has virial velocity $V_{vir}>55$ km
s-1.
Evidence also suggests that bulges with low $B/T$ and low Sérsic index $n$ may
be common even in high mass and/or early-type spirals. Balcells et al. (2003)
report that early-type disk galaxies tend to have $n<$ 3 and often from 1 to
2. Laurikainen et al. (2007) find barred and unbarred early-type disk galaxies
to have mean $B/T$ between 0.25 and 0.35, while later Hubble types have
$B/T<0.2$; they also find mean bulge Sérsic indices to be $\sim 2.5$ or less
across the Hubble sequence. Graham & Worley (2008) report low $B/D$ ratios
across the Hubble sequence based on bulge-disk decomposition of $K$-band
images of local spiral galaxies. They suggest that these low values are
problematic for $\Lambda$CDM simulations, but no quantitative assessment of
the extent of the problems is presented.
These emerging statistics on the fraction of bulgeless ($B/T\sim 0$) galaxies,
and galaxies with low $B/T$ and low $n$ bulges provide important first
constraints. More work is needed to fully explore the the distribution of
bulge properties in both high and low mass galaxies. In particular, we need to
explore how the observed distributions of bulge $B/T$ and $n$ compare with the
predictions from $\Lambda$CDM-based simulations of galaxy evolution. To the
best of our knowledge, few such quantitative comparisons have been attempted,
so that it remains unclear how serious the problem of low $B/T$ galaxies is.
This study is an attempt to derive robust observational constraints on bulge
properties in high mass spirals and to attempt such a comparison with models.
Completely resolving the issue of low $B/T$ systems will require understanding
the different types of bulges and their formation pathways. Bulges are
commonly divided in several groups: classical bulges, boxy/peanut bulges, and
‘pseudobulges’ or disky bulges. Classical bulges are believed to be built by
major mergers ($M_{1}/M_{2}\geq 1/4$) and the associated violent relaxation of
stars. They are associated with modest-to-high bulge Sérsic indices, in the
range $2<n<6$ (Hopkins et al. 2008; Springel et al. 2005; Robertson et al.
2006; $\S$ 5.8). Boxy/peanut bulges are believed to be the result of vertical
resonances and buckling instabilities in bars, which are viewed at high
inclination (Combes & Sanders 1981; Combes et al. 1990; Pfenniger & Norman
1990; Bureau & Athanassoula 2005; Athanassoula 2005; Martinez-Valpuesta et al.
2006). Pseudobulges or disky bulges are believed to form as a result of gas
inflow into the central kiloparsec and subsequent star formation building a
compact disky, high $v/\sigma$ stellar component (Kormendy 1993; Jogee 1999;
Kormendy & Kennicutt 2004, hereafter KK04; Jogee, Scoville, & Kenney 2005;
Athanassoula 2005; Kormendy & Fisher 2005). Pseudobulges tend to have a bulge
$n<2.5$ (Kormendy & Fisher 2005; Fisher & Drory 2008).
One possibility for the formation of disky bulges or pseudobulges is the idea
of secular evolution (Kormendy 1993; KK04; Jogee, Scoville, & Kenney 2005),
where a stellar bar or globally oval structure in a non-interacting galaxy
drives the gas inflow into the inner kpc via shocks and gravitational torque.
Another idea for building disky bulges is that the gas inflow into the inner
kiloparsec is driven by external non-secular processes, such as tidal
interaction and minor mergers. The gas inflow in such cases can be caused by a
tidally induced non-axisymmetric feature, such as a bar (e.g., Quinn et al.
1993; Hernquist & Mihos 1995), and by tidal torques from the companion. The
subsequent central star formation can still form a compact high $v/\sigma$
stellar component, $aka$ a pseudobulge.
Throughout this paper, we avoid making any a priori assumptions about the
origin of different types of bulges by simply referring to them according to
their bulge Sérsic index $n$ or bulge-to-total mass ratio ($B/T$). We consider
bulges of high ($n\geq$ 4), intermediate ($2<n<4$) and low ($n\leq 2$) index,
as well as those of low or high $B/T$.
The structural properties of galaxy components, such as bulges, disks, and
bars can be derived through the decomposition of the $2D$ light distribution,
taking into account the PSF. Many early studies have performed only two
component $2D$ bulge-disk decomposition (e.g., Allen et al. 2006; Byun &
Freeman 1995; de Jong 1996; Simard 1998; Wadadekar et al. 1999), ignoring the
contribution of the bar, even in strongly barred galaxies. However, recent
work has shown that it is important to include the bar in $2D$ decomposition
of barred galaxies, else the $B/T$ ratio can be artificially inflated, and
bulge properties skewed (e.g., Laurikainen et al. 2005, 2007). Furthermore,
since most ($\geq 60\%$) bright spiral galaxies are barred in the NIR
(Eskridge et al. 2000; Laurikainen et al. 2004; Marinova & Jogee 2007,
hereafter MJ07; Menendez-Delmestre et al. 2007), the inclusion of the bar is
quite important. This has led to several recent studies, where $2D$ bulge-
disk-bar decomposition are being performed (e.g. Laurikainen et al. 2007;
Reese et al. 2007; Gadotti & Kauffmann 2007).
Another advantage of bulge-disk-bar decomposition over bulge-disk
decomposition is that the former allows us to constrain the properties of the
bar itself. Bars provide the most important internal mechanism for
redistributing angular momentum in baryonic and dark matter components (e.g.
Weinberg 1985; Debattista & Sellwood 1998, 2000; Athanassoula 2002; Berentzen,
Shlosman, & Jogee 2006). They efficiently drive gas inflows into the central
kpc, feed central starbursts (Elmegreen 1994; Knapen et al. 1995; Hunt &
Malakan 1999; Jogee et al. 1999; Jogee, Scoville, & Kenney 2005; Jogee 2006)
and lead to the formation of disky or pseudobulges (see above). Furthermore,
the prominence of strong bars out to $z\sim 1$ over the last 8 Gyr (Jogee et
al. 2004; Sheth et al. 2008) suggest that bars have been present over
cosmological times and can shape the dynamical and secular evolution of disks.
Thus, quantifying bar properties, such as the fractional light and mass ratio
(Bar/$T$), can yield insight into these processes.
In this paper, we constrain the properties of bulges and bars along the Hubble
sequence, and compare our results with $\Lambda$CDM-based simulations of
galaxy evolution. In $\S$ 2, we define our complete sample of 143 bright
($M_{B}\leq-19.3$) low-to-moderately inclined ($i\leq 70^{\circ}$) spirals
from the Ohio State University Bright Spiral Galaxy Survey (OSUBSGS; Eskridge
et al. 2002), which is widely used as the local reference sample for bright
spirals by numerous studies (e.g., Eskridge et al. 2000; Block et al. 2002;
Buta et al. 2005; MJ07 ; Laurikainen et al. 2004, 2007). In $\S$ 3, we perform
$2D$ bulge-disk and bulge-disk-bar decompositions of $H$-band images using
GALFIT (Peng et al. 2002), and derive fractional light ratios ($B/T$, Bar/$T$,
Disk/$T$), as well as Sérsic indices and half light radii or scale lengths.
Tests to verify the robustness of our decompositions are presented in $\S$ 4.
In $\S$ 5, we present our results. Specifically, the total stellar mass
present in bulges, disks, and bars is calculated ($\S$ 5.2). In $\S$ 5.3, the
distribution of bulge Sérsic index $n$ and $B/T$ as a function of galaxy
Hubble type and stellar mass is presented, and the surprising prevalence of
bulges with low Sérsic index $n$ and low $B/T$ is established. A comparison
with other works is presented in $\S$ 5.4. We examine how Bar/$T$ and bar
fraction (the fraction barred disks) change as a function of host galaxy
properties in $\S$ 5.5. In $\S$ 5.8, we compare our observed distribution of
bulge $B/T$ and $n$ in high mass ($M_{\star}\geq 1.0\times 10^{10}M_{\odot}$)
spirals with predictions from $\Lambda$CDM cosmological semi-analytical
models. $\S$ 6 summarizes our results.
## 2 Sample Properties
### 2.1 OSUBSGS
Our dataset is derived from the 182 $H$-band images from the public data
release of the Ohio State University Bright Spiral Galaxy Survey (OSUBSGS;
Eskridge et al. 2002). These galaxies are a subset of the RC3 catalog that
have $m_{B}\leq 12$, Hubble types $0\leq T\leq 9$ (S0/a to Sm), $D_{25}\leq
6^{\prime}.5$, and $-80^{\circ}<\delta<+50^{\circ}$. Imaging of OSUBSGS
galaxies spans optical and near infrared (NIR) wavelengths with $BVRJHK$
images available for most galaxies. OSUBSGS images were acquired on a wide
range of telescopes with apertures ranging from 0.9-2.4m. The $JHK$ data were
acquired with a variety of telescopes and detectors, but mainly with the 1.8 m
Perkins reflector at Lowell Observatory and the CTIO 1.5 m telescope with the
OSIRIS detector, having 18.5 micron pixels (Depoy et al. 1993). Pixel scale is
dependent on the telescope and for these observations ranged between
$1-1.50^{\prime\prime}$/pix. Exposure times were heterogeneous, but the total
observing time per object was typically between 10-15 minutes in $H$. The
resulting limiting $H$-band surface brightness is $\sim 20$ mag arcsec-2. The
typical limiting surface brightnesses of the images $\sim 26$ mag arcsec-2 in
$B$-band and $\sim 20$ mag arcsec-2 in $H$-band (Eskridge et al. 2002). The
seeing depends on observing time and location. We find the $H$-band images
have seeing of $\sim 3^{\prime\prime}$.
We choose to use the NIR images rather than optical ones for several reasons.
Firstly, NIR images are better tracers of the stellar mass than optical
images, and the mass-to-light ratio is less affected by age or dust gradients.
Secondly, obscuration by dust and SF are minimized in the NIR, compared to the
optical. As the $K$-band images are of poor quality, we settle on using the
$H$-band images.
The OSUBSGS is widely used as the local reference sample for bright spirals by
numerous studies (e.g. Eskridge et al. 2000; Block et al. 2002; Buta et al.
2005; MJ07; Laurikainen et al. 2004, 2007). Thus, there are numerous
complementary results that we can use or compare with. In particular, MJ07
have identified bars in this sample using quantitative criteria based on
ellipse fitting, and characterized their sizes, position angles, and
ellipticities.
OSUBSGS is a magnitude-limited survey with objects whose distances range up to
$\sim 60$ Mpc. Faint galaxies are inevitably missed at larger distances,
resulting in the absolute magnitude distribution in Figure 1. We compare the
$B$-band LF of this sample with a Schechter (Schechter 1976) LF (SLF) with
$\Phi^{*}=5.488\times 10^{-3}$ Mpc-3, $\alpha=-1.07$, and $M^{*}_{B}=-20.5$
(Efstathiou, Ellis & Peterson 1988) in Figure 2. The volume used to determine
the number density in each magnitude bin is
$V_{max}=\frac{4\pi}{3}d^{3}_{max}(M),$ (1)
where
$d_{max}(M)=10^{1+0.2(m_{c}-M)}$ (2)
is the maximum distance out to which a galaxy of absolute magnitude $M$ can be
observed given the cutoff magnitude $m_{c}$. If the SLF is representative of
the true LF, then Figure 2 suggests that the OSUBSGS sample is seriously
incomplete at $M_{B}>-19.3$, while at the brighter end (-19.3 to -23) the
shape of its LF matches fairly well the SLF. We thus conclude that the sample
is reasonably complete for bright ($M_{B}\leq-19.3$ or $L_{B}>$ 0.33 $L^{*}$)
galaxies.
We exclude highly inclined ($i>70^{\circ}$) galaxies for which structural
decomposition does not yield accurate results. Thus, our final sample S1
consists of 143 bright ($M_{B}\leq-19.3$) low-to-moderately inclined ($i\leq
70^{\circ}$) spirals with Hubble types mainly in the range S0/a to Sc (Figure
1). Of the 126 for which we could derive stellar masses (see $\S$ 2.2, most
have stellar masses $M_{\star}\geq 1.0\times 10^{10}M_{\odot}$ (Figure 3).
Table 1 summarizes the morphologies, luminosities, and stellar masses of the
sample. Note that there are few galaxies of late Hubble types (Scd or later)
and we do not draw any conclusions on such systems from our study. In a future
paper, we will tackle galaxies of lower mass and later Hubble types.
### 2.2 Stellar Masses
We derive global stellar masses for most of the OSUBSGS sample galaxies using
the relation between stellar mass and rest-frame $B-V$ color from Bell et al.
(2003). Using population synthesis models, the latter study calculates stellar
$M/L$ ratio as a function of color using functions of the form
$log_{10}(M/L)=a_{\lambda}+b_{\lambda}\times Color+C$, where $a_{\lambda}$ and
$b_{\lambda}$ are bandpass dependent constants and C is a constant that
depends on the stellar initial mass function (IMF). For the $V$ band Bell et
al. (2003) find $a_{\lambda}=-0.628$ and $b_{\lambda}=1.305$; assuming a
Kroupa (1993) IMF, they find $C=-0.10$. This yields an expression for the
stellar mass in $M_{\odot}$ for a given $B-V$ color:
$M_{\star}=v_{lum}10^{-0.628+1.305(B-V)-0.10},$ (3)
where
$v_{lum}=10^{-0.4(V-4.82)}.$ (4)
Here, $v_{lum}$ is the luminosity parametrized in terms of absolute $V$
magnitude.
How reliable are stellar masses determined from this procedure? Clearly, the
above relationship between $M_{\star}$ and $B-V$ cannot apply to all galaxies,
and must depend on the assumed stellar IMF, and range of ages, dust, and
metallicity. However, it is encouraging to note that several studies (Bell et
al. 2003; Drory et al. 2004; Salucci, Yegorova, & Drory 2008) find generally
good agreement between masses based on broad-band colors and those from
spectroscopic (e.g. Kauffmann et al. 2003) and dynamical (Drory et al. 2004)
techniques. Typical errors are within a factor of two to three. Salucci et al.
(2008) derive disk masses with both photometric and kinematic methods and find
the two methods are equivalent on average. For a sample of 18, they find an
rms scatter of 0.23 dex, while on an individual basis the deviation can be as
high as 0.5 dex.
We used this relation to compute stellar masses for 126 of 143 (88%) objects.
The remainder did not have $B-V$ colors available in the Hyperleda database or
RC3. The mass distribution is summarized in Figure 3. Individual masses are
listed in Table 1. This sample of 126 galaxies is referenced henceforth as
sample S2.
## 3 Method and Analysis
The structural properties of galaxy components, such as bulges, disks, and
bars can be derived through the decomposition of the 2D light distribution,
taking into account the PSF. There are several algorithms for $2D$ luminosity
decomposition, including GIM2D (Simard et al. 2002), GALFIT (Peng et al.
2002), and BUDDA (de Souza et al. 2004). The latter two allow bulge-disk-bar
decomposition, while the former only allows bulge-disk decomposition.
Most previous work has addressed $2D$ bulge-disk decomposition only. Allen et
al. (2006), for example, performed bulge-disk decomposition of $B$-band images
with GIM2D on 10,095 galaxies from the Millennium Galaxy Catalog (Liske et al.
2003; Driver et al. 2005). However, recent work (e.g., Laurikainen et al.
2005; Graham & Balcells, in preparation) has shown that the $B/T$ ratio can be
artificially inflated in a barred galaxy unless the bar component is included
in the $2D$ decomposition. The fact that most ($\geq 60\%$) bright spiral
galaxies are barred in the NIR (Eskridge et al. 2000; Laurikainen et al. 2004;
MJ07; Menendez-Delmestre et al. 2007), further warrants the inclusion of the
bar. Another advantage of bulge-disk-bar decomposition is that it allows us to
constrain the properties of the bar itself, and to constrain scenarios of bar-
driven evolution (see $\S$ 1).
Motivated by these considerations, several studies have tackled the problem of
$2D$ bulge-disk-bar decomposition. Laurikainen et al. (2005, 2007) have
developed a $2D$ multicomponent decomposition code designed to model bulges,
disks, primary and secondary bars, and lenses; they apply Sérsic functions to
bulges and use either Sérsic or Ferrers functions to describe bars and lenses.
Reese et al. (2007) have written a non-parametric algorithm to model bars in
$\sim 70$ $I$-band images. Gadotti & Kauffmann (2007) are performing $2D$
bulge-disk-bar and bulge-disk decomposition of 1000 barred and unbarred
galaxies from SDSS with the BUDDA software.
In this study, we perform $2D$ two-component bulge-disk decomposition and
three-component bulge-disk-bar decomposition of the OSUBSGS sample with
GALFIT. We note that Laurikainen et al. (2007) have also performed bulge-disk-
bar decomposition on the OSUBSGS sample. However, there are also important
complementary differences between our study and theirs. The decomposition
algorithm and tests on the robustness performed in our study are different
(see $\S$ 3 and $\S$ 4). Furthermore, unlike Laurikainen et al. (2007), we
also compare the bulge-to-total ratio ($B/T$) with predictions from
hierarchical models of galaxy evolution ($\S$ 5), and also present the
distribution of bar-to-total ratio (Bar/$T$).
### 3.1 Image Preparation
Running GALFIT on an image requires initial preparation. The desired fitting
region and sky background must be known, and the PSF image, bad pixel mask (if
needed), and pixel noise map must be generated. We addressed these issues as
follows: (1) The GALFIT fitting region must be large enough to include the
outer galaxy disk, as well as some sky region. Since cutting out empty regions
of sky can drastically reduce GALFIT run-time, a balance was sought between
including the entire galaxy and some decent sky region, while excluding large
extraneous blank sky areas. (2) It is possible for GALFIT to fit the sky
background, but this is not recommended. When the sky is a free parameter, the
wings of the bulge Sérsic profile can become inappropriately extended,
resulting in a Sérsic index that is too high. Sky backgrounds were measured
separately and designated as fixed parameters; (3) GALFIT requires a PSF image
to correct for seeing effects. Statistics of many stars in each frame can be
used to determine an average PSF. However, many of our images contain merely a
few stars. Instead, a high S/N star from each frame was used as a PSF. (4) We
used ordered lists of pixel coordinates to make bad pixel masks, which are
useful for blocking out bright stars and other image artifacts. (5) We had
GALFIT internally calculate pixel noise maps for an image from the noise
associated with each pixel. Noise values are determined from image header
information concerning gain, read noise, exposure time, and the number of
combined exposures.
### 3.2 Decomposition Steps
Figure 4 summarizes our method of decomposition, which we now detail. GALFIT
requires initial guesses for each component it fits. It uses a Levenberg-
Marquardt downhill-gradient algorithm to determine the minimum $\chi^{2}$
based on the input guesses. GALFIT continues iterating until the $\chi^{2}$
changes by less than 5e-04 for five iterations (Peng et al. 2002). We
recognize that a drawback to any least-squares method is that a local minimum,
rather than a global minimum, in $\chi^{2}$ space may be converged upon. We
explore this possibility with multiple tests described in § 4. We adopted an
iterative process, involving three separate invocations of GALFIT, to perform
1-component, 2-component, and 3-component decomposition:
1. 1.
Stage 1 (single Sérsic): In Stage 1, a single Sérsic component is fitted to
the galaxy. This serves the purpose of measuring the total luminosity, which
is conserved in later Stages, and the centroid of the galaxy, which is
invariant in later fits.
2. 2.
Stage 2 (exponential plus Sérsic): In Stage 2, the image is fit with the sum
of an exponential disk and a Sérsic component. During the Stage 2 fit, the
disk $b/a$ and $PA$ are held constant at values, which we take from the
published ellipse fits of MJ07, as well as ellipse fits of our own. This
procedure reduces the number of free parameters in the fit by fixing the disk
$b/a$ and $PA$, which are easily measurable parameters. It also prevents
GALFIT from confusing the disk and bar, and artificially stretching the disk
along the bar PA in an attempt to mimic the bar. As initial guesses for the
Sérsic component in Stage 2, the output of Stage 1 is used. The Sérsic
component in Stage 2 usually represents the bulge, in which case Stage 2
corresponds to a standard bulge-disk decomposition.
However, in a few rare cases, where the galaxy has just a bar and a disk, the
Sérsic component in Stage 2 represents a bar. The latter is recognizable by a
low Sérsic index and large half-light radius.
3. 3.
Stage 3 (exponential plus two Sérsic components): In Stage 3, a three-
component model consisting of an exponential disk, a Sérsic bulge, and a
Sérsic bar is fit. As suggested by Peng et al. (2002), the bar can be well
described by an elongated, low-index Sérsic ($n<1$) profile. As in Stage 2,
the disk $b/a$ and $PA$ are held constant at values predetermined from ellipse
fits. We provide initial guesses for the bar $b/a$ and $PA$, based on ellipse
fits of the images from MJ07 or analysis of the images in DS9. We provide
GALFIT with input guesses for the bulge parameters, based on the output from
Stage 2. In principle, it is also possible to generate reasonable guess
parameters for the bulge and disk from a bulge-disk decomposition on a $1D$
profile taken along a select PA. As described in $\S$ 4.3, we also experiment
with initial guesses derived in this way, and find that the final convergence
solution is the same. We also note that GALFIT fixes the bulge $b/a$ and does
not allow it to vary with radius, while real bulges may have a varying $b/a$.
We tested the impact of fixed and varying bulge $b/a$ on the derived $B/T$
($\S$ 4.1) and find that there is no significant change in $B/T$.
For objects with central point sources, the bulge Sérsic index in the Stage 2
and Stage 3 models can be inadvertently overestimated unless an extra nuclear
component is added to the model. Balcells et al. (2003) note that for galaxies
imaged both from the ground and $HST$, the combination of unresolved central
sources and seeing effects mimic high-index bulge Sérsic profiles in the
ground images. Depending on sample and resolution, the frequency of central
sources can range from 50% to 90% (Ravindranath et al. 2001; Böker et al.
2002; Balcells et al. 2003). Ravindranath et al. (2001) find a frequency of
50% in early type (E, S0, S0/a) galaxies, while Böker et al. (2002) measure a
frequency of 75% for spirals with Hubble types Scd to Sm. Balcells et al.
(2003) determine a frequency of 84% for S0-Sbc galaxies imaged with $HST$. Our
dataset most closely resembles the latter sample, so we might expect that, as
an upper limit, a similar fraction of our galaxies will need to be corrected
with an extra compact component.
We added point sources as third or fourth components to the initial models.
For those cases where the model successfully converged with the extra
component, the images were visually inspected to verify the presence of a
central bright source. Sometimes, the model converged to significantly
different and unreasonable parameters for all components. Other times, the
model would converge to a very dim point source without changing any of the
other model parameters. Where new model parameters were not unreasonable and
not identical to the case without the point source, the new model was adopted.
This was the case for 111 of 143 (78%) of our sample. The point sources
contribute less than 1%, 3%, and 5% of the total luminosity 55%, 86%, and 95%
of the time, respectively. As the point sources take up such a small fraction
of the light distribution, their contribution is folded back into $B/T$ in all
cases where a point source was modeled. Inclusion of the point source reduced
bulge index by $\sim 0.8$ for both barred and unbarred galaxies. Such a change
is expected based on the above discussion. The decline in bulge index caused a
minor decrease in $B/T$; on the mean, this change was 1.04% for barred
galaxies and 0.32% for unbarred galaxies. For barred galaxies, this light most
primarily added to $D/T$ rather than Bar/$T$.
It is important to recognize the physical significance of the added nuclear
components. We began by determining which objects show evidence for AGN
activity. The sample was checked against the catalog of Ho et al. (1997), the
Véron Catalog of Quasars & AGN, 12th Edition (Véron-Cetty & Véron 2006), and
NED. Of the 111 objects fit with point sources, 43 (39%) contain AGN. An
additional 20 (18%) possess HII nuclei according to Ho et al. (1997) and
visibly show bright compact nuclei. The remaining 48 (43%) probably contain
neither AGN nor HII nuclei, but could house nuclear star clusters. For these
objects, we visually inspected the images to ensure there was a bright compact
source at the center. We are confident that the fitted point source components
have physical counterparts in the data images.
GALFIT also allows a diskiness/boxiness parameter to be added to any Sérsic or
exponential profile. We did not use this parameter for any bulge or disk
profiles. Bars in general have boxy isophotes, and we could have included the
diskiness/boxiness parameter in the bar profiles. However, it was found that
adding boxiness to the bar did not change any model parameters, including
fractional luminosities $B/T$, $D/T$, and Bar/$T$, by more than a small
percentage, even though the appearance of the residual images improved in some
cases due to the change in bar shape. Consequently, we chose to neglect bar
boxiness altogether.
### 3.3 Choosing the Best Fit Between Stage 2 and Stage 3
All objects in our sample were subjected to Stages 1, 2, and 3. Depending on
whether a galaxy with a bulge is unbarred or barred, its best fit should be
taken from the Stage 2 bulge-disk decomposition or the Stage 3 bulge-disk-bar
decomposition, respectively. For objects with prominent bars, it is obvious
that the Stage 3 model provides the best fit. However, it is more difficult to
decide between Stage 2 versus Stage 3 fits in galaxies which host weak bars
with no strong visual signature. In practice, we therefore applied the set of
criteria below to each galaxy in order to select between the Stage 2 bulge-
disk decomposition and Stage 3 bulge-disk-bar decomposition. Table 1 lists the
model chosen for each galaxy. Table 2 summarizes the model parameters from the
best fits.
For completeness, we note that for the few rare galaxies (see $\S$ 3.2) that
have just a bar and a disk, the choice of a final solution is between the
Stage 2 bar-disk decomposition and Stage 3 bulge-disk-bar decomposition. The
same guidelines below can be used to identify the best model.
1. 1.
GALFIT calculates a $\chi^{2}$ and $\chi^{2}_{\nu}$ for each model. It was
found that $\chi^{2}$ almost universally declines between the Stage 2 and
Stage 3 fits for a given object. This is because in the Stage 3 fit, five
extra free parameters (bar luminosity, $r_{e}$, Sérsic index, $b/a$, and $PA$)
are added with the Sérsic bar component, allowing GALFIT to almost always make
a lower $\chi^{2}$ model during Stage 3. However, this does not necessarily
mean that the solution in Stage 3 is more correct physically. Thus, an
increasing $\chi^{2}$ was interpreted as a sign that the Stage 3 fit should
not be adopted, but a decreasing $\chi^{2}$ was not considered as a sufficient
condition to adopt Stage 3.
2. 2.
In cases with prominent bars, a symmetric light distribution due to
unsubtracted bar light was often found in the Stage 1 and Stage 2 bulge-disk
residuals. This was strong evidence that the Stage 3 bulge-disk-bar fit be
selected. NGC 4643 is shown in Figure 5 because it has a particularly striking
bar residual; the corresponding fit parameters appear in Table 3. Note that in
all figures and tables, we adopt the convention that PA values are
positive/negative if they are measured from North counterclockwise/clockwise.
3. 3.
The Stage 2 and Stage 3 models were selected only so long as the model
parameters were all well behaved. In unbarred galaxies, the Stage 3 model
parameters might be unphysically large or small, in which case the Stage 2 fit
was favored. Conversely, in galaxies with prominent bars, the bulge component
of the Stage 2 bulge-disk fit tends to grow too extended in size. Addition of
a bar in the Stage 3 bulge-disk-bar fit removes this artifact, giving a more
physical solution. An extreme example of this situation is the barred galaxy
NGC 4548, which has a prominent bar and a faint disk. The Stage 2 fit, based
on a Sérsic bulge and exponential disk, is highly inadequate to describe the
bulge, disk, and the bar because it leads to an extremely extended bulge. The
Stage 3 bulge-disk-bar fit, however, yields a believable fit with a prominent
bar. The results of Stage 1, Stage 2, and Stage 3 are displayed in Figure 6
and Table 4.
4. 4.
Not all barred galaxies had unphysical Stage 2 models. Instead, the bulge
could be stretched along the $PA$ of the bar, giving the bulge a lower Sérsic
index and larger effective radius. A Stage 3 model that returned the bulge to
a size and shape more representative of the input image was favored over the
Stage 2 fit. Figure 7 and Table 5 demonstrate this behavior in NGC 4902. We
distinguish this effect from cases like NGC 4548 (Figure 6 and Table 4) where
the Stage 2 fit is completely wrong.
5. 5.
In cases where there was no bar, GALFIT can sometimes be enticed into fitting
a bar to any existing spiral arms, rings, or the clumpy disks of late-type
spirals. Stage 3 fits in these cases could be discarded by noting the
resulting discrepancies in appearance between the galaxy images and the Stage
3 model images. Examples of false bars are shown in Figure 8.
After fitting the whole sample and picking the best fit from either the Stage
2 bulge-disk decomposition or the Stage 3 bulge-disk-bar decomposition, we
also performed the following extra tests. For our sample S1 of 143 bright
($M_{B}\leq-19.3$) low-to-moderately inclined ($i\leq 70^{\circ}$) spirals in
the OSUBSGS survey, we determine the fraction (75 of 143 or $\sim 52\%$) of
spiral galaxies where a bulge-disk-bar decomposition was picked as the best
fit for the $H$-band image. There are also eight galaxies with pure bar-disk
fits. The $H$-band bar fraction, which is defined as the fraction of disk
galaxies that are barred, is therefore $58.0\pm 4.13\%$ (83 of 143). We then
compared our results ($58.0\%\pm 4.13\%$) with the $H$-band bar fraction (60%)
determined from ellipse fits of the OSUBSGS sample by MJ07, with a slightly
more conservative inclination cut ($i\leq 60^{\circ}$). The two numbers are in
excellent agreement. As a further check to our fits, we compare the bar and
unbarred classification for individual galaxies from our fits with those from
MJ07, which were based on ellipse fits. Of the 73 galaxies that we classify as
barred, and that are mutually fitted by MJ07, 54 (74%) are also classified as
barred by MJ07. The remaining 19 (26%) galaxies are mainly weakly barred (with
Bar/$T$ below 0.08). Their RC3 optical types are weakly barred AB (10), barred
B (7), and unbarred A (2).
In most previous bulge-disk and bulge-disk-bar $2D$ decomposition, the issue
of parameter coupling and the systematic exploration of local versus global
minima in $\chi^{2}$ have been ignored. Quantifying how the parameters are
coupled is important in measuring error bars for the model parameters. With
$2D$ models containing several free parameters, this is not an easy task.
Although we also do not address this problem in rigorous detail, we describe
in § 4.4 simple test that explores parameter coupling in our models.
## 4 Extra Tests to Verify Correctness of Fits
### 4.1 Varying $b/a$ as a Function of Radius
Models generated with GALFIT do not allow the $b/a$ of the bulge, disk, or bar
to vary with radius. Since real bulges may have a varying $b/a$, it is
legitimate to investigate the impact of fixing the bulge $b/a$ on the
estimated $B/T$. We therefore performed the following test on NGC 4548. To
mimic a model bulge of varying $b/a$, we fitted the bulge light of NGC 4548
with ten concentric Sérsic profiles of increasing $r_{e}$ and varying $b/a$.
The $r_{e}$ of the outermost profile comes from the original bulge model (see
Table 4) where $b/a$ was kept constant with radius. The separation in $r_{e}$
between adjacent profiles is 0.5 pixels (0.75”). The luminosity, Sérsic index,
$b/a$, and $PA$ of each profile were free parameters. The disk and bar
components were fixed to the values in Table 4, as the emphasis was on the
change in the bulge.
Figure 9 compares the $B/T$ obtained by fitting the bulge of NGC 4548 with a
Sérsic model of constant $b/a$ as opposed to a Sérsic model varying $b/a$. The
bulge $b/a$ (0.88), $PA$ (-66.5), and $B/T$ (13%) from the original Sérsic fit
of constant $b/a$ (Table 4) are indicated with horizontal lines on the 3
panels. The top two panels show the run of $b/a$ (0.85 to 1.0) and $PA$
($-90^{\circ}$ to $+90^{\circ}$) of the ten concentric Sérsic profiles. It can
be seen that the Sérsic indices of the ten bulge models were generally higher
toward the center and declined at larger $r_{e}$, indicating that the ‘fitted
bulge is more concentrated at the center. The bottom panel shows the
cumulative $B/T$ calculated by summing all models with $r\leq r_{e}$. The last
point representing the summed $B/T$ from all ten components is 14.5%, in good
agreement with the 13.0% value from the Sérsic fit of constant $b/a$. Thus,
using a Sérsic model of constant $b/a$, does not have any significant adverse
impact on our derived $B/T$ in NGC 4548.
### 4.2 Fitting Artificially Simulated Images
An elementary test is to determine if GALFIT can recover the known parameters
of artificially simulated noisy images. The images were simulated by taking
parametric model images produced by GALFIT, and adding noise to the images
with the PyFITS module for Python (Barrett & Bridgman 1999). Noise in each
pixel was calculated by adding in quadrature the noise due to the source, sky,
and read noise. The standard deviation of pixel noise in electrons was
computed as
$\sigma=\sqrt{T_{source}+T_{sky}+T_{read}^{2}},$ (5)
where $T_{source}$ is the number of electrons due to the source, $T_{sky}$ is
the number of electrons due to the sky, and $T_{read}$ is the detector read
noise. The contribution due to detector dark current was very small and
therefore neglected. The offset added to each pixel was drawn from a normal
distribution centered at zero with standard deviation $\sigma$.
Our test sample consisted of 40 models (20 bulge-disk and 20 bulge-disk-bar)
with noise added as described. Thirty of the images included point sources as
extra components. The range explored for each parameter in the model images is
a fair representation of the parameter space covered by our full sample (e.g.,
$B/T$ ranges from $0.02-0.70$, the bulge index ranges from $\sim 0-5$, and the
full range of possible bulge and bar PA was also tested ($-90^{\circ}$ to
$+90^{\circ}$)). In terms of surface brightness, the models span five
magnitudes in mean surface brightness inside the disk scalelength. Examples of
the noise-added models are shown in Figure 10.
The noisy images were subjected to the $2D$ decomposition procedure outlined
in Figure 4. GALFIT reproduced the model (bulge, disk, bar, and point source)
parameters quite closely for the majority of the test cases. Figure 11
compares the recovered versus original model parameters. Except for some
extreme cases where the images were highly distorted by noise, all parameters
were recovered to within a few percent. Figure 12 plots the ratio of model-to-
recovered parameter against mean surface brightness inside the disk
scalelength; there is no strong trend in error with dimming surface
brightness. This suggests our decompositions are effective across the
parameter space spanned by our sample. The overall success of this test is
evidence that GALFIT is able to converge to the absolute minimum in $\chi^{2}$
space for our bulge-disk and bulge-disk-bar decompositions when the input is
the sum of parametric functions.
### 4.3 Using $1D$ Decomposition To Generate Guesses for Bulge Parameters
It is important to verify that GALFIT converges to the same solution even if
the initial guesses for the bulge parameters in Stage 2 and 3 are different.
Bulge-disk decomposition from $1D$ profiles provides an alternative means of
generating initial guesses. While $1D$ bulge-disk decompositions of radial
profiles along the bar major axis can be influenced by the bar, decomposition
of cuts along the bar minor axis will not be influenced as heavily. The
resulting bulge and disk parameters should be adequate guesses for Stage 3 of
our $2D$ decomposition method.
We tested the robustness of our Stage 3 fits by extracting initial guesses for
the bulge and disk using $1D$ decomposition along the bar minor axis. The
nonlinear least-squares algorithm designed to perform the $1D$ decomposition
simultaneously fits the sky-subtracted profiles with the sum of a Sérsic bulge
and an exponential disk, while ignoring the PSF. The results from the $1D$
decomposition include a bulge magnitude, $r_{e}$, Sérsic index, disk
magnitude, and disk scalelength.
The robustness of several bulge-disk-bar fits were tested by using the results
of the $1D$ decomposition as input to Stage 3. The $1D$ decompositions do not
provide information about the axis ratio ($b/a$) or $PA$, so these parameters
for the bulge were estimated by eye; for the disk, the $b/a$ and $PA$ were
fixed to the values determined by ellipse fitting, as described in $\S$ 3.2.
The initial bar parameters were unchanged from the earlier Stage 3 fits. In
all cases, the new models were identical to the Stage 3 models. As an example,
Table 6 compares Stage 3 input derived from 1D decomposition and GALFIT for
NGC 4548 and NGC 4643. In each case, both sets of input reproduced the same
results.
### 4.4 Parameter Coupling
Assessing the coupling between model parameters is complicated when models
have a large number of free parameters. A standard approach is to calculate
confidence regions using multi-dimensional ellipsoids for a given
$\Delta\chi^{2}$ contour. As the errors in the GALFIT models are not normally
distributed, but are instead dominated by the systematics of galaxy structure,
this approach does not yield meaningful results because of the ambiguity in
assessing which $\Delta\chi^{2}$ contour levels are statistically significant.
We carry out a simple test for representative galaxies to determine not only
the effects of parameter coupling, but also the the effect parameter coupling
has on model parameter errors, paying particularly close attention to $B/T$,
$D/T$, and Bar/$T$, as they are of primary interest. We perform this test on
four representative galaxies (NGC 3885, NGC 4151, NGC 4643, and NGC 7213). Two
are barred (NGC 4151 and NGC 4643). Two have high ($n>2$) bulge indices (NGC
4643 and NGC 7213), and the other two have low ($n<2$) bulge indices.
We fit two and three-component bulge-disk and bulge-disk-bar models using
fixed bulge indices of $n=1$ and $n=4$. The initial inputs to these fits were
the same as those used to generate the model in which bulge $n$ is a free
parameter. We then had GALFIT re-fit the models with bulge index as a free
parameter using these output model parameters as input initial guesses. We
compare the output of these two fits with the the best fit as selected in §
3.3, in which bulge $n$ is a free parameter. Ideally, the re-fits should
converge to the same final parameters as the selected best fit.
Table 7 displays the outcome of this test for the four representative systems.
For barred galaxies NGC 4643 and NGC 4151, the $n=1$ and $n=4$ re-fits
converged to the same $\chi^{2}$ as when bulge $n$ is initially left a free
parameter. $B/T$, $D/T$, and Bar/$T$ are precisely equal for NGC 4643, while
for NGC 4151 there is a small dispersion of 0.1-0.2%. For unbarred galaxies
NGC 3885 and NGC 7213, the $n=1$ re-fits again converged to the same
$\chi^{2}$ and model parameters as when bulge $n$ is initially left a free
parameter. This is not true for the $n=4$ re-fits. When the $n=4$ condition is
enforced, the bulges in these cases become too extended and luminous while the
$\chi^{2}$ drop below those when bulge $n$ is initially left a free parameter.
During the $n=4$ re-fits, $B/T$ increases further at the expense of $D/T$ and
the $\chi^{2}$ remain unchanged or decrease further. The $B/D$ ratio from the
$n=4$ re-fit for NGC 3885 is 1.3, roughly 3.5 times higher than when $n$ is
free. Given that NGC 3885 is an S0/a galaxy with a bulge embedded in a smooth
extended disk, the latter $B/D$ is arguably too large. For NGC 7213, the $n=4$
re-fit yields starkly unphysical values. The $B/D$ ratio is 11.3, and
$r_{e}/h$, the ratio between bulge effective radius and disk scalelength is
9.0. The lower $\chi^{2}$ in these cases cannot be taken seriously as the
bulges are too luminous and the resulting $B/D$ ratios do not match the data
images.
As illustrated by the above discussion and Table 7, this test shows that in
some cases (e.g., NGC 4643 and NGC 4151) GALFIT converges to the similar model
parameters and $B/T$, $D/T$, and Bar/$T$ while starting from highly different
initial input guesses (e.g. bulge $n=1$, bulge $n=4$, and bulge $n$ based on
the Stage 2 fits) in different regions of parameter space. For NGC 3885 and
NGC 7213, however, during the $n=4$ re-fits GALFIT converged to models that
were unphysical and different compared with the reasonable models generated
with input guesses corresponding to the bulge fixed at $n=1$ or the bulge $n$
based on output from Stage 1. In effect, when the initial input guesses were
very different from the data images, the resulting models were found, in spite
of the lower $\chi^{2}$, to be unphysical through comparison with the input
data images. We emphasize that for all sources analyzed in the paper, the
data, converged model output, and residuals are always inspected before
adopting the best final fit (see § 3.3).
Table 7 also provides hints as to how the model parameters are coupled to
bulge index. As suggested in the above discussion, fixing the bulge index to
$n=4$ leads to a more extended and luminous bulge, causing bulge $r_{e}$ and
$B/T$ to rise without fail for increasing bulge index. The disk is coupled
with the bulge such that increasing bulge index, bulge $r_{e}$, and $B/T$
yields a reduction in $D/T$. At the same time, disk scalelength either
increases (NGC 3885 and NGC 4643) or decreases (NGC 3885 and NGC 4151); in the
latter two cases, the disk becomes very compact and the bulge quite extended.
The behavior of the bar is coupled with _both_ the bulge and disk. In the case
of NGC 4643, as bulge index is raised to $n=4$, bar $r_{e}$ becomes slightly
larger, but Bar/$T$ falls by a factor of 2.6 as light is redistributed from
the bar and disk to the bulge. Bar index also declines as the bar assumes a
flatter profile. For NGC 4151, bar $r_{e}$ again increases slightly, but this
time Bar/$T$ rises by a factor of 1.7 as light is transferred from the disk to
the bulge and bar.
Based on the above test, we stress that GALFIT was able to overcome this
parameter coupling in the cases where the input guess parameters well-
reflected the data images.
## 5 Results and Discussion
### 5.1 Impact of Bars in $2D$ Decomposition
From the Stage 2 bulge-disk decomposition and Stage 3 bulge-disk-bar
decompositions, which we performed on all objects ($\S$ 3.2) we saw firsthand
the effects of adding a bar to the fit of a barred galaxy. We summarize below
some of these effects in order to underscore the importance of including a bar
component in the $2D$ luminosity decomposition of barred galaxies
1. 1.
During the Stage 2 bulge-disk decomposition of a barred galaxy, the luminosity
which comes from the galaxy’s disk, bulge, and bar gets distributed only
between two model components: the model bulge and disk. Since the disk $b/a$
and $PA$ are measured independently and held constant during the fits, the
Stage 2 model tends to distort the bulge in order to fit the bar. Thus, the
bulge in the Stage 2 bulge-disk decomposition of a barred galaxy can be
artificially long or too bright and extended. When a model bar component is
added in the Stage 3 bulge-disk-bar decomposition of a barred galaxy, it
forces a reshuffling of the luminosity between the three components.
Generally, the bulge declines in luminosity, whereas light can be either taken
from, or added back, to the disk.
2. 2.
We find that the inclusion of a bar component in the Stage 3 bulge-disk-bar
decomposition of a barred galaxy reduces the bulge fractional luminosity
$B/T$, compared with the Stage 2 bulge-disk decomposition. For our 75 barred
galaxies, the reductions correspond to factors of less than two, 2 to 4, and
above 4, in 36%, 25%, and 39% of barred galaxies, respectively. The larger
changes in $B/T$ occur in very strongly barred galaxies, where a prominent bar
cause the Stage 2 bulge-disk decomposition to overestimate the bulge. For
instance, $B/T$ declines in both of NGC 4643 (Figure 5 and Table 3) and NGC
4548 (Figure 6 and Table 4). In the latter case, $B/T$ is reduced by a factor
of $\sim 5$ between Stage 2 and Stage 3. These examples underscore the
importance of including bars in $2D$ luminosity decomposition of very strongly
barred galaxies.
3. 3.
The scalelength of the disk is generally unchanged by including the bar. NGC
4548 (Figure 6 and Table 4) is a good example. Sometimes, however, the disk
from the Stage 2 bulge-disk decomposition of a barred galaxy is erroneous due
to a poor fit. The disk parameters from the Stage 3 bulge-disk-bar
decomposition are quite different in such cases. NGC 4643 (Figure 5 and Table
3) illustrates this behavior.
We find that for our sample S1 of 143 bright ($M_{B}\leq-19.3$) low-to-
moderately inclined ($i\leq 70^{\circ}$) spirals (Figure 1) in the OSUBSGS
survey, 75 of 143 or $\sim 52\%$ are better fit with a Stage 3 bulge-disk-bar
decomposition than a Stage 2 bulge-disk decomposition. There are also eight
galaxies with pure bar-disk fits. As stated in $\S$ 3.3, the resulting
$H$-band bar fraction ($58.0\%\pm 4.13\%$ or 83 of 143) is in excellent
agreement with the $H$-band bar fraction of 60% reported by MJ07 based on
ellipse fits of the OSUBSGS sample, with a slightly more conservative
inclination cut ($i\leq 60^{\circ}$).
### 5.2 Mass in Bulges, Disks, and Bars
The fractional $H$-band luminosities in the bulge, disk, and bar ($B/T$,
$D/T$, Bar/$T$) of each galaxy can be considered as a fractional mass if we
assume that the same mass-to-light ($M/L$) ratio can be used to convert the
$H$-band luminosities of both the numerator ($B$, $D$, or Bar) and the
denominator ($T$) terms into a stellar mass. This is not an unreasonable
assumption as the $H$-band $M/L$ ratio is not very sensitive to differences in
dust or age that might exist between the bulge, disk, and bar. The
uncertainties in $M/L$ can be estimated by looking at population synthesis
models. Charlot, Worthey, & Bressan (1996) find that for idealized galaxies
with a single generation of stars, the uncertainties in $M/L$ ratio due to
different input stellar models and spectra are roughly $\pm 35\%$ for a fixed
metallicity and IMF. Furthermore, as the age of a stellar population varies
from $\sim 0.5$ Gyr to 10 Gyr, the $K$-band $M/L$ ratio rises by a factor of
$\sim$ 2 to 3 (Charlot 1996). Asymptotic giant branch (AGB) stars dominate the
NIR light for ages between 0.1 and 1 Gyr, while red giant branch (RGB) and
super-giant branch (SGB) stars dominate between 1 Gyr and 10 Gyr.
In this paper, we convert the $B/T$ light ratio determined from $H$-band
images to a $B/T$ mass ratio by assuming a constant mass-to-light ($M/L$) in
the $H$-band for both the bulge and the rest of the galaxy. However, Moorthy &
Holtzman (2006) present line strengths of bulges and inner disks for 38
spirals with Hubble types S0 to Sc. They show 76% of spirals have negative
metallicity gradients. $B-K$ color gradients are shown to largely match
metallicity gradients and are likewise negative outward, indicating bulge
$M/L$ is higher than in the disk and bar. If bulges are much older than the
disks, then our prescription would underestimate the true $B/T$ mass ratio. If
we assume an extreme case where bulges are $\sim 12$ Gyr and the disk light is
dominated by a young 3 Gyr population, our assumption of a constant $H$-band
$M/L$ ratio would underestimate the true $B/T$ by a factor of $\leq 2$
(Charlot 1996). In several sections of the paper (e.g., $\S$ 5.3, $\S$ 5.8),
we illustrate how our main results would change if the true $B/T$ was higher
by up to a factor of two. On the other hand, central regions of galaxies may
harbor intense episodes of star formation. If the bulge is younger than the
disk and happens to have star formation and a significant young population of
massive stars, then our prescription could overestimate the true $B/T$ mass
ratio. This would make our current results on the high fraction of low $B/T$
bulges even stronger.
Using the total galaxy stellar mass from $\S$ 2.2, the fractional masses can
be converted into absolute masses. (We do not convert the $H$-band luminosity
directly into a mass as the $H$-band images do not have photometric
calibration). The results are shown in Table 8. For our sample S1 of 143
bright ($M_{B}\leq-19.3$) low-to-moderately inclined ($i\leq 70^{\circ}$)
spirals with a mass-weighted mean Hubble type of Sab-Sb, we find that $\sim
70\%$ of the stellar mass is in disks, $\sim$ 10% is in stellar bars and $\sim
20\%$ is in bulges (with $\sim 11\%$ in $n>2$ bulges and $\sim 9\%$ in $n\leq
2$ bulges). Thus while bulges with $n\leq 2$ are highly ubiquitous (see next
section), they only account for a small fraction of the total stellar mass
budget.
Figure 13 shows the stellar mass for bulges, disks, and bars along the Hubble
sequence.
It is useful to compare our results with those of Driver et al. (2006), who
performed bulge-disk decomposition of $B$-band images with GIM2D on 10,095
galaxies from the Millennium Galaxy Catalog (Liske et al. 2003; Driver et al.
2005). They found 68.6% of the stellar mass to be in disks, and 32.6% in
bulges (with 30.8% in high $n$ bulges, and 1.8% in low $n\leq 2$ bulges).
Their study thus finds a higher stellar mass fraction in all bulges (32.6% vs
our 18.9%), and in high $n$ bulges (30.8% vs our 10.4%), and a lower fraction
in low $n\leq 2$ bulges (1.8% vs our 8.4%), and disks+bars (68.6% vs our 71.6%
+ 9.6%). This difference can be attributed to the fact that the Driver et al.
(2006) study did not perform bulge-disk-bar fits and thus, their $B/T$ ratios
may be skewed to higher values.
### 5.3 Distribution of Bulge Index and $B/T$
Figure 14 shows the individual and mean $B/T$ and bulge Sérsic index, plotted,
as a function of Hubble type and galaxy stellar mass. Barred and unbarred
galaxies are shown separately. Figure 15 shows the relationship between bulge
index and $B/T$.
We first consider the $B/T$ values in Figure 14. The mean $B/T$ in barred
galaxies is lower than in unbarred galaxies, but there is a large overlap in
the individual values. The offset in the mean $B/T$ of barred and unbarred
galaxies reported here, agrees with the result of Laurikainen et al. (2007;
see $\S$ 5.4) on the same sample. We also note that $B/T$ does not correlate
with Bar/$T$ (Fig. 16): aside from the six galaxies with large Bar/$T$
($>0.3$), most galaxies have low-to-moderate Bar/$T$ and a wide range of $B/T$
is seen at each Bar/$T$. This is reassuring and suggests that the bar fit is
not arbitrarily biasing the $B/T$ values. The distribution of Bar/$T$ is
further discussed in $\S$ 5.5.
How does the $B/T$ vary as a function of Hubble type and galaxy stellar mass?
Bulges with very high $B/T$ ($>0.4$) exist primarily in galaxies with high
mass ($M_{\star}>6\times 10^{10}M_{\odot}$) and early types (S0/a to Sab).
Bulges with very low $B/T$ ($<0.1$) lie primarily in lower mass galaxies with
later morphologies (Sb to Sc). It is striking that $\sim 69\%$ of bright
($M_{B}\leq-19.3$) low-to-moderately inclined ($i\leq 70^{\circ}$) spirals
have $B/T\leq 0.2$: these bulges are pervasive and exist across the whole
spectrum of S0/a to Sd. The results are summarized in Table 9. We shall
return to this point in § 5.8. We note again that these $B/T$ mass ratios were
calculated assuming a constant $M/L$ ratio in the $H$-band for the bulge and
disk components. As noted in $\S$ 5.2, if the bulge in these high mass spirals
is much younger (older) than the disk and bar, then the $B/T$ can be
overestimated (underestimated) by up to a factor of two, and the limiting
value of 0.2 for the $B/T$ cited in the above fraction, would have to be
modified in the extreme case to 0.1 and 0.4, respectively.
Some of the low $B/T\leq 0.2$ values for six barred S0/a and Sa galaxies in
Figure 14 may at first look suspicious. Balcells et al. (2007) report the mean
$B/T$ for S0 galaxies to be 0.25, so much smaller $B/T$ are potentially
worrisome. OSUBSGS $H$-band images of these objects in Figure 17 show a smooth
extended disk around the bulge. It should be noted that Hubble types were
originally assigned on a combination of criteria including disk smoothness and
spiral arm topology in addition to the prominence of the bulge. It is likely
that these galaxies were assigned early Hubble types due to their smooth
extended disks, in spite of their low bulge-to-disk ratio.
Similarly, some of the high $B/T\sim 0.4$ bulges in three of the Sc galaxies
may at first seem odd. However, again, visual inspection of their image
(Figure 17) reveals prominent spiral arms and clumpiness, which may explain
why they were assigned late Hubble types.
How does the bulge Sérsic index $n$ vary as a function of Hubble type, and
galaxy stellar mass (Figure 14), as well as $B/T$ (Figure 15)? The results are
summarized in Table 9. Only a small fraction ($\sim 1\%$) of bright spirals
have high $n\geq 4$ bulges; such bulges lie primarily in S0/a to Sab, and have
a large $B/T>0.2$. A moderate fraction ($\sim 22\%$) have intermediate $2<n<4$
bulges; these exist in barred and unbarred S0/a to Sd, and their $B/T$ spans a
wide range (0.03 to 0.5) with a mean of 0.29. A strikingly large fraction
($\sim 76\%$) of bright spirals have low $n\leq 2$ bulges; such bulges exist
in barred and unbarred galaxies across a wide range of Hubble types, and their
$B/T$ varies from 0.01 to 0.4, with most having $B/T\leq 0.2$.
### 5.4 Comparison With Other Work
As an independent check of our decomposition method, we compare our results
with independently published decompositions.
We find our mean $H$-band $B/D$ (Figure 18) ratios are comparable to the
$K$-band $B/D$ derived with the $1D$ bulge-disk decompositions of Graham
(2001) and Trujillo et al. (2002). Like Graham (2001) and Trujillo et al.
(2002), we find $B/D$ is widely variable with Hubble type and that mean $B/D$
steadily declines from Sa through Scd galaxies. Graham (2001) and Trujillo et
al. (2002) find bulge indices are widely variable within a Hubble type, but
they are in general $>1$ for early types and $<1$ for late types. We likewise
find wide scatter in bulge index with $n<1$ bulges existing in both early and
late types. Figure 18 is also in good agreement with the more recent results
of Graham & Worley (2008), who have calculated, as a function of morphology,
inclination and dust-corrected $B/D$ and bulge Sérsic indices. They find $B/D$
values are typically $<1/3$.
Another meaningful comparison can be made with Laurikainen et al. (2007) who,
using their own $2D$ decomposition code, fit a hybrid sample containing some
OSUBSGS galaxies combined with additional S0 galaxies. One difference between
their work and ours is that they typically model bars with a Ferrers function,
but may sometimes use a Sérsic profile, while we use only the latter. Also,
they include additional components to model secondary bars or inner disks.
They report a distinct offset in the mean $B/T$ between barred and unbarred
galaxies, which we confirm in Figure 14. Their mean $B/T$ are similar to ours,
and they conclude that pseudobulges exist throughout the Hubble sequence. The
Sérsic indices derived by Laurikainen et al. (2007) are likewise similar, on
the mean, to ours for both barred and unbarred systems. They likewise find
$n\leq 2$ bulges across early and late Hubble type galaxies.
Balcells et al. (2003) emphasized that bulges typically have indices $\sim 3$
or lower. Our results in Figure 14 are consistent. We find a low frequency
($\sim 1\%$) of high $n\geq 4$ bulges, with most bulges having $n\leq 3$.
### 5.5 Bar Strength
Stellar bars exert gravitational torques on the gaseous component and are
particularly efficient in driving gas from the outer disk into the inner
kiloparsec (see $\S$ 1). Thus, it would be natural to have a measure of bar
strength that is sensitive to the strength of the gravitational torque and
hence measures both the shape and mass of the bar.
Many measures of bar strength have been formulated. The $Q_{b}$ method of
(Block et al. 2002; Buta et al. 2003; Buta et al. 2005) measures directly the
gravitational torque at a single point along the bar. This method requires a
value of scaleheight for the disk and a model of the potential to be made from
the image. In the bar/inter-bar contrast method of Elmegreen & Elmegreen
(1985) and Elmegreen et al. (1996), bar strength is parameterized as the ratio
between peak surface brightness in the bar region and the minimum surface
brightness in the inter-bar region. Elmegreen & Elmegreen (1985) and Elmegreen
et al. (1996) also characterize bar strength with the maximum amplitude of the
$m=2$ mode from Fourier decomposition. When ellipse fitting is applied, the
maximum ellipticity of the bar, $e_{bar}$, can be used to characterize bar
strength (e.g. MJ07). This constitutes a partial measure of bar strength only,
however, as it offers no information about mass of the bar.
Bulge-disk-bar decomposition in the $H$-band provides another _partial_
measure of bar strength through the $H$-band Bar/$T$ light ratio, which is a
measure of the Bar/$T$ mass ratio under the assumption that the $H$-band $M/L$
ratio is the same for the bar and the rest of the galaxy, as discussed in $\S$
5.2. Figures 19 and 20 explore the derived bar properties.
The upper left panel of Figure 19 plots the individual and mean Bar/$T$
against Hubble type. There is a wide range ($\sim 0.03$ to $\sim 0.47$) in the
individual Bar/$T$ at a given Hubble type. The mean Bar/$T$ remains fairly
constant with Hubble type from Sa to Sb, but shows a possible weak decline by
about 0.1 from Sb to Sc. Their number statistics are too small to make any
robust statement for later Hubble types. We also note that six systems have
high Bar/$T$ above $0.3$; these are displayed in Figure 21.
Bar Sérsic indices are mostly below unity. Neither the individual, nor the
mean bar Sérsic index, show any trend with Hubble type or with stellar mass,
for Sa to Sc galaxies (Fig. 19). Thus, the steepness of the bar profile does
not seem to depend on the Hubble type. Is the bar mass ratio and its mass
profile related? There is a wide range in the individual Bar/$T$ at a given
bar Sérsic index (Fig. 20). The mean Bar/$T$ rises with bar index out to a bar
index of $\sim 0.6$, and then flattens out. This suggests that on the mean,
bars of lower Bar/$T$ have flatter profiles.
Is there a relation between the bar strength and the bulge present in a
galaxy? There is a wide range in the individual Bar/$T$ at a given $B/T$, and
at a given bulge Sérsic index (Fig. 20). The mean Bar/$T$ shows a weak decline
for bulge Sérsic indices above two. Similarly the mean Bar/$T$ shows a weak
rise from 0.1 to 0.25 as $B/T$ rises out to 0.15, after which the trend
flattens or reverses.
Both Bar/$T$ and maximum bar ellipticity $e_{bar}$ are partial measures of bar
strength. Figure 19 shows mean Bar/$T$ may scale weakly with Hubble type. The
bars with highest $e_{bar}$ (i.e, thin bars) are often termed strong bars, and
$e_{bar}$ has been shown to correlate with $Q_{b}$. Total bar strength should
scale with both bar mass and bar ellipticity. Does bar strength have a
dependence on Hubble type? The upper left panel of Figure 20 plots the product
of Bar/$T$ and $e_{bar}$, as determined by MJ07 for galaxies mutually
classified as barred, against Hubble type. There is a wide range in
Bar/$T\times e_{bar}$ in each bin, and mean bar strength shows no definite
trend with Hubble type. We note that bars with high Bar/$T$ and high $e_{bar}$
should exert the largest gravitational torque and be most effective at driving
gas inflows. A nice example is the oval or lens galaxy NGC 1317 (Figure 21);
the bar has a low ellipticity, but its $B/T$ is large as it is extended and
massive. Such bars or lenses may exert significant gravitational torques
although they are not very elongated.
### 5.6 Bar Fraction as a Function of $B/T$ and Bulge Index
As outlined in $\S$ 5.3, we found that as many as $\sim 76\%$ of bright
spirals have bulges with $n\leq 2$; such bulges exist in barred and unbarred
galaxies across a wide range of Hubble types, and their spread in $B/T$ is
from 0.01 to 0.4, with most having $B/T\leq$ 0.2. The variation of the bar
fraction as a function of $B/T$ and bulge $n$ can provide important
constraints on bulge formation scenarios ($\S$ 5.8). Table 10 shows our
results. The bar fraction declines with bulge index; $\sim 65\%$ of the
spirals with low $n\leq 2$ bulges host bars while intermediate $2<n<4$ bulges
have a lower bar fraction ($\sim 38\%$). The high $n\geq 4$ bulges in the
sample are unbarred, so the bar fraction is $0\%$. Systems with low $B/T$ are
more likely to be barred. For $B/T\leq 0.2$, the bar fraction is high ($\sim
68\%$). Systems with $0.2<B/T<0.4$ and $B/T\geq 0.4$ have lower bar fraction
($\sim 42\%$ and $\sim 17\%$).
Overall, Table 10 shows bulges with low $n\leq 2$ and low $B/T\leq 0.2$
preferentially exist in barred galaxies. This is consistent with earlier work
(Odewahn 1996; Barazza et al. 2008; Marinova et al. 2008; Aguerri et al. 2008,
in prep.) where an enhanced optical bar fraction is seen is galaxies with late
Hubble types or low $B/D$. It may be tempting to infer this result to mean the
formation pathway of two-thirds of low-$B/T$ bulges is related to bars in that
spontaneous or/and tidally induced bars play a role in bulge formation (with
the remaining one-third of such bulges may have been formed either by
mechanisms like retrograde minor mergers or short-lived bars). We caution that
this type of cause-effect relationship is not the only scenario consistent
with this result. It may also be possible that bar instabilities are favored
in galaxies with low $B/T$ and no inner Lindblad resonances (ILR). Under these
conditions, the swing amplifier model with a feedback loop (Julian & Toomre
1966; Toomre 1981; Binney & Tremaine 1987) may be responsible for bar
formation and partly account for the high bar fraction in galaxies of low
$B/T$.
### 5.7 Formation of Bulges
Our observational results provide some interesting challenges for models of
galaxy evolution that try to address the origin of present-day bulges. Any
successful model must be able to account for the observed distribution of
bulge $B/T$ and $n$ in high mass ($M_{\star}\geq 1.0\times 10^{10}M_{\odot}$)
spirals, as shown in Table 9 and Table 10. In particular, the following
results must be reproduced:
1. 1.
In terms of the overall distribution of bulge $n$, as much as ($\sim 74\%$) of
high mass spirals have bulge $n\leq 2$: such bulges exist in barred and
unbarred galaxies and their $B/T$ ranges from 0.01 to 0.4, with most having
$B/T\leq 0.2$ (Table 9). A moderate fraction ($\sim 24\%$) of high mass
spirals have $2<n<4$, and just ($\sim 2\%$) have $n\geq 4$.
2. 2.
Theoretical models often make more robust predictions on the bulge-to-total
mass ratio $B/T$ than on the bulge index $n$, so we consider the empirical
$B/T$ distribution in detail. We note that as much as $\sim 66\%$ of high mass
spirals have bulges with $B/T\leq 0.2$ (Table 9). In terms of bar fraction,
$\sim 68\%$ are barred (Table 10).
3. 3.
The fraction of bars rises among spirals with low bulge index $n$. About 63%
of spirals with low $n\leq 2$ bulges host bars, while the bar fraction in
spirals with $2<n<4$ bulges (44%) is two-thirds as large (Table 10).
In a hierarchical Universe, there are several physical processes that
contribute to the assembly of bulges: major mergers, minor mergers, and
secular evolution. We briefly describe these, expanding on our introduction in
$\S$ 1.
Major mergers, defined as those with mass ratio $M_{1}/M_{2}\geq 1/4$,
typically destroy the extended outer stellar disks during violent relaxation,
leaving behind a classical bulge. Such bulges are associated with modest-to-
high bulge Sérsic indices, in the range $2<n<6$ (Hopkins et al. 2008; Springel
et al. 2005; Robertson et al. 2006; $\S$ 5.8) in simulations. This trend is
also consistent with the fact that among ellipticals, high luminosity ones
tend to have a Sérsic index $n>4$, while low luminosity ones tend to have
$2\leq n\leq 3.5$ (Caon et al. 1993; Kormendy et al. 2008). The final Sérsic
index depends on the amount of residual gas the settles into a somewhat disky
component. Simulations by Hopkins et al. (2008) find that the Sérsic indices
of remnants from 1:1 gas-rich major mergers lie in the range of $2<n<4$, with
most above 2.5 (see Fig. 22). This body of evidence strongly suggests that
many bulges with $n>2$ might have a major merger origin.
Minor mergers, typically defined as those with mass ratio
$1/10<M_{1}/M_{2}<1/4$, do not destroy the stellar disk of the primary system,
but can contribute to building bulges via three pathways. Firstly, a fraction
$F_{sat}$ of the satellite’s stellar mass can end up in the central region of
the primary galaxy. The value of $F_{sat}$ depends on how centrally
concentrated the in-falling satellite is. Typically, the more diffuse outer
stellar body is tidally stripped, while the central core sinks by dynamical
friction to the central region (e.g., Quinn et al. 1993; Walker et al. 1996).
Secondly, a non-axisymmetric feature (e.g., a stellar bar or bar-like feature)
can be induced in the main disk, and gravitational torques exerted by the
feature can drive gas into the inner kpc (e.g., Hernquist & Mihos 1995; Jogee
2006 and references therein), where subsequent SF forms a compact high
$v/\sigma$ stellar component, or disky pseudobulge. Most of the gas inflow
happens during the merger phase and large gas inflow rates (e.g., $\gg 1$
$M_{\odot}$ per year) may be generated. Thirdly, gas inflow can also be caused
by direct tidal torques from the companion (e.g., Hernquist & Mihos 1995;
Eliche-Moral et al. 2006). It is to be noted that in the simulations by
Hernquist & Mihos (1995), the gas inflow generated by non-axisymmetric
features (e.g., bar-like features) in the inner part of the disk is much
larger than that caused by direct tidal torques from the satellite. In the
recent work of Eliche-Moral et al. (2006), N-body simulations of minor mergers
followed by fits of 1D Sérsic+exponential models to the remnants, suggest that
the bulge Sérsic index and $B/D$ ratio can grow as a result of the central re-
concentration of stellar disk material in the primary system by tidal forces.
Minor mergers are frequent under $\Lambda$CDM, and the likelihood of multiple
successive minor mergers occurring during the formation of a galaxy is high.
Bournaud, Jog, & Combes (2007) study the effects of repeated minor mergers on
galaxy structure. They show that a disk galaxy undergoing successive minor
mergers will eventually transform into an elliptical galaxy with an
$r^{1/4}$-law profile and high $V/\sigma$. However, galaxy growth is not
completely merger-driven and the efficiency of minor mergers at creating
ellipticals must be regulated by other mechanisms (e.g., cold gas accretion).
In addition, the process of secular evolution can build a disky bulge
(pseudobulge) between merger events. Here a stellar bar or globally oval
structure in a non-interacting galaxy drives gas inflow into the inner kpc,
where subsequent star formation forms a compact high $v/\sigma$ stellar
component (e.g., Kormendy 1993; Jogee 1999; KK04; Jogee, Scoville, & Kenney
2005; Athanassoula 2005; Kormendy & Fisher 2005; Kormendy 2008). This process
is different from that of minor mergers in the sense that it happens in the
quiescent phase of the galaxy, between minor or major merger events. The
prevalence of pseudobulges in galaxies of different Hubble types is discussed
in KK04, and select examples of S0 galaxies with pseudobulges are also shown
in Kormendy & Cornell (2004) and KK04.
The present-day bulge mass can be written as the sum of mass contributed from
each process:
$M_{bulge}=M_{bulge}\times(f_{\rm maj}+f_{\rm min1}+f_{\rm min2}+f_{\rm
min3}+f_{\rm sec}),$ (6)
where
* •
$f_{\rm maj}$ is the percentage of the bulge stellar mass, which is built by
major mergers,
* •
$f_{\rm min1}$ is the percentage of the bulge stellar mass, which is built
during minor mergers from stars of the satellite. This depends on the fraction
$F_{sat}$ of the satellite’s stellar mass, which ends up in the central region
of the primary galaxy during each minor merger.
* •
$f_{\rm min2}$ is the percentage of the bulge stellar mass, which is built
during minor mergers from gas inflow caused by a tidally induced bar.
* •
$f_{\rm min3}$ is the percentage of the stellar mass, which is built during
minor mergers from gas inflow caused by tidal torques from the companion.
* •
$f_{\rm sec}$ is the percentage of the stellar mass, which is built secularly
between merger events from gas inflow caused by bars or ovals
In § 5.8, we compare our derived distribution of bulge $n$ and $B/T$ with
hierarchical models that model major and minor mergers, but not secular
evolution. The main goal of the model is to see whether bulges built via major
mergers can account for the large fraction of high mass spirals with bulges of
low $B/T$ or/and low $n$. A secondary goal is to see if a first order
simplified prescription for minor mergers can broadly account for the
observations. We stress here that bulge-building during minor mergers is
modeled in a very simple way: all the stars in the satellite are assumed to
contribute to the bulge of the larger galaxy (i.e., $F_{sat}=100\%$), and
bulge-building via gas inflow driven through tidal torques and via
gravitational torques from induced bars are ignored (i.e., $f_{\rm min2}$ = 0,
and $f_{\rm min3}$ = 0). Furthermore, the models entirely ignore secular
evolution between mergers. In a future paper, these extra terms will be
addressed and a comprehensive picture built of the relative importance of
minor mergers and secular processes in making present-day bulges.
### 5.8 Comparison of $B/T$ With Hierarchical Models of Galaxy Evolution
We compare our data with the predictions from cosmological semi-analytical
models based on Khochfar & Burkert (2005) and Khochfar & Silk (2006). We
briefly describe the models first. The merger trees of dark matter (DM) halos
are derived by using the extended Press-Schechter formalism (Press & Schechter
1974), as in Somerville & Kolatt (1999). When two DM halos merge, the merger
time scale of the galaxies is calculated by considering the timescale it would
take the satellite galaxies to reach the central galaxy at the center of the
halo via dynamical friction (e.g., Kauffmann et al. 1999; Springel et al.
2001). The baryonic physics, which includes radiative cooling, star formation,
and feedback from supernovae, is treated via semi-analytic prescriptions (see
Khochfar & Silk (2006) and references therein). Baryonic mass inside the dark
matter halos is divided between hot gas, cold gas, and stars. The hot gas is
initially shock-heated to the halo virial temperature. As the gas radiatively
cools, it settles down into a rotationally supported disk at the halo center.
Cold disk gas is allowed to fragment and subsequently form stars according to
the Schmidt-Kennicutt law (Kennicutt 1998). Star formation is regulated by
feedback from supernovae using the prescription in Kauffmann et al. (1999).
Major mergers are typically considered as those with stellar mass ratio
$M_{1}/M_{2}\geq 1/4$. In the simulations, one assumes that during a major
merger any existing stellar disk is destroyed, gas is converted to stars with
some star formation efficiency (SFE), and all stars present undergo violent
relaxation to form a bulge. Therefore, the bulge-to-total stellar mass ratio
($B/T$) of a bulge immediately after a major merger is always one. Note that
the SFE during a major merger is not assumed to be 100% as there is mounting
evidence from SPH simulations (Springel & Hernquist 2005; Cox et al. 2008)
that not all cold gas is converted to stars. Instead, the burst efficiency
defined by Cox et al. (2008) is applied to control the fraction of stars
formed due to the interaction. This efficiency is dependent on the relative
masses of merging galaxies and is expressed as
$e=e_{1:1}{\left(\frac{M_{Satellite}}{M_{Primary}}\right)}^{\gamma},$ (7)
where $e_{1:1}$ is the burst efficiency for a 1:1 merger and $\gamma$ fixes
the dependence on mass ratio; Cox et al. (2008) find $e_{1:1}=0.55$ and
$\gamma=0.69$. The remaining fraction (1-$e$) of gas is added to the gaseous
disk and can start making stars.
As stated above, immediately after a major merger, the remnant is a bulge with
a $B/T$ equal to one. As time proceeds, $B/T$ falls because a stellar disk
grows around the bulge as hot gas in the halo cools, settles into a disk, and
forms stars. The formation of stars by any residual cold gas left at the end
of the major merger also helps to grow the disk. Thus $B/T$ falls until the
next major merger happens, at which point $B/T$ is reset to one in the models.
The bulge may also grow in stellar mass due to minor mergers. Minor mergers
are defined as mergers with mass ratio $1/10<M_{1}/M_{2}<1/4$, and the stellar
disk of the large companion is not destroyed during such mergers. The models
assume that during minor mergers, all the stars in the satellite are added to
the bulge of the host, while the gas settles in the disk. When DM halos grow
by accretion or minor mergers, the hot gas that comes in with a satellite is
immediately stripped and added to the hot gas component of the host. The cold
gas in the disk of the satellite is only added to the cold gas of the host if
they merge. Until they merge the satellite is using up its own cold gas to
make stars.
Fig. 23 shows the relationship between the present-day $B/T$ of a a high mass
($M_{\star}\geq 1.0\times 10^{10}M_{\odot}$) spiral and the redshift $z_{\rm
last}$ of its last major merger. As expected, systems where the last major
merger occurred at earlier times, have had more time to grow a disk and have a
lower $B/T$. The dispersion in the present-day $B/T$ at a given $z_{\rm last}$
is due to the different times spent by a galaxy in terms of being a satellite
versus a central galaxy in a DM halo, since the cooling of gas and the growth
of a disk is stopped when a galaxy becomes a satellite. Thus, galaxies that
became a satellite galaxy shortly after their last major merger stayed at high
$B/T$. Conversely, galaxies that continued to be a central galaxy for a long
time after their last major merger will have low $B/T$.
The present-day $B/T$ of a high mass ($M_{\star}\geq 1.0\times
10^{10}M_{\odot}$) spiral depends on its major merger history. In particular,
we note from Fig. 23 that a high mass ($M_{\star}\geq 1.0\times
10^{10}M_{\odot}$) galaxy, which has undergone a past major merger since
$z\leq 2$ will end up hosting a present-day $B/T>0.2$. In effect, a high mass
spiral can have a present-day $B/T\leq 0.2$ only if its last major merger
occurred at $z>2$ (lookback times $>10$ Gyr).
The predicted distribution of present-day $B/T$ depends on the galaxy merger
history in the models and it is relevant to ask how well the latter is
constrained observationally. Over the redshift range $z\sim 0.24$ to 0.80
(lookback times of 3 to 7 Gyr), recent studies by Jogee et al. (2008, 2009)
find that among high mass ($M_{\star}\geq 2.5\times 10^{10}M_{\odot}$)
galaxies, $\sim$ 10% of galaxies are undergoing mergers of mass ratio $>1/10$,
and $\sim$ 3% are undergoing major mergers of mass ratio $>1/4$. These
findings agree within a factor of less than $\sim 2$ with the merger rates
from the models of Khochfar & Burkert (2001) over $z\sim$ 0.24 to 0.80. At
higher redshifts, the empirical merger rate/fraction is uncertain due to
relatively modest volumes and bandpass shifting effects, but there is a
general trend towards higher merger fractions at higher redshifts (e.g.,
Conselice et al. 2003). The models used here (Khochfar & Burkert 2001) agree
with this trend and predict that $\sim$ 13.5% and $\sim$ 20% of high mass
($M_{\star}\geq 1.0\times 10^{10}M_{\odot}$) spirals have undergone major
mergers since $z\leq 2$ and $z\leq 4$, respectively (see Table 11).
The contribution of galaxies with different merger histories to the present-
day $B/T$ distribution are shown in Table 11. The top and middle parts of the
table describe systems with and without major mergers since $z\leq 2$ and
$z\leq 4$, respectively. In the model, $\sim$ 13.5% of present-day high mass
($M_{\star}\geq 1.0\times 10^{10}M_{\odot}$) spirals, experienced a major
merger since $z\leq 2$, causing most of them ($\sim$ 11.2%) to have a present-
day high $B/T>0.4$ and a negligible fraction ($\sim$ 0.1%) to have a low
present-day $B/T\leq 0.2$. In contrast, the remaining $\sim$ 86.5% spirals
experienced no major merger at $z\leq$ 2, and most (67.2%) of them have a
present-day low $B/T\leq 0.2$. If the comparisons are extended to systems
without a major merger since $z\leq 4$, the numbers are very similar (see
middle part of Table 11).
Table 11 shows that there is good agreement between the model and data for the
fraction of high mass spirals with present-day low $B/T\leq 0.2$ ($\sim$ 67%
in the model versus $\sim$ 66% in the data). The model contribution to low
$B/T\leq 0.2$ comes almost entirely from galaxies, which have not had a major
merger since $z\leq 2$ (see column 4 in Table 11). In fact, most of these
galaxies have not even had a major merger since $z\leq 4$, as illustrated by
the bottom part of Table 11. In the model, the fraction ($\sim 1.6\%$; column
3 of Table 11) of high mass spirals, which have undergone a major merger since
$z\leq 4$ and host a bulge with a present-day $B/T\leq 0.2$, is a factor of
over thirty smaller than the observed fraction ($\sim 66\%$) of high mass
spirals with $B/T\leq 0.2$. Thus, bulges built via major mergers since $z\leq
4$ seriously fail to account for the bulges present in $\sim 66\%$ of high
mass spirals. These results are also illustrated in Fig. 24, which shows the
comparison between data and models for the cumulative fraction of high mass
spirals as a function of present-day $B/T$.
It is also interesting to note from Table 11 that although the models
reproduce well the frequency of bulges with present-day low $B/T\leq 0.2$,
they tend to over-produce the frequency of present-day high $B/T>0.4$ systems
by nearly a factor of two ($\sim$ 14% in the model versus $\sim$ 8% in the
data; see columns 6 and 2 in middle part of Table 11). Most of this
overproduction stems from major mergers at $z\leq 4$ ($\sim$ 13%; column 4 in
middle part of Table 11). This suggests that major mergers, as currently
modeled here, are building bulges too efficiently.
One possible solution to this problem might relate to the suggestion by
Hopkins et al. (2009) that the efficiency of bulge-building during the major
merger of two spirals depends not only on the mass ratio $M_{\rm 1}/M_{\rm 2}$
of the merger, but also depends on the cold gas mass fraction $f_{\rm gas}$ in
the disk. In their semi-analytic models, the entire stellar mass of the
satellite violently relaxes, but the fraction of stellar mass in the primary
disk that violently relaxes and adds to the bulge is $M_{1}/M_{2}$. This
differs with our models where the entire primary stellar disk is always
destroyed in a major merger. Furthermore, the fraction $F$ of the total gas
mass, which loses angular momentum, falls to the nucleus, and is transformed
into stellar mass in a nuclear starburst, is $\sim(1-f_{\rm
gas})\times(M_{1}/M_{2})$. In particular, $F$ is lower for more gas-rich
disks, causing a suppression of the burst efficiency in gas-rich systems, and
a reduction in the stellar mass that ends up in the bulge built during the
major merger. The predictions for the distribution of $B/T$ from the Hopkins
et al. (2009) models are shown in the bottom of Table 11, with major mergers
considered as those with baryonic galaxy-galaxy mass ratio $>$ /1/3. Due to
the reduced stellar mass that ends up in the bulge after a major merger, the
Hopkins et al. (2009) models tend to yield lower $B/T$ after such a merger.
Thus, the models produce overall fewer high $B/T>0.4$ systems, and more
intermediate 0.2 $<B/T<$ 0.4 and low $B/T\leq 0.2$ systems (see column 6 in
Table 11). Nonetheless, the predictions are not very different from the models
by Khochfar & Burkert (2005) and Khochfar & Silk (2006), which we use in this
paper. In particular, the conclusion that the large frequency of high mass
($M_{\star}\geq 1.0\times 10^{10}M_{\odot}$) spirals with low $B/T\leq 0.2$
bulges can only be accounted for by spirals without a major merger since
$z\leq 2$ also holds with the Hopkins et al. (2009) models.
Thus, we conclude that the observed large frequency ($\sim$ 66%) of high mass
($M_{\star}\geq 1.0\times 10^{10}M_{\odot}$) spirals with low present-day
$B/T\leq 0.2$ can be accounted for in our and other hierarchical models by
high mass spirals, which have not undergone a major merger since $z\leq 2$,
and most of which have not even experienced a major merger over the last 12
Gyr since $z\leq 4$. In the models, most of these present-day low $B/T\leq
0.2$ bulges are built by minor mergers since $z\leq 4$. As noted earlier, our
models explore bulge-building via minor and major mergers, but do not
explicitly incorporate secular processes (see $\S$ 5.7). In practice, secular
processes may contribute to the building of present-day low $B/T\leq 0.2$
bulges, and are particularly relevant at $z\leq 2$, where major mergers cannot
build such bulges.
For completeness, we further explore how sensitive are our results to other
assumptions made in the data and models of Khochfar & Burkert (2005) and
Khochfar & Silk (2006):
* •
How sensitive are the results to the mass ratio used to separate major and
minor mergers? Fig. 25 is similar to Fig. 24 except that the model now defines
major mergers as those with mass ratio $M_{1}/M_{2}\geq 1/6$. In this case,
about 30% of the model spirals undergo major mergers since $z\leq 4$ rather
than $\sim 20\%$. The overall model $F$ (black dashed line) now under-predicts
the data $F$ by about 10% for $B/T>$ 0.2. However, the main conclusion that
bulges built by major mergers since $z\leq 4$ cannot account for most of the
low $B/T\leq 0.2$ bulges, present in a large percentage ($\sim 66\%$) of
spirals still holds.
* •
How sensitive are the results to the $B/T$ cut used to define spirals? Fig. 26
is similar to Fig. 24 except that here spirals are considered to be systems
with a $B/T\leq 0.55$ rather than 0.75 in the models, and a corresponding cut
is applied to the data points. The results are similar to Fig. 24
* •
How sensitive are the results to our assumed constant $H$-band mass-to-light
($M/L$) for the bulge, disk, and bar? Fig. 27 is similar to Fig. 24 except
that the $B/T$ of all the observed galaxies has been multiplied by a factor of
two, in order to test what would happen in the case where the $M/L$ ratio of
the bulge in $H$-band is twice as high as that of the disk and bar. This could
happen in an extreme example where the dominant bulge stellar population was
much older (e.g. 12 Gyr) than the age of the dominant disk stellar population
(e.g., 3 Gyr). In such a case, the fraction of high mass spirals with
$B/T\leq$ 0.2 would change from $\sim 66\%$ in Fig. 24 to $\sim 55\%$, and
deviate from the model overall model $F$ (black dashed line) by $\sim 20\%$.
However, the main conclusion that bulges built by major mergers since $z\leq
4$ cannot account for most of the low $B/T\leq 0.2$ bulges, present in a large
percentage ($\sim 55\%$) of spirals still holds.
Finally, it is important to note that so far we have compared the data and
model only in terms of bulge $B/T$, but not in terms of bulge index $n$ or in
terms of bar fraction. In effect, we have only shown that the models reproduce
a subset of the results outlined in points (1)-(3) of $\S$ 5.7. Since the
semi-analytic models do not produce a distribution of bulge index $n$, we
resort to presenting only an indirect comparison in Table 12. We assume that
bulges, which form in major mergers have a bulge $n>2.5$. This assumption is
based the evidence presented in $\S$ 5.7. Thus, in Test 1 of Table 12, we
compare the fraction ($\sim$ 66%) of galaxies in the semi-analytic models
having $B/T\leq 0.2$ and no major merger since $z\leq 4$, to the observed
fraction ($\sim$ 65%) of galaxies with $B/T\leq 0.2$ and bulge $n\leq 2.5$.
There is close agreement between the two values. In Test 2 of Table 12, the
fraction ($\sim$ 12.7%) of model galaxies with $B/T>0.4$ and no major merger
since $z\leq 4$ is a factor of $\sim$ 3 higher than the fraction ($\sim$ 3.5%)
of high mass spirals with $B/T>0.4$ and bulge $n>2.5$. Similarly, the fraction
($\sim$ 1.6%) of model galaxies with $B/T\leq 0.2$ and a past major merger
since $z\leq 4$ is also a factor of $\sim$ 2 higher than the observed fraction
($\sim$ 0.9%) of high mass spirals with $B/T\leq 0.2$ and bulge $n>2.5$ (Test
3, Table 12). Thus, in terms of bulge $B/T$ and $n$, there is good agreement
between data and model for Test 1 (involving model galaxies with no major
mergers since $z\leq 4$). This suggests that the vast majority of bulges with
$B/T\leq 0.2$ and $n\leq 2.5$ likely formed in galaxies having had no major
merger since $z\leq 4$. However, the agreement is less good for Tests 2 and 3
(involving model galaxies with major mergers since $z\leq 4$) and this
suggests that the models may be building bulges a little bit too efficiently
during a major merger, in agreement with the conclusion reached earlier.
What about the role of bars in the formation of these bulges of low $B/T$ and
low $n$? A detailed direct comparison with the semi-analytic models is not
possible as the role of bars is not yet modeled, but related comparisons are
possible. First, it is important to note that bar-driven gas inflow into the
inner kpc and the subsequent building of disky stellar components or
‘pseudobulges’ (see $\S$ 1) can happen in both isolated galaxies and in minor
mergers ($\S$ 5.7), since bars can be spontaneously induced in an isolated
disk or tidally induced during an interaction or minor merger. The triggering
of a bar is favored in a prograde interaction or minor merger. Thus, bulge-
building via induced bars is more likely to happen in prograde rather than
retrograde minor mergers. Statistically about half of minor mergers might be
prograde or prograde-like, and half retrograde. Thus, one would expect bars to
be induced in only half of the minor mergers. If this assumption is correct
and if most of the mass in bulges with present-day $B/T\leq 0.2$ is formed in
minor mergers, then one would expect only about half of these bulges to host
bars. This is close to what is observed, as shown by Table 10. We see that
$\sim 63\%$ of high mass spirals with low $n\leq 2$ bulges and $\sim 68\%$ of
spirals with low $B/T\leq 0.2$ bulges host bars. This suggests that in high
mass spirals, spontaneous and/or tidally induced bars _may_ play a part in
forming up to two-thirds of $B/T\leq 0.2$ or $n\leq 2$ bulges. The remaining
one-third of such bulges may have been formed either by mechanisms that do not
involve bars (e.g., retrograde minor mergers) or by bars that are not long-
lived.
## 6 Summary
The properties of galaxy components (bulges, disks, and bars) in the local
Universe provide key constraints for models of galaxy evolution. Most previous
$2D$ decompositions have focused on two-component bulge-disk decomposition,
and ignored the contribution of the bar even in strongly barred galaxies.
However, as shown by this work and other recent studies (e.g., Laurikainen et
al. 2005; Laurikainen et al. 2007; Reese et al. 2007), it is important to
include the bar component in the $2D$ decomposition, in order to correctly
estimate the bulge-to-total ratio ($B/T$) and disk properties. In this paper
we have developed an iterative $2D$, bulge-disk-bar decomposition technique
using GALFIT and applied it to $H$-band images to a complete sample (S1) of
143 bright ($M_{B}\leq-19.3$) low-to-moderately inclined ($i\leq 70^{\circ}$)
spirals from the OSU Bright Spiral Galaxy Survey (OSUBSGS). The sample has
primarily spirals with Hubble type S0/a to Sc and stellar mass $M_{\star}\geq
1.0\times 10^{10}M_{\odot}$. We performed two-component bulge-disk
decomposition, as well as three-component bulge-disk-bar decomposition on the
$2D$ light distribution of all galaxies, taking into account the PSF. We use
an exponential profile for the disk, and Sérsic profiles for the bulge and
bar. A number of quantitative indicators, including bar classification from
ellipse fits, are used to pick either the bulge-disk-bar decomposition or
bulge-disk decomposition, as the best final fit for a galaxy. Our main results
are the following.
1. 1.
We find that it is necessary to include the bar component in $2D$
decomposition of barred galaxies, otherwise, the bulge-to-total ratio ($B/T$)
will be overestimated and the disk properties may be skewed. Examples of the
effect of including the bar are shown for the prominently barred galaxies NGC
4643 (Figure 5, Table 3) and NGC 4548 (Figure 6, Table 4).
2. 2.
We find that out of the 143 low-to-moderately inclined ($i\leq 70^{\circ}$)
spirals in our sample, 75 of 143 or $\sim 52\%$ are better fit with a Stage 3
bulge-disk-bar decomposition than a Stage 2 bulge-disk decomposition. There
are also eight galaxies with pure bar-disk fits. The resulting $H$-band bar
fraction, defined as the fraction of disk galaxies that are barred, is
$58.0\pm 4.13\%$ (84 of 143). This fraction is in excellent agreement with the
$H$-band bar fraction of 60% reported by MJ07, based on ellipse fits of the
same OSUBSGS sample, with a more conservative inclination cut ($i\leq
60^{\circ}$).
3. 3.
$H$-band images tend to trace the overall mass fairly well and are not overly
impacted by extinction and age gradients. We therefore assume a constant mass-
to-light ($M/L$) in the $H$-band for the bulge, disk, and bar, and assume
their $H$-band light fraction is a measure of their mass fraction. For our
sample S1 of 143 bright ($M_{B}\leq-19.3$) low-to-moderately inclined ($i\leq
70^{\circ}$) spirals with a mass-weighted mean Hubble type of Sab-Sb, we find
that 71.6% of the stellar mass is in disks, 9.6% is in stellar bars and 18.9%
is in bulges (with 10.4% in $n>2$ bulges and 8.4% in $n\leq 2$ bulges).
If disks and bars are much younger (e.g., $\sim 3$ Gyr old ) than bulges
(e.g., $\sim 12$ Gyr old), then our prescription would underestimate the true
$B/T$ by a factor of $\leq$ 2\. On the other hand, if the bulge is younger
than the disk and happens to harbor a significant young population of massive
stars, then our prescription will overestimate the true $B/T$ mass ratio, and
make our current results on the high fraction of low $B/T$ bulges (see point 4
below) even stronger.
4. 4.
We explore the relationship between $B/T$, bulge Sérsic index, and Hubble
types (Fig. 14 & Fig. 15). Only a small fraction ($\sim 1\%$) of bright
spirals have high $n\geq 4$ bulges; such bulges lie primarily in S0/a to Sab,
and have a large $B/T>0.2$. A moderate fraction ($\sim 22\%$) have
intermediate $2<n<4$ bulges; these exist in barred and unbarred S0/a to Sd,
and their $B/T$ spans a wide range (0.03 to 0.5) with a mean of 0.29. Finally,
a strikingly large fraction ($\sim 76$%) of bright spirals have low $n\leq 2$
bulges; such bulges exist in barred and unbarred galaxies across a wide range
of Hubble types, and their $B/T$ varies from 0.01 to 0.4, with most having
$B/T\leq 0.2$.
5. 5.
Bulges with very high $B/T$ ($>0.4$) exist primarily in galaxies with high
mass ($M_{\star}>6\times 10^{10}M_{\odot}$) and early types (S0/a to Sab).
Bulges with very low $B/T$ ($<0.1$) lie primarily in lower mass galaxies with
later morphologies (Sb to Sc). As many as $\sim 69\%$ of bright spirals have
bulges with $B/T\leq 0.2$: these bulges are pervasive and exist across the
whole spectrum of S0/a to Sd (Figure 14).
6. 6.
Modeling bars with $2D$ decomposition allows us to measure bar properties and
the bar-to-total ratio (Bar/$T$), which is a measure of bar strength. There is
a wide range ($\sim 0.03$ to $\sim 0.47$) in the individual Bar/$T$ at a given
Hubble type. The mean Bar/$T$ remains fairly constant with Hubble type from Sa
to Sb, but shows a possible weak decline by about 0.1 from Sb to Sc (See
Figure 19 and Figure 20). The bar fraction (Table 10) declines with $B/T$; it
is highest ($\sim 68\%$) for bright spirals with $B/T\leq 0.2$, and lower
($\sim 36\%$) by nearly a factor of two in spirals with $B/T>0.2$.
It may be tempting to infer this to mean the formation of two-thirds of
low-$B/T$ bulges is related to spontaneous or/and tidally induced bars. Such a
cause-effect relationship is not the only scenario consistent with this
result. It may also be possible that bar instabilities are favored in galaxies
with low $B/T$ and no inner Lindblad resonances. Under these conditions, the
swing amplifier model with a feedback loop may be responsible for bar
formation and partly account for the high bar fraction in galaxies of low
$B/T$.
7. 7.
We compare the observed distribution of bulge $B/T$ and $n$ in high mass
($M_{\star}\geq 1.0\times 10^{10}M_{\odot}$) spirals with predictions from a
set of $\Lambda$CDM cosmological semi-analytical models (Table 11, Table 12,
and Figs. 24 to 27). Major mergers are considered as those with stellar mass
ratio $M_{1}/M_{2}\geq 1/4$. In the models, a high mass spiral can have a
bulge with a present-day low $B/T\leq$ 0.2 only if it did not undergo a major
merger since $z\leq 2$ (Fig. 23). The model merger history shows that only
$\sim$ 13.5% and $\sim$ 20% of the high mass spirals experience major mergers
since $z\leq 2$ and $z\leq 4$, respectively. The fraction ($\sim$ 1.6%) of
high mass spirals which have undergone a major merger since $z\leq 4$ and host
a bulge with a present-day low $B/T\leq 0.2$ is a factor of over thirty
smaller than the observed fraction ($\sim 66\%$) of high mass spirals with
$B/T\leq 0.2$(Table 11). Thus, bulges built via major mergers since $z\leq 4$,
over the last 12 Gyr, seriously fail to account for most of the low $B/T\leq
0.2$ bulges present in two-thirds of high-mass spirals. The overall picture
that emerges is that the observed large frequency ($\sim$ 66%) of high mass
($M_{\star}\geq 1.0\times 10^{10}M_{\odot}$) spirals with low present-day
$B/T\leq 0.2$ can be accounted for in our hierarchical models by high mass
spirals, which have not undergone a major merger since $z\leq 2$, and most of
which have not even experienced a major merger since $z\leq 4$. Most of these
present-day low $B/T\leq 0.2$ bulges are likely to have been built by a
combination of minor mergers and/or secular processes since $z\leq 4$.
S.J. and T.W. acknowledge support from the National Aeronautics and Space
Administration (NASA) LTSA grant NAG5-13063, NSF grant AST-0607748, and $HST$
grants GO-10395 and GO-10861 from STScI, which is operated by AURA, Inc., for
NASA, under NAS5-26555. We thank Chien Peng for technical assistance in the
operation of GALFIT. We acknowledge the usage of the Hyperleda database
(http://leda.univ-lyon1.fr). This research has made use of the NASA/IPAC
Extragalactic Database (NED) which is operated by the Jet Propulsion
Laboratory, California Institute of Technology, under contract with the
National Aeronautics and Space Administration.
## References
* Allen et al. (2006) Allen, P. D., Driver, S. P., Graham, A. W., Cameron, E., Liske, J., & de Propris, R. 2006, MNRAS, 371, 2
* Athanassoula (2002) Athanassoula, E. 2002, ApJL, 569, L83
* Athanassoula (2003) Athanassoula, E. 2003, MNRAS, 341, 1179
* Athanassoula (2005) Athanassoula, E. 2005, MNRAS, 358, 1477
* Balcells et al. (2003) Balcells, M., Graham, A. W., Domínguez-Palmero, L., & Peletier, R. F. 2003, ApJL, 582, L79
* Balcells et al. (2007) Balcells, M., Graham, A. W., & Peletier, R. F. 2007, ApJ, 665, 1084
* Barazza et al. (2007) Barazza, F. D., Jogee, S., & Marinova, I. 2007, in IAU Symposium 235, Galaxy Evolution across the Hubble Time, ed. F. Combes & J. Palous (Cambridge: Cambridge University Press), 76
* Barazza et al. (2008) Barazza, F. D., Jogee, S., & Marinova, I. 2008, ApJ, 675, 1194
* Barrett & Bridgman (1999) Barrett, P. E., & Bridgman, W. T. 1999, in ASP Conf. Ser. 172, Astronomical Data Analysis Software and Systems VIII, ed. D. M. Mehringer, R. L. Plante, & D. A. Roberts (San Francisco: Astronomical Society of the Pacific), 483
* Bell et al. (2003) Bell, E. F., McIntosh, D. H., Katz, N., & Weinberg, M. D. 2003, ApJS, 149, 289
* Benedict et al. (1996) Benedict, F. G., Smith, B. J., & Kenney, J. D. P. 1996, AJ, 111, 1861
* Berentzen et al. (2006) Berentzen, I., Shlosman, I., & Jogee, S. 2006, ApJ, 637, 582
* (13) Binney, J., & Tremaine, S. 1987, Galactic Dynamics (Princeton: Princeton Univ. Press)
* Blanton et al. (2005) Blanton, M. R., Lupton, R. H., Schlegel, D. J., Strauss, M. A., Brinkmann, J., Fukugita, M., & Loveday, J. 2005, ApJ, 631, 208
* Block et al. (2002) Block, D. L., Bournaud, F., Combes, F., Puerari, I., & Buta, R. 2002, A&A, 394, L35
* Böker et al. (2002) Böker, T., Laine, S., van der Marel, R. P., Sarzi, M., Rix, H.-W., Ho, L. C., & Shields, J. C. 2002, AJ, 123, 1389
* Bottinelli et al. (1995) Bottinelli, L., Gouguenheim, L., Paturel, G., & Teerikorpi, P. 1995, A&A, 296, 64
* Bournaud et al. (2007) Bournaud, F., Jog, C. J., & Combes, F. 2007, A&A, 476, 1179
* Bureau & Athanassoula (2005) Bureau, M., & Athanassoula, E. 2005, ApJ, 626, 159
* Burkert & D’Onghia (2004) Burkert, A. M., & D’Onghia, E. 2004, in Penetrating Bars Through Masks of Cosmic Dust: the Hubble Tuning Fork Strikes a New Note, ed. D. L. Block, I. Puerari, K. C. Freeman, R. Groess, & E. K. Block (Dordrecht: Kluwer Academic Publishers), 341
* Burkert & Naab (2004) Burkert, A., & Naab, T. 2004, in Carnegie Observatories Astrophysics Series, Vol. 1: Coevolution of Black Holes and Galaxies, ed. L. C. Ho (Cambridge: Cambridge Univ. Press), 421
* Buta et al. (2003) Buta, R., Block, D. L., & Knapen, J. H. 2003, AJ, 126, 1148
* Buta et al. (2005) Buta, R., Vasylyev, S., Salo, H., & Laurikainen, E. 2005, AJ, 130, 506
* Byun & Freeman (1995) Byun, Y. I., & Freeman, K. C. 1995, ApJ, 448, 563
* Caon et al. (1993) Caon, N., Capaccioli, M., & D’Onofrio, M. 1993, MNRAS, 265, 1013
* Charlot (1996) Charlot, S. 1996, in The Universe at High-z, Large-Scale Structure and the Cosmic Microwave Background, ed. E. Martinez-Gonzalez & J. L. Sanz (New York: Springer-Verlag), 53
* Charlot et al. (1996) Charlot, S., Worthey, G., & Bressan, A. 1996, ApJ, 457, 625
* Cole et al. (2000) Cole, S., Lacey, C. G., Baugh, C. M., & Frenk, C. S. 2000, MNRAS, 319, 168
* Conselice et al. (2003) Conselice, C. J., Bershady, M. A., Dickinson, M., & Papovich, C. 2003, AJ, 126, 1183
* Combes et al. (1990) Combes, F., Debbasch, F., Friedli, D., & Pfenniger, D. 1990, A&A, 233, 82
* Combes & Sanders (1981) Combes, F., & Sanders, R. H. 1981, A&A, 96, 164
* Cox et al. (2008) Cox, T. J., Jonsson, P., Somerville, R. S., Primack, J. R., & Dekel, A. 2008, MNRAS, 384, 386
* Debattista & Sellwood (1998) Debattista, V. P., & Sellwood, J. A. 1998, ApJL, 493, L5
* Debattista & Sellwood (2000) Debattista, V. P., & Sellwood, J. A. 2000, ApJ, 543, 704
* de Jong (1996) de Jong, R. S. 1996, A&A Suppl., 118, 557
* Depoy et al. (1993) Depoy, D. L., Atwood, B., Byard, P. L., Frogel, J., & O’Brien, T. P. 1993, Proc. SPIE, 1946, 667
* de Souza et al. (2004) de Souza, R. E., Gadotti, D. A., & dos Anjos, S. 2004, ApJS, 153, 411
* de Vaucouleurs et al. (1976) de Vaucouleurs, G., de Vaucouleurs, A., & Corwin, H. G. 1976, Second Reference Catalogue of Bright Galaxies, University of Texas Monographs in Astronomy, (Austin: University of Texas Press)
* de Vaucouleurs et al. (1991) de Vaucouleurs, G., de Vaucouleurs, A., Corwin, H. G., Jr., Buta, R. J., Paturel, G., & Fouque, P. 1991, Third Reference Catalogue of Bright Galaxies, Volume 1-3, (New York: Springer-Verlag)
* D’Onghia et al. (2006) D’Onghia, E., Burkert, A., Murante, G., & Khochfar, S. 2006, MNRAS, 372, 1525
* Driver et al. (2005) Driver, S. P., Liske, J., Cross, N. J. G., De Propris, R., & Allen, P. D. 2005, MNRAS, 360, 81
* Driver et al. (2006) Driver, S. P., et al. 2006, MNRAS, 368, 414
* Drory et al. (2004) Drory, N., Bender, R., & Hopp, U. 2004, ApJL, 616, L103
* Efstathiou et al. (1988) Efstathiou, G., Ellis, R. S., & Peterson, B. A. 1988, MNRAS, 232, 431
* Eliche-Moral et al. (2006) Eliche-Moral, M. C., Balcells, M., Aguerri, J. A. L., & González-García, A. C. 2006, A&A, 457, 91
* Elmegreen (1994) Elmegreen, B. G. 1994, ApJL, 425, L73
* Elmegreen & Elmegreen (1985) Elmegreen, B. G., & Elmegreen, D. M. 1985, ApJ, 288, 438
* Elmegreen et al. (1996) Elmegreen, B. G., Elmegreen, D. M., Chromey, F. R., Hasselbacher, D. A., & Bissell, B. A. 1996, AJ, 111, 2233
* Eskridge et al. (2000) Eskridge, P. B., et al. 2000, AJ, 119, 536
* Eskridge et al. (2002) Eskridge, P. B., et al. 2002, ApJS, 143, 73
* Fisher & Drory (2008) Fisher, D. B., & Drory, N. 2008, AJ, 136, 773
* Gadotti & Kauffmann (2007) Gadotti, D., & Kauffmann, G. 2007, in IAU Symposium 241, Stellar Populations as Building Blocks of Galaxies, ed. A. Vazdekis & R. F. Peletier (Cambridge: Cambridge University Press), 507
* Graham (2001) Graham, A. W. 2001, AJ, 121, 820
* Graham & Worley (2008) Graham, A. W., & Worley, C. C. 2008, MNRAS, 752
* (55) Hernquist, L. & Mihos, J. C. 1995, ApJ, 448, 41
* Ho et al. (1997) Ho, L. C., Filippenko, A. V., & Sargent, W. L. W. 1997, ApJS, 112, 315
* Hopkins et al. (2008) Hopkins, P. F., Cox, T. J., Dutta, S. N., Hernquist, L., Kormendy, J., & Lauer, T. R. 2008, ArXiv e-prints, 805, arXiv:0805.3533
* Hopkins et al. (2009) Hopkins, P. F., et al. 2009, ApJ, submitted (arXiv:0901.4111)
* Hunt & Malkan (1999) Hunt, L. K., & Malkan, M. A. 1999, ApJ, 516, 660
* Jogee (1999) Jogee, S. 1999, Ph.D. thesis, Yale University
* Jogee (2006) Jogee, S. 2006, in Physics of Active Galactic Nuclei at all Scales, ed. D. Alloin, R. Johnson, & P. Lira (Berlin: Springer), 143
* Jogee et al. (1999) Jogee, S., Kenney, J. D. P., & Smith, B. J. 1999, ApJ, 526, 665
* Jogee et al. (2005) Jogee, S., Scoville, N., & Kenney, J. D. P. 2005, ApJ, 630, 837
* Jogee et al. (2004) Jogee, S., et al. 2004, ApJL, 615, L105
* Jogee et al. (2008) Jogee, S. 2008, in ‘The Galaxy Disk in a Cosmological Context’, Proceedings of IAU Symposium 254, Eds. J. Andersen, J. Bland-Hawthorn & B. Nordstrom, held in Copenhagen in June 2008, in press (arXiv:0810.5617)
* Jogee et al. (2009) Jogee, S., et al. 2009, ApJ, accepted
* Julian & Toomre (1966) Julian, W. H., & Toomre, A. 1966, ApJ, 146, 810
* Kauffmann et al. (1999) Kauffmann, G., Colberg, J. M., Diaferio, A., & White, S. D. M. 1999, MNRAS, 303, 188
* Kauffmann et al. (2003) Kauffmann, G., et al. 2003, MNRAS, 341, 33
* Kautsch et al. (2006) Kautsch, S. J., Grebel, E. K., Barazza, F. D., & Gallagher, J. S., III 2006, A&A, 445, 765
* Kenney et al. (1993) Kenney, J. D. P., Carlstrom, J. E., & Young, J. S. 1993, ApJ, 418, 687
* Kennicutt (1998) Kennicutt, R. C., Jr. 1998, ApJ, 498, 541
* Khochfar & Burkert (2001) Khochfar, S., & Burkert, A. 2001, ApJ, 561, 517
* Khochfar & Burkert (2005) Khochfar, S., & Burkert, A. 2005, MNRAS, 359, 1379
* Khochfar & Silk (2006) Khochfar, S., & Silk, J. 2006, MNRAS, 370, 902
* Knapen et al. (1995) Knapen, J. H., Beckman, J. E., Heller, C. H., Shlosman, I., & de Jong, R. S. 1995, ApJ, 454, 623
* Knapen et al. (2000) Knapen, J. H., Shlosman, I., & Peletier, R. F. 2000, ApJ, 529, 93
* Koda et al. (2007) Koda, J., Milosavljevic, M., & Shapiro, P. R. 2007, preprint (arXiv:0711.3014)
* Kormendy (1993) Kormendy, J. 1993, in IAU Symposium 153, Galactic Bulges, ed. H. Dejonghe & H. J. Habing (Dordrecht: Kluwer), 209
* Kormendy (2008) Kormendy, J. 2008, in IAU Symposium 245, Formation and Evolution of Galaxy Bulges, ed. M. Bureau, E. Athanassoula, & B. Barbury (Cambridge: Cambridge Univ. Press), in press
* Kormendy & Cornell (2004) Kormendy, J., & Cornell, M. E. 2004, in Penetrating Bars Through Masks of Cosmic Dust: the Hubble Tuning Fork Strikes a New Note, ed. D. L. Block, I. Puerari, K. C. Freeman, R. Groess, & E. K. Block (Dordrecht: Kluwer Academic Publishers), 261
* Kormendy & Fisher (2005) Kormendy, J., & Fisher, D. B. 2005, Rev. Mex. AA Ser. Conf., 23, 101 (arXiv:0507525)
* Kormendy & Fisher (2008) Kormendy, J., & Fisher, D. B. 2008, in Formation and Evolution of Galaxy Disks, ed. J. G. Funes, S. J., & E. M. Corsini (San Francisco: ASP), 297
* (84) Kormendy, J., Fisher, D. B., Cornell, M. E., & Bender, R. 2008, ApJS, in press (arXiv:0810.1681)
* Kormendy & Kennicutt (2004) Kormendy, J., & Kennicutt, R. C., Jr. 2004, ARAA, 42, 603 (KK04)
* Kroupa et al. (1993) Kroupa, P., Tout, C. A., & Gilmore, G. 1993, MNRAS, 262, 545
* Laine et al. (2002) Laine, S., Shlosman, I., Knapen, J. H., & Peletier, R. F. 2002, ApJ, 567, 97
* Laurikainen et al. (2004) Laurikainen, E., Salo, H., & Buta, R. 2004, ApJ, 607, 103
* Laurikainen et al. (2005) Laurikainen, E., Salo, H., & Buta, R. 2005, MNRAS, 362, 1319
* Laurikainen et al. (2007) Laurikainen, E., Salo, H., Buta, R., & Knapen, J. H. 2007, MNRAS, 381, 401
* Liske et al. (2003) Liske, J., Lemon, D. J., Driver, S. P., Cross, N. J. G., & Couch, W. J. 2003, MNRAS, 344, 307
* Marinova & Jogee (2007) Marinova, I., & Jogee, S. 2007, ApJ, 659, 1176 (MJ07)
* (93) Marinova, I., et al. . 2008, ApJ, submitted
* Martinez-Valpuesta et al. (2006) Martinez-Valpuesta, I., Shlosman, I., & Heller, C. 2006, ApJ, 637, 214
* Menéndez-Delmestre et al. (2007) Menéndez-Delmestre, K., Sheth, K., Schinnerer, E., Jarrett, T. H., & Scoville, N. Z. 2007, ApJ, 657, 790
* Mihos & Hernquist (1996) Mihos, J. C., & Hernquist, L. 1996, ApJ, 464, 641
* Moorthy & Holtzman (2006) Moorthy, B. K., & Holtzman, J. A. 2006, MNRAS, 371, 583
* Mulchaey & Regan (1997) Mulchaey, J. S., & Regan, M. W. 1997, ApJL, 482, L135
* Naab & Burkert (2001) Naab, T., & Burkert, A. 2001, ApJL, 555, L91
* Naab & Burkert (2003) Naab, T., & Burkert, A. 2003, ApJ, 597, 893
* Navarro & Steinmetz (2000) Navarro, J. F., & Steinmetz, M. 2000, ApJ, 538, 477
* (102) Odewahn, S. C. 1996, in Barred Galaxies, IAU Coll. 157, ed. R. Buta, D. A. Crocker & B. G. Elmegreen, ASP Conf. Ser. 91 (San Francisco: ASP), 30
* Paturel et al. (2003) Paturel G., Petit C., Prugniel P., Theureau G., Rousseau J., Brouty M., Dubois P., Cambrésy L., 2003, A&A, 412, 45
* Peng et al. (2002) Peng, C. Y., Ho, L. C., Impey, C. D., & Rix, H.-W. 2002, AJ, 124, 266
* Pfenniger & Norman (1990) Pfenniger, D., & Norman, C. 1990, ApJ, 363, 391
* Press & Schechter (1974) Press, W. H., & Schechter, P. 1974, ApJ, 187, 425
* Ravindranath et al. (2001) Ravindranath, S., Ho, L. C., Peng, C. Y., Filippenko, A. V., & Sargent, W. L. W. 2001, AJ, 122, 653
* Reese et al. (2007) Reese, A. S., Williams, T. B., Sellwood, J. A., Barnes, E. I., & Powell, B. A. 2007, AJ, 133, 2846
* Robertson et al. (2006) Robertson, B., Bullock, J. S., Cox, T. J., Di Matteo, T., Hernquist, L., Springel, V., & Yoshida, N. 2006, ApJ, 645, 986
* Robertson et al. (2004) Robertson, B., Yoshida, N., Springel, V., & Hernquist, L. 2004, ApJ, 606, 32
* Salucci et al. (2008) Salucci, P., Yegorova, I. A., & Drory, N. 2008, MNRAS, 388, 159
* Schechter (1976) Schechter, P. 1976, ApJ, 203, 297
* Schlegel et al. (1998) Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525
* Sheth et al. (2008) Sheth, K., et al. 2008, ApJ, 675, 1141
* Simard (1998) Simard, L. 1998, in ASP Conf. Ser. 145, Astronomical Data Analysis Software and Systems VII, ed. R. Albrecht, R. N. Hook, & H. A. Bushouse (San Francisco: Astronomical Society of the Pacific), 108
* Simard et al. (2002) Simard, L., et al. 2002, ApJS, 142, 1
* Somerville & Kolatt (1999) Somerville, R. S., & Kolatt, T. S. 1999, MNRAS, 305, 1
* Springel et al. (2005) Springel, V., Di Matteo, T., & Hernquist, L. 2005, MNRAS, 361, 776
* Springel & Hernquist (2005) Springel, V., & Hernquist, L. 2005, ApJL, 622, L9
* Springel et al. (2001) Springel, V., White, S. D. M., Tormen, G., & Kauffmann, G. 2001, MNRAS, 328, 726
* Steinmetz & Navarro (2002) Steinmetz, M., & Navarro, J. F. 2002, NewA, 7, 155
* (122) Toomre, A. 1981, in The Structure and Evolution of Normal Galaxies, ed. S. M. Fall & D. Lynden-Bell (New York: Cambridge University Press), 111
* Trujillo et al. (2002) Trujillo, I., Asensio Ramos, A., Rubiño-Martín, J. A., Graham, A. W., Aguerri, J. A. L., Cepa, J., & Gutiérrez, C. M. 2002, MNRAS, 333, 510
* (124) Tully, R. B. 1988, Nearby Galaxies Catalog (Cambridge: Cambridge University Press)
* (125) Quinn, P. J., Hernquist, L., & Fullagar, D. P. 1993, ApJ, 403, 74
* Véron-Cetty & Véron (2006) Véron-Cetty, M.-P., & Véron, P. 2006, A&A, 455, 773
* Wadadekar et al. (1999) Wadadekar, Y., Robbason, B., & Kembhavi, A. 1999, AJ, 117, 1219
* (128) Walker, I. R., Mihos, J. C., & Hernquist, L. 1996, ApJ, 460, 121
* Warren et al. (2007) Warren, S. J., et al. 2007, MNRAS, 375, 213
* Weinberg (1985) Weinberg, M. D. 1985, MNRAS, 213, 451
Figure 1: Our final sample S1 consists of 143 bright ($M_{B}\leq-19.3$) low-
to-moderately inclined ($i\leq 70^{\circ}$) spirals in the OSUBSGS survey. The
distribution of absolute $B$-band magnitude for the sample of bright spirals
in the OSUBSGS survey is shown in the top panel before (solid line) and after
(shaded greyscale) the cut to remove highly inclined ($i>70^{\circ}$) spirals.
The distribution of Hubble types for the sample is shown in the bottom panel
before (solid line) and after (shaded greyscale) the cut to remove highly
inclined ($i>70^{\circ}$) spirals.
Figure 2: The luminosity function of the full OSUBSGS sample is compared with
the $B$-band Schechter luminosity function (SLF). The former is calculated as
described in § 2.1 using equation (1). The parameters for the SLF are
$\Phi^{*}=5.488\times 10^{-3}$ Mpc-3, $\alpha=-1.07$, and $M^{*}_{B}=-20.5$
(Efstathiou, Ellis & Peterson 1988), corresponding to $H_{0}$=70 km/s Mpc-1.
Figure 3: Out of our final sample S1 of 143 bright ($M_{B}\leq-19.3$) low-to-
moderately inclined ($i\leq 70^{\circ}$) OSUBSGS spirals, stellar masses could
be estimated for 126 galaxies. Their stellar mass distribution is shown, as
determined in $\S$ 2.2. Most have stellar masses $M_{\star}\geq 1.0\times
10^{10}M_{\odot}$. This sample of 126 galaxies is referenced henceforth as the
sample S2.
Figure 4: An overview of the method of decomposition. All images are subjected
to Stages 1, 2, and 3. Either the best fit of Stage 2 or Stage 3 is chosen as
the best model.
Figure 5: Complete $2D$ decomposition for NGC 4643. Note the prominent bar
residuals in the residual for the Stage 1 and Stage 2 bulge-disk
decomposition. This is a case where the prominent bar causes the Stage 2
bulge-disk fit to artificially extend the bulge and inflate the $B/T$. The
disk fitted in Stage 2 has a low surface brightness and is very extended, well
beyond the real disk: the $b/a$ and $PA$ of the fitted disk is shown as
contours. Stage 3 bulge-disk-bar decomposition provides the best model. The
$\chi^{2}$ for the Stage 1, Stage 2, and Stage 3 residual images are 7360.7,
7284.8, and 2111.59, respectively. See Table 3 for the fit parameters.
Figure 6: The complete $2D$ decomposition for NGC 4548. This is an extreme
example where the prominent bar results in an extended bulge and inflated
$B/T$ in the Stage 2 bulge-disk fit. Like NGC 4643 in Figure 5, the disk
fitted in Stage 2 has a low surface brightness and is very extended: its $b/a$
and $PA$ are shown as contours. Stage 3 bulge-disk-bar decomposition provides
the best model. The $\chi^{2}_{\nu}$ for the Stage 1, Stage 2, and Stage 3
residual images are 7076.1, 6301.3, and 3260.4, respectively. See Table 4 for
the fit parameters.
Figure 7: This plot shows the data image, Stage 2 model, and Stage 3 model for
NGC 4902. The Stage 2 bulge is too bright and is extended along the major axis
of the bar ($B/T$=31.2% and $b/a$=0.45). In Stage 3, the bulge and bar are fit
with distinct components ($B/T$=5.59%, bulge $b/a$=0.68, Bar/$T$=9.97%, bar
$b/a$=0.22). All other fit parameters appear in Table 5.
Figure 8: The data images and Stage 3 bulge-disk-bar decomposition models of
NGC 5427 and NGC 7412 are shown. The Stage 3 models each distinctly show a
false bar component, which is not present in the data images. The false
components can be inspired by prominent spiral arms, such as those present in
these galaxies. Such cases are flagged during the visual inspection of fits
and the Stage 3 bulge-disk-bar decomposition is discarded in favor of the
Stage 2 bulge-disk decomposition.
Figure 9: This plot compares the $B/T$ obtained by fitting the bulge of NGC
4548 with a Sérsic model of constant $b/a$ as opposed to a Sérsic model
varying $b/a$. To mimic a Sérsic model with varying $b/a$ in GALFIT, the bulge
was fitted with ten concentric Sérsic profiles with fixed $r_{e}$, each
separated by 0.75”. The top two panels show the run of $b/a$ and $PA$ of the
ten concentric Sérsic profiles. The bottom panel shows the cumulative $B/T$
calculated by summing all models with $r\leq r_{e}$. The bulge $b/a$ (0.88),
$PA$ (-66.5), and $B/T$ (13%) from the original Sérsic fit of constant $b/a$
(Table 4) are indicated with horizontal lines on the 3 panels.
Figure 10: An elementary test is to determine if GALFIT can recover the known
parameters of artificial noisy images. Noisy images were simulated by taking
parametric model images (left panels) produced by GALFIT, and adding noise and
sky background (right panels). The noisy images were then fitted to see if the
original known parameters can be recovered. See § 4.2 for details.
Figure 11: The plots compare recovered versus original model parameters for
the simulated images discussd in § 4.2. The vertical axis limits demonstrate
the range explored for each parameter. The dotted line shows $y=x$ for
comparison. Except for some extreme cases where the images were highly
distorted by noise, all parameters were recovered to within a few percent.
Figure 12: The ratio of model-to-recovered parameter is plotted against mean
surface brightness inside the disk scalelength, $\mu=mag+2.5\times
log_{10}(2\times\pi\times b/a\times h^{2})$, for the simulated images
discussed in § 4.2. Surface brightness is not photometrically calibrated and
is shown for a zeropoint of 0.
Figure 13: The top, middle, and bottom panels show stellar mass for bulges,
disks, and bars, respectively, along the Hubble sequence. Values are shown for
sample S2 of 126 galaxies in Fig. 3. The legend in each panel indicates the
type of decomposition used for each data point.
Figure 14: The individual and mean $B/T$ (left panels) and bulge Sérsic index
(right panels) are plotted as a function of Hubble type for the sample S1 of
bright galaxies, and as a function of galaxy stellar mass for sample S2. The
error bars indicate the standard deviation of the population around the mean
in each bin. The legend in each panel indicates the type of decomposition used
for each data point. The mean $B/T$ and bulge index in barred galaxies differ
systematically from unbarred galaxies, but there is a large overlap in the
individual values. As many as $\sim 69\%$ of bright spiral galaxies have
$B/T\leq 0.2$; these bulges are pervasive and exist across the Hubble
sequence. Furthermore, as many as $\sim 76\%$ of bright spirals have low
$n\leq 2$ bulges. Such bulges exist in barred and unbarred galaxies across a
wide range of Hubble types.
Figure 15: The relation between $B/T$ and bulge index is shown. In the top
panel, galaxies are coded according to bar class. The legend indicates the
type of decomposition used for each data point. In the lower panel, galaxies
are coded according to Hubble type. A striking $\sim 76\%$ of bright spirals
have low $n\leq 2$ bulges. Such bulges exist in barred and unbarred galaxies
across a wide range of Hubble types, and their $B/T$ range from 0.01 to 0.4,
with most having $B/T\leq$ 0.2. A moderate fraction ($\sim 22\%$) have
intermediate $2<n<4$ bulges. These exist in barred and unbarred S0/a to Sd
galaxies, and their $B/T$ spans a wide range (0.05 to 0.5). Only ($\sim 1\%$)
have $n\geq 4$.
Figure 16: $B/T$ is plotted against Bar/$T$ and sorted by bulge Sérsic index.
There are six galaxies with Bar/$T$ $\geq 0.3$.
Figure 17: The top two rows show $H$-band images of barred galaxies, which
have early RC3 Hubble types, but yet have $B/T\leq 0.2$. The bottom row shows
$H$-band images of unbarred galaxies, which have late RC3 Hubble types, but
yet have $B/T\sim 0.4$. The Hubble types assigned to these objects more
reflect disk smoothness and spiral arm topology than $B/T$. All images are
from OSUBSGS with characteristics as described in § 2.
Figure 18: $B/D$ is plotted against Hubble type. The legend indicates the
type of decomposition used for each data point. The mean values for barred and
unbarred together in each bin are shown.
Figure 19: The properties of bars are shown. The error bars indicate the
standard deviation of the population around the mean in each bin. The legend
in each panel indicates the type of decomposition used for each data point.
Upper left: Mean and individual Bar/$T$ plotted against Hubble type. Upper
right: Mean and individual bar Sérsic indices plotted against Hubble type.
Lower left: Bar/$T$ plotted against total galaxy stellar mass. The mean
Bar/$T$ in bins of stellar mass is indicated. Lower right: Bar Sérsic index
plotted against total galaxy stellar mass.
Figure 20: Bar strength is plotted against Hubble type and the properties of
bars are compared with bulges. The legend in each panel indicates the type of
decomposition used for each data point. Upper left: Bar strength, the product
of Bar/$T$ and peak bar ellipticity $e_{bar}$ from MJ07 is plotted against
Hubble type. Upper right: Bar/$T$ is plotted against bar Sérsic index. Lower
left: Bar/$T$ is plotted against bulge Sérsic index. Lower right: Bar/$T$ is
plotted against $B/T$. In the first plot, mean bar strong is calculated for
each Hubble type. In the latter three plots, mean Bar/$T$ is calculated for
bins along the ordinate axis. The error bars indicate the standard deviation
of the population around the mean in each bin.
Figure 21: $H$-band OSUBSGS images of spirals with prominent bars of Bar/$T$
($>0.3$) are shown. An interesting example is the oval or lens galaxy NGC
1317: the bar has a low ellipticity, but its $B/T$ is large because it is
extended and massive. Such bars/lenses may exert significant gravitational
torques although they are not very elongated.
Figure 22: Shown is the distribution of Sérsic indices $n$ for remnants of
1:1 gas-rich major mergers in the simulations of Hopkins et al. 2008: they lie
in the range of $2<n<6$. Specifically, $\sim$ 22% of the remnants have
classical $n\geq 4$, as much as 20% have low $n\leq 2.5$, while 50% have
$n\leq 3$. Almost none have $n\leq 2$. [Figure: courtesy of Phil Hopkins]
Figure 23: For those high mass ($M_{\star}\geq 1.0\times 10^{10}M_{\odot}$)
galaxies in the theoretical models that experienced a major merger (see $\S$
5.8), the $B/T$ of the remnant at $z\sim 0$ is plotted against the redshift
$z_{\rm last}$ of the last major merger. Systems where the last major merger
occurred at earlier times have had more time to grow a disk and thus have a
lower $B/T$ at $z\sim$ 0\. The dispersion in the present-day $B/T$ at a given
$z_{\rm last}$ is due to the different times spent by a galaxy in terms of
being a satellite versus a central galaxy in a DM halo, since the cooling of
gas and the growth of a disk is stopped when a galaxy becomes a satellite. In
the model, a high mass galaxy that has undergone a major merger at $z\leq 2$
has a present-day $B/T>0.2$. In effect, a high mass spiral can have a present-
day $B/T\leq 0.2$ only if its last major merger occurred at $z>2$ (lookback
times $>10$ Gyr).
Figure 24: The cumulative fraction $F$ of high mass ($M_{\star}\geq 1.0\times
10^{10}M_{\odot}$) spirals with present-day $B/T\leq$ the x-axis value is
shown for the data (colored lines/ points) and for the theoretical model
(black lines/points) described in $\S$ 5.8. Model and data spirals are defined
as systems with $B/T\leq 0.75$. The magenta line shows $F$ from the data,
while the other two colored lines break this $F$ in terms of bar class (top
panel) or bulge $n$ (lower panel). The black dashed line shows $F$ from all
model galaxies, while the black dotted line and black dots show the
contribution of model galaxies that experienced, respectively, no major merger
and one or more major mergers since $z\leq 4$. Major mergers are defined here
as those with $M_{1}/M_{2}\geq 1/4$. In the model, the fraction ($\sim$ 1.6%;
see Table 11) of high mass spirals, which have undergone a major merger since
$z\leq 4$ and host a bulge with a present-day $B/T\leq 0.2$ is a factor of
over 20 smaller than the observed fraction ($\sim$ 66%) of high mass spirals
with present-day $B/T\leq 0.2$. Thus, bulges built via major mergers since
$z\leq 4$ seriously fail to account for most of the low $B/T\leq 0.2$ bulges
present in two-thirds of high-mass spirals.
Figure 25: This figure is similar to Fig. 24, except that the model now
defines major mergers as those with mass ratio $M_{1}/M_{2}\geq 1/6$.
Figure 26: This figure is similar to Fig. 24, except that here spirals are
considered to be systems with a $B/T\leq$ 0.55 rather than 0.75 in the models,
and a corresponding cut is applied to the data points.
Figure 27: This figure is similar to Fig. 24, except that $B/T$ of all the
observed galaxies has been multiplied by a factor of two, in order to test
what would happen in the case where the $M/L$ ratio of the bulge in $H$-band
is twice as high as that of the disk and bar. This could happen in an extreme
example where the dominant bulge stellar population was much older (e.g. 12
Gyr) than the age of the dominant disk stellar population (e.g., 3 Gyr). In
such a case, the fraction of high mass spirals with $B/T\leq 0.2$ would change
from $\sim 66\%$ in Fig. 24 to $\sim 50\%$. However, the main conclusion that
bulges built by major mergers since $z\leq 4$ cannot account for most of the
low $B/T\leq 0.2$ bulges, present in a large percentage ($\sim 55\%$) of
spirals still holds.
Table 1: Sample S1 of Bright ($M_{B}\leq-19.3$) Low-to-Moderately Inclined
($i\leq 70^{\circ}$) Spirals in OSUBSGS (N=143)
Galaxy Name | Best Fit | Hubble Type | Bar Type | $D$ | $M_{B}$ | $B-V$ | $M_{\star}$
---|---|---|---|---|---|---|---
| | (RC3) | (RC3) | (Mpc) | (mag) | (mag) | ($M_{\odot}$)
(1) | (2) | (3) | (4) | (5) | (6) | (7) | (8)
ESO138-10 | PSF+Bulge+Disk | SA(s)cd | A | 15.75 | -20.2 | - | -
IC0239 | PSF+Bulge+Disk+Bar | SAB(rs)cd | AB | 15.21 | -19.3 | 0.70 | 1.34e+10
IC4444 | Bulge+Disk | SAB(rs)bc | AB | 28.82 | -20.9 | 0.64 | 4.35e+10
IC5325 | PSF+Bulge+Disk | SAB(rs)bc | AB | 19.39 | -19.6 | 0.56 | 9.78e+09
NGC0150 | PSF+Bulge+Disk+Bar | SB(rs)bc | B | 20.57 | -20.1 | 0.64 | 2.17e+10
NGC0157 | PSF+Bulge+Disk | SAB(rs)bc | AB | 22.39 | -21.4 | 0.59 | 5.81e+10
NGC0210 | Bulge+Disk+Bar | SAB(s)b | AB | 21.75 | -20.4 | 0.71 | 3.77e+10
NGC0278 | PSF+Bulge+Disk | SAB(rs)b | AB | 12.64 | -19.5 | 0.64 | 1.24e+10
NGC0289 | PSF+Bulge+Disk+Bar | SAB(rs)bc | AB | 20.79 | -20.2 | 0.73 | 3.33e+10
NGC0428 | PSF+Bulge+Disk+Bar | SAB(s)m | AB | 15.96 | -19.4 | 0.44 | 5.28e+09
NGC0488 | PSF+Bulge+Disk | SA(r)b | A | 31.39 | -21.8 | 0.87 | 2.44e+11
NGC0578 | PSF+Bulge+Disk+Bar | SAB(rs)c | AB | 20.89 | -20.5 | 0.51 | 1.87e+10
NGC0613 | PSF+Bulge+Disk+Bar | SB(rs)bc | B | 18.75 | -20.8 | 0.68 | 4.92e+10
NGC0685 | Bar+Disk | SAB(r)c | AB | 16.29 | -19.4 | 0.46 | 5.47e+09
NGC0779 | Bulge+Disk+Bar | SAB(r)b | AB | 18.54 | -20.0 | 0.79 | 3.64e+10
NGC0864 | PSF+Bulge+Disk+Bar | SAB(rs)c | AB | 21.43 | -20.6 | 0.55 | 2.31e+10
NGC0908 | PSF+Bulge+Disk | SA(s)c | A | 19.07 | -21.3 | 0.65 | 6.60e+10
NGC1042 | PSF+Bulge+Disk | SAB(rs)cd | AB | 17.89 | -20.2 | 0.54 | 1.63e+10
NGC1073 | PSF+Bulge+Disk+Bar | SB(rs)c | B | 16.29 | -19.9 | 0.50 | 1.04e+10
NGC1084 | PSF+Bulge+Disk | SA(s)c | A | 18.32 | -20.6 | 0.58 | 2.64e+10
NGC1087 | Bar+Disk | SAB(rs)c | AB | 20.36 | -20.6 | 0.52 | 2.09e+10
NGC1187 | PSF+Bulge+Disk+Bar | SB(r)c | B | 17.46 | -20.2 | 0.56 | 1.66e+10
NGC1241 | PSF+Bulge+Disk+Bar | SB(rs)b | B | 28.50 | -21.7 | 0.85 | 2.05e+11
NGC1300 | PSF+Bulge+Disk+Bar | SB(rs)bc | B | 20.14 | -20.9 | 0.68 | 5.39e+10
NGC1302 | PSF+Bulge+Disk+Bar | (R)SB(r)0 | B | 21.43 | -20.2 | 0.89 | 6.37e+10
NGC1309 | PSF+Bulge+Disk | SA(s)bc | A | 27.86 | -20.5 | 0.44 | 1.46e+10
NGC1317 | Bulge+Disk+Bar | SAB(r)a | AB | 18.11 | -20.2 | 0.89 | 6.42e+10
NGC1350 | PSF+Bulge+Disk+Bar | (R’)SB(r)ab | B | 18.11 | -21.1 | 0.87 | 1.38e+11
NGC1371 | PSF+Bulge+Disk+Bar | SAB(rs)a | AB | 18.32 | -20.0 | 0.90 | 5.43e+10
NGC1385 | PSF+Bar+Disk | SB(s)cd | B | 18.75 | -20.3 | 0.51 | 1.61e+10
NGC1511 | Bulge+Disk | SAa;pec | A | 16.18 | -19.7 | 0.57 | 1.15e+10
NGC1559 | PSF+Bar+Disk | SB(s)cd | B | 15.32 | -20.4 | 0.35 | 9.33e+09
NGC1703 | PSF+Bulge+Disk | SB(r)b | B | 18.64 | -19.4 | 0.56 | 8.43e+09
NGC1792 | PSF+Bulge+Disk | SA(rs)bc | A | 14.57 | -20.5 | 0.68 | 3.71e+10
NGC1808 | PSF+Bulge+Disk+Bar | (R)SAB(s)a | AB | 11.57 | -19.9 | 0.81 | 3.57e+10
NGC1964 | PSF+Bulge+Disk+Bar | SAB(s)b | AB | 21.43 | -20.8 | 0.77 | 6.54e+10
NGC2090 | Bulge+Disk+Bar | SA(rs)c | A | 10.93 | -19.3 | 0.79 | 1.90e+10
NGC2139 | Bulge+Disk | SAB(rs)cd | AB | 24.00 | -20.3 | 0.36 | 8.35e+09
NGC2196 | PSF+Bulge+Disk | (R’)SA(s)a | A | 30.86 | -21.0 | 0.81 | 9.22e+10
NGC2442 | PSF+Bulge+Disk+Bar | SAB(s)bc;pec | AB | 18.32 | -20.8 | 0.82 | 8.28e+10
NGC2559 | Bulge+Disk+Bar | SB(s)bc;pec | B | 21.43 | -21.0 | - | -
NGC2566 | Bulge+Disk | (R’)SB(rs)ab;pec | B | 22.61 | -20.7 | 0.81 | 6.98e+10
NGC2775 | PSF+Bulge+Disk | SA(r)ab | A | 18.21 | -20.6 | 0.90 | 9.46e+10
NGC3059 | Bar+Disk | SB(rs)c | B | 15.86 | -20.1 | 0.68 | 2.52e+10
NGC3166 | Bulge+Disk+Bar | SAB(rs)0 | AB | 23.57 | -20.2 | 0.93 | 7.41e+10
NGC3169 | PSF+Bulge+Disk | SA(s)a;pec | A | 21.11 | -20.3 | 0.85 | 6.09e+10
NGC3223 | PSF+Bulge+Disk | SA(s)b | A | 40.82 | -21.9 | 0.82 | 2.36e+11
NGC3227 | PSF+Bulge+Disk+Bar | SAB(s)a;pec | AB | 22.07 | -20.1 | 0.82 | 4.42e+10
NGC3261 | PSF+Bulge+Disk+Bar | SB(rs)b | B | 35.79 | -21.3 | - | -
NGC3275 | PSF+Bulge+Disk+Bar | SB(r)ab | B | 45.43 | -21.2 | - | -
NGC3319 | PSF+Bar+Disk | SB(rs)cd | B | 12.32 | -19.4 | 0.41 | 4.76e+09
NGC3338 | PSF+Bulge+Disk+Bar | SA(s)c | A | 24.43 | -20.6 | 0.59 | 2.67e+10
NGC3423 | Bulge+Disk | SA(s)cd | A | 11.68 | -19.5 | 0.45 | 5.89e+09
NGC3504 | Bulge+Disk+Bar | (R)SAB(s)ab | AB | 21.43 | -20.5 | 0.72 | 4.06e+10
NGC3513 | Bulge+Disk+Bar | SB(rs)c | B | 18.21 | -19.9 | 0.43 | 7.52e+09
NGC3583 | Bulge+Disk+Bar | SB(s)b | B | 36.43 | -20.8 | - | -
NGC3596 | PSF+Bulge+Disk | SAB(rs)c | AB | 24.64 | -19.6 | - | -
NGC3646 | PSF+Bulge+Disk | Ring | - | 59.79 | -22.9 | 0.65 | 2.82e+11
NGC3675 | PSF+Bulge+Disk | SA(s)b | A | 13.71 | -20.2 | - | -
NGC3684 | PSF+Bulge+Disk | SA(rs)bc | A | 25.07 | -19.4 | 0.62 | 1.00e+10
NGC3686 | PSF+Bulge+Disk+Bar | SB(s)bc | B | 25.18 | -19.6 | 0.57 | 9.97e+09
NGC3705 | PSF+Bulge+Disk+Bar | SAB(r)ab | AB | 18.21 | -19.9 | 0.79 | 3.15e+10
NGC3726 | PSF+Bulge+Disk+Bar | SAB(r)c | AB | 18.21 | -20.6 | 0.49 | 1.93e+10
NGC3810 | PSF+Bulge+Disk | SA(rs)c | A | 18.11 | -20.1 | 0.58 | 1.72e+10
NGC3885 | PSF+Bulge+Disk | SA(s)0 | A | 29.79 | -19.6 | 0.95 | 4.62e+10
NGC3887 | PSF+Bulge+Disk+Bar | SB(r)bc | B | 20.68 | -20.4 | - | -
NGC3893 | PSF+Bulge+Disk | SAB(rs)c | AB | 18.21 | -19.9 | - | -
NGC3938 | PSF+Bulge+Disk | SA(s)c | A | 18.21 | -20.0 | 0.52 | 1.23e+10
NGC3949 | PSF+Bulge+Disk | SA(s)bc | AB | 18.21 | -19.9 | 0.45 | 8.66e+09
NGC4027 | Bar+Disk | SB(s)dm | B | 27.43 | -20.6 | 0.54 | 2.25e+10
NGC4030 | PSF+Bulge+Disk | SA(s)bc | A | 27.75 | -20.8 | - | -
NGC4051 | PSF+Bulge+Disk+Bar | SAB(rs)bc | AB | 18.21 | -20.0 | 0.65 | 1.95e+10
NGC4062 | PSF+Bulge+Disk | SA(s)c | A | 10.39 | -19.5 | 0.76 | 2.07e+10
NGC4123 | PSF+Bulge+Disk+Bar | SB(r)c | B | 17.68 | -19.8 | 0.61 | 1.50e+10
NGC4145 | PSF+Bulge+Disk | SAB(rs)d | AB | 22.18 | -20.1 | 0.51 | 1.34e+10
NGC4151 | PSF+Bulge+Disk+Bar | (R’)SAB(rs)ab | AB | 21.75 | -20.1 | 0.73 | 2.93e+10
NGC4212 | PSF+Bulge+Disk | SAc | A | 18.00 | -20.4 | 0.67 | 3.28e+10
NGC4254 | PSF+Bulge+Disk | SA(s)c | A | 18.00 | -22.6 | 0.57 | 1.61e+11
NGC4293 | PSF+Bulge+Disk+Bar | (R)SB(s)0 | B | 18.21 | -20.1 | 0.90 | 5.94e+10
NGC4303 | Bulge+Disk+Bar | SAB(rs)bc | AB | 16.29 | -21.8 | 0.53 | 6.76e+10
NGC4314 | PSF+Bulge+Disk+Bar | SB(rs)a | B | 10.39 | -19.8 | 0.85 | 3.69e+10
NGC4394 | PSF+Bulge+Disk+Bar | (R)SB(r)b | B | 18.00 | -19.4 | 0.85 | 2.61e+10
NGC4414 | PSF+Bulge+Disk+Bar | SA(rs)c | A | 10.39 | -20.0 | 0.84 | 4.38e+10
NGC4450 | PSF+Bulge+Disk+Bar | SA(s)ab | A | 18.00 | -21.9 | 0.82 | 2.22e+11
NGC4487 | PSF+Bulge+Disk+Bar | SAB(rs)cd | AB | 21.32 | -19.6 | - | -
NGC4490 | PSF+Bulge+Disk | SB(s)d;pec | B | 8.36 | -21.7 | 0.43 | 4.10e+10
NGC4527 | Bulge+Disk+Bar | SAB(s)bc | AB | 14.46 | -21.5 | 0.86 | 1.87e+11
NGC4548 | PSF+Bulge+Disk+Bar | SB(rs)b | B | 18.00 | -20.8 | 0.81 | 7.85e+10
NGC4593 | PSF+Bulge+Disk+Bar | (R)SB(rs)b | B | 42.32 | -20.8 | - | -
NGC4618 | PSF+Bar+Disk | SB(rs)m | B | 7.82 | -19.3 | 0.44 | 4.62e+09
NGC4643 | Bulge+Disk+Bar | SB(rs)0 | B | 27.54 | -19.9 | 0.96 | 6.49e+10
NGC4647 | Bulge+Disk | SAB(rs)c | AB | 18.00 | -19.8 | 0.65 | 1.67e+10
NGC4651 | PSF+Bulge+Disk | SA(rs)c | A | 18.00 | -19.6 | 0.57 | 1.05e+10
NGC4654 | Bulge+Disk | SAB(rs)cd | AB | 18.00 | -20.6 | 0.60 | 2.99e+10
NGC4665 | PSF+Bulge+Disk+Bar | SB(s)0 | B | 19.18 | -19.3 | - | -
NGC4689 | PSF+Bulge+Disk | SA(rs)bc | A | 18.00 | -20.7 | 0.65 | 3.84e+10
NGC4691 | PSF+Bulge+Disk+Bar | (R)SB(s)0;pec | B | 24.11 | -19.6 | 0.58 | 1.04e+10
NGC4698 | PSF+Bulge+Disk | SA(s)ab | A | 18.00 | -19.9 | 0.91 | 5.15e+10
NGC4699 | Bulge+Disk+Bar | SAB(rs)b | AB | 27.54 | -21.5 | 0.89 | 2.06e+11
NGC4772 | PSF+Bulge+Disk+Bar | SA(s)a | A | 17.46 | -19.4 | 0.92 | 3.30e+10
NGC4775 | PSF+Bulge+Disk | SA(s)d | A | 28.50 | -20.4 | - | -
NGC4781 | PSF+Bulge+Disk | SB(rs)d | B | 24.11 | -20.8 | - | -
NGC4818 | PSF+Bulge+Disk+Bar | SAB(rs)ab;pec | AB | 23.04 | -19.7 | 0.89 | 3.88e+10
NGC4856 | PSF+Bulge+Disk+Bar | SB(s)0 | B | 22.61 | -20.2 | 0.99 | 9.04e+10
NGC4902 | PSF+Bulge+Disk+Bar | SB(r)b | B | 42.00 | -21.4 | 0.69 | 8.33e+10
NGC4930 | PSF+Bulge+Disk+Bar | SB(rs)b | B | 37.50 | -21.2 | 0.90 | 1.61e+11
NGC4939 | PSF+Bulge+Disk | SA(s)bc | A | 47.46 | -22.2 | 0.64 | 1.43e+11
NGC4941 | PSF+Bulge+Disk | (R)SAB(r)ab | AB | 6.86 | -19.4 | 0.84 | 2.50e+10
NGC4995 | PSF+Bulge+Disk+Bar | SAB(rs)b | AB | 30.00 | -20.6 | 0.87 | 8.25e+10
NGC5054 | Bulge+Disk | SA(s)bc | A | 29.25 | -21.1 | 0.76 | 8.31e+10
NGC5085 | Bulge+Disk | SA(s)c | A | 30.96 | -19.3 | 0.87 | 2.54e+10
NGC5101 | PSF+Bulge+Disk+Bar | (R)SB(rs)0 | B | 29.36 | -20.8 | 1.00 | 1.69e+11
NGC5121 | PSF+Bulge+Disk | (R’)SA(s)a | A | 23.68 | -19.4 | 0.95 | 3.68e+10
NGC5161 | PSF+Bulge+Disk | SA(s)c | A | 35.89 | -21.7 | 0.79 | 1.65e+11
NGC5247 | PSF+Bulge+Disk | SA(s)bc | A | 23.79 | -21.2 | 0.54 | 3.86e+10
NGC5371 | PSF+Bulge+Disk+Bar | SAB(rs)bc | AB | 40.50 | -22.1 | 0.70 | 1.74e+11
NGC5427 | PSF+Bulge+Disk | SA(s)c;pec | A | 40.82 | -21.2 | 0.57 | 4.61e+10
NGC5483 | Bulge+Disk+Bar | SA(s)c | A | 26.46 | -20.3 | - | -
NGC5643 | PSF+Bulge+Disk+Bar | SAB(rs)c | AB | 18.11 | -20.9 | 0.74 | 6.68e+10
NGC5676 | PSF+Bulge+Disk | SA(rs)bc | A | 36.96 | -21.5 | 0.68 | 9.01e+10
NGC5701 | PSF+Bulge+Disk+Bar | (R)SB(rs)0 | B | 27.96 | -19.9 | 0.88 | 4.67e+10
NGC5713 | PSF+Bulge+Disk | SAB(rs)bc;pec | AB | 32.57 | -21.1 | 0.64 | 5.21e+10
NGC5850 | PSF+Bulge+Disk+Bar | SB(r)b | B | 30.54 | -21.5 | 0.79 | 1.37e+11
NGC5921 | PSF+Bulge+Disk+Bar | SB(r)bc | B | 27.00 | -20.6 | 0.66 | 3.51e+10
NGC5962 | Bulge+Disk+Bar | SA(r)c | A | 34.07 | -20.9 | 0.64 | 4.43e+10
NGC6215 | PSF+Bulge+Disk | SA(s)c | A | 21.96 | -20.4 | 0.54 | 1.84e+10
NGC6221 | Bulge+Disk+Bar | SB(s)bc;pec | B | 20.79 | -21.6 | 0.74 | 1.31e+11
NGC6300 | PSF+Bulge+Disk+Bar | SB(rs)b | B | 15.32 | -20.3 | 0.78 | 4.49e+10
NGC6384 | Bulge+Disk | SAB(r)bc | AB | 28.50 | -21.5 | 0.72 | 1.05e+11
NGC6753 | PSF+Bulge+Disk+Bar | (R)SA(r)b | A | 43.82 | -21.6 | 0.83 | 1.81e+11
NGC6782 | PSF+Bulge+Disk+Bar | (R)SAB(r)a | AB | 54.43 | -21.4 | - | -
NGC6902 | PSF+Bulge+Disk | SA(r)b | A | 38.25 | -21.2 | 0.71 | 8.12e+10
NGC6907 | PSF+Bulge+Disk+Bar | SB(s)bc | B | 46.07 | -21.8 | 0.69 | 1.21e+11
NGC7083 | PSF+Bulge+Disk | SA(s)bc | A | 41.46 | -21.8 | 0.65 | 1.05e+11
NGC7205 | PSF+Bulge+Disk | SA(s)bc | A | 21.96 | -20.7 | 0.60 | 3.11e+10
NGC7213 | PSF+Bulge+Disk | SA(s)a | A | 23.57 | -21.0 | 0.89 | 1.26e+11
NGC7217 | PSF+Bulge+Disk | (R)SA(r)ab | A | 17.14 | -20.5 | 0.90 | 8.38e+10
NGC7412 | PSF+Bulge+Disk+Bar | SB(s)b | B | 22.61 | -20.2 | 0.53 | 1.50e+10
NGC7479 | PSF+Bulge+Disk+Bar | SB(s)c | B | 34.71 | -21.7 | 0.75 | 1.44e+11
NGC7552 | Bulge+Disk+Bar | (R’)SB(s)ab | B | 20.89 | -20.5 | 0.68 | 3.49e+10
NGC7723 | PSF+Bulge+Disk+Bar | SB(r)b | B | 25.39 | -20.5 | 0.73 | 4.32e+10
NGC7727 | PSF+Bulge+Disk | SAB(s)a;pec | AB | 24.96 | -20.8 | 0.91 | 1.13e+11
NGC7741 | Bulge+Disk+Bar | SB(s)cd | B | 13.18 | -19.3 | 0.53 | 6.67e+09
Note. — Columns are: (1) Galaxy name. (2) The best fit chosen based on the
criteria outlined in § 3.3. (3) Hubble type from RC3 (de Vaucouleurs et al.
1991). (4) RC3 bar type, which is based on visual inspection of optical images
and runs as ‘B’=‘strongly barred’, ‘AB’=‘weakly barred’, and ‘A’=‘unbarred’.
(5) The distance in Mpc from the Nearby Galaxies Catalog (Tully 1988) scaled
to a Hubble constant of 70 km s-1 Mpc-1. Distances are adjusted assuming the
Galaxy is retarded by 300 km $s^{-1}$ from the universal expansion by the mass
of the Virgo Cluster. Exceptions are NGC 6753, NGC 6782, and NGC 6907 where
distances are from RC3, NGC 3504 where the distance is from Kenney et al.
(1993), and NGC 4314 where the distance is from Benedict et al. (1996). (6)
Absolute $B$-band magnitude from Hyperleda. Corrections have been applied for
galactic and internal extinction. Galactic absorption is computed based on
Schlegel et al. (1998). Internal extinction is corrected for following
Bottinelli et al. (1995). K-correction co-efficients are drawn from RC2 (de
Vaucouleurs et al. 1976). (7) $B-V$ color from Hyperleda. Corrections for
extinction and k-correction have been applied as in column 6. (8) Stellar
mass, calculated as outlined in § 2.2.
Table 2: Parameters From $2D$ Luminosity Decomposition for Sample S1 (N=143)
Galaxy Name | Best Fit | $B/T$ | $D/T$ | Bar/$T$ | $B/D$ | Bulge $r_{e}$ | Bulge $r_{e}$ | Bulge $n$ | Disk h | Disk h | Bar $r_{e}$ | Bar $r_{e}$ | Bar $n$
---|---|---|---|---|---|---|---|---|---|---|---|---|---
| | % | % | % | | ′′ | kpc | | ′′ | kpc | ′′ | kpc |
(1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | (9) | (10) | (11) | (12) | (13) | (14)
ESO138-10 | PSF+Bulge+Disk | 16.25 | 83.75 | - | 0.19 | 10.94 | 0.86 | 2.07 | 27.10 | 2.14 | - | - | -
IC0239 | PSF+Bulge+Disk+Bar | 2.45 | 92.65 | 4.91 | 0.03 | 6.63 | 0.41 | 0.23 | 34.56 | 2.15 | 13.95 | 0.87 | 0.19
IC4444 | Bulge+Disk | 31.97 | 68.03 | - | 0.47 | 7.94 | 1.06 | 2.33 | 17.67 | 2.36 | - | - | -
IC5325 | PSF+Bulge+Disk | 6.79 | 93.21 | - | 0.07 | 15.17 | 1.56 | 1.75 | 23.18 | 2.39 | - | - | -
NGC0150 | PSF+Bulge+Disk+Bar | 5.72 | 81.18 | 13.11 | 0.07 | 4.22 | 0.44 | 0.05 | 33.57 | 3.52 | 17.82 | 1.87 | 0.39
NGC0157 | PSF+Bulge+Disk | 2.29 | 97.71 | - | 0.02 | 3.44 | 0.41 | 0.44 | 31.82 | 3.78 | - | - | -
NGC0210 | Bulge+Disk+Bar | 30.07 | 48.54 | 21.39 | 0.62 | 4.81 | 0.57 | 1.71 | 83.21 | 9.92 | 29.41 | 3.51 | 0.31
NGC0278 | PSF+Bulge+Disk | 4.23 | 95.77 | - | 0.04 | 3.46 | 0.15 | 0.57 | 13.48 | 0.58 | - | - | -
NGC0289 | PSF+Bulge+Disk+Bar | 8.83 | 85.99 | 5.18 | 0.10 | 4.91 | 0.57 | 0.46 | 20.28 | 2.36 | 18.90 | 2.19 | 0.04
NGC0428 | PSF+Bulge+Disk+Bar | 5.50 | 71.64 | 22.86 | 0.08 | 9.27 | 0.67 | 0.38 | 32.55 | 2.34 | 30.70 | 2.21 | 0.37
NGC0488 | PSF+Bulge+Disk | 21.72 | 78.28 | - | 0.28 | 9.91 | 1.52 | 3.07 | 38.94 | 5.96 | - | - | -
NGC0578 | PSF+Bulge+Disk+Bar | 2.05 | 93.63 | 4.32 | 0.02 | 4.19 | 0.46 | 0.56 | 40.73 | 4.47 | 14.95 | 1.64 | 0.18
NGC0613 | PSF+Bulge+Disk+Bar | 13.14 | 56.97 | 29.90 | 0.23 | 6.12 | 0.64 | 1.05 | 44.69 | 4.64 | 62.49 | 6.49 | 0.53
NGC0685 | Bar+Disk | - | 96.53 | 3.47 | - | - | - | - | 41.24 | 4.28 | 20.68 | 2.15 | 0.16
NGC0779 | Bulge+Disk+Bar | 16.18 | 57.69 | 26.13 | 0.28 | 6.04 | 0.59 | 2.31 | 38.74 | 3.79 | 29.00 | 2.84 | 0.25
NGC0864 | PSF+Bulge+Disk+Bar | 2.77 | 86.88 | 10.35 | 0.03 | 3.52 | 0.38 | 0.66 | 39.54 | 4.22 | 21.07 | 2.25 | 0.37
NGC0908 | PSF+Bulge+Disk | 8.86 | 91.14 | - | 0.10 | 7.18 | 0.84 | 1.62 | 48.96 | 5.72 | - | - | -
NGC1042 | PSF+Bulge+Disk | 3.00 | 97.00 | - | 0.03 | 5.91 | 0.57 | 0.09 | 39.83 | 3.86 | - | - | -
NGC1073 | PSF+Bulge+Disk+Bar | 2.36 | 79.58 | 18.06 | 0.03 | 5.47 | 0.46 | 0.41 | 50.02 | 4.17 | 34.24 | 2.85 | 0.83
NGC1084 | PSF+Bulge+Disk | 5.23 | 94.77 | - | 0.06 | 3.24 | 0.32 | 0.77 | 18.68 | 1.82 | - | - | -
NGC1087 | Bar+Disk | - | 93.13 | 6.87 | - | - | - | - | 28.55 | 2.78 | 7.79 | 0.76 | 1.27
NGC1187 | PSF+Bulge+Disk+Bar | 6.94 | 64.96 | 28.10 | 0.11 | 4.81 | 0.51 | 1.57 | 38.20 | 4.06 | 51.85 | 5.51 | 0.72
NGC1241 | PSF+Bulge+Disk+Bar | 11.70 | 69.10 | 19.21 | 0.17 | 2.61 | 0.70 | 0.91 | 27.69 | 7.43 | 17.89 | 4.80 | 0.49
NGC1300 | PSF+Bulge+Disk+Bar | 11.04 | 74.94 | 14.02 | 0.15 | 4.63 | 0.51 | 1.19 | 63.53 | 6.96 | 65.89 | 7.21 | 0.25
NGC1302 | PSF+Bulge+Disk+Bar | 17.65 | 63.79 | 18.57 | 0.28 | 4.96 | 0.59 | 2.01 | 44.92 | 5.34 | 20.44 | 2.43 | 0.76
NGC1309 | PSF+Bulge+Disk | 27.07 | 72.93 | - | 0.37 | 10.86 | 1.68 | 2.27 | 16.67 | 2.58 | - | - | -
NGC1317 | Bulge+Disk+Bar | 13.86 | 41.48 | 44.65 | 0.33 | 3.99 | 0.53 | 1.94 | 47.37 | 6.31 | 20.22 | 2.69 | 2.19
NGC1350 | PSF+Bulge+Disk+Bar | 19.52 | 69.99 | 10.50 | 0.28 | 7.41 | 0.95 | 1.59 | 52.77 | 6.73 | 51.07 | 6.51 | 0.31
NGC1371 | PSF+Bulge+Disk+Bar | 12.46 | 80.16 | 7.38 | 0.16 | 5.17 | 0.52 | 1.11 | 29.34 | 2.93 | 18.92 | 1.89 | 0.36
NGC1385 | PSF+Bar+Disk | - | 75.54 | 24.46 | - | - | - | - | 28.69 | 2.87 | 15.63 | 1.56 | 1.28
NGC1511 | Bulge+Disk | 28.47 | 71.53 | - | 0.40 | 27.78 | 2.55 | 1.35 | 23.99 | 2.20 | - | - | -
NGC1559 | PSF+Bar+Disk | - | 97.06 | 2.94 | - | - | - | - | 31.16 | 2.86 | 9.84 | 0.90 | 0.53
NGC1703 | PSF+Bulge+Disk | 6.32 | 93.68 | - | 0.07 | 2.83 | 0.30 | 0.56 | 18.24 | 1.91 | - | - | -
NGC1792 | PSF+Bulge+Disk | 2.65 | 97.35 | - | 0.03 | 3.72 | 0.31 | 0.84 | 38.50 | 3.23 | - | - | -
NGC1808 | PSF+Bulge+Disk+Bar | 3.92 | 74.27 | 21.82 | 0.05 | 2.78 | 0.19 | 0.45 | 38.78 | 2.71 | 5.71 | 0.40 | 0.67
NGC1964 | PSF+Bulge+Disk+Bar | 41.78 | 49.87 | 8.35 | 0.84 | 8.31 | 0.97 | 2.58 | 18.59 | 2.17 | 8.05 | 0.94 | 0.05
NGC2090 | Bulge+Disk+Bar | 14.67 | 58.39 | 26.94 | 0.25 | 17.68 | 1.12 | 2.19 | 160.02 | 10.17 | 33.74 | 2.14 | 0.43
NGC2139 | Bulge+Disk | 15.16 | 84.84 | - | 0.18 | 8.27 | 1.04 | 1.53 | 18.12 | 2.27 | - | - | -
NGC2196 | PSF+Bulge+Disk | 46.38 | 53.62 | - | 0.86 | 13.42 | 2.12 | 2.38 | 28.38 | 4.48 | - | - | -
NGC2442 | PSF+Bulge+Disk+Bar | 13.33 | 62.47 | 24.19 | 0.21 | 4.78 | 0.46 | 1.68 | 78.46 | 7.55 | 60.15 | 5.79 | 0.24
NGC2559 | Bulge+Disk+Bar | 6.98 | 81.64 | 11.37 | 0.09 | 4.01 | 0.43 | 1.38 | 34.65 | 3.72 | 21.40 | 2.30 | 0.24
NGC2566 | Bulge+Disk | 22.71 | 77.29 | - | 0.29 | 1.74 | 0.20 | 4.42 | 23.05 | 2.60 | - | - | -
NGC2775 | PSF+Bulge+Disk | 60.87 | 39.13 | - | 1.56 | 47.35 | 4.37 | 4.85 | 28.09 | 2.59 | - | - | -
NGC3059 | Bar+Disk | - | 90.68 | 9.32 | - | - | - | - | 64.22 | 5.92 | 23.94 | 2.21 | 1.43
NGC3166 | Bulge+Disk+Bar | 24.97 | 50.74 | 24.29 | 0.49 | 3.25 | 0.30 | 0.81 | 20.14 | 1.84 | 13.85 | 1.26 | 0.53
NGC3169 | PSF+Bulge+Disk | 38.53 | 61.47 | - | 0.63 | 11.01 | 0.96 | 2.41 | 55.12 | 4.79 | - | - | -
NGC3223 | PSF+Bulge+Disk | 11.71 | 88.29 | - | 0.13 | 5.39 | 1.06 | 1.74 | 29.05 | 5.73 | - | - | -
NGC3227 | PSF+Bulge+Disk+Bar | 10.78 | 45.23 | 44.00 | 0.24 | 4.84 | 0.44 | 0.32 | 44.67 | 4.07 | 39.51 | 3.60 | 1.27
NGC3261 | PSF+Bulge+Disk+Bar | 15.06 | 70.15 | 14.79 | 0.21 | 2.95 | 0.52 | 1.96 | 28.89 | 5.08 | 14.19 | 2.49 | 1.05
NGC3275 | PSF+Bulge+Disk+Bar | 14.15 | 63.44 | 22.41 | 0.22 | 2.21 | 0.48 | 1.85 | 28.25 | 6.20 | 20.75 | 4.55 | 0.94
NGC3319 | PSF+Bar+Disk | - | 91.41 | 8.59 | - | - | - | - | 66.08 | 14.49 | 14.60 | 3.20 | 0.43
NGC3338 | PSF+Bulge+Disk+Bar | 5.29 | 81.96 | 12.75 | 0.06 | 4.39 | 0.39 | 0.76 | 36.42 | 3.25 | 20.62 | 1.84 | 0.39
NGC3423 | Bulge+Disk | 28.10 | 71.90 | - | 0.39 | 24.09 | 1.39 | 2.39 | 31.21 | 1.80 | - | - | -
NGC3504 | Bulge+Disk+Bar | 23.15 | 39.49 | 37.37 | 0.59 | 2.37 | 0.25 | 1.04 | 29.55 | 3.09 | 26.02 | 2.72 | 0.78
NGC3513 | Bulge+Disk+Bar | 4.45 | 92.08 | 3.47 | 0.05 | 10.73 | 0.88 | 1.46 | 33.18 | 2.72 | 18.32 | 1.50 | 0.08
NGC3583 | Bulge+Disk+Bar | 8.62 | 53.40 | 37.98 | 0.16 | 1.29 | 0.19 | 1.28 | 21.78 | 3.19 | 14.89 | 2.18 | 0.68
NGC3596 | PSF+Bulge+Disk | 12.15 | 87.85 | - | 0.14 | 5.34 | 0.43 | 0.67 | 18.54 | 1.50 | - | - | -
NGC3646 | PSF+Bulge+Disk | 13.53 | 86.47 | - | 0.16 | 3.60 | 1.04 | 1.20 | 26.01 | 7.54 | - | - | -
NGC3675 | PSF+Bulge+Disk | 44.42 | 55.58 | - | 0.80 | 28.41 | 1.42 | 2.66 | 47.12 | 2.35 | - | - | -
NGC3684 | PSF+Bulge+Disk | 26.92 | 73.08 | - | 0.37 | 11.43 | 1.10 | 2.15 | 15.96 | 1.53 | - | - | -
NGC3686 | PSF+Bulge+Disk+Bar | 3.02 | 95.00 | 1.98 | 0.03 | 4.39 | 0.31 | 0.66 | 27.57 | 1.96 | 21.32 | 1.52 | 0.03
NGC3705 | PSF+Bulge+Disk+Bar | 27.73 | 59.20 | 13.07 | 0.47 | 8.62 | 0.63 | 1.83 | 37.77 | 2.74 | 30.00 | 2.18 | 0.07
NGC3726 | PSF+Bulge+Disk+Bar | 3.09 | 95.10 | 1.81 | 0.03 | 5.64 | 0.37 | 0.53 | 49.58 | 3.24 | 41.09 | 2.69 | 0.05
NGC3810 | PSF+Bulge+Disk | 38.43 | 61.57 | - | 0.62 | 12.35 | 0.82 | 1.28 | 23.78 | 1.57 | - | - | -
NGC3885 | PSF+Bulge+Disk | 27.10 | 72.90 | - | 0.37 | 3.13 | 0.41 | 0.46 | 11.94 | 1.57 | - | - | -
NGC3887 | PSF+Bulge+Disk+Bar | 4.19 | 83.01 | 12.80 | 0.05 | 5.27 | 0.44 | 1.01 | 42.69 | 3.55 | 37.62 | 3.13 | 0.62
NGC3893 | PSF+Bulge+Disk | 54.61 | 45.39 | - | 1.20 | 20.05 | 1.31 | 2.05 | 23.40 | 1.52 | - | - | -
NGC3938 | PSF+Bulge+Disk | 7.30 | 92.70 | - | 0.08 | 6.43 | 0.34 | 1.18 | 32.84 | 1.75 | - | - | -
NGC3949 | PSF+Bulge+Disk | 7.75 | 92.25 | - | 0.08 | 4.67 | 0.22 | 0.64 | 15.22 | 0.72 | - | - | -
NGC4027 | Bar+Disk | - | 79.48 | 20.52 | - | - | - | - | 43.11 | 2.03 | 26.50 | 1.25 | 1.80
NGC4030 | PSF+Bulge+Disk | 47.29 | 52.71 | - | 0.90 | 17.18 | 1.73 | 2.22 | 20.67 | 2.08 | - | - | -
NGC4051 | PSF+Bulge+Disk+Bar | 14.99 | 46.99 | 38.02 | 0.32 | 11.29 | 0.54 | 1.20 | 65.81 | 3.12 | 69.69 | 3.31 | 0.40
NGC4062 | PSF+Bulge+Disk | 1.69 | 98.31 | - | 0.02 | 3.62 | 0.19 | 0.50 | 34.98 | 1.79 | - | - | -
NGC4123 | PSF+Bulge+Disk+Bar | 6.87 | 79.23 | 13.90 | 0.09 | 6.55 | 0.56 | 0.59 | 56.76 | 4.84 | 42.98 | 3.66 | 0.46
NGC4145 | PSF+Bulge+Disk | 7.75 | 92.25 | - | 0.08 | 12.90 | 0.82 | 0.68 | 63.59 | 4.05 | - | - | -
NGC4151 | PSF+Bulge+Disk+Bar | 41.32 | 50.26 | 8.42 | 0.82 | 9.36 | 0.62 | 0.44 | 37.74 | 2.49 | 57.00 | 3.76 | 0.10
NGC4212 | PSF+Bulge+Disk | 3.67 | 96.33 | - | 0.04 | 1.44 | 0.01 | 1.38 | 22.56 | 0.13 | - | - | -
NGC4254 | PSF+Bulge+Disk | 38.98 | 61.02 | - | 0.64 | 30.46 | 5.03 | 2.68 | 34.19 | 5.64 | - | - | -
NGC4293 | PSF+Bulge+Disk+Bar | 8.49 | 61.70 | 29.81 | 0.14 | 8.00 | 0.40 | 0.99 | 58.26 | 2.88 | 55.17 | 2.73 | 0.46
NGC4303 | Bulge+Disk+Bar | 8.28 | 82.81 | 8.91 | 0.10 | 2.67 | 0.30 | 1.55 | 44.58 | 4.93 | 32.35 | 3.58 | 0.55
NGC4314 | PSF+Bulge+Disk+Bar | 26.49 | 52.44 | 21.07 | 0.51 | 11.22 | 0.74 | 2.05 | 62.35 | 4.14 | 48.98 | 3.25 | 0.35
NGC4394 | PSF+Bulge+Disk+Bar | 15.20 | 69.58 | 15.22 | 0.22 | 4.20 | 0.22 | 1.63 | 41.74 | 2.22 | 28.70 | 1.53 | 0.59
NGC4414 | PSF+Bulge+Disk+Bar | 11.48 | 83.86 | 4.65 | 0.14 | 3.50 | 0.17 | 1.72 | 21.27 | 1.05 | 17.66 | 0.87 | 0.05
NGC4450 | PSF+Bulge+Disk+Bar | 16.86 | 71.49 | 11.65 | 0.24 | 8.19 | 1.10 | 2.26 | 47.16 | 6.33 | 36.13 | 4.85 | 0.33
NGC4487 | PSF+Bulge+Disk+Bar | 3.10 | 93.77 | 3.13 | 0.03 | 6.66 | 0.47 | 0.83 | 27.92 | 1.99 | 14.52 | 1.04 | 0.27
NGC4490 | PSF+Bulge+Disk | 6.64 | 93.36 | - | 0.07 | 21.41 | 0.88 | 0.35 | 44.02 | 1.81 | - | - | -
NGC4527 | Bulge+Disk+Bar | 18.24 | 63.84 | 17.91 | 0.29 | 5.17 | 0.61 | 1.99 | 55.52 | 6.59 | 41.19 | 4.89 | 0.50
NGC4548 | PSF+Bulge+Disk+Bar | 12.96 | 68.65 | 18.39 | 0.19 | 6.97 | 0.23 | 1.56 | 58.22 | 1.96 | 44.91 | 1.51 | 0.51
NGC4593 | PSF+Bulge+Disk+Bar | 25.13 | 28.15 | 46.72 | 0.89 | 5.27 | 0.97 | 1.04 | 33.07 | 6.11 | 41.94 | 7.75 | 0.64
NGC4618 | PSF+Bar+Disk | - | 86.11 | 13.89 | - | - | - | - | 46.22 | 8.54 | 17.05 | 3.15 | 0.66
NGC4643 | Bulge+Disk+Bar | 24.96 | 54.10 | 20.95 | 0.46 | 5.43 | 0.52 | 2.53 | 48.22 | 4.64 | 21.30 | 2.05 | 0.62
NGC4647 | Bulge+Disk | 33.39 | 66.61 | - | 0.50 | 12.90 | 1.26 | 1.61 | 42.63 | 4.17 | - | - | -
NGC4651 | PSF+Bulge+Disk | 41.67 | 58.33 | - | 0.71 | 18.52 | 1.01 | 0.63 | 30.46 | 1.66 | - | - | -
NGC4654 | Bulge+Disk | 1.74 | 98.26 | - | 0.02 | 3.78 | 0.27 | 0.93 | 30.16 | 2.18 | - | - | -
NGC4665 | PSF+Bulge+Disk+Bar | 14.74 | 67.38 | 17.89 | 0.22 | 6.92 | 0.37 | 1.12 | 50.48 | 2.73 | 30.68 | 1.66 | 0.70
NGC4689 | PSF+Bulge+Disk | 5.22 | 94.78 | - | 0.06 | 6.97 | 0.73 | 1.14 | 41.85 | 4.38 | - | - | -
NGC4691 | PSF+Bulge+Disk+Bar | 15.80 | 69.09 | 15.11 | 0.23 | 11.62 | 0.89 | 0.61 | 33.90 | 2.59 | 24.99 | 1.91 | 0.42
NGC4698 | PSF+Bulge+Disk | 21.72 | 78.28 | - | 0.28 | 6.78 | 0.48 | 1.78 | 28.34 | 2.02 | - | - | -
NGC4699 | Bulge+Disk+Bar | 19.93 | 76.48 | 3.59 | 0.26 | 2.62 | 0.27 | 2.08 | 15.93 | 1.62 | 13.49 | 1.37 | 0.02
NGC4772 | PSF+Bulge+Disk+Bar | 34.25 | 51.55 | 14.20 | 0.66 | 9.29 | 0.67 | 1.49 | 58.16 | 4.17 | 52.58 | 3.77 | 0.50
NGC4775 | PSF+Bulge+Disk | 32.85 | 67.15 | - | 0.49 | 21.43 | 2.31 | 1.75 | 18.48 | 1.99 | - | - | -
NGC4781 | PSF+Bulge+Disk | 10.18 | 89.82 | - | 0.11 | 12.02 | 1.04 | 0.94 | 32.82 | 2.85 | - | - | -
NGC4818 | PSF+Bulge+Disk+Bar | 7.04 | 73.74 | 19.22 | 0.10 | 2.82 | 0.21 | 0.47 | 36.31 | 2.69 | 16.38 | 1.21 | 0.51
NGC4856 | PSF+Bulge+Disk+Bar | 27.44 | 61.22 | 11.35 | 0.45 | 5.95 | 0.51 | 1.53 | 30.61 | 2.64 | 17.07 | 1.47 | 0.44
NGC4902 | PSF+Bulge+Disk+Bar | 5.59 | 84.44 | 9.97 | 0.07 | 4.31 | 0.80 | 0.85 | 29.24 | 5.45 | 14.13 | 2.64 | 0.39
NGC4930 | PSF+Bulge+Disk+Bar | 23.74 | 61.29 | 14.98 | 0.39 | 6.46 | 1.14 | 1.37 | 45.02 | 7.93 | 32.96 | 5.80 | 0.38
NGC4939 | PSF+Bulge+Disk | 21.37 | 78.63 | - | 0.27 | 9.52 | 2.01 | 2.78 | 32.14 | 6.79 | - | - | -
NGC4941 | PSF+Bulge+Disk | 15.08 | 84.92 | - | 0.18 | 4.30 | 0.25 | 0.84 | 23.23 | 1.36 | - | - | -
NGC4995 | PSF+Bulge+Disk+Bar | 6.62 | 87.18 | 6.20 | 0.08 | 4.14 | 0.50 | 0.05 | 19.56 | 2.34 | 18.03 | 2.16 | 0.34
NGC5054 | Bulge+Disk | 9.64 | 90.36 | - | 0.11 | 5.28 | 0.65 | 2.42 | 52.65 | 6.48 | - | - | -
NGC5085 | Bulge+Disk | 6.04 | 93.96 | - | 0.06 | 5.53 | 0.74 | 1.48 | 33.88 | 4.55 | - | - | -
NGC5101 | PSF+Bulge+Disk+Bar | 32.63 | 48.31 | 19.06 | 0.68 | 7.61 | 0.96 | 2.28 | 28.39 | 3.59 | 37.85 | 4.78 | 0.31
NGC5121 | PSF+Bulge+Disk | 36.43 | 63.57 | - | 0.57 | 4.68 | 0.48 | 2.41 | 15.62 | 1.61 | - | - | -
NGC5161 | PSF+Bulge+Disk | 27.14 | 72.86 | - | 0.37 | 20.54 | 3.38 | 1.47 | 43.98 | 7.24 | - | - | -
NGC5247 | PSF+Bulge+Disk | 7.54 | 92.46 | - | 0.08 | 8.46 | 0.79 | 1.04 | 50.85 | 4.78 | - | - | -
NGC5371 | PSF+Bulge+Disk+Bar | 8.60 | 76.08 | 15.32 | 0.11 | 3.13 | 0.55 | 1.29 | 55.81 | 9.84 | 23.03 | 4.06 | 1.05
NGC5427 | PSF+Bulge+Disk | 7.76 | 92.24 | - | 0.08 | 4.74 | 0.45 | 0.60 | 24.59 | 2.31 | - | - | -
NGC5483 | Bulge+Disk+Bar | 0.98 | 91.28 | 7.73 | 0.01 | 3.23 | 0.39 | 1.22 | 23.25 | 2.83 | 8.74 | 1.06 | 0.32
NGC5643 | PSF+Bulge+Disk+Bar | 8.21 | 81.78 | 10.01 | 0.10 | 5.68 | 0.46 | 2.14 | 45.30 | 3.63 | 48.58 | 3.89 | 0.43
NGC5676 | PSF+Bulge+Disk | 6.67 | 93.33 | - | 0.07 | 3.71 | 0.54 | 1.29 | 23.10 | 3.39 | - | - | -
NGC5701 | PSF+Bulge+Disk+Bar | 24.43 | 63.47 | 12.09 | 0.38 | 11.13 | 1.19 | 2.41 | 70.67 | 7.56 | 26.01 | 2.78 | 0.40
NGC5713 | PSF+Bulge+Disk | 33.40 | 66.60 | - | 0.50 | 15.59 | 2.02 | 1.84 | 18.52 | 2.40 | - | - | -
NGC5850 | PSF+Bulge+Disk+Bar | 14.76 | 63.74 | 21.50 | 0.23 | 6.22 | 1.06 | 1.58 | 72.24 | 12.29 | 46.05 | 7.83 | 0.90
NGC5921 | PSF+Bulge+Disk+Bar | 10.43 | 70.50 | 19.06 | 0.15 | 2.55 | 0.26 | 1.97 | 40.88 | 4.10 | 35.05 | 3.51 | 0.92
NGC5962 | Bulge+Disk+Bar | 10.30 | 78.82 | 10.88 | 0.13 | 2.33 | 0.32 | 1.37 | 14.56 | 1.99 | 13.85 | 1.89 | 0.02
NGC6215 | PSF+Bulge+Disk | 6.95 | 93.05 | - | 0.07 | 2.55 | 0.27 | 0.96 | 14.02 | 1.47 | - | - | -
NGC6221 | Bulge+Disk+Bar | 5.27 | 81.98 | 12.76 | 0.06 | 3.16 | 0.29 | 2.73 | 51.47 | 4.78 | 19.33 | 1.80 | 0.87
NGC6300 | PSF+Bulge+Disk+Bar | 5.98 | 86.55 | 7.47 | 0.07 | 5.43 | 0.42 | 1.44 | 47.08 | 3.60 | 31.64 | 2.42 | 0.30
NGC6384 | Bulge+Disk | 27.92 | 72.08 | - | 0.39 | 15.66 | 1.82 | 3.13 | 40.05 | 4.65 | - | - | -
NGC6753 | PSF+Bulge+Disk+Bar | 5.08 | 72.91 | 22.02 | 0.07 | 1.50 | 0.32 | 0.94 | 19.17 | 4.12 | 7.88 | 1.69 | 0.80
NGC6782 | PSF+Bulge+Disk+Bar | 26.27 | 51.75 | 21.98 | 0.51 | 3.90 | 1.03 | 1.10 | 27.14 | 7.20 | 19.27 | 5.11 | 0.38
NGC6902 | PSF+Bulge+Disk | 40.55 | 59.45 | - | 0.68 | 12.68 | 2.43 | 3.03 | 30.13 | 5.77 | - | - | -
NGC6907 | PSF+Bulge+Disk+Bar | 10.32 | 63.12 | 26.56 | 0.16 | 3.49 | 0.75 | 1.10 | 28.99 | 6.19 | 23.37 | 4.99 | 0.33
NGC7083 | PSF+Bulge+Disk | 13.23 | 86.77 | - | 0.15 | 5.87 | 1.23 | 1.37 | 23.26 | 4.86 | - | - | -
NGC7205 | PSF+Bulge+Disk | 6.08 | 93.92 | - | 0.06 | 4.70 | 0.48 | 0.88 | 31.03 | 3.16 | - | - | -
NGC7213 | PSF+Bulge+Disk | 65.74 | 34.26 | - | 1.92 | 19.55 | 2.42 | 2.68 | 63.19 | 7.83 | - | - | -
NGC7217 | PSF+Bulge+Disk | 53.71 | 46.29 | - | 1.16 | 21.07 | 1.36 | 2.21 | 26.84 | 1.73 | - | - | -
NGC7412 | PSF+Bulge+Disk+Bar | 4.09 | 68.60 | 27.31 | 0.06 | 2.84 | 0.34 | 1.28 | 23.58 | 2.78 | 35.06 | 4.13 | 1.20
NGC7479 | PSF+Bulge+Disk+Bar | 8.76 | 63.51 | 27.72 | 0.14 | 6.00 | 0.98 | 1.09 | 37.59 | 6.17 | 40.63 | 6.67 | 0.47
NGC7552 | Bulge+Disk+Bar | 23.44 | 61.08 | 15.48 | 0.38 | 2.70 | 0.31 | 0.64 | 17.54 | 1.99 | 42.05 | 4.78 | 0.24
NGC7723 | PSF+Bulge+Disk+Bar | 5.11 | 85.03 | 9.85 | 0.06 | 2.61 | 0.34 | 0.54 | 21.70 | 2.82 | 21.81 | 2.84 | 0.90
NGC7727 | PSF+Bulge+Disk | 41.77 | 58.23 | - | 0.72 | 8.31 | 1.05 | 2.02 | 24.46 | 3.09 | - | - | -
NGC7741 | Bulge+Disk+Bar | 3.09 | 89.02 | 7.90 | 0.03 | 9.51 | 0.50 | 0.31 | 60.98 | 3.18 | 30.02 | 1.56 | 0.40
Note. — Columns are: (1) Galaxy name. (2) The best fit chosen based on the
criteria outlined in § 3.3. (3) Bulge-to-total light ratio. If a PSF is
included in the model, the contribution from the PSF is also added in. (4)
Disk-to-total light ratio. (5) Bar-to-total light ratio. (6) Bulge-to-disk
ratio (7) Bulge effective radius in arcseconds. (8) Bulge effective radius in
kpc. (9) Bulge Sérsic index. (10) Disk scalelength in arcseconds. (11) Disk
scalelength in kpc. (12) Bar effective radius in arcseconds. (13) Bar Bulge
effective radius in kpc. (14) Bar Sérsic index.
Table 3: Decomposition For NGC 4643
Fit | | $r_{e}$ or h | $r_{e}$ or h | $n$ | $b/a$ | Position Angle | Fractional light
---|---|---|---|---|---|---|---
| | (′′) | (kpc) | | | |
(1) | (2) | (3) | (4) | (5) | (6) | (7) | (8)
Stage 1 | Sérsic | 27.90 | 2.66 | 4.44 | 0.80 | -51.03 | 100%
Stage 2 | Bulge | 23.86 | 2.30 | 4.16 | 0.80 | -51.08 | 34.6%
| Disk | 335.88 | 32.33 | 1.00 | 0.84 | 66.94 | 65.4%
Stage 3 | Bulge | 5.43 | 0.52 | 2.53 | 0.90 | 60.52 | 25.0%
| Disk | 48.22 | 4.64 | 1.00 | 0.84 | 66.94 | 54.1%
| Bar | 21.30 | 2.05 | 0.62 | 0.37 | -45.84 | 20.9%
Note. — Columns are: (1) Indicates whether the model is for Stage 1 (pure
Sérsic profile), Stage 2 (bulge+disk or bar+disk), or Stage 3
(bulge+disk+bar). (2) One of pure Sérsic profile, Sérsic bulge, exponential
disk, or Sérsic bar. (3) Effective radius of the bulge/bar or disk scalelength
in arcseconds. (4) Effective radius of the bulge/bar or disk scalelength in
kpc. (5) The Sérsic index of the profile. (6) Axis ratio of the component. (7)
Position angle in degrees of the component. Positive angles are East of North.
(8) The percentage of total galaxy light contributed by the component.
Table 4: Decomposition For NGC 4548
Fit | | $r_{e}$ or h | $r_{e}$ or h | $n$ | $b/a$ | Position Angle | Fractional light
---|---|---|---|---|---|---|---
| | (′′) | (kpc) | | | |
(1) | (2) | (3) | (4) | (5) | (6) | (7) | (8)
Stage 1 | Sérsic | 154.59 | 5.19 | 5.19 | 0.80 | 78.31 | 100%
Stage 2 | Bulge | 57.86 | 1.94 | 4.32 | 0.76 | 75.77 | 61.5%
| Disk | 60.39 | 2.03 | 1.00 | 0.75 | -32.54 | 38.5%
Stage 3 | Bulge | 6.98 | 0.23 | 1.56 | 0.88 | -66.50 | 13.0%
| Disk | 58.22 | 1.96 | 1.00 | 0.75 | -32.54 | 68.6%
| Bar | 44.91 | 1.51 | 0.51 | 0.35 | 66.65 | 18.4%
Note. — Columns are: (1) Indicates whether the model is for Stage 1 (pure
Sérsic profile), Stage 2 (bulge+disk or bar+disk), or Stage 3
(bulge+disk+bar). (2) One of pure Sérsic profile, Sérsic bulge, exponential
disk, or Sérsic bar. (3) Effective radius of the bulge/bar or disk scalelength
in arcseconds. (4) Effective radius of the bulge/bar or disk scalelength in
kpc. (5) The Sérsic index of the profile. (6) Axis ratio of the component. (7)
Position angle in degrees of the component. Positive angles are East of North.
(8) The percentage of total galaxy light contributed by the component.
Table 5: Decomposition For NGC 4902
Fit | | $r_{e}$ or h | $r_{e}$ or h | $n$ | $b/a$ | Position Angle | Fractional light
---|---|---|---|---|---|---|---
| | (′′) | (kpc) | | | |
(1) | (2) | (3) | (4) | (5) | (6) | (7) | (8)
Stage 1 | Sérsic | 154.4 | 28.8 | 5.11 | 0.54 | 69.0 | 100%
Stage 2 | Bulge | 11.5 | 2.15 | 1.02 | 0.38 | 68.5 | 17.4%
| Disk | 32.8 | 6.12 | 1.00 | 0.84 | 81.1 | 82.6%
Stage 3 | Bulge | 4.31 | 0.80 | 0.85 | 0.68 | -52.96 | 5.59%
| Disk | 29.2 | 5.45 | 1.00 | 0.84 | 81.1 | 82.6%
| Bar | 14.1 | 2.64 | 0.39 | 0.22 | 66.37 | 9.97%
Note. — Columns are: (1) Indicates whether the model is for Stage 1 (pure
Sérsic profile), Stage 2 (bulge+disk or bar+disk), or Stage 3
(bulge+disk+bar). (2) One of pure Sérsic profile, Sérsic bulge, exponential
disk, or Sérsic bar. (3) Effective radius of the bulge/bar or disk scalelength
in arcseconds. (4) Effective radius of the bulge/bar or disk scalelength in
kpc. (5) The Sérsic index of the profile. (6) Axis ratio of the component. (7)
Position angle in degrees of the component. Positive angles are East of North.
(8) The percentage of total galaxy light contributed by the component.
Table 6: Checking GALFIT robustness with different input guesses for Stage 3
| $B/T$ | Bulge $r_{e}$ | Bulge $n$ | $D/T$ | Disk $h$ | Bar/$T$ | Bar $r_{e}$ | Bar $n$
---|---|---|---|---|---|---|---|---
| | (”) | | | (”) | | (”) |
(1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | (9)
NGC 4548
Initial guesses from 1D decomposition | 17.5% | 7.39 | 1.17 | 63.5% | 28.4 | 19.0% | 37.5 | 0.54
Stage 3 Output | 13.0% | 6.98 | 1.56 | 68.6% | 58.2 | 18.4% | 44.9 | 0.51
Initial guesses from from Stage 2 | 11.1% | 7.50 | 1.70 | 69.9% | 64.5 | 19.1% | 37.5 | 0.54
Stage 3 Output | 13.0% | 6.98 | 1.56 | 68.6% | 58.2 | 18.4% | 44.9 | 0.51
NGC 4643
Initial guesses from 1D decomposition | 33.6% | 7.18 | 0.86 | 40.4% | 37.5 | 26.0% | 22.0 | 0.60
Stage 3 Output | 25.0% | 5.43 | 2.53 | 54.1% | 48.2 | 20.9% | 21.3 | 0.62
Initial guesses from Stage 2 | 24.1% | 5.30 | 2.5 | 51.8% | 46.4 | 24.1% | 22.0 | 0.60
Stage 3 Output | 25.0% | 5.43 | 2.53 | 54.1% | 48.2 | 20.9% | 21.3 | 0.62
Note. — For each galaxy, the rows are: (1) The initial guesses to Stage 3
provided by 1D decomposition. (2) The corresponding output for the input of
row (1). (3) The initial guesses to Stage 3 provided by Stage 2. (4) The
corresponding output for the input of row (3).
The columns are: (2) Bulge-to-total luminosity fraction. (3) Effective radius
of the bulge in arcseconds. (4) The Sérsic index of the profile. (5) Disk-to-
total luminosity fraction. (6) Disk scalelength in arcseconds. (7) Bar-to-
total luminosity fraction. (8) Bar effective radius in arcseconds. (9) Bar
Sérsic index.
Table 7: Parameter Coupling
Type of Fit | $\chi^{2}$ | Bulge $r_{e}$ | Bulge $n$ | $B/T$ | Disk $h$ | $D/T$ | Bar $r_{e}$ | Bar $n$ | Bar/$T$
---|---|---|---|---|---|---|---|---|---
| | (”) | | | (”) | | (”) | |
(1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | (9) | (10)
NGC 3885
initial input | 11241.33 | 2.32 | 2.00 | 27.7% | 10.44 | 72.3% | - | - | -
n=free | 8210.12 | 3.13 | 0.46 | 27.1% | 11.94 | 72.9% | - | - | -
n=1 fixed | [8307.82] | [2.96] | [1.00] | [29.1%] | [12.90] | [70.9%] | - | - | -
n=4 fixed | [7373.55] | [6.28] | [4.00] | [53.7%] | [20.45] | [46.3%] | - | - | -
n=1 re-fit | 8210.12 | 3.13 | 0.46 | 27.1% | 11.94 | 72.9% | - | - | -
n=4 re-fit | 7258.46 | 7.11 | 4.49 | 57.3% | 19.87 | 42.7% | - | - | -
NGC 7213
initial input | 605757.39 | 5.80 | 4.00 | 56.4% | 9.25 | 43.6% | - | - | -
n=free | 71170.10 | 19.55 | 2.68 | 65.7% | 63.19 | 34.3% | - | - | -
n=1 fixed | [96312.29] | [7.81] | [1.00] | [31.4%] | [27.79] | [68.6%] | - | - | -
n=4 fixed | [70998.69] | [41.11] | [4.00] | [95.4%] | [6.75] | [4.6%] | - | - | -
n=1 re-fit | 71170.10 | 19.53 | 2.68 | 65.7% | 63.16 | 34.3% | - | - | -
n=4 re-fit | 69970.51 | 72.04 | 5.60 | 91.9% | 7.98 | 8.1% | - | - | -
NGC 4643
initial input | 7841.46 | 5.34 | 2.50 | 24.1% | 46.40 | 51.8% | 22.04 | 0.60 | 24.1%
n=free | 2111.59 | 5.43 | 2.53 | 25.0% | 48.22 | 54.1% | 21.30 | 0.62 | 20.9%
n=1 fixed | [3001.98] | [2.31] | [1.00] | [9.6%] | [53.37] | [57.0%] | [15.05] | [0.88] | [33.4%]
n=4 fixed | [2491.06] | [12.08] | [4.00] | [35.4%] | [87.22] | [51.7%] | [23.37] | [0.61] | [12.9%]
n=1 re-fit | 2111.59 | 5.43 | 2.53 | 25.0% | 48.21 | 54.1% | 21.30 | 0.62 | 20.9%
n=4 re-fit | 2111.59 | 5.43 | 2.53 | 25.0% | 48.22 | 54.1% | 21.30 | 0.62 | 20.9%
NGC 4151
initial input | 1388891.11 | 7.50 | 0.50 | 24.6% | 33.00 | 65.0% | 55.50 | 0.10 | 10.3%
n=free | 16391.66 | 9.36 | 0.44 | 41.3% | 37.74 | 50.3% | 57.00 | 0.10 | 8.4%
n=1 fixed | [18100.67] | [11.58] | [1.00] | [42.2%] | [86.28] | [41.1%] | [56.70] | [0.17] | [16.7%]
n=4 fixed | [20051.12] | [108.39] | [4.00] | [51.4%] | [8.21] | [20.4%] | [68.23] | [0.28] | [28.2%]
n=1 re-fit | 16391.58 | 9.38 | 0.44 | 41.2% | 38.31 | 49.7% | 55.57 | 0.12 | 9.0%
n=4 re-fit | 16390.93 | 9.36 | 0.44 | 41.3% | 38.20 | 49.8% | 55.88 | 0.11 | 8.8%
Note. — For each galaxy, the rows are: (1) The model outputs (columns 2 to 10)
from GALFIT based on the Stage 1 or Stage 2 fits. These outputs are used as
initial guesses in the fits for rows 2 and 3. (2) The model outputs (columns 2
to 10) from GALFIT when the bulge $n$ is allowed to vary freely, and the input
guesses are based on the parameters in columns 2 to 10 of row 1. (3) The model
outputs (columns 2 to 10) from GALFIT when the bulge Sérsic index is held
fixed at $n=1$, and the input guesses are based on the parameters in columns 2
to 10 of row 1. The model is not always physically meaningful because the
bulge index is fixed. (4) As in row 3, except that the bulge Sérsic index is
now held fixed at $n=4$. The model is not always physically meaningful because
the bulge index is fixed. (5) The model outputs (columns 2 to 10) from GALFIT,
when the bulge $n$ is allowed to vary freely, but the input guesses are now
based on the parameters in columns 2 to 10 of row 3. (6) The model outputs
(columns 2 to 10) from GALFIT, when the bulge $n$ is allowed to vary freely,
but the input guesses are now based on the parameters in columns 2 to 10 of
row 4. The columns are: (1) The type of fit whose model outputs are shown in
columns 2 to 10. (2) The $\chi^{2}$ of the fit. (3) Effective radius of the
bulge in arcseconds. (4) The Sérsic index of the bulge (in the case of Stage 2
bulge-disk fits in row 1 and in the case of all fits in rows 2 to 6), or the
Sérsic index of a single component (in the case of Stage 1 fits in row 1). (5)
Bulge-to-total luminosity fraction. (6) Disk scalelength in arcseconds. (7)
Disk-to-total luminosity fraction. (8) Bar effective radius in arcseconds. (9)
Bar Sérsic index. (10) Bar-to-total luminosity fraction.
Table 8: Mass Breakdown of Galactic Structures Structure | Mass
---|---
Bulges | $18.9\%\pm 3.49\%$
Disks | $71.6\%\pm 4.02\%$
Bars | $9.58\%\pm 2.62\%$
Bulges with $n>2$ | $10.4\%\pm 2.72\%$
Bulges with $n\leq 2$ | $8.42\%\pm 2.47\%$
Table 9: $B/T$ and bulge $n$ in bright/high mass spirals Sample
($M_{B}\leq-19.3$)
---
Fraction of spirals with bulge $n\geq 4$ | $1.40\%\pm 0.98\%$
Fraction of spirals with bulge $2<n<4$ | $22.4\%\pm 3.49\%$
Fraction of spirals with bulge $n\leq 2$ | $76.2\%\pm 3.56\%$
Fraction of spirals with bulge $n>2$ | $23.8\%\pm 3.56\%$
Fraction of spirals with $B/T\leq 0.2$ | $68.5\%\pm 3.88\%$
Fraction of spirals with $B/T>0.2$ | $31.1\%\pm 3.88\%$
Fraction of spirals with $0.2<B/T<0.4$ | $23.1\%\pm 3.52\%$
Fraction of spirals with $B/T\geq 0.4$ | $8.39\%\pm 2.32\%$
Sample $M_{\star}\geq 1\times 10^{10}M_{\odot}$
Fraction of spirals with bulge $n\geq 4$ | $1.77\%\pm 1.24\%$
Fraction of spirals with bulge $2<n<4$ | $23.9\%\pm 4.01\%$
Fraction of spirals with bulge $n\leq 2$ | $74.3\%\pm 4.11\%$
Fraction of spirals with bulge $n>2$ | $25.7\%\pm 4.11\%$
Fraction of spirals with $B/T\leq 0.2$ | $66.4\%\pm 4.44\%$
Fraction of spirals with $B/T>0.2$ | $33.6\%\pm 4.44\%$
Fraction of spirals with $0.2<B/T<0.4$ | $25.7\%\pm 4.11\%$
Fraction of spirals with $B/T\geq 0.4$ | $7.96\%\pm 2.55\%$
Table 10: Bar Fraction as a Function of $B/T$ and Bulge Index Sample
($M_{B}\leq-19.3$)
---
Bar fraction in spirals with bulge $n\geq 4$ | $0.00\%\pm 0.00\%$
Bar fraction in spirals with bulge $2<n<4$ | $37.5\%\pm 8.56\%$
Bar fraction in spirals with bulge $n\leq 2$ | $65.1\%\pm 4.56\%$
Bar fraction of spirals with bulge $n>2$ | $35.3\%\pm 8.20\%$
Bar fraction of spirals with $B/T\leq 0.2$ | $68.4\%\pm 4.70\%$
Bar fraction of spirals with $B/T>0.2$ | $35.6\%\pm 7.14\%$
Bar fraction of spirals with $0.2<B/T<0.4$ | $42.4\%\pm 8.60\%$
Bar fraction of spirals with $B/T\geq 0.4$ | $16.7\%\pm 10.8\%$
Sample $M_{\star}\geq 1\times 10^{10}M_{\odot}$
Bar fraction of spirals with bulge $n\geq 4$ | $0.00\%\pm 0.00\%$
Bar fraction of spirals with bulge $2<n<4$ | $44.4\%\pm 9.56\%$
Bar fraction of spirals with bulge $n\leq 2$ | $63.1\%\pm 5.27\%$
Bar fraction of spirals with bulge $n>2$ | $41.4\%\pm 9.15\%$
Bar fraction of spirals with $B/T\leq 0.2$ | $68.0\%\pm 5.39\%$
Bar fraction of spirals with $B/T>0.2$ | $36.8\%\pm 7.83\%$
Bar fraction of spirals with $0.2<B/T<0.4$ | $41.4\%\pm 9.15\%$
Bar fraction of spirals with $B/T\geq 0.4$ | $22.2\%\pm 13.9\%$
Table 11: $B/T$: Data versus Hierarchical Models of Galaxy Evolution
| Data | Model | Model spirals with | Model spirals with | All model
---|---|---|---|---|---
| | | major mergers | no major merger | spirals
| | | since redshift $z_{1}$ | since redshfit $z_{1}$ |
(1) | (2) | (3) | (4) | (5) | (6)
| For $z_{\rm 1}\leq 2$ | For $z_{\rm 1}\leq 2$ |
Fraction of spirals with $B/T\leq 0.2$ | 66.4% $\pm$ 4.44% | SK | 0.1% | 67.2% | 67.3%
Fraction of spirals with $0.2<B/T\leq 0.4$ | 25.7% $\pm$ 4.11% | SK | 2.2% | 16.6% | 18.8%
Fraction of spirals with $0.4<B/T\leq 0.75$ | 8.0% $\pm$ 2.5% | SK | 11.2% | 2.6% | 13.8%
Fraction of spirals with $B/T\leq 0.75$ | 100% | SK | 13.5% | 86.5% | 100.0%
| For $z_{\rm 1}\leq 4$ | For $z_{\rm 1}\leq 4$ |
Fraction of spirals with $B/T\leq 0.2$ | 66.4% $\pm$ 4.44% | SK | 1.6% | 65.7% | 67.3%
Fraction of spirals with $0.2<B/T\leq 0.4$ | 25.7% $\pm$ 4.11% | SK | 5.5% | 13.3% | 18.8%
Fraction of spirals with $0.4<B/T\leq 0.75$ | 8.0% $\pm$ 2.5% | SK | 12.6% | 1.3% | 13.9%
Fraction of spirals with $B/T\leq 0.75$ | 100% | SK | 19.8% | 80.2% | 100.0%
| For $z_{\rm 1}\leq 2$ | For $z_{\rm 1}\leq 2$ |
Fraction of spirals with $B/T\leq 0.2$ | 66.4% $\pm$ 4.44% | PH | 3.0% | 67.7% | 70.7%
Fraction of spirals with $0.2<B/T\leq 0.4$ | 25.7% $\pm$ 4.11% | PH | 7.1% | 13.1% | 20.1%
Fraction of spirals with $0.4<B/T\leq 0.75$ | 8.0% $\pm$ 2.5% | PH | 5.9% | 3.1% | 9.0%
Fraction of spirals with $B/T\leq 0.75$ | 100% | PH | 15.7% | 84.0% | 99.8%
Note. — Column 2 shows the empirical fraction of high mass spirals with a
given $B/T$. Columns 3 to 6 show the fraction of high mass spirals with a
given $B/T$ in the models, which define spirals as systems with $B/T\leq
0.75$. Column 3 indicates the model used: SK refers to the models by Khochfar
& Burkert (2005) and Khochfar & Silk (2006), while PH refers to those by
Hopkins et al. (2009). The models are described in §5.8 and have different
prescriptions for the amount of mass that gets converted into bulge stars.
Column 4 shows the fraction of spirals, which have a given $B/T$ and have had
one or more major mergers since redshift $z_{1}$. Column 5 shows the fraction
of spirals, which have a given $B/T$ and have had no major merger since
redshift $z_{1}$. Column 6 is the sum of column 4 and 5, and shows the
fraction of spirals with a given $B/T$ in the models. It can be directly
compared to the data in column 2.
Table 12: Comparison of model merger fraction since $z\leq 4$ and bulge Sérsic
index Stellar mass limit: $1.0\times 10^{10}M_{\odot}$
---
Model galaxies: 7674
Observed galaxies: 113
Test 1
% of model galaxies w/ $B/T\leq 0.2$ with no major merger since $z\leq 4$: | 5040/7674 = $65.7\%\pm 0.54\%$
% of observed galaxies w/ $B/T\leq 0.2$ and bulge $n\leq 3$: | 75/113 = $66.4\%\pm 4.44\%$
% of observed galaxies w/ $B/T\leq 0.2$ and bulge $n\leq 2.5$: | 74/113 = $65.5\%\pm 4.47\%$
% of observed galaxies w/ $B/T\leq 0.2$ and bulge $n\leq 2$: | 67/113 = $59.3\%\pm 4.62\%$
Test 2
% of model galaxies w/ $B/T\geq 0.4$ and a major merger since $z\leq 4$: | 971/7674 = $12.7\%\pm 0.38\%$
% of observed spirals w/ $B/T\geq 0.4$ and bulge $n>3$: | 2/113 = $1.77\%\pm 1.24\%$
% of observed spirals w/ $B/T\geq 0.4$ and bulge $n>2.5$: | 4/113 = $3.54\%\pm 1.74\%$
% of observed spirals w/ $B/T\geq 0.4$ and bulge $n>2$: | 7/113 = $6.19\%\pm 2.27\%$
Test 3
% of model galaxies w/ $B/T\leq 0.2$ and a major merger since $z\leq 4$: | 126/7674 = $1.64\%\pm 0.15\%$
% of observed galaxies w/ $B/T\leq 0.2$ and bulge $n>3$: | 0/113 = $0.00\%\pm 0.00\%$
% of observed galaxies w/ $B/T\leq 0.2$ and bulge $n>2.5$ | 1/113 = $0.88\%\pm 0.88\%$
% of observed galaxies w/ $B/T\leq 0.2$ and bulge $n>2$ | 8/113 = $7.08\%\pm 2.41\%$
| arxiv-papers | 2008-07-01T17:28:42 | 2024-09-04T02:48:56.476876 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Tim Weinzirl (1), Shardha Jogee (1), Sadegh Khochfar (2)(3), Andreas\n Burkert (4), John Kormendy (1) ((1) Univ. of Texas at Austin, USA, (2) Univ.\n of Oxford, UK (3), Max Planck Institut fur extraterrestrische Physik,\n Germany, (4) Universitats-Sternwarte Munchen, Germany)",
"submitter": "Tim Weinzirl",
"url": "https://arxiv.org/abs/0807.0040"
} |
0807.0087 | 11institutetext: Department of Mathematics and Computer Science, University of
the Balearic Islands, E-07122 Palma de Mallorca, Spain 22institutetext:
Research Institute of Health Science (IUNICS), E-07122 Palma de Mallorca,
Spain 33institutetext: Algorithms, Bioinformatics, Complexity and Formal
Methods Research Group, Technical University of Catalonia, E-08034 Barcelona,
Spain
# Path lengths in tree-child time consistent hybridization networks
Gabriel Cardona 11 Mercè Llabrés 1122 Francesc Rosselló 1122 Gabriel Valiente
2233
###### Abstract
Hybridization networks are representations of evolutionary histories that
allow for the inclusion of reticulate events like recombinations,
hybridizations, or lateral gene transfers. The recent growth in the number of
hybridization network reconstruction algorithms has led to an increasing
interest in the definition of metrics for their comparison that can be used to
assess the accuracy or robustness of these methods. In this paper we establish
some basic results that make it possible the generalization to tree-child time
consistent (TCTC) hybridization networks of some of the oldest known metrics
for phylogenetic trees: those based on the comparison of the vectors of path
lengths between leaves. More specifically, we associate to each hybridization
network a suitably defined vector of ‘splitted’ path lengths between its
leaves, and we prove that if two TCTC hybridization networks have the same
such vectors, then they must be isomorphic. Thus, comparing these vectors by
means of a metric for real-valued vectors defines a metric for TCTC
hybridization networks. We also consider the case of fully resolved
hybridization networks, where we prove that simpler, ‘non-splitted’ vectors
can be used.
## 1 Introduction
An evolutionary history is usually modelled by means of a rooted phylogenetic
tree, whose root represents a common ancestor of all species under study (or
whatever other taxonomic units are considered: genes, proteins,…), the leaves,
the extant species, and the internal nodes, the ancestral species. But
phylogenetic trees can only cope with speciation events due to mutations,
where each species other than the universal common ancestor has only one
parent in the evolutionary history (its parent in the tree). It is clearly
understood now that other speciation events, which cannot be properly
represented by means of single arcs in a tree, play an important role in
evolution [10]. These are _reticulation events_ like genetic recombinations,
hybridizations, or lateral gene transfers, where a species is the result of
the interaction between two parent species. This has lead to the introduction
of _networks_ as models of phylogenetic histories that capture these
reticulation events side by side with the classical mutations.
Contrary to what happens in the phylogenetic trees literature, where the basic
concepts are well established, there is still some lack of consensus about
terminology in the field of ‘phylogenetic networks’ [16]. Following [23], in
this paper we use the term _hybridization network_ to denote the most general
model of reticulated evolutionary history: a directed acyclic graph with only
one root, which represents the last universal common ancestor and which we
assume, thus, of out-degree greater than 1. In such a graph, nodes represent
species (or any other taxonomy unit) and arcs represent direct descendance. A
node with only one parent (a _tree node_) represents a species derived from
its parent species through mutation, and a node with more than one parent (a
_hybrid node_) represents a species derived from its parent species through
some reticulation event.
The interest in representing phylogenetic histories by means of networks has
lead to many hybridization network reconstruction methods [13, 14, 17, 18, 19,
21, 25, 28]. These reconstruction methods often search for hybridization
networks satisfying some restriction, like for instance to have as few hybrid
nodes as possible (in _perfect phylogenies_), or to have their reticulation
cycles satisfying some structural restrictions (in _galled trees_ and
_networks_). Two popular and biologically meaningful such restrictions are the
_time consistency_ [1, 18], the possibility of assigning times to the nodes in
such a way that tree children exist later than their parents and hybrid
children coexist with their parents (and in particular, the parents of a
hybrid species coexist in time), and the _tree child condition_ [8, 31], that
imposes that every non-extant species has some descendant through mutation
alone. The _tree-child time consistent_ (TCTC) hybridization networks have
been recently proposed as the class where meaningful phylogenetic networks
should be searched [30]. Recent simulations (reported in [27]) have shown that
over 64% of 4132 hybridization networks obtained using the coalescent model
[15] under various population and sample sizes, sequence lengths, and
recombination rates, were TCTC: the percentage of TCTC networks among the time
consistent networks obtained in these simulations increases to 92.8%.
The increase in the number of available hybridization networks reconstruction
algorithms has made it necessary the introduction of methods for the
comparison of hybridization networks to be used in their assessment, for
instance by comparing inferred networks with either simulated or true
phylogenetic histories, and by evaluating the robustness of reconstruction
algorithms when adding new species [18, 32]. This has lead recently to the
definition of several metrics defined on different classes of hybridization
networks [4, 5, 6, 9, 8, 18, 20]. All these metrics generalize in one way or
another well-known metrics for phylogenetic trees.
Some of the most popular metrics for phylogenetic trees are based on the
comparison of the vectors of path lengths between leaves [3, 11, 12, 22, 26,
29]. Introduced in the early seventies, with different names depending on the
author and the way these vectors are compared, they are globally known as
_nodal distances_. Actually, these vectors of paths lengths only _separate_
(in the sense that equal vectors means isomorphic trees), on the one hand,
unrooted phylogenetic trees, and, on the other hand, fully resolved rooted
phylogenetic trees, and therefore, as far as rooted phylogenetic trees goes,
the distances defined through these vectors are only true metrics for fully
resolved trees. These metrics were recently generalized to arbitrary rooted
phylogenetic trees [7]. In this generalization, each path length between two
leaves was replaced by the pair of distances from the leaves to their least
common ancestor, and the vector of paths lengths between leaves was replaced
by the _splitted path lengths_ matrix obtained in this way. These matrices
separate arbitrary rooted phylogenetic trees, and therefore the _splitted
nodal distances_ defined through them are indeed metrics on the space of
rooted phylogenetic trees.
In a recent paper [6] we have generalized these splitted nodal distances to
TCTC hybridization networks with all their hybrid nodes of out-degree 1. The
goal of this paper is to go one step beyond in two directions: to generalize
to the TCTC hybridization networks setting both the classical nodal distances
for fully resolved rooted phylogenetic trees and the new splitted nodal
distances for rooted phylogenetic trees. Thus, on the one hand, we introduce a
suitable generalization of the vectors of path lengths between leaves that
separate _fully resolved_ (where every non extant species has exactly two
children, and every reticulation event involves exactly two parent species)
TCTC hybridization networks. On the other hand, we show that if we split these
new path lengths in a suitable way and we add a bit of extra information, the
resulting vectors separate arbitrary TCTC hybridization networks. Then, the
vectors obtained in both cases can be used to define metrics that generalize,
respectively, the nodal distances for fully resolved rooted phylogenetic trees
and the splitted nodal distances for rooted phylogenetic trees.
The key ingredient in the proofs of our main results is the use of sets of
suitable reductions that applied to TCTC hybridization networks with $n$
leaves and $m$ internal nodes produce TCTC hybridization networks with either
$n-1$ leaves or with $n$ leaves and $m-1$ internal nodes (in the fully
resolved case, the reductions we use are specifically tailored to make them
remove always one leaf). Similar sets of reductions have already been
introduced for TCTC hybridization networks with all their hybrid nodes of out-
degree 1 [6] and for _tree sibling_ (where every hybrid node has a tree
sibling) time consistent hybridization networks with all their hybrid nodes of
in-degree 2 and out-degree 1 [4], and they have been proved useful in those
contexts not only to establish properties of the corresponding networks by
algebraic induction, but also to generate in a recursive way all networks of
the type under consideration. We hope that the reductions introduced in this
paper will find similar applications elsewhere.
## 2 Preliminaries
### 2.1 Notations on DAGs
Let $N=(V,E)$ denote in this subsection a directed acyclic (non-empty, finite)
graph; a _DAG_ , for short. A node $v\in V$ is a _child_ of $u\in V$ if
$(u,v)\in E$; we also say in this case that $u$ is a _parent_ of $v$. All
children of the same parent are said to be _sibling_ of each other.
Given a node $v\in V$, its _in-degree_ $\deg_{in}(v)$ and its _out-degree_
$\deg_{out}(v)$ are, respectively, the number of its parents and the number of
its children. The _type_ of $v$ is the ordered pair
$(\deg_{in}(v),\deg_{out}(v))$. A node $v$ is a _root_ when $\deg_{in}(v)=0$,
a _tree node_ when $\deg_{in}(v)\leqslant 1$, a _hybrid node_ when
$\deg_{in}(v)\geqslant 2$, a _leaf_ when $\deg_{out}(v)=0$, _internal_ when
$\deg_{out}(v)\geqslant 1$, and _elementary_ when $\deg_{in}(v)\leqslant 1$
and $\deg_{out}(v)=1$. A _tree arc_ (respectively, a _hybridization arc_) is
an arc with head a tree node (respectively, a hybrid node). A DAG $N$ is
_rooted_ when it has only one root.
A _path_ on $N$ is a sequence of nodes $(v_{0},v_{1},\dots,v_{k})$ such that
$(v_{i-1},v_{i})\in E$ for all $i=1,\dots,k$. We call $v_{0}$ the _origin_ of
the path, $v_{1},\ldots,v_{k-1}$ its _intermediate nodes_ , and $v_{k}$ its
_end_. The _length_ of the path $(v_{0},v_{1},\dots,v_{k})$ is $k$, and it is
_non-trivial_ if $k\geqslant 1$. The _acyclicity_ of $N$ means that it does
not contain _cycles_ : non-trivial paths from a node to itself.
We denote by $u\\!\rightsquigarrow\\!{}v$ any path with origin $u$ and end
$v$. Whenever there exists a path $u\\!\rightsquigarrow\\!{}v$, we shall say
that $v$ is a _descendant_ of $u$ and also that $u$ is an _ancestor_ of $v$.
When the path $u\\!\rightsquigarrow\\!{}v$ is non-trivial, we say that $v$ is
a _proper_ descendant of $u$ and that $u$ is an _proper_ ancestor of $v$. The
_distance_ from a node $u$ to a descendant $v$ is the length of a shortest
path from $u$ to $v$.
The _height_ $h(v)$ of a node $v$ in a DAG $N$ is the largest length of a path
from $v$ to a leaf. The absence of cycles implies that the nodes of a DAG can
be stratified by means of their heights: the nodes of height 0 are the leaves,
the nodes of height 1 are those nodes all whose children are leaves, the nodes
of height 2 are those nodes all whose children are leaves and nodes of height
1, and so on. If a node has height $m>0$, then all its children have height
smaller than $m$, and at least one of them has height exactly $m-1$.
A node $v$ of $N$ is a _strict descendant_ of a node $u$ if it is a descendant
of it, and every path from a root of $N$ to $v$ contains the node $u$: in
particular, we understand every node as a strict descendant of itself. When
$v$ is a strict descendant of $u$, we also say that $u$ is a _strict ancestor_
of $v$.
The following lemma will be used several times in this paper.
###### Lemma 1
Let $u$ be a proper strict ancestor of a node $v$ in a DAG $N$, and let $w$ be
an intermediate node in a path $u\\!\rightsquigarrow\\!{}v$. Then, $u$ is also
a strict ancestor of $w$.
###### Proof
Let $r\\!\rightsquigarrow\\!{}w$ be a path from a root of $N$ to $w$, and
concatenate to it the piece $w\\!\rightsquigarrow\\!{}v$ of the path
$u\\!\rightsquigarrow\\!{}v$ under consideration. This yields a path
$r\\!\rightsquigarrow\\!{}v$ that must contain $u$. Since $u$ does not appear
in the piece $w\\!\rightsquigarrow\\!{}v$, we conclude that it is contained in
the path $r\\!\rightsquigarrow\\!{}w$. This proves that every path from a root
of $N$ to $w$ contains the node $u$.
For every pair of nodes $u,v$ of $N$:
* •
$CSA(u,v)$ is the set of all common ancestors of $u$ and $v$ that are strict
ancestors of at least one of them;
* •
the _least common semi-strict ancestor_ (_LCSA_) of $u$ and $v$, in symbols
$[u,v]$, is the node in $CSA(u,v)$ of minimum height.
The LCSA of two nodes $u,v$ in a phylogenetic network is well defined and it
is unique: it is actually the unique element of $CSA(u,v)$ that is a
descendant of all elements of this set [5]. The following result on LCSAs will
be used often. It is the generalization to DAGs of Lemma 6 in [6], and we
include its easy proof for the sake of completeness.
###### Lemma 2
Let $N$ be a DAG and let $u,v$ be a pair of nodes of $N$ such that $v$ is not
a descendant of $u$. If $u$ is a tree node with parent $u^{\prime}$, then
$[u,v]=[u^{\prime},v]$.
###### Proof
We shall prove that $CSA(u,v)=CSA(u^{\prime},v)$.
Let $x\in CSA(u,v)$. Since $u$ is not an ancestor of $v$, $x\neq u$ and hence
any path $x\\!\rightsquigarrow\\!{}u$ is non-trivial. Then, since $u^{\prime}$
is the only parent of $u$, it appears in this path, and therefore $x$ is also
an ancestor of $u^{\prime}$. This shows that $x$ is a common ancestor of
$u^{\prime}$ and $v$. Now, if $x$ is a strict ancestor of $v$, we already
conclude that $x\in CSA(u^{\prime},v)$, while if $x$ is a strict ancestor of
$u$, it will be also a strict ancestor of $u^{\prime}$ by Lemma 1, and hence
$x\in CSA(u^{\prime},v)$, too. This proves that $CSA(u,v)\subseteq
CSA(u^{\prime},v)$
Conversely, let $x\in CSA(u^{\prime},v)$. Since $u^{\prime}$ is the parent of
$u$, it is clear that $x$ is a common ancestor of $u$ and $v$, too. If $x$ is
a strict ancestor of $v$, this implies that $x\in CSA(u,v)$. If $x$ is a
strict ancestor of $u^{\prime}$, then it is also a strict ancestor of $u$
(every path $r\\!\rightsquigarrow\\!{}u$ must contain the only parent
$u^{\prime}$ of $u$, and then $x$ will belong to the piece
$r\\!\rightsquigarrow\\!{}u^{\prime}$ of the path
$r\\!\rightsquigarrow\\!{}u$), and therefore $x\in CSA(u,v)$, too. This
finishes the proof of the equality.
Let $S$ be any non-empty finite set of _labels_. We say that the DAG $N$ is
_labeled in $S$_, or that it is an _$S$ -DAG_, for short, when its leaves are
bijectively labeled by elements of $S$. Although in real applications the set
$S$ would correspond to a given set of extant taxa, for the sake of simplicity
we shall assume henceforth that $S=\\{1,\ldots,n\\}$, with $n=|S|$. We shall
always identify, usually without any further notice, each leaf of an $S$-DAG
with its label in $S$.
Two $S$-DAGs $N,N^{\prime}$ are _isomorphic_ , in symbols $N\cong N^{\prime}$,
when they are isomorphic as directed graphs and the isomorphism maps each leaf
in $N$ to the leaf with the same label in $N^{\prime}$.
### 2.2 Path lengths in phylogenetic trees
A _phylogenetic tree_ on a set $S$ of taxa is a rooted $S$-DAG without hybrid
nodes and such that its root is non-elementary. A phylogenetic tree is _fully
resolved_ , or _binary_ , when every internal node has out-degree 2. Since all
ancestors of a node in a phylogenetic tree are strict, the LCSA $[u,v]$ of two
nodes $u,v$ in a phylogenetic tree is simply their _least common ancestor_ :
the unique common ancestor of them that is a descendant of every other common
ancestor of them.
Let $T$ be a phylogenetic tree on the set $S=\\{1,\ldots,n\\}$. For every
$i,j\in S$, we shall denote by $\ell_{T}(i,j)$ and $\ell_{T}(j,i)$ the lengths
of the paths $[i,j]\\!\rightsquigarrow\\!{}i$ and
$[i,j]\\!\rightsquigarrow\\!{}j$, respectively. In particular,
$\ell_{T}(i,i)=0$ for every $i=1,\ldots,n$.
###### Definition 1
Let $T$ be a phylogenetic tree on the set $S=\\{1,\ldots,n\\}$. The _path
length_ between two leaves $i$ and $j$ is
$L_{T}(i,j)=\ell_{T}(i,j)+\ell_{T}(j,i).$
The _path lengths vector_ of $T$ is the vector
$L(T)=\big{(}L_{T}(i,j)\big{)}_{1\leqslant i<j\leqslant
n}\in\mathbb{N}^{n(n-1)/2}$
with its entries ordered lexicographically in $(i,j)$.
The following result is a special case of Prop. 2 in [7].
###### Proposition 1
Two fully resolved phylogenetic trees on the same set $S$ of taxa are
isomorphic if, and only if, they have the same path lengths vectors. ∎
The thesis in the last result is false for arbitrary phylogenetic trees.
Consider for instance the phylogenetic trees with Newick strings (1,2,(3,4));
and ((1,2),3,4); depicted111Henceforth, in graphical representations of DAGs,
hybrid nodes are represented by squares, tree nodes by circles, and
indeterminate nodes, that is, nodes that can be of tree or hybrid type, by
squares with rounded corners. in Fig. 1. It is straightforward to check that
they have the same path lengths vectors, but they are not isomorphic.
$1$$2$$3$$4$
$1$$2$$3$$4$
Figure 1: Two non-isomorphic phylogenetic trees with the same path lengths
vectors.
This problem was overcome in [7] by replacing the path lengths vectors by the
following matrices of distances.
###### Definition 2
The _splitted path lengths matrix_ of $T$ is the $n\times n$ square matrix
$\ell(T)=\big{(}\ell_{T}(i,j)\big{)}_{i=1,\ldots,n\atop
j=1,\ldots,n}\in\mathcal{M}_{n}(\mathbb{N}).$
Now, the following result is (again, a special case of) Theorem 11 in [7].
###### Proposition 2
Two phylogenetic trees on the same set $S$ of taxa are isomorphic if, and only
if, they have the same splitted path lengths matrices. ∎
## 3 TCTC networks
While the basic notion of phylogenetic tree is well established, the notion of
phylogenetic network is much less well defined [16]. The networks we consider
in this paper are the (almost) most general possible ones: rooted $S$-DAGs
with non-elementary root. Following [23], we shall call them _hybridization
networks_. In these hybridization networks, every node represents a different
species, and the arcs represent direct descendance, be it through mutation
(tree arcs) or through some reticulation event (hybridization arcs).
It is usual to forbid elementary nodes in hybridization networks [23], mainly
because they cannot be reconstructed. We allow them here for two reasons. On
the one hand, because allowing them simplifies considerably some proofs, as it
will be hopefully clear in Section 5. On the other hand, because, as Moret et
al point out [18, §4.3], they can be useful both from the biological point of
view, to include auto-polyploidy in the model, as well as from the formal
point of view, to make a phylogeny satisfy other constraints, like for
instance time consistency (see below) or the impossibility of successive
hybridizations. Of course, our main results apply without any modification to
hybridization networks without elementary nodes as well.
Following [5], by a _phylogenetic network_ on a set $S$ of taxa we understand
a rooted $S$-DAG $N$ with non-elementary root where every hybrid node has
exactly one child, and it is a tree node. Although, from the mathematical
point of view, phylogenetic networks are a special case of hybridization
networks, from the point of view of modelling they represent in a different
way evolutive histories with reticulation events: in a phylogenetic network,
every tree node represents a different species and every hybrid node, a
reticulation event that gives rise to the species represented by its only
child.
A hybridization network $N=(V,E)$ is _time consistent_ when it allows a
_temporal representation_ [1]: a mapping
$\tau:V\to\mathbb{N}$
such that $\tau(u)<\tau(v)$ for every tree arc $(u,v)$ and $\tau(u)=\tau(v)$
for every hybridization arc $(u,v)$. Such a temporal representation can be
understood as an assignment of times to nodes that strictly increases from
parents to tree children and so that the parents of each hybrid node coexist
in time.
###### Remark 1
Let $N=(V,E)$ be a time consistent hybridization network, and let
$N_{1}=(V_{1},E_{1})$ be a hybridization network obtained by removing from $N$
some nodes and all their descendants (as well as all arcs pointing to any
removed node). Then $N_{1}$ is still time consistent, because the restriction
of any temporal representation $\tau:V\to\mathbb{N}$ of $N$ to $V_{1}$ yields
a temporal representation of $N_{1}$.
A hybridization network satisfies the _tree-child condition_ , or it is _tree-
child_ , when every internal node has at least one child that is a tree node
(a _tree child_). So, tree-child hybridization networks can be understood as
general models of reticulate evolution where every species other that the
extant ones, represented by the leaves, has some descendant through mutation.
Tree-child hybridization networks include galled trees [13, 14] as a
particular case [8].
A _tree path_ in a tree-child hybridization network is a non-trivial path such
that its end and all its intermediate nodes are tree nodes. A node $v$ is a
_tree descendant_ of a node $u$ when there exists a tree path from $u$ to $v$.
By [9, Lem. 2], every internal node $u$ of a tree-child hybridization network
has some tree descendant leaf, and by [9, Cor. 4] every tree descendant $v$ of
$u$ is a strict descendant of $u$ and the path $u\\!\rightsquigarrow\\!{}v$ is
unique.
To simplify the notations, we shall call _TCTC-networks_ the tree-child time
consistent hybridization networks: these include the tree-child time
consistent phylogenetic networks, which were the objects dubbed TCTC-networks
in [5, 6]. Every phylogenetic tree is also a TCTC-network. Let
$\mathrm{TCTC}_{n}$ denote the class of all TCTC-networks on
$S=\\{{1},\ldots,{n}\\}$.
We prove now some basic properties of TCTC-networks that will be used later.
###### Lemma 3
Let $u$ be a node of a TCTC-network $N$, and let $v$ be a child of $u$. The
node $v$ is a tree node if, and only if, it is a strict descendant of $u$.
###### Proof
Assume first that $v$ is a tree child of $u$. Since $u$ is the only parent of
$v$, every non-trivial path ending in $v$ must contain $u$. This shows that
$u$ is a strict ancestor of $v$.
Assume now that $v$ is a hybrid child of $u$ that is also a strict descendant
of it, and let us see that this leads to a contradiction. Indeed, in this case
the set $H(u)$ of hybrid children of $u$ that are strict descendants of it is
non-empty, and we can choose a node $v_{0}$ in it of largest height. Let
$v_{1}$ be any parent of $v_{0}$ other than $u$. Since $u$ is a strict
ancestor of $v_{0}$, it must be an ancestor of $v_{1}$, and since $u$ and
$v_{1}$ have the hybrid child $v_{0}$ is common, they must have the same
temporal representation, and therefore $v_{1}$ as well as all intermediate
nodes in any path $u\\!\rightsquigarrow\\!{}v_{1}$ must be hybrid. Moreover,
since $u$ is a strict ancestor of $v_{0}$, it is also a strict ancestor of
$v_{1}$ as well as of any intermediate node in any path
$u\\!\rightsquigarrow\\!{}v_{1}$ (by Lemma 1). In particular, the child of $u$
in a path $u\\!\rightsquigarrow\\!{}v_{1}$ will belong to $H(u)$ and its
height will be larger than the height of $v_{0}$, which is impossible.
###### Corollary 1
All children of the root of a TCTC-network are tree nodes.
###### Proof
Every node in a hybridization network is a strict descendant of the root.
Then, Lemma 3 applies.
The following result is the key ingredient in the proofs of our main results;
it generalizes to hybridization networks Lemma 3 in [6], which referred to
phylogenetic networks. A similar result was proved in [4] for _tree-sibling_
(that is, where every hybrid node has a sibling that is a tree node) time
consistent phylogenetic networks with all its hybrid nodes of in-degree 2.
###### Lemma 4
Every TCTC-network with more than one leaf contains at least one node $v$
satisfying one of the following properties:
1. (a)
$v$ is an internal tree node and all its children are tree leaves.
2. (b)
$v$ is a hybrid internal node, all its children are tree leaves, and all its
siblings are leaves or hybrid nodes.
3. (c)
$v$ is a hybrid leaf, and all its siblings are leaves or hybrid nodes.
###### Proof
Let $N$ be a TCTC-network and $\tau$ a temporal representation of it. Let
$v_{0}$ be an internal node of highest $\tau$-value and, among such nodes, of
smallest height. The tree children of $v_{0}$ have strictly higher
$\tau$-value than $v$, and therefore they are leaves. And the hybrid children
of $v_{0}$ have the same $\tau$-value than $v_{0}$ but smaller height, and
therefore they are also leaves.
Now:
* •
If $v_{0}$ is a tree node all whose children are tree nodes, taking $v=v_{0}$
we are in case (a).
* •
If $v_{0}$ is a hybrid node all whose children are tree nodes, then its
parents have its same $\tau$-value, which, we recall, is the highest one. This
implies that their children ($v_{0}$’s siblings) cannot be internal tree
nodes, and hence they are leaves or hybrid nodes. So, taking $v=v_{0}$, we are
in case (b).
* •
If $v_{0}$ has some hybrid child, take as the node $v$ in the statement this
hybrid child: it is a leaf, and all its parents have the same $\tau$-value as
$v_{0}$, which implies, arguing as in the previous case, that all siblings of
$v$ are leaves or hybrid nodes. Thus, $v$ satisfies (c).
We introduce now some reductions for TCTC-networks. Each of these reductions
applied to a TCTC-network with $n$ leaves and $m$ internal nodes produces a
TCTC-network with either $n-1$ leaves and $m$ internal nodes or with $n$
leaves and $m-1$ internal nodes, and given any TCTC-network with more than two
leaves, it will always be possible to apply to it some of these reductions.
This lies at the basis of the proofs by algebraic induction of the main
results in this paper.
Let $N$ be a TCTC-network with $n\geqslant 3$ leaves.
1. (U)
Let $i$ be one tree leaf of $N$ and assume that its parent has only this
child. The _$U(i)$ reduction_ of $N$ is the network $N_{U(i)}$ obtained by
removing the leaf $i$, together with its incoming arc, and labeling with $i$
its former parent; cf. Fig. 2. This reduction removes the only child of a
node, and thus it is clear that $N_{U(i)}$ is still a TCTC-network, with the
same number of leaves but one internal node less than $N$.
$i$
$\Longrightarrow$
$i$
Figure 2: The $U(i)$-reduction.
2. (T)
Let $i,j$ be two sibling tree leaves of $N$ (that may, or may not, have other
siblings). The _$T(i;j)$ reduction_ of $N$ is the network $N_{T(i;j)}$
obtained by removing the leaf $i$, together with its incoming arc; cf. Fig. 3.
This reduction procedure removes one tree leaf, but its parent $u$ keeps at
least another tree child, and if $u$ was the root of $N$ then it would not
become elementary after the reduction, because $n\geqslant 3$ and therefore,
since $j$ is a leaf, $u$ should have at least another child. Therefore,
$N_{T(i;j)}$ is a TCTC-network with the same number of internal nodes as $N$
and $n-1$ leaves.
$j$$i$$\cdots$
$\Longrightarrow$
$j$$\cdots$
Figure 3: The $T(i;j)$-reduction.
3. (H)
Let $i$ be a hybrid leaf of $N$, let $v_{1},\ldots,v_{k}$, with $k\geqslant
2$, be its parents, and assume that each one of these parents has (at least)
one tree leaf child: for every $l=1,\ldots,k$, let $j_{l}$ be a tree leaf
child of $v_{l}$. The _$H(i;j_{1},\ldots,j_{k})$ reduction_ of $N$ is the
network $N_{H(i;j_{1},\dots,j_{k})}$ obtained by removing the hybrid leaf $i$
and its incoming arcs; cf. Fig. 4. This reduction procedure preserves the time
consistency and the tree-child condition (it removes a hybrid leaf), and the
root does not become elementary: indeed, the only possibility for the root to
become elementary is to be one of the parents of $i$, which is impossible by
Corollary 1. Therefore, $N_{H(i;j_{1},\dots,j_{k})}$ is a TCTC-network with
the same number of internal nodes as $N$ and $n-1$ leaves.
$i$$v_{1}$$j_{1}$$\cdots$$v_{k}$$j_{k}$$\cdots$$\cdots$$\Longrightarrow$
$v_{1}$$j_{1}$$\cdots$$v_{k}$$j_{k}$$\cdots$$\cdots$
Figure 4: The $H(i;j_{1},\ldots,j_{k})$-reduction.
We shall call the inverses of the U, T, and H reduction procedures,
respectively, the $\textrm{U}^{-1}$, $\textrm{T}^{-1}$, and $\textrm{H}^{-1}$
_expansions_ , and we shall denote them by $U^{-1}(i)$, $R^{-1}(i;j)$, and
$H^{-1}(i;j_{1},\ldots,j_{k})$. More specifically, for every TCTC-network $N$:
* •
If $N$ has some leaf labeled $i$, the expansion $U^{-1}(i)$ can be applied to
$N$ and the resulting network $N_{U^{-1}(i)}$ is obtained by unlabeling the
leaf $i$ and adding to it a tree leaf child labeled with $i$. $N_{U^{-1}(i)}$
is always a TCTC-network.
* •
If $N$ has no leaf labeled with $i$ and some tree leaf labeled with $j$, the
expansion $T^{-1}(i;j)$ can be applied to $N$, and the resulting network
$N_{T^{-1}(i;j)}$ is obtained by adding to the parent of the leaf $j$ an new
tree leaf child labeled with $i$. $N_{T^{-1}(i;j)}$ is always a TCTC-network.
* •
If $N$ has no leaf labeled with $i$ and some tree leaves labeled with
$j_{1},\ldots,j_{k}$, $k\geqslant 2$, that are not sibling of each other, the
expansion $H^{-1}(i;j_{1},\ldots,j_{k})$ can be applied to $N$ and the
resulting network $N_{H^{-1}(i;j_{1},\ldots,j_{k})}$ is obtained by adding a
new hybrid node labeled with $i$ and arcs from the parents of
$j_{1},\ldots,j_{k}$ to $i$. $N_{H^{-1}(i;j_{1},\ldots,j_{k})}$ is always a
tree child hybridization network, but it need not be time consistent, as the
parents of $j_{1},\ldots,j_{k}$ may have different temporal representations in
$N$ (for instance, one of them could be a tree descendant of another one).
The following result is easily deduced from the explicit descriptions of the
reduction and expansion procedures, and the fact that isomorphisms preserve
labels and parents.
###### Lemma 5
Let $N$ and $N^{\prime}$ be two TCTC-networks. If $N\cong N^{\prime}$, then
the result of applying to both $N$ and $N^{\prime}$ the same U reduction
(respectively, T reduction, H reduction, $\textrm{U}^{-1}$ expansion,
$\textrm{T}^{-1}$ expansion, or $\textrm{H}^{-1}$ expansion) are again two
isomorphic hybridization networks.
Moreover, if we apply an U reduction (respectively, T reduction, or H
reduction) to a TCTC-network $N$, and then we apply to the resulting TCTC-
network the inverse $\textrm{U}^{-1}$ expansion (respectively,
$\textrm{T}^{-1}$ expansion, or $\textrm{H}^{-1}$ expansion), we obtain a
TCTC-network isomorphic to $N$. ∎
As we said above, every TCTC-network with at least 3 leaves allows the
application of some reduction.
###### Proposition 3
Let $N$ be a TCTC-network with more than two leaves. Then, at least one U, R,
or H reduction can be applied to $N$.
###### Proof
By Lemma 4, $N$ contains either an internal (tree or hybrid) node $v$ all
whose children are tree leaves, or a hybrid leaf $i$ all whose siblings are
leaves or hybrid nodes. In the first case, we can apply to $N$ either the
reduction $U(i)$ (if $v$ has only one child, and it is the tree leaf $i$) or
$T(i;j)$ (if $v$ has at least two tree leaf children, $i$ and $j$). In the
second case, let $v_{1},\ldots,v_{k}$, with $k\geqslant 2$, be the parents of
$i$. By the tree child condition, each $v_{l}$, with $l=1,\ldots,k$, has some
tree child, and by the assumption on $i$, it will be a leaf, say $j_{l}$.
Then, we can apply to $N$ the reduction $H(i;j_{1},\ldots,j_{k})$.
Therefore, every TCTC-network with $n\geqslant 3$ leaves and $m$ internal
nodes is obtained by the application of an $\textrm{U}^{-1}$,
$\textrm{T}^{-1}$, or $\textrm{H}^{-1}$ expansion to a TCTC-network with
either $n-1$ leaves or $n$ leaves and $m-1$ internal nodes. This allows the
recursive construction of all TCTC-networks from TCTC-networks (actually,
phylogenetic trees) with 2 leaves and 1 internal node.
###### Example 1
Fig. 5 shows how a sequence of reductions transforms a certain TCTC-network
$N$ with 4 leaves into a phylogenetic tree with 2 leaves. The sequence of
inverse expansions would then generate $N$ from this phylogenetic tree. This
sequence of expansions generating $N$ is, of course, not unique.
$1$$2$$3$$4$
$\Longrightarrow$$U(3)$
$1$$2$$4$$3$
$\Longrightarrow$$H(3;1,4)$
$1$$2$$4$
$\Longrightarrow$$T(2;1)$
$1$$4$
$\Longrightarrow$$U(1)$
$4$$1$
$\Longrightarrow$$U(4)$
$1$$4$
Figure 5: A sequence of reductions.
## 4 Path lengths vectors for fully resolved networks
Let $N$ be a hybridization network on $S=\\{1,\dots,n\\}$. For every pair of
leaves ${i},{j}$ of $N$, let $\ell_{N}(i,j)$ and $\ell_{N}(j,i)$ be the
distance from $[i,j]$ to $i$ and to $j$, respectively.
###### Definition 3
The _LCSA-path length_ between two leaves $i$ and $j$ in $N$ is
$L_{N}(i,j)=\ell_{N}(i,j)+\ell_{N}(j,i).$
The _LCSA-path lengths vector_ of $N$ is
$L(N)=\big{(}L_{N}(i,j)\big{)}_{1\leqslant i<j\leqslant
n}\in\mathbb{N}^{n(n-1)/2},$
with its entries ordered lexicographically in $(i,j)$.
Notice that $L_{N}(i,j)=L_{N}(j,i)$, for every pair of leaves $i,j\in S$.
If $N$ is a phylogenetic tree, the LCSA-path length between two leaves is the
path length between them as defined in §2.2, and therefore the vectors $L(N)$
defined therein and here are the same. But, contrary to what happens in
phylogenetic trees, the LCSA-path length between two leaves $i$ and $j$ in a
hybridization network need not be the smallest sum of the distances from a
common ancestor of $i$ and $j$ to these leaves (that is, the distance between
these leaves in the undirected graph associated to the network).
###### Example 2
Consider the TCTC-network $N$ depicted in Fig. 6. Table 1 gives, in its upper
triangle, the LCSA of every pair of different leaves, and in its lower
triangle, the LCSA-path length between every pair of different leaves.
Notice that, in this network, $[3,5]=r$, because the root is the only common
ancestor of 3 and 5 that is strict ancestor of some of them, and hence
$L_{N}(3,5)=8$, but $e$ is a common ancestor of both leaves and the length of
both paths $e\\!\rightsquigarrow\\!{}3$ and $e\\!\rightsquigarrow\\!{}5$ is 3.
Similarly, $f$ is also a common ancestor of both leaves and the length of both
paths $f\\!\rightsquigarrow\\!{}3$ and $f\\!\rightsquigarrow\\!{}5$ is 3. This
is an example of LCSA-path length between two leaves that is largest than the
smallest sum of the distances from a common ancestor of these leaves to each
one of them.
$1$$2$$3$$4$$5$$6$$5$$A$$B$$a$$b$$c$$d$$e$$f$$r$ Figure 6: The network $N$ in Example 2. Table 1: For every $1\leqslant i<j\leqslant 6$, the entry $(i,j)$ of this table is $[i,j]$, and the entry $(j,i)$ is $L_{N}(i,j)$, with $N$ the network in Fig. 6. | 1 | 2 | 3 | 4 | 5 | 6
---|---|---|---|---|---|---
1 | | $e$ | $e$ | $r$ | $a$ | $r$
2 | 4 | | $b$ | $r$ | $e$ | $r$
3 | 5 | 3 | | $c$ | $r$ | $f$
4 | 6 | 6 | 3 | | $f$ | $f$
5 | 3 | 5 | 8 | 5 | | $d$
6 | 6 | 6 | 5 | 4 | 3 |
In a _fully resolved_ phylogeny with reticulation events, every non extant
species should have two direct descendants, and every reticulation event
should involve two parent species, as such an event corresponds always to the
exchange of genetic information between two parents: as Semple points out
[23], hybrid nodes with in-degree greater than 2 actually represent “an
uncertainty of the exact order of ‘hybridization’.” Depending on whether we
use hybridization or phylogenetic networks to model phylogenies, we
distinguish between:
* •
_Fully resolved hybridization networks_ : hybridization networks with all
their nodes of types $(0,2)$, $(1,0)$, $(1,2)$, $(2,0)$, or $(2,2)$.
* •
_Fully resolved phylogenetic networks_ : phylogenetic networks with all their
nodes of types $(0,2)$, $(1,0)$, $(1,2)$, or $(2,1)$.
To simplify the language, we shall say that a hybridization network is _quasi-
binary_ when all its nodes are of types $(0,2)$, $(1,0)$, $(1,2)$, $(2,0)$,
$(2,1)$, or $(2,2)$. These quasi-binary networks include as special cases the
fully resolved hybridization and phylogenetic networks.
Our main result in this section establishes that the LCSA-path lengths vectors
separate fully resolved (hybridization or phylogenetic) TCTC-networks, thus
generalizing Proposition 1 from trees to networks. To prove this result, we
shall use the same strategy as the one developed in [4] or [6] to prove that
the metrics introduced therein were indeed metrics: algebraic induction based
on reductions. Now, we cannot use the reductions defined in the last section
as they stand, because they may generate elementary nodes that are forbidden
in fully resolved networks. Instead, we shall use certain suitable
combinations of them that always reduce in one the number of leaves.
So, consider the following reduction procedures for quasi-binary TCTC networks
$N$ with $n$ leaves:
1. ($\mathbf{R}$)
Let $i,j$ be two sibling tree leaves of $N$. The _$R(i;j)$ reduction_ of $N$
is the quasi-binary TCTC-network $N_{R(i;j)}$ obtained by applying first the
$T(i;j)$ reduction to $N$ and then the $U(j)$ reduction to the resulting
network. The final result is that the leaves $i$ and $j$ are removed, together
with their incoming arcs, and then their former common parent, which now has
become a leaf, is labeled with $j$; cf. Fig. 7.
$u$$j$$i$
$\Longrightarrow$
$j$
Figure 7: The $R(i;j)$-reduction.
2. ($\mathbf{H}_{0}$)
Let $i$ be a hybrid leaf, let $v_{1}$ and $v_{2}$ be its parents and assume
that the other children of these parents are tree leaves $j_{1}$ and $j_{2}$,
respectively. The _$H_{0}(i;j_{1},j_{2})$ reduction_ of $N$ is the quasi-
binary TCTC-network $N_{H_{0}(i;j_{1},j_{2})}$ obtained by applying first the
reduction $H(i;j_{1},j_{2})$ to $N$ and then the reductions $U(j_{1})$ and
$U(j_{2})$ to the resulting network. The overall effect is that the hybrid
leaf $i$ and the tree leaves $j_{1},j_{2}$ are removed, together with their
incoming arcs, and then the former parents $v_{1},v_{2}$ of $j_{1}$ and
$j_{2}$ are labeled with $j_{1}$ and $j_{2}$, respectively; cf. Fig. 8.
$i$$v_{1}$$j_{1}$$v_{2}$$j_{2}$
$\Longrightarrow$
$j_{1}$$j_{2}$
Figure 8: The $H_{0}(i;j_{1},j_{2})$-reduction.
3. ($\mathbf{H}_{1}$)
Let $A$ be a hybrid node with only one child $i$, that is a tree node. Let
$v_{1}$ and $v_{2}$ be the parents of $A$ and assume that the other children
of these parents are tree leaves $j_{1}$ and $j_{2}$, respectively. The
_$H_{1}(i;j_{1},j_{2})$ reduction_ of $N$ is the TCTC-network
$N_{H_{1}(i;j_{1},j_{2})}$ obtained by applying first the reduction $U(i)$ to
$N$, followed by the reduction $H_{0}(i;j_{1},j_{2})$ to the resulting
network. The overall effect is that the leaf $i$, its parent $A$ and the
leaves $j_{1},j_{2}$ are removed, together with their incoming arcs, and then
the former parents $v_{1},v_{2}$ of $j_{1}$ and $j_{2}$ are labeled with
$j_{1}$ and $j_{2}$, respectively; cf. Fig. 9.
$A$$i$$v_{1}$$j_{1}$$v_{2}$$j_{2}$
$\Longrightarrow$
$j_{1}$$j_{2}$
Figure 9: The $H_{1}(i;j_{1},j_{2})$-reduction.
We use $\textrm{H}_{0}$ and $\textrm{H}_{1}$ instead of H and U because, for
our purposes in this section, it has to be possible to decide whether or not
we can apply a given reduction to a given fully resolved network $N$ from the
knowledge of $L(N)$, and this cannot be done for the U reduction, while, as we
shall see below, it is possible for $\textrm{H}_{0}$ and $\textrm{H}_{1}$.
$\textrm{H}_{0}$ reductions cannot be applied to fully resolved phylogenetic
networks (they don’t have hybrid leaves) and $\textrm{H}_{1}$ reductions
cannot be applied to fully resolved hybridization networks (they don’t have
out-degree 1 hybrid nodes). The application of an R or an $\textrm{H}_{0}$
reduction to a fully resolved TCTC hybridization network is again a fully
resolved TCTC hybridization network, and the application of an R or an
$\textrm{H}_{1}$ reduction to a fully resolved TCTC phylogenetic network is
again a fully resolved TCTC phylogenetic network.
We shall call the inverses of the R, H0 and H1 reduction procedures,
respectively, the $\textrm{R}^{-1}$, $\textrm{H}_{0}^{-1}$ and
$\textrm{H}_{1}^{-1}$ _expansions_ , and we shall denote them by
$R^{-1}(i;j)$, $H_{1}^{-1}(i;j_{1},j_{2})$ and $H_{0}^{-1}(i;j_{1},j_{2})$.
More specifically, for every quasi-binary TCTC-network $N$ with no leaf
labeled $i$:
* •
the expansion $R^{-1}(i;j)$ can be applied to $N$ if it has a leaf labeled
$j$, and the resulting network $N_{R^{-1}(i;j)}$ is obtained by unlabeling the
leaf $j$ and adding to it two leaf tree children labeled with $i$ and $j$;
* •
the expansion $H_{0}^{-1}(i;j_{1},j_{2})$ can be applied to $N$ if it has a
pair of leaves labeled $j_{1},j_{2}$, and the resulting network
$N_{H_{0}^{-1}(i;j_{1},j_{2})}$ is obtained by adding a new hybrid leaf
labeled with $i$, and then, for each $l=1,2$, unlabeling the leaf $j_{l}$ and
adding to it a new tree leaf child labeled with $j_{l}$ and an arc to $i$.
* •
the expansion $H_{1}^{-1}(i;j_{1},j_{2})$ can be applied to $N$ if it has a
pair of leaves labeled $j_{1},j_{2}$, and the resulting network
$N_{H_{1}^{-1}(i;j_{1},j_{2})}$ is obtained by adding a new node $A$, a tree
leaf child $i$ to it, and then, for each $l=1,2$, unlabeling the leaf $j_{l}$
and adding to it a new tree leaf child labeled with $j_{l}$ and an arc to $A$.
A $\textrm{R}^{-1}(i;j)$ expansion of a quasi-binary TCTC-network is always a
quasi-binary TCTC-network, but an $\textrm{H}_{0}^{-1}(i;j_{1},j_{2})$ or an
$\textrm{H}_{1}^{-1}(i;j_{1},j_{2})$ expansion of a quasi-binary TCTC-network,
while still being always quasi-binary and tree child, needs not be time
consistent: for instance, the leaves $j_{1}$ and $j_{2}$ could be a hybrid
leaf and a tree sibling of it. Moreover, we have the following result, which
is a direct consequence of Lemma 5 and we state it for further reference.
###### Lemma 6
Let $N$ and $N^{\prime}$ be two quasi-binary TCTC-networks. If $N\cong
N^{\prime}$, then the result of applying to both $N$ and $N^{\prime}$ the same
R reduction (respectively, $\textrm{H}_{0}$ reduction, $\textrm{H}_{1}$
reduction, $\textrm{R}^{-1}$ expansion, $\textrm{H}_{0}^{-1}$ expansion, or
$\textrm{H}_{1}^{-1}$ expansion) is again two isomorphic networks.
Moreover, if we apply an R reduction (respectively, $\textrm{H}_{0}$ reduction
or $\textrm{H}_{1}$ reduction) to a quasi-binary TCTC-network $N$, and then we
apply to the resulting network the inverse $\textrm{R}^{-1}$ expansion
(respectively, $\textrm{H}_{0}^{-1}$ expansion or $\textrm{H}_{1}^{-1}$
expansion), we obtain a quasi-binary TCTC-network isomorphic to $N$. ∎
We have moreover the following result.
###### Proposition 4
Let $N$ be a quasi-binary TCTC-network with more than one leaf. Then, at least
one R, $\textrm{H}_{0}$, or $\textrm{H}_{1}$ reduction can be applied to $N$.
###### Proof
If $N$ contains some internal node with two tree leaf children $i$ and $j$,
then the reduction $R(i;j)$ can be applied. If $N$ does not contain any node
with two tree leaf children, then, by Lemma 4, it contains a hybrid node $v$
that is either a leaf (say, labeled with $i$) or it has only one child, which
is a tree leaf (say, labeled with $i$), and such that all siblings of $v$ are
leaves or hybrid nodes. Now, the quasi-binarity of $N$ and the tree child
condition entail that $v$ has two parents, that each one of them has exactly
one child other than $v$, and that this second child is a tree node. So, $v$
has exactly two siblings, and they are tree leaves, say $j_{1}$ and $j_{2}$.
Then, the reduction $H_{0}(i;j_{1},j_{2})$ (if $v$ is a leaf) or
$H_{1}(i;j_{1},j_{2})$ (if $v$ is not a leaf) can be applied.
###### Corollary 2
1. (a)
If $N$ is a fully resolved TCTC hybridization network with more than one leaf,
then, at least one R or $\textrm{H}_{0}$ reduction can be applied to it.
2. (b)
If $N$ is a fully resolved TCTC phylogenetic network with more than one leaf,
then, at least one R or $\textrm{H}_{1}$ reduction can be applied to it.∎
We shall prove now that the application conditions for the reductions
introduced above can be read from the LCSA-path lengths vector of a fully
resolved TCTC-network and that they modify in a specific way the LCSA-path
lengths of the network which they are applied to. This will entail that if two
fully resolved (hybridization or phylogenetic) TCTC-networks have the same
LCSA-path lengths vectors, then the same reductions can be applied to both
networks and the resulting fully resolved TCTC-networks still have the same
LCSA-path lengths vectors. This will be the basis of the proof by induction on
the number of leaves that two TCTC hybridization or phylogenetic networks with
the same LCSA-path lengths vectors are always isomorphic.
###### Lemma 7
Let $i,j$ be two leaves of a quasi-binary TCTC-network $N$. Then, $i$ and $j$
are siblings if, and only if, $L_{N}(i,j)=2$.
###### Proof
If $L_{N}(i,j)=2$, then the paths $[i,j]\\!\rightsquigarrow\\!{}i$ and
$[i,j]\\!\rightsquigarrow\\!{}j$ have length 1, and therefore $[i,j]$ is a
parent of $i$ and $j$. Conversely, if $i$ and $j$ are siblings and $u$ is a
parent in common of them, then, by the quasi-binarity of $N$, they are the
only children of $u$, and by the tree-child condition, one of them, say $i$,
is a tree node. But then, $u$ is a strict ancestor of $i$, an ancestor of $j$,
and no proper descendant of $u$ is an ancestor of both $i$ and $j$. This
implies that $u=[i,j]$ and hence that $L_{N}(i,j)=2$.
###### Lemma 8
Let $N$ be a quasi-binary TCTC-network on a set $S$ of taxa.
1. (1)
The reduction $R(i;j)$ can be applied to $N$ if, and only if, $L_{N}(i,j)=2$
and, for every $k\in S\setminus\\{i,j\\}$, $L_{N}(i,k)=L_{N}(j,k)$.
2. (2)
If the reduction $R(i;j)$ can be applied to $N$, then
$\begin{array}[]{l}L_{N_{R(i;j)}}(j,k)=L_{N}(j,k)-1\quad\mbox{for every $k\in
S\setminus\\{i,j\\}$}\\\ L_{N_{R(i;j)}}(k,l)=L_{N}(k,l)\quad\mbox{for every
$k,l\in S\setminus\\{i,j\\}$}\end{array}$
###### Proof
As far as (1) goes, $R(i;j)$ can be applied to $N$ if, and only if, the leaves
$i$ and $j$ are siblings and of tree type. Now, if $i$ and $j$ are two tree
sibling leaves and $u$ is their parent, then on the one hand, $L_{N}(i,j)=2$
by Lemma 7, and on the other hand, since, by Lemma 2, $[i,k]=[u,k]=[j,k]$ for
every leaf $k\neq i,j$, we have that
$\begin{array}[]{l}\ell_{N}(i,k)=\ell_{N}(j,k)=1+\mbox{distance from $[u,k]$
to $u$}\\\ \ell_{N}(k,i)=\ell_{N}(k,j)=\mbox{distance from $[u,k]$ to
$k$}\end{array}$
and therefore $L_{N}(i,k)=L_{N}(j,k)$ for every $k\in S\setminus\\{i,j\\}$.
Conversely, assume that $L_{N}(i,j)=2$ and that $L_{N}(i,k)=L_{N}(j,k)$ for
every $k\in S\setminus\\{i,j\\}$. The fact that $L_{N}(i,j)=2$ implies that
$i$ and $j$ share a parent $u$. If one of these leaves, say $i$, is hybrid,
then the tree child condition implies that the other, $j$, is of tree type.
Let now $v$ be the other parent of $i$ and $k$ a tree descendant leaf of $v$,
and let $h$ be the length of the unique path $v\\!\rightsquigarrow\\!{}k$.
Then $v$ is a strict ancestor of $k$ and an ancestor of $i$, and no proper
tree descendant of $v$ can possibly be an ancestor of $i$: otherwise, there
would exist a path from a proper tree descendant of $v$ to $u$, and then the
time consistency property would forbid $u$ and $v$ to have a hybrid child in
common. Therefore $v=[i,k]$ and $L_{N}(i,k)=h+1$. Now, the only possibility
for the equality $L_{N}(j,k)=h+1$ to hold is that some intermediate node in
the path $v\\!\rightsquigarrow\\!{}k$ is an ancestor of the only parent $u$ of
$j$, which, as we have just seen, is impossible. This leads to a
contradiction, which shows that $i$ and $j$ are both tree sibling leaves. This
finishes the proof of (1).
As far as (2) goes, in $N_{R(i;j)}$ we remove the leaf $i$ and we replace the
leaf $j$ by its parent. By Lemma 2, this does not modify the LCSA $[j,k]$ of
$j$ and any other remaining leaf $k$, and since we have shortened in 1 any
path ending in $j$, we deduce that $L_{N_{R(i;j)}}(j,k)=L_{N}(j,k)-1$ for
every $k\in S\setminus\\{i,j\\}$. On the other hand, for every $k,l\in
S\setminus\\{i,j\\}$, the reduction $R(i;j)$ has affected neither the LCSA
$[k,l]$ of $k$ and $l$, nor the paths $[k,l]\\!\rightsquigarrow\\!{}k$ or
$[k,l]\\!\rightsquigarrow\\!{}l$, which implies that
$L_{N_{R(i;j)}}(k,l)=L_{N}(k,l)$
###### Lemma 9
Let $N$ be a fully resolved TCTC hybridization network on a set $S$ of taxa.
1. (1)
The reduction $H_{0}(i;j_{1},j_{2})$ can be applied to $N$ if, and only if,
$L_{N}(i,j_{1})=L_{N}(i,j_{2})=2$.
2. (2)
If the reduction $H_{0}(i;j_{1},j_{2})$ can be applied to $N$, then
$\begin{array}[]{l}L_{N_{H_{0}(i;j_{1},j_{2})}}(j_{1},j_{2})=L_{N}(j_{1},j_{2})-2\\\
L_{N_{H_{0}(i;j_{1},j_{2})}}(j_{1},k)=L_{N}(j_{1},k)-1\quad\mbox{for every
$k\in S\setminus\\{i,j_{1},j_{2}\\}$}\\\
L_{N_{H_{0}(i;j_{1},j_{2})}}(j_{2},k)=L_{N}(j_{2},k)-1\quad\mbox{for every
$k\in S\setminus\\{i,j_{1},j_{2}\\}$}\\\
L_{N_{H_{0}(i;j_{1},j_{2})}}(k,l)=L_{N}(k,l)\quad\mbox{for every $k,l\in
S\setminus\\{i,j_{1},j_{2}\\}$}\end{array}$
###### Proof
As far as (1) goes, the reduction $H_{0}(i;j_{1},j_{2})$ can be applied to $N$
if, and only if, $i$ is a hybrid sibling of the tree leaves $j_{1}$ and
$j_{2}$. If this last condition happens, then $L_{N}(i,j_{1})=2$ and
$L_{N}(i,j_{2})=2$ by Lemma 7. Conversely, $L_{N}(i,j_{1})=L_{N}(i,j_{2})=2$
implies that $i,j_{1}$ and $i,j_{2}$ are pairs of sibling leaves. Since no
node of $N$ can have more than 2 children, and at least one of its children
must be of tree type, this implies that $i$ is a hybrid node (with two
different parents), and $j_{1}$ and $j_{2}$ are tree nodes.
As far as (2) goes, the tree leaves $j_{1}$ and $j_{2}$ are replaced by their
parents. By Lemma 7, this does not affect any LCSA and it only shortens in 1
the paths ending in $j_{1}$ or $j_{2}$. Thus, the $H_{0}(i;j_{1},j_{2})$
reduction does not affect the LCSA-path length between any pair of remaining
leaves other than $j_{1}$ and $j_{2}$, it shortens in 1 the LCSA-path length
between $j_{1}$ or $j_{2}$ and any remaining leaf other than $j_{1}$ or
$j_{2}$, and it shortens in 2 the LCSA-path length between $j_{1}$ and
$j_{2}$.
###### Lemma 10
Let $N$ be a fully resolved TCTC phylogenetic network on a set $S$ of taxa.
1. (1)
The reduction $H_{1}(i;j_{1},j_{2})$ can be applied to $N$ if, and only if,
* •
$L_{N}(i,j_{1})=L_{N}(i,j_{2})=3$,
* •
$L_{N}(j_{1},j_{2})\geqslant 4$,
* •
if $L_{N}(j_{1},j_{2})=4$, then $L_{N}(j_{1},k)=L_{N}(j_{2},k)$ for every
$k\in S\setminus\\{j_{1},j_{2},i\\}$.
2. (2)
If the reduction $H_{1}(i;j_{1},j_{2})$ can be applied to $N$, then
$\begin{array}[]{l}L_{N_{H_{1}(i;j_{1},j_{2})}}(j_{1},j_{2})=L_{N}(j_{1},j_{2})-2\\\
L_{N_{H_{1}(i;j_{1},j_{2})}}(j_{1},k)=L_{N}(j_{1},k)-1\ \mbox{for every $k\in
S\setminus\\{i,j_{1},j_{2}\\}$}\\\
L_{N_{H_{1}(i;j_{1},j_{2})}}(j_{2},k)=L_{N}(j_{2},k)-1\ \mbox{for every $k\in
S\setminus\\{i,j_{1},j_{2}\\}$}\\\
L_{N_{H_{1}(i;j_{1},j_{2})}}(k,l)=L_{N}(k,l)\ \mbox{for every $k\in
S\setminus\\{i,j_{1},j_{2}\\}$}\end{array}$
###### Proof
As far as (1) goes, the reduction $H_{1}(i;j_{1},j_{2})$ can be applied to $N$
if, and only if, $j_{1}$ and $j_{2}$ are tree leaves that are not siblings and
they share a sibling hybrid node that has the tree leaf $i$ as its only child.
Now, if this application condition for $H_{1}(i;j_{1},j_{2})$ is satisfied,
then $L_{N}(i,j_{1})=3$, because the parent of $j_{1}$ is an ancestor of $i$,
a strict ancestor of $j_{1}$, and clearly no proper descendant of it is an
ancestor of $i$ and $j_{1}$; by a similar reason, $L_{N}(i,j_{2})=3$.
Moreover, since $j_{1}$ and $j_{2}$ are not sibling,
$L_{N}(j_{1},j_{2})\geqslant 3$. But if $L_{N}(j_{1},j_{2})=3$, then there
would exist an arc from the parent of $j_{1}$ to the parent of $j_{2}$, or
vice versa, which would entail a node of out-degree 3 that cannot exist in the
fully resolved network $N$. Therefore, $L_{N}(j_{1},j_{2})\geqslant 4$.
Finally, if $L_{N}(j_{1},j_{2})=4$, this means that the parents $x$ and $y$ of
$j_{1}$ and $j_{2}$ (that are tree nodes, because they have out-degree 2 and
$N$ is a phylogenetic network) are sibling: let $u$ be their parent in common.
In this case, no leaf other than $j_{1},j_{2},i$ is a descendant of $u$, and
therefore, for every $k\in S\setminus\\{j_{1},j_{2},i\\}$,
$[j_{1},k]=[x,k]=[u,k]=[y,k]=[j_{2},k]$
by Lemma 4, and thus
$\begin{array}[]{l}\ell_{N}(j_{1},k)=\ell_{N}(j_{2},k)=2+\mbox{distance from
$[u,k]$ to $u$}\\\ \ell_{N}(k,j_{1})=\ell_{N}(k,j_{2})=\mbox{distance from
$[u,k]$ to $k$},\end{array}$
which implies that $L_{N}(j_{1},k)=L_{N}(j_{2},k)$.
Conversely, assume that $L_{N}(i,j_{1})=L_{N}(i,j_{2})=3$, that
$L_{N}(j_{1},j_{2})\geqslant 4$, and that if $L_{N}(j_{1},j_{2})=4$, then
$L_{N}(j_{1},k)=L_{N}(j_{2},k)$ for every $k\in
S\setminus\\{j_{1},j_{2},i\\}$. Let $x$, $y$ and $z$ be the parents of
$j_{1}$, $j_{2}$ and $i$, respectively. Notice that these parents are pairwise
different (otherwise, the LCSA-path length between a pair among
$j_{1},j_{2},i$ would be 2). Moreover, since $N$ is a phylogenetic network,
$j_{1}$, $j_{2}$ and $i$ are tree nodes. Then,
$L_{N}(i,j_{1})=L_{N}(i,j_{2})=3$ implies that there must exist an arc between
the nodes $x$ and $z$ and an arc between the nodes $y$ and $z$.
Now, if these arcs are $(z,x)$ and $(z,y)$, the node $z$ would have out-degree
3, which is impossible. Assume now that $(x,z)$ and $(z,y)$ are arcs of $N$.
In this case, both $z$ and $x$ have out-degree 2, which implies (recall that
$N$ is a phylogenetic network) that they are tree nodes. Then,
$x=[j_{1},j_{2}]$ (it is an ancestor of $j_{2}$, a strict ancestor of $j_{1}$,
and no proper descendant of it is an ancestor of $j_{1}$ and $j_{2}$) and
therefore $L_{N}(j_{1},j_{2})=4$. In this case, we assume that
$L_{N}(j_{1},k)=L_{N}(j_{2},k)$ for every $k\in
S\setminus\\{j_{1},j_{2},i\\}$. Now we must distinguish two cases, depending
on the type of node $y$:
* •
If $y$ is a tree node, let $p$ be its child other than $j_{2}$, and let $k$ be
a tree descendant leaf of $p$. In this case, $[j_{1},k]=x$ and $[j_{2},k]=y$
(by the same reason why $x$ is $[j_{1},j_{2}]$), and hence
$L_{N}(j_{1},k)=L_{N}(j_{2},k)+2$, against the assumption
$L_{N}(j_{1},k)=L_{N}(j_{2},k)$.
* •
If $y$ is a hybrid node, let $p$ be its parent other than $z$, and let $k$ be
a tree descendant leaf of $p$ ($k\neq{j_{2}}$, because $j_{2}$ is not a tree
descendant of $p$). In this case, $[j_{2},k]=p$ (because $p$ is an ancestor of
$j_{2}$ and a strict ancestor of $k$, and the time consistency property
implies that no intermediate node in the path $p\\!\rightsquigarrow\\!{}k$ can
be an ancestor of $y$). Now, if the length of the (only) path
$p\\!\rightsquigarrow\\!{}k$ is $h$, then $L_{N}(j_{2},k)=h+2$, and for the
equality $L_{N}(j_{1},k)=h+2$ to hold, either the arc $(x,p)$ belongs to $N$,
which is impossible because $x$ would have out-degree 3, or a node in the path
$p\\!\rightsquigarrow\\!{}k$ is an ancestor of $x$, which is impossible
because of the time consistency property.
In both cases we reach a contradiction that implies that the arcs
$(x,z),(z,y)$ do not exist in $N$. By symmetry, the arcs $(y,z),(z,x)$ do not
exist in $N$, either. Therefore, the only possibility is that $N$ contains the
arcs $(x,z),(y,z)$, that is, that $z$ is hybrid child of the nodes $x$ and
$y$. This finishes the proof of (1).
As far as (2) goes, it is proved as in Lemma 9.
Now we can prove the main results in this section.
###### Proposition 5
Let $N$ and $N^{\prime}$ be two fully resolved TCTC hybridization networks on
the same set $S$ of taxa. Then, $L(N)=L(N^{\prime})$ if, and only if, $N\cong
N^{\prime}$.
###### Proof
The ‘if’ implication is obvious. We prove the ‘only if’ implication by
induction on the number $n$ of elements of $S$.
The cases $n=1$ and $n=2$ are straightforward, because there exist only one
TCTC-network on $S=\\{1\\}$ and one TCTC-network on $S=\\{1,2\\}$: the one-
node graph and the phylogenetic tree with leaves 1,2, respectively.
Assume now that the thesis is true for fully resolved TCTC hybridization
networks with $n$ leaves, and let $N$ and $N^{\prime}$ be two fully resolved
TCTC hybridization networks on the same set $S$ of $n+1$ labels such that
$L(N)=L(N^{\prime})$. By Corollary 2.(a), an $R(i;j)$ or a
$H_{0}(i;j_{1},j_{2})$ can be applied to $N$. Moreover, since the possibility
of applying one such reduction depends on the LCSA-path lengths vector by
Lemmas 8.(1) and 9.(1), and $L(N)=L(N^{\prime})$, it will be possible to apply
the same reduction to $N^{\prime}$. So, let $N_{1}$ and $N_{1}^{\prime}$ be
the fully resolved TCTC hybridization networks obtained by applying the same R
or $\textrm{H}_{0}$ reduction to $N$ and $N^{\prime}$.
From Lemmas 8.(2) and 9.(2) we deduce that $L(N_{1})=L(N_{1}^{\prime})$ and
hence, by the induction hypothesis, $N_{1}\cong N_{1}^{\prime}$. Finally, if
we apply to $N_{1}$ and $N_{1}^{\prime}$ the $\mathrm{R}^{-1}$ or
$\mathrm{H}_{0}^{-1}$ expansion that is inverse to the reduction applied to
$N$ and $N^{\prime}$, then, by Lemma 6, we obtain again $N$ and $N^{\prime}$
and they are isomorphic.
A similar argument, using Lemmas 8 and 10, proves the following result.
###### Proposition 6
Let $N$ and $N^{\prime}$ be two fully resolved TCTC phylogenetic networks on
the same set $S$ of taxa. Then, $L(N)=L(N^{\prime})$ if, and only if, $N\cong
N^{\prime}$.∎
###### Remark 2
The LCSA-path lengths vectors do not separate quasi-binary TCTC-networks.
Indeed, consider the TCTC-networks $N,N^{\prime}$ depicted in Fig. 10. They
are quasi-binary (but neither fully resolved phylogenetic networks nor fully
resolved hybridization networks), and a simple computation shows that
$L(N)=L(N^{\prime})=(3,6,3,3,6,3).$
The network $N$ in Fig. 10 also shows that Lemma 10.(1) is false for quasi-
binary hybridization networks.
$1$$2$$3$$4$$N$
$1$$2$$3$$4$$N^{\prime}$
Figure 10: These two quasi-binary TCTC-networks have the same LCSA-path
length vectors.
Let $\mathrm{FRH}_{n}$ (respectively, $\mathrm{FRP}_{n}$) denote the classes
of fully resolved TCTC hybridization (respectively, phylogenetic) networks on
$S=\\{1,\ldots,n\\}$. We have just proved that the mappings
$L:\mathrm{FRH}_{n}\to\mathbb{R}^{n(n-1)/2},\quad
L:\mathrm{FRP}_{n}\to\mathbb{R}^{n(n-1)/2}$
are injective, and therefore they can be used to induce metrics on
$\mathrm{FRH}_{n}$ and $\mathrm{FRP}_{n}$ from metrics on
$\mathbb{R}^{n(n-1)/2}$.
###### Proposition 7
For every $n\geqslant 1$, let $D$ be any metric on $\mathbb{R}^{n(n-1)/2}$.
The mappings $d:\mathrm{FRH}_{n}\times\mathrm{FRH}_{n}\to\mathbb{R}$ and
$d:\mathrm{FRP}_{n}\times\mathrm{FRP}_{n}\to\mathbb{R}$ defined by
$d(N_{1},N_{2})=D(L(N_{1}),L(N_{2}))$ satisfy the axioms of metrics up to
isomorphisms:
1. (1)
$d(N_{1},N_{2})\geqslant 0$,
2. (2)
$d(N_{1},N_{2})=0$ if, and only if, $N_{1}\cong N_{2}$,
3. (3)
$d(N_{1},N_{2})=d(N_{2},N_{1})$,
4. (4)
$d(N_{1},N_{3})\leqslant d(N_{1},N_{2})+d(N_{2},N_{3})$.
###### Proof
Properties (1), (3) and (4) are direct consequences of the corresponding
properties of $D$, while property (2) follows from the separation axiom for
$D$ (which says that $D(M_{1},M_{2})=0$ if, and only if, $M_{1}=M_{2}$) and
Proposition 5 or 6, depending on the case.
For instance, using as $D$ the Manhattan distance on $\mathbb{R}^{n(n-1)/2}$,
we obtain the metric on $\mathrm{FRH}_{n}$ or $\mathrm{FRP}_{n}$
$d_{1}(N_{1},N_{2})=\sum_{1\leqslant i<j\leqslant
n}|L_{N_{1}}(i,j)-L_{N_{2}}(i,j)|,$
and using as $D$ the Euclidean distance we obtain the metric
$d_{2}(N_{1},N_{2})=\sqrt{\sum_{1\leqslant i<j\leqslant
n}(L_{N_{1}}(i,j)-L_{N_{2}}(i,j))^{2}}.$
These metrics generalize to fully resolved TCTC (hybridization or
phylogenetic) networks the classical distances for fully resolved phylogenetic
trees introduced by Farris [11] and Clifford [29] around 1970.
## 5 Splitted path lengths vectors for arbitrary networks
As we have seen in §2.2 and Remark 2, the path lengths vectors do not separate
arbitrary TCTC-networks. Since to separate arbitrary phylogenetic trees we
splitted the path lengths (Definition 2), we shall use the same strategy in
the networks setting. In this connection, we already proved in [6] that the
matrix
$\ell(N)=\big{(}\ell_{N}(i,j)\big{)}_{i=1,\ldots,n\atop j=1,\ldots,n}$
separates TCTC _phylogenetic_ networks on $S=\\{1,\ldots,n\\}$ with tree nodes
of arbitrary out-degree and hybrid nodes of arbitrary in-degree. But it is not
true for TCTC hybridization networks, as the following example shows.
$1$$2$$3$$4$$5$$6$$N$
$1$$2$$3$$4$$5$$6$$N^{\prime}$
Figure 11: These two hybridization TCTC-networks are such that
$\ell_{N}(i,j)=\ell_{N^{\prime}}(i,j)$, for every pair of leaves $i,j$.
###### Example 3
Consider the pair of non-isomorphic TCTC-networks $N$ and $N^{\prime}$
depicted in Fig. 11. A simple computation shows that
$\ell(N)=\ell(N^{\prime})=\left(\begin{array}[]{cccccc}0&1&2&2&1&2\\\\[-4.30554pt]
1&0&1&1&2&2\\\\[-4.30554pt] 2&1&0&2&2&2\\\\[-4.30554pt]
2&1&2&0&1&2\\\\[-4.30554pt] 1&2&2&1&0&1\\\\[-4.30554pt]
2&2&2&2&1&0\end{array}\right)$
So, in order to separate arbitrary TCTC-networks we need to add some extra
information to the distances $\ell_{N}(i,j)$ from LCSAs to leaves. The extra
information we shall use is whether the LCSA of each pair of leaves is a
strict ancestor of one leaf or the other (or both). So, for every pair of
different leaves $i,j$ of $N$, let $h_{N}(i,j)$ be $-1$ if $[i,j]$ is a strict
ancestor of $i$ but not of $j$, $1$ if $[i,j]$ is a strict ancestor of $j$ but
not of $i$, and $0$ if $[i,j]$ is a strict ancestor of both $i$ and $j$.
Notice that $h_{N}(j,i)=-h_{N}(i,j)$.
###### Definition 4
Let $N$ be a hybridization network on the set $S=\\{1,\ldots,n\\}$.
For every $i,j\in S$, the _splitted LCSA-path length_ from $i$ to $j$ is the
ordered 3-tuple
$L_{N}^{s}(i,j)=(\ell_{N}(i,j),\ell_{N}(j,i),h_{N}(i,j)).$
The _splitted LCSA-path lengths vector_ of $N$ is
$L^{s}(N)=\big{(}L^{s}_{N}(i,j)\big{)}_{1\leqslant i<j\leqslant
n}\in(\mathbb{N}\times\mathbb{N}\times\\{-1,0,1\\})^{n(n-1)/2}$
with its entries ordered lexicographically in $(i,j)$.
###### Example 4
Consider the quasi-binary TCTC-networks $N$ and $N^{\prime}$ depicted in Fig.
10. Then
$\begin{array}[]{rl}L^{s}(N)&\\!\\!=\\!\\!\big{(}(2,1,-1),(3,3,0),(1,2,-1),(1,2,1),(2,4,0),(1,2,-1)\big{)}\\\
L^{s}(N^{\prime})&\\!\\!=\\!\\!\big{(}(1,2,1),(2,4,0),(1,2,1),(1,2,-1),(3,3,0),(2,1,1)\big{)}\end{array}$
###### Example 5
Consider the TCTC-networks $N$ and $N^{\prime}$ depicted in Fig. 11. Then
$\begin{array}[]{rl}L^{s}(N)&\\!\\!=\\!\\!\big{(}(1,1,-1),(2,2,0),(2,2,0),(1,1,-1),(2,2,0),(1,1,1),(1,1,1),\\\
&\ (2,2,0),(2,2,0),(2,2,0),(2,2,0),(2,2,0),(1,1,-1),(2,2,0),(1,1,1)\big{)}\\\
L^{s}(N^{\prime})&\\!\\!=\\!\\!\big{(}(1,1,1),(2,2,0),(2,2,0),(1,1,1),(2,2,0),(1,1,-1),(1,1,-1),\\\
&\
(2,2,0),(2,2,0),(2,2,0),(2,2,0),(2,2,0),(1,1,1),(2,2,0),(1,1,-1)\big{)}\end{array}$
###### Remark 3
If $N$ is a phylogenetic tree on $S$, then $h_{N}(i,j)=0$ for every $i,j\in
S$.
We shall prove now that these splitted LCSA-path lengths vectors separate
arbitrary hybridization TCTC-networks. The master plan for proving it is
similar to the one used in the proof of Proposition 5: induction based on the
fact that the application conditions for the reductions introduced in Section
3 can be read in the splitted LCSA-path lengths vectors of TCTC-networks and
that these reductions modify in a controlled way these vectors.
###### Lemma 11
Let $N$ be a TCTC-network on a set $S$ of taxa.
1. (1)
The reduction $U(i)$ can be applied to $N$ if, and only if,
$\ell_{N}(i,j)\geqslant 2$ for every $j\in S\setminus\\{i\\}$.
2. (2)
If the reduction $U(i)$ can be applied to $N$, then
$\begin{array}[]{l}L_{N_{U(i)}}(i,j)=L_{N}(i,j)-(1,0,0)\quad\mbox{for every
$j\in S\setminus\\{i\\}$}\\\ L_{N_{U(i)}}(j,k)=L_{N}(j,k)\quad\mbox{for every
$j,k\in S\setminus\\{i\\}$}\end{array}$
###### Proof
As far as (1) goes, the reduction $U(i)$ can be applied to $N$ if, and only
if, the leaf $i$ is a tree node and the only child of its parent. Let us check
now that this last condition is equivalent to $\ell_{N}(i,j)\geqslant 2$ for
every $j\in S\setminus\\{i\\}$. To do this, we distinguish three cases:
* •
Assume that $i$ is a tree node and the only child of its parent $x$. Then, for
every $j\in S\setminus\\{i\\}$, the LCSA of $i$ and $j$ is a proper ancestor
of $x$, and therefore $\ell_{N}(i,j)\geqslant 2$.
* •
Assume that $i$ is a tree node and that it has a sibling $y$. Let $x$ be the
parent of $i$ and $y$ and let $j$ be a tree descendant leaf of $y$. Then
$[i,j]=x$, because $x$ is a strict ancestor of $i$, an ancestor of $j$ and
clearly no descendant of $x$ is an ancestor of both $i$ and $j$. Therefore, in
this case, $\ell_{N}(i,j)=1$ for this leaf $j$.
* •
Assume that $i$ is a hybrid node. Let $x$ be any parent of $i$ and let $j$ be
a tree descendant of $x$. Then, $[i,j]=x$, because $x$ is a strict ancestor of
$j$, an ancestor of $i$, and no intermediate node in the unique path
$x\\!\rightsquigarrow\\!{}j$ is an ancestor of $i$ (it would violate the time
consistency property). Therefore, in this case, $\ell_{N}(i,j)=1$ for this
leaf $j$, too.
Since these three cases cover all possibilities, we conclude that $i$ is a
tree node without siblings if, and only if, $\ell_{N}(i,j)\geqslant 2$ for
every $j\in S\setminus\\{i\\}$. This finishes the proof of (1).
As far as (2) goes, in $N_{U(i)}$ we replace the tree leaf $i$ by its parent.
By Lemma 2, this does not modify any LCSA, and it only shortens in 1 any path
ending in $i$. Therefore
$\begin{array}[]{l}\ell_{N_{U(i)}}(i,j)=\ell_{N}(i,j)-1,\
\ell_{N_{U(i)}}(j,i)=\ell_{N}(j,i)\quad\mbox{for every $j\in
S\setminus\\{i\\}$}\\\ \ell_{N_{U(i)}}(j,k)=\ell_{N}(j,k),\
\ell_{N_{U(i)}}(k,j)=\ell_{N}(k,j)\quad\mbox{for every $j,k\in
S\setminus\\{i\\}$}\\\ \end{array}$
As far as the $h$ component of the splitted LCSA-path lengths goes, notice
that a node $u$ is a strict ancestor of a tree leaf $i$ if, and only if, it is
a strict ancestor of its parent $x$ (because every path ending in $i$ contains
$x$). Therefore, an internal node of $N_{U(i)}$ is a strict ancestor of the
leaf $i$ in $N_{U(i)}$ if, and only if, it is a strict ancestor of the leaf
$i$ in $N$. On the other hand, replacing a tree leaf without siblings by its
only parent does not affect any path ending in another leaf, and therefore an
internal node of $N_{U(i)}$ is a strict ancestor of a leaf $j\neq i$ in
$N_{U(i)}$ if, and only if, it is a strict ancestor of the leaf $j$ in $N$.
So, by Lemma 2, the LCSA of a pair of leaves in $N$ and in $N_{U(i)}$ is the
same, and we have just proved that this LCSA is a strict ancestor of exactly
the same leaves in both networks: this implies that
$\begin{array}[]{l}h_{N_{U(i)}}(i,j)=h_{N}(i,j)\quad\mbox{for every $j\in
S\setminus\\{i\\}$}\\\ h_{N_{U(i)}}(j,k)=h_{N}(j,k)\quad\mbox{for every
$j,k\in S\setminus\\{i\\}$}\\\ \end{array}$
###### Lemma 12
Let $N$ be a TCTC-network on a set $S$ of taxa.
1. (1)
The reduction $T(i;j)$ can be applied to $N$ if, and only if,
$L_{N}^{s}(i,j)=(1,1,0)$.
2. (2)
If the reduction $T(i;j)$ can be applied to $N$, then
$L^{s}_{N_{T(i;j)}}(k,l)=L^{s}_{N}(k,l)\quad\mbox{for every $k,l\in
S\setminus\\{i\\}$}$
###### Proof
As far as (1) goes, $T(i;j)$ can be applied to $N$ if, and only if, the leaves
$i$ and $j$ are tree nodes and sibling. Let us prove that this last condition
is equivalent to $\ell_{N}(i,j)=\ell_{N}(j,i)=1$ and $h_{N}(i,j)=0$. Indeed,
if the leaves $i$ and $j$ are tree nodes and sibling, then their parent is
their LCSA and moreover it is a strict ancestor of both of them, which implies
that $\ell_{N}(i,j)=\ell_{N}(j,i)=1$ and $h_{N}(i,j)=0$. Conversely, assume
that $\ell_{N}(i,j)=\ell_{N}(j,i)=1$ and $h_{N}(i,j)=0$. The equalities
$\ell_{N}(i,j)=\ell_{N}(j,i)=1$ imply that $[i,j]$ is a parent of $i$ and $j$,
and $h_{N}(i,j)=0$ implies that this parent of $i$ and $j$ is a strict
ancestor of both of them, and therefore, by Lemma 3, that $i$ and $j$ are tree
nodes. This finishes the proof of (1).
As far as (2) goes, in $N_{T(i;j)}$ we simply remove the leaf $i$ without
removing anything else. Therefore, no path ending in a remaining leaf is
affected, and as a consequence no $L^{s}(k,l)$ with $k,l\neq i$, is modified.
###### Lemma 13
Let $N$ be a TCTC-network on a set $S$ of taxa.
1. (1)
The reduction $H(i;j_{1},\ldots,j_{k})$ can be applied to $N$ if, and only if,
* •
$L_{N}^{s}(i,j_{l})=(1,1,1)$, for every $l\in\\{1,\ldots,k\\}$.
* •
$\ell_{N}(j_{a},j_{b})\geqslant 2$ or $\ell_{N}(j_{b},j_{a})\geqslant 2$ for
every $a,b\in\\{1,\ldots,k\\}$.
* •
For every $s\notin\\{j_{1},\ldots,j_{k}\\}$, if $\ell_{N}(i,s)=1$ and
$h_{N}(i,s)=1$, then $\ell_{N}(j_{l},s)=1$ and $h_{N}(j_{l},s)=0$ for some
$l\in\\{1,\ldots,k\\}$.
2. (2)
If the reduction $H(i;j_{1},\ldots,j_{k})$ can be applied to $N$, then
$L_{N_{H(i;j_{1},\ldots,j_{k})}}(s,t)=L_{N}(s,t)\quad\mbox{for every $s,t\in
S\setminus\\{i\\}$}$
###### Proof
As far as (1) goes, $H(i;j_{1},\ldots,j_{k})$ can be applied to $N$ if, and
only if, $j_{1},\ldots,j_{k}$ are tree leaves that are not sibling of each
other, the leaf $i$ is a hybrid sibling of $j_{1},\ldots,j_{k}$, and the only
parents of $i$ are those of $j_{1},\ldots,j_{k}$. Now:
* •
For each $l=1,\ldots,k$, the condition $L_{N}^{s}(i,j_{l})=(1,1,1)$ says that
$i$ and $j_{l}$ are sibling, and that their parent in common is a strict
ancestor of $j_{l}$ but not of $i$. Using Lemma 3, we conclude that this
condition is equivalent to the fact that $i$ and $j_{l}$ are sibling, $j_{l}$
is a tree node, and $i$ a hybrid node.
* •
Assume that $j_{1},\ldots,j_{k}$ are tree leaves, with parents
$v_{1},\ldots,v_{k}$, respectively. In this case, the condition
$\ell_{N}(j_{a},j_{b})\geqslant 2$ or $\ell_{N}(j_{b},j_{a})\geqslant 2$ is
equivalent to the fact that $j_{a},j_{b}$ are not sibling. Indeed, if $j_{a}$
and $j_{b}$ are sibling, then $\ell_{N}(j_{a},j_{b})=\ell_{N}(j_{b},j_{a})=1$.
Conversely, if $j_{a}$ and $j_{b}$ are not sibling, then there are two
possibilities: either $v_{a}$ is an ancestor of $j_{b}$, but not its parent,
in which case $v_{a}=[j_{a},j_{b}]$ and $\ell_{N}(j_{b},j_{a})\geqslant 2$, or
$v_{a}$ is not an ancestor of $j_{b}$, in which case $[j_{a},j_{b}]$ is a
proper ancestor of $v_{a}$ and hence $\ell_{N}(j_{a},j_{b})\geqslant 2$.
* •
Assume that $j_{1},\ldots,j_{k}$ are tree leaves, with parents
$v_{1},\ldots,v_{k}$, respectively, and that $i$ is a hybrid sibling of them.
Let us see that the only parents of $i$ are $v_{1},\ldots,v_{k}$ if, and only
if, for every $s\notin\\{j_{1},\ldots,j_{k}\\}$, $\ell_{N}(i,s)=1$ and
$h_{N}(i,s)=1$ imply that $\ell_{N}(j_{l},s)=1$ and $h_{N}(j_{l},s)=0$ for
some $l=1,\ldots,k$.
Indeed, assume that the only parents of $i$ are $v_{1},\ldots,v_{k}$, and let
$s\notin\\{j_{1},\ldots,j_{k}\\}$ be a leaf such that $\ell_{N}(i,s)=1$ and
$h_{N}(i,s)=1$. Since $\ell_{N}(i,s)=1$, some parent of $i$, say $v_{l}$, is
the LCSA of $i$ and $s$, and $h_{N}(i,s)=1$ implies that $v_{l}$ is a strict
ancestor of $s$. But then $v_{l}$ will be the LCSA of its tree leaf $j_{l}$
and $s$ and strict ancestor of both of them, and thus $\ell_{N}(j_{l},s)=1$
and $h_{N}(j_{l},s)=0$.
Conversely, assume that, for every $s\notin\\{j_{1},\ldots,j_{k}\\}$,
$\ell_{N}(i,s)=1$ and $h_{N}(i,s)=1$ imply that $\ell_{N}(j_{l},s)=1$ and
$h_{N}(j_{l},s)=0$ for some $l=1,\ldots,k$. Let $v$ be a parent of $i$, and
let $s$ be a tree descendant leaf of $v$. Then, $v=[i,s]$ ($v$ is a strict
ancestor of $s$, an ancestor of $i$, and no intermediate node in the unique
path $v\\!\rightsquigarrow\\!{}s$ is an ancestor of $i$, by the time
consistency property) and thus $\ell_{N}(i,s)=1$; moreover, $h_{N}(i,s)=1$ by
Lemma 3. Now, if $s=j_{l}$, for some $l=1,\ldots,k$, then $v=v_{l}$. On the
other hand, if $s\notin\\{j_{1},\ldots,j_{k}\\}$, then by assumption, there
will exist some $j_{l}$ such that $\ell_{N}(j_{l},s)=1$ and
$h_{N}(j_{l},s)=0$, that is, such that $v_{l}$ is a strict ancestor of $s$.
This implies that $v=v_{l}$. Indeed, if $v\neq v_{l}$, then either $v_{l}$ is
an intermediate node in the path $v\\!\rightsquigarrow\\!{}s$, and in
particular a tree descendant of $v$, which is forbidden by the time
consistency because $v$ and $v_{l}$ have the hybrid child $i$ in common, or
$v$ is a proper descendant of $v_{l}$ through a path where $v_{l}$ and all the
intermediate nodes are hybrid (if some of these nodes were of tree type, the
temporal representation of $v$ would be greater than that of $v_{l}$,
contradicting again the time consistency), in which case the child of $v_{l}$
in this path would be a hybrid child of $v_{l}$ that is a strict descendant of
it (because it is intermediate in the path
$v_{l}\\!\rightsquigarrow\\!{}v\\!\rightsquigarrow\\!{}s$ and $s$ is a strict
descendant of $v_{l}$), which is impossible by Lemma 3.
This finishes the proof of (1).
As far as (2) goes, in $N_{H(i;j_{1},\ldots,j_{k})}$ we simply remove the
hybrid leaf $i$ without removing anything else, and therefore no splitted
LCSA-path length of a pair of remaining leaves is affected.
###### Theorem 5.1
Let $N$ and $N^{\prime}$ be two TCTC-networks on the same set $S$ of taxa.
Then, $L^{s}(N)=L^{s}(N^{\prime})$ if, and only if, $N\cong N^{\prime}$.
###### Proof
The ‘if’ implication is obvious. We prove the ‘only if’ implication by double
induction on the number $n$ of elements of $S$ and the number $m$ of internal
nodes of $N$.
As in Proposition 5, the cases $n=1$ and $n=2$ are straightforward, because
both $\mathrm{TCTC}_{1}$ and $\mathrm{TCTC}_{2}$ consist of a single network.
On the other hand, the case when $m=1$, for every $n$, is also
straightforward: assuming $S=\\{1,\ldots,n\\}$, the network $N$ is in this
case the phylogenetic tree with Newick string (1,2,…,n);, consisting only of
the root and the leaves, and in particular $L^{s}_{N}(i,j)=(1,1,0)$ for every
$1\leqslant i<j\leqslant n$. If $L^{s}(N)=L^{s}(N^{\prime})$, we have that
$L^{s}_{N^{\prime}}(i,j)=(1,1,0)$ for every $1\leqslant i<j\leqslant n$, and
therefore all leaves in $N^{\prime}$ are tree nodes and sibling of each other
by Lemma 3. Since the root of a hybridization network cannot be elementary,
this says that $N^{\prime}$ is also a phylogenetic tree with Newick string
(1,2,…,n); and hence it is isomorphic to $N$.
Let now $N$ and $N^{\prime}$ two TCTC-networks with $n\geqslant 3$ leaves such
that $L^{s}(N)=L^{s}(N^{\prime})$ and $N$ has $m\geqslant 2$ internal nodes.
Assume as induction hypothesis that the thesis in the theorem is true for
pairs of TCTC-networks $N_{1},N_{1}^{\prime}$ with $n-1$ leaves or with $n$
leaves and such that $N_{1}$ has $m-1$ internal nodes.
By Proposition 3, a reduction $U(i)$, $T(i;j)$ or $H(i;j_{1},\ldots,j_{k})$
can be applied to $N$. Since the application conditions for such a reduction
depend only on the splitted LCSA-path lengths vectors by Lemmas 11.(1), 12.(1)
and 13.(1), and $L^{s}(N)=L^{s}(N^{\prime})$, we conclude that we can apply
the same reduction to $N^{\prime}$.
Now, we apply the same reduction to $N$ and $N^{\prime}$ to obtain new TCTC-
networks $N_{1}$ and $N_{1}^{\prime}$, respectively. If the reduction was of
the form $U(i)$, $N_{1}$ and $N_{1}^{\prime}$ have $n$ leaves and $N_{1}$ has
$m-1$ internal nodes; if the reduction was of the forms $T(i;j)$ or
$H(i;j_{1},\ldots,j_{k})$, $N_{1}$ and $N_{1}^{\prime}$ have $n-1$ leaves. In
all cases, $L^{s}(N_{1})=L^{s}(N_{1}^{\prime})$ by Lemmas 11.(2), 12.(2) and
13.(2), and therefore, by the induction hypothesis, $N_{1}\cong
N_{1}^{\prime}$.
Finally, by Lemma 5, $N$ and $N^{\prime}$ are obtained from $N_{1}$ and
$N_{1}^{\prime}$ by applying the same expansion $\textrm{U}^{-1}$,
$\textrm{T}^{-1}$, or $\textrm{H}^{-1}$, and they are isomorphic.
The vectors of splitted LCSA-path lengths vectors do not separate
hybridization networks much more general than the TCTC, as we following
examples show.
###### Remark 4
The vectors of splitted distances do not separate arbitrary (that, is,
possibly time inconsistent) tree-child phylogenetic networks. Indeed, the non-
isomorphic tree-child binary phylogenetic networks $N$ and $N^{\prime}$
depicted in Fig. 12 have the same $L^{s}$ vectors:
$L^{s}(N)=L^{s}(N^{\prime})=\big{(}(2,1,1),(4,1,1),(3,1,1)\big{)}.$
$1$$2$$3$$N$
$1$$2$$3$$N^{\prime}$
Figure 12: These two tree-child binary phylogenetic networks have the same
splitted LCSA-path lengths vectors.
###### Remark 5
The splitted LCSA-path lengths vectors do not separate tree-sibling time
consistent phylogenetic networks, either. Consider for instance the tree-
sibling time consistent fully resolved phylogenetic networks $N$ and
$N^{\prime}$ depicted in Figure 13. A simple computation shows that they have
the same $L^{s}$ vectors, but they are not isomorphic.
$1$$2$$3$$4$$5$$6$$7$$8$$N$
$1$$2$$7$$8$$5$$6$$3$$4$$N^{\prime}$
Figure 13: These two tree-sibling time consistent binary phylogenetic
networks have the same splitted LCSA-path lengths vectors.
As in the fully resolved case, the injectivity of the mapping
$L^{s}:\mathrm{TCTC}_{n}\to\mathbb{R}^{3n(n-1)/2}$
makes it possible to induce metrics on $\mathrm{TCTC}_{n}$ from metrics on
$\mathbb{R}^{3n(n-1)/2}$. The proof of the following result is similar to that
of Proposition 7.
###### Proposition 8
For every $n\geqslant 1$, let $D$ be any metric on $\mathbb{R}^{3n(n-1)/2}$.
The mapping $d^{s}:\mathrm{TCTC}_{n}\times\mathrm{TCTC}_{n}\to\mathbb{R}$
defined by $d(N_{1},N_{2})=D(L^{s}(N_{1}),L^{s}(N_{2}))$ satisfies the axioms
of metrics up to isomorphisms. ∎
For instance, using as $D$ the Manhattan distance or the Euclidean distance,
we obtain, respectively, the metrics on $\mathrm{TCTC}_{n}$
$\begin{array}[]{rl}\displaystyle
d^{s}_{1}(N_{1},N_{2})&\displaystyle=\sum_{1\leqslant i<j\leqslant
n}\big{(}|\ell_{N_{1}}(i,j)-\ell_{N_{2}}(i,j)|+|\ell_{N_{1}}(j,i)-\ell_{N_{2}}(j,i)|\\\\[-4.30554pt]
&\displaystyle\qquad\qquad\qquad\qquad\qquad\qquad\qquad+|h_{N_{1}}(i,j)-h_{N_{2}}(i,j)|\big{)}\\\\[4.30554pt]
&\displaystyle=\sum_{1\leqslant i\neq j\leqslant
n}\big{(}|\ell_{N_{1}}(i,j)-\ell_{N_{2}}(i,j)|+\frac{1}{2}|h_{N_{1}}(i,j)-h_{N_{2}}(i,j)|\big{)}\\\\[12.91663pt]
\displaystyle d^{s}_{2}(N_{1},N_{2})&\displaystyle=\Big{(}\sum_{1\leqslant
i<j\leqslant
n}\big{(}(\ell_{N_{1}}(i,j)-\ell_{N_{2}}(i,j))^{2}+(\ell_{N_{1}}(j,i)-\ell_{N_{2}}(j,i))^{2}\\\\[-4.30554pt]
&\displaystyle\qquad\qquad\qquad\qquad\qquad\qquad\qquad+(h_{N_{1}}(i,j)-h_{N_{2}}(i,j))^{2}\big{)}\Big{)}^{\frac{1}{2}}\\\\[4.30554pt]
&\displaystyle=\sqrt{\sum_{1\leqslant i\neq j\leqslant
n}\big{(}(\ell_{N_{1}}(i,j)-\ell_{N_{2}}(i,j))^{2}+\frac{1}{2}(h_{N_{1}}(i,j)-h_{N_{2}}(i,j))^{2}\big{)}}\end{array}$
These metrics generalize to TCTC-networks the splitted nodal metrics for
arbitrary phylogenetic trees defined in [7]. and the nodal metric for TCTC
_phylogenetic_ networks defined in [6].
## 6 Conclusions
A classical result of Smolenskii [24] establishes that the vectors of
distances between pairs of leaves separate unrooted phylogenetic trees on a
given set of taxa. This result generalizes easily to fully resolved rooted
phylogenetic trees [7], and it lies at the basis of the classical definitions
of _nodal distances_ for unrooted as well as for fully resolved rooted
phylogenetic trees based on the comparison of these vectors [3, 11, 12, 22,
26, 29]. But these vectors do not separate arbitrary rooted phylogenetic
trees, and therefore they cannot be used to compare the latter in a sound way.
This problem was overcome in [7] by introducing the _splitted path lengths_
matrices and showing that they separate arbitrary rooted phylogenetic trees on
a given set of taxa. It is possible then to define _splitted nodal_ metrics
for arbitrary rooted phylogenetic trees by comparing these matrices.
In this paper we have generalized these results to the class
$\mathrm{TCTC}_{n}$ of tree-child time consistent hybridization networks
(TCTC-networks) with $n$ leaves. For every pair $i,j$ of leaves in a TCTC-
network $N$, we have defined the _LCSA-path length_ $L_{N}(i,j)$ and the
_splitted LCSA-path length_ $L_{N}^{s}(i,j)$ between $i$ and $j$ and we have
proved that the vectors $L(N)=(L_{N}(i,j))_{1\leqslant i<j\leqslant n}$
separate fully resolved networks in $\mathrm{TCTC}_{n}$ and the vectors
$L^{s}(N)=(L_{N}^{s}(i,j))_{1\leqslant i<j\leqslant n}$ separate arbitrary
TCTC-networks.
The vectors $L(N)$ and $L^{s}(N)$ can be computed in low polynomial time by
means of simple algorithms that do not require the use of sophisticated data
structures. Indeed, let $n$ be the number of leaves and $m$ the number of
internal nodes in $N$. As we explained in [5, §V.D], for each internal node
$v$ and for each leaf $i$, it can be decided whether $v$ is a strict or a non-
strict ancestor of $i$, or not an ancestor of it at all, by computing by
breadth-first search the shortest paths from the root to each leaf before and
after removing each of the $m$ nodes in turn, because a non-strict descendant
of a node will still be reachable from the root after removing that node,
while a strict descendant will not. All this information can be computed in
$O(m(n+m))$ time, and once it has been computed the least common semi-strict
ancestor of two leaves can be computed in $O(m)$ time by selecting the node of
least height among those which are ancestors of the two leaves and strict
ancestors of at least one of them. This allows the computation of $L(N)$ and
$L^{s}(N)$ in $O(m^{2}+n^{2}m)$ time.
These vectors $L(N)$ and $L^{s}(N)$ can be used then to define metrics for
fully resolved and arbitrary TCTC-networks, respectively, from metrics for
real-valued vectors. The metrics obtained in this way can be understood as
generalizations to $\mathrm{TCTC}_{n}$ of the (non-splitted or splitted) nodal
metrics for phylogenetic trees and they can be computed in low polynomial time
if the metric used to compare the vectors can be done so: this is the case,
for instance, when this metric is the Manhattan or the Euclidean metric (in
the last case, computing the square root with $O(10^{m+n})$ significant digits
[2], which should be more than enough).
It remains to study the main properties of the metrics defined in this way,
like for instance their diameter or the distribution of their values. It is
important to recall here that these are open problems even for the classical
nodal distances for fully resolved rooted phylogenetic trees.
## Acknowledgment
The research reported in this paper has been partially supported by the
Spanish DGI projects MTM2006-07773 COMGRIO and MTM2006-15038-C02-01.
## References
* [1] M. Baroni, C. Semple, M. Steel, Hybrids in real time, Syst. Biol. 55 (2006) 46–56.
* [2] P. Batra, Newton’s method and the computational complexity of the fundamental theorem of algebra, Electron. Notes Theor. Comput. Sci. 202 (2008) 201–218.
* [3] J. Bluis, D.-G. Shin, Nodal distance algorithm: Calculating a phylogenetic tree comparison metric, in: Proc. 3rd IEEE Symp. BioInformatics and BioEngineering, 2003.
* [4] G. Cardona, M. Llabrés, F. Rosselló, G. Valiente, A distance metric for a class of tree-sibling phylogenetic networks, Bioinformatics 24 (13) (2008) 1481–1488.
* [5] G. Cardona, M. Llabrés, F. Rosselló, G. Valiente, Metrics for phylogenetic networks I: Generalizations of the Robinson-Foulds metric, submitted (2008).
* [6] G. Cardona, M. Llabrés, F. Rosselló, G. Valiente, Metrics for phylogenetic networks II: Nodal and triplets metrics, submitted (2008).
* [7] G. Cardona, M. Llabrés, F. Rosselló, G. Valiente, Nodal metrics for rooted phylogenetic trees, submitted, available at arxiv.org/abs/0806.2035 (2008).
* [8] G. Cardona, F. Rosselló, G. Valiente, Comparison of tree-child phylogenetic networks, IEEE T. Comput. Biol. preprint, 30 June 2008 , doi:10.1109/TCBB.2007.70270.
* [9] G. Cardona, F. Rosselló, G. Valiente, Tripartitions do not always discriminate phylogenetic networks, Math. Biosci. 211 (2) (2008) 356–370.
* [10] W. F. Doolittle, Phylogenetic classification and the universal tree, Science 284 (5423) (1999) 2124–2128.
* [11] J. S. Farris, A successive approximations approach to character weighting, Syst. Zool. 18 (1969) 374–385.
* [12] J. S. Farris, On comparing the shapes of taxonomic trees, Syst. Zool. 22 (1973) 50–54.
* [13] D. Gusfield, S. Eddhu, C. Langley, The fine structure of galls in phylogenetic networks, INFORMS J. Comput, 16 (4) (2004) 459–469.
* [14] D. Gusfield, S. Eddhu, C. Langley, Optimal, efficient reconstruction of phylogenetic networks with constrained recombination, J. Bioinformatics Comput. Biol. 2 (1) (2004) 173–213.
* [15] J. Hein, M. H. Schierup, C. Wiuf, Gene Genealogies, Variation and Evolution: A Primer in Coalescent Theory, Oxford University Press, 2005.
* [16] D. H. Huson, D. Bryant, Application of Phylogenetic Networks in Evolutionary Studies, Mol. Biol. Evol. 23 (2) (2006) 254–267.
* [17] D. H. Huson, T. H. Klöpper, Beyond galled trees - decomposition and computation of galled networks, in: Proceedings RECOMB 2007, vol. 4453 of Lecture Notes in Computer Science, Springer-Verlag, 2007.
* [18] B. M. E. Moret, L. Nakhleh, T. Warnow, C. R. Linder, A. Tholse, A. Padolina, J. Sun, R. Timme, Phylogenetic networks: Modeling, reconstructibility, and accuracy, IEEE T. Comput. Biol. 1 (1) (2004) 13–23.
* [19] L. Nakhleh, J. Sun, T. Warnow, C. R. Linder, B. M. E. Moret, A. Tholse, Towards the development of computational tools for evaluating phylogenetic network reconstruction methods, in: Proc. 8th Pacific Symp. Biocomputing, 2003.
* [20] L. Nakhleh, J. Sun, T. Warnow, C. R. Linder, B. M. E. Moret, A. Tholse, Towards the development of computational tools for evaluating phylogenetic network reconstruction methods, in: Proc. 8th Pacific Symp. Biocomputing, 2003.
* [21] L. Nakhleh, T. Warnow, C. R. Linder, K. S. John, Reconstructing reticulate evolution in species: Theory and practice, J. Comput. Biol. 12 (6) (2005) 796–811.
* [22] J. B. Phipps, Dendrogram topology, Syst. Zool. 20 (1971) 306–308.
* [23] C. Semple, Hybridization networks, in: O. Gascuel, M. Steel (eds.), Reconstructing evolution: New mathematical and computational advances, Oxford University Press, 2008, pp. 277–314.
* [24] Y. A. Smolenskii, A method for the linear recording of graphs, USSR Computational Mathematics and Mathematical Physics 2 (1963) 396–397.
* [25] Y. S. Song, J. Hein, Constructing minimal ancestral recombination graphs, J. Comput. Biol. 12 (2) (2005) 147–169.
* [26] M. A. Steel, D. Penny, Distributions of tree comparison metrics—some new results, Syst. Biol. 42 (2) (1993) 126–141.
* [27] G. Valiente, Phylogenetic networks, course at the Int. Summer School on Bioinformatics and Computational Biology Lipari (June 14–21, 2008).
* [28] L. Wang, K. Zhang, L. Zhang, Perfect phylogenetic networks with recombination, J. Comput. Biol. 8 (1) (2001) 69–78.
* [29] W. T. Williams, H. T. Clifford, On the comparison of two classifications of the same set of elements, Taxon 20 (4) (1971) 519–522.
* [30] S. J. Willson, Restrictions on meaningful phylogenetic networks, contributed talk at the EMBO Workshop on Current Challenges and Problems in Phylogenetics (Isaac Newton Institute for Mathematical Sciences, Cambridge, UK, 3–7 September 2007).
* [31] S. J. Willson, Reconstruction of certain phylogenetic networks from the genomes at their leaves, J. Theor. Biol. 252 (2008) 338–349.
* [32] S. M. Woolley, D. Posada, K. A. Crandall, A comparison of phylogenetic network methods using computer simulation, Plos ONE 3 (4) (2008) e1913.
| arxiv-papers | 2008-07-01T09:13:32 | 2024-09-04T02:48:56.493761 | {
"license": "Public Domain",
"authors": "Gabriel Cardona, Merce Llabres, Francesc Rossello, Gabriel Valiente",
"submitter": "Francesc Rossell\\'o",
"url": "https://arxiv.org/abs/0807.0087"
} |
0807.0149 | # Charged Lepton-Flavor Violation in Beyond-Standard Models
Junji Hisano Institute for Cosmic Ray Research (ICRR), University of Tokyo,
Kashiwa, Chiba 277-8582, Japan
###### Abstract
I discuss charged lepton-flavor violation in physics beyond the standard model
and review topics related to it.
## I Introduction
It is found from discovery of the neutrino oscillation that the lepton-flavor
symmetries are not exact in nature, while the charged lepton-flavor violation
(cLFV) has not been observed yet. The cLFV is suppressed even if the tiny
neutrino masses observed in the experiments are introduced in the standard
model (SM). The cLFV processes are proportional to the fourth power of the
neutrino masses due to the GIM mechanism, similar to the quark sector. The
predicted branching ratio of $\mu\rightarrow e\gamma$ is smaller than
$10^{-54}$. However, the lepton-flavor symmetries are accidental in the SM,
and it is a big mystery in particle physics why the cLFV processes are not
discovered, when considering beyond-standard models (BSMs) Raidal:2008jk .
Searches for cLFV processes have long history. Search for $\mu\rightarrow
e\gamma$ was performed soon after muons were discovered, and it was found that
muons are not an excited state of electrons. On 60’s, the two-neutrino
hypothesis, in which the lepton flavor conservations were introduced, was
proposed to suppress $\mu\rightarrow e\gamma$. The tau-lepton flavor
conservation has been also tested after tau lepton was discovered on 75’.
However, on 98’, the neutrino oscillation was discovered at SuperKamiokande
experiments.
Now the bounds on the cLFV processes are significantly improved by efforts of
experimentalists. The current experimental bounds on the representative cLFV
processes of muon are given as follows,
$\displaystyle{\rm Br}(\mu\rightarrow e\gamma)$ $\displaystyle<$
$\displaystyle 1.2\times 10^{-11}~{}\mbox{\cite[cite]{\@@bibref{Authors
Phrase1YearPhrase2}{Brooks:1999pu}{\@@citephrase{(}}{\@@citephrase{)}}}},$ (1)
$\displaystyle{\rm Br}(\mu\rightarrow eee)$ $\displaystyle<$ $\displaystyle
1.0\times 10^{-12}~{}\mbox{\cite[cite]{\@@bibref{Authors
Phrase1YearPhrase2}{Bellgardt:1987du}{\@@citephrase{(}}{\@@citephrase{)}}}},$
(2) $\displaystyle{\rm R}(\mu\rightarrow e;{\rm Ti})$ $\displaystyle<$
$\displaystyle 6.1\times 10^{-13}~{}\mbox{\cite[cite]{\@@bibref{Authors
Phrase1YearPhrase2}{Dohmen:1993mp}{\@@citephrase{(}}{\@@citephrase{)}}}},$ (3)
$\displaystyle{\rm R}(\mu\rightarrow e;{\rm Au})$ $\displaystyle<$
$\displaystyle 7.0\times 10^{-13}~{}\mbox{\cite[cite]{\@@bibref{Authors
Phrase1YearPhrase2}{Bertl:2006up}{\@@citephrase{(}}{\@@citephrase{)}}}}.$ (4)
$\mu$-$e$ conversion rates in nuclei, given in Eqs. 3 and 4, are normalized by
the muon capture rates. Those for tau-lepton’s processes come from the Belle
and Babar experiments, and they are111 The bounds on Eqs. 5 and 6 are combined
results of Belle Abe:2006sf and Babar Aubert:2005ye ; Aubert:2005wa .
$\displaystyle{\rm Br}(\tau\rightarrow\mu\gamma)$ $\displaystyle<$
$\displaystyle 1.6\times 10^{-8}~{}\mbox{\cite[cite]{\@@bibref{Authors
Phrase1YearPhrase2}{Banerjee:2007rj}{\@@citephrase{(}}{\@@citephrase{)}}}},$
(5) $\displaystyle{\rm Br}(\tau\rightarrow e\gamma)$ $\displaystyle<$
$\displaystyle 9.4\times 10^{-8}~{}\mbox{\cite[cite]{\@@bibref{Authors
Phrase1YearPhrase2}{Banerjee:2007rj}{\@@citephrase{(}}{\@@citephrase{)}}}},$
(6) $\displaystyle{\rm Br}(\tau\rightarrow 3l)$ $\displaystyle<$
$\displaystyle\sim 10^{-8}~{}\mbox{\cite[cite]{\@@bibref{Authors
Phrase1YearPhrase2}{Banerjee:2007rj}{\@@citephrase{(}}{\@@citephrase{)}}}},$
(7) $\displaystyle{\rm Br}(\tau\rightarrow l+{\rm hadron(s)})$
$\displaystyle<$ $\displaystyle\sim
10^{-8}~{}\mbox{\cite[cite]{\@@bibref{Authors
Phrase1YearPhrase2}{Banerjee:2007rj}{\@@citephrase{(}}{\@@citephrase{)}}}}.$
(8)
A new experiment for $\mu\rightarrow e\gamma$ search, MEG, has been started at
PSI this year meg . It is argued tmori that a successful physics run for 5-6
months reaches to $(7$-$8)\times 10^{-13}$, which is improvement of more than
10 compared with the current bound derived by the MEGA experiment on 98’
Brooks:1999pu , and that a goal of the first phase of the experiment is
$(1$-$2)\times 10^{-13}$ after another two years. Further improvement down to
$\sim 10^{-14}$ in the second phase may be possible after some upgrades.
The signal in $\mu$-$e$ conversion experiments is monochromatic electron with
energy $m_{\mu}-E_{\rm bound}$ ($E_{\rm bound}$ bounding energy), and then the
backgrounds (BGs) are quite suppressed. While the conversion rates are
suppressed by $10^{-(2-3)}$ compared with $Br(\mu\rightarrow e\gamma)$ in
typical BSMs such as supersymmetric models, the high sensitivities to the BSMs
are expected. Two experiments for $\mu$-$e$ conversion searches with
sensitivities $10^{-(16-17)}$, Mu2e in Fermilab mu2e and COMET in J-parc
comet , are planed. Furthermore, the PRISM/PRIME experiment, in which very
intensive pulsed beam is produced by the FFAG muon storage ring, is also
planed. It has ultimate sensitivity as $10^{-(18-19)}$ comet .
The searches for cLFV in the tau decay will be continued by the Belle
experiment; however, further improvement requires higher luminosity ${\cal
L}$. Now, two experiments, the Super KEKB Akeroyd:2004mj and the Super flavor
factory Giorgi:2006qj , are planed. The processes $\tau\rightarrow\mu\gamma$
and $\tau\rightarrow e\gamma$ already suffer from irreducible BGs from
$e^{-}e^{+}\rightarrow\tau^{+}\tau^{-}\gamma$, Improvements of the
sensitivities are scaled as $\sqrt{\cal L}$ without reduction of the BGs, and
the reaches are argued $10^{-(8-9)}$. Other cLFV processes in the tau decay
are almost BG-free, and the reaches are argued $10^{-(9-10)}$.
## II CLFV in BSMs
Following are effective operators with cLFV of muon up to $D=6$,
$\displaystyle{\cal
L}=\frac{m_{\mu}}{\Lambda^{2}}\bar{e}\sigma^{\mu\nu}F_{\mu\nu}\mu+\frac{1}{\Lambda_{F}^{2}}\bar{e}\mu\bar{e}e+\frac{1}{\Lambda_{F}^{2}}\bar{e}\mu\bar{q}q.$
(9)
The first term is for $\mu\rightarrow e\gamma$, and the second and third ones
are for $\mu\rightarrow 3e$ and $\mu$-$e$ conversion in nuclei, respectively.
Here, other LFV terms with different tensor structures are omitted for
simplicity. Roughly speaking, the branching ratios or conversion rates are
$\sim(m_{W}/\Lambda)^{4}$ or $(m_{W}/\Lambda_{F})^{4}$. Thus, $\Lambda$ and
$\Lambda_{F}$ should be larger than $\sim 10^{(5-6)}$ GeV from the
experimental bounds. This implies that the cLFV searches have quite
sensitivities to the BSMs.
We have several motivations to consider the BSMs at TeV scale. The naturalness
argument for the Higgs boson mass terms is one of them. The dark matter in the
universe may be related to physics at TeV scale. In addition, some people
consider origin of the neutrino masses at TeV scale.
Why does nature hide clues of the BSMs from FCNC processes including cLFV
processes? Some of BSMs, such as the two- or multi-Higgs doublet models and
the left-right symmetric models, introduce new Higgs and/or gauge bosons at
TeV scale. Some models, such as supersymmetric models, extra dimension models,
and the little Higgs model with T parity, introduce partners of leptons and
quarks. Those new fields introduce new sources of FCNCs. While FCNCs in the SM
are suppressed by small quark/lepton masses or small mixing angle due to the
GIM mechanism, the suppression is not necessarily automatic in the BSMs.
First, let us consider the cLFV processes in the supersymmetric standard model
(SUSY SM), since it is the leading candidate among the BSMs and also a
prototype of the BSMs. Supersymmetry is a symmetry between bosons and
fermions, and superpartners for each particles in the SM are introduced to the
SUSY SM.
Supersymmetry is not exact in nature. We need to introduce SUSY-breaking mass
terms for SUSY particles, which have not yet been discovered. The SUSY-
breaking terms are new sources of the flavor violation, since the squark and
slepton mass matrices are not necessarily simultaneously diagonalized with
those of quarks and leptons.222 The R parity is assumed here. When it is
violating, new LFV sources are introduced in the SUSY SM. See Ref.
deGouvea:2000cf for the detail discussion. This leads to so-called the SUSY
flavor problem. We will discuss this problem from a viewpoint of cLFV in the
following.
In the SUSY SM $\mu\rightarrow e\gamma$ is generated by one-loop diagrams, and
the branching ratio is approximately given as
$\displaystyle{\rm Br}(\mu\rightarrow
e\gamma)\sim\frac{\alpha}{4\pi}\left(\frac{m_{W}}{m_{SUSY}}\right)^{4}\sin^{2}\theta_{\tilde{e}\tilde{\mu}}\left(\frac{\Delta
m_{\tilde{l}}^{2}}{m^{2}_{SUSY}}\right)^{2},$
where $m_{SUSY}$ the SUSY breaking scale, $\sin\theta_{\tilde{e}\tilde{\mu}}$
the slepton mixing angle, ${\Delta m_{\tilde{l}}^{2}}$ is the slepton mass
square difference.
While the branching ratio of $\mu\rightarrow e\gamma$ is suppressed by one-
loop factor, we need much more suppression in order to make it below the
experimental bound. Three directions are proposed. First is the universal
scalar mass hypothesis, in which the squarks and sleptons with common quantum
numbers are degenerate in mass (${\Delta m_{\tilde{l}}^{2}}\ll m_{SUSY}^{2}$).
Many models are constructed following this direction; the gravity mediation
Nilles:1983ge , the gauge mediation gaugemed , the gaugino mediation ginom ,
and the anomaly mediation ams . Second is the alignment hypothesis
($\sin\theta_{\tilde{e}\tilde{\mu}}\ll 1$). It is assumed that squark and
slepton mass matrices can be diagonalized in the same basis as those of quarks
and leptons due to some flavor symmetries or some mechanism Nir:1993mx . Third
is the decoupling hypothesis ($m_{SUSY}\gg m_{W}$). Squarks and sleptons in
the first and second generations are so heavy ($O(10^{4-5})$ GeV) that the
flavor violation in the first and second generations are suppressed effsusy .
In these three hypothesis, the cLFV processes are suppressed. However, the
improvements of the experimental sensitivities may probe the origin of the
SUSY breaking terms and also physics beyond the SUSY SM.
In the universal scalar mass hypothesis, if some physics has LFV interactions
below the SUSY-breaking mediation scale, the LFV slepton mass terms are
induced radiatively by the renormalization-group effect Hall:1985dx . In this
case the LFV mass terms are not suppressed by powers of the energy scale for
the LFV interactions. The seesaw mechanism and the GUTs are nowadays ones of
attractive models from the phenomenological and theoretical points of view. In
these models, LFV Yukawa interactions are introduced. Thus, if the SUSY
breaking mediation scale is higher than the GUT Barbieri:1994pv or the right-
handed neutrino mass scale Borzumati:1986qx , sizable LFV processes might be
predicted. In Fig. 1, ${\rm Br}(\mu\rightarrow e\gamma)$ in the SUSY seesaw
model is shown. The observed large neutrino mixing angles enhances the cLFV
processes of muon and tau lepton Hisano:1997tc ; Hisano:1998fj . Various
studies are performed under the universal scalar mass hypothesis. See
references in Ref. Raidal:2008jk .
Figure 1: ${\rm Br}(\mu\rightarrow e\gamma)$ in the SUSY seesaw model. This
figure comes from Ref. Hisano:1998fj .
In the decoupling hypothesis, squarks and sleptons in the third generation may
have large flavor violation. In addition, even when SUSY particle masses
larger than $O(1$-$10)$ TeV, the Higgs boson exchange may give sizable
contributions to the cLFV processes Babu:2002et . The SUSY SM has two doublet
Higgs bosons ($H_{1}$ and $H_{2}$), and their effective couplings for leptons
are
$\displaystyle-{\cal
L}_{Y}=\bar{e}_{Ri}Y_{li}L_{i}H_{1}+\bar{e}_{Ri}\Delta_{ij}L_{j}H_{2}^{\dagger}+h.c.,$
(11)
where $Y_{li}$ is the tree-level Yukawa coupling and $\Delta_{ij}$ is non-
holomorphic correction generated at one-loop level. $\Delta_{ij}$ is not
suppressed by the SUSY breaking scale, in addition to enhancement of
$\tan\beta$. Then, when the SUSY particles are heavy enough, the Higgs boson
exchange dominates in the cLFV processes. Fig. 2 shows the branching ratios
for cLFV processes of muon, induced by the Higgs boson exchange in the SUSY
SM. This figure comes from Ref. Paradisi:2006jp . $\mu\rightarrow e\gamma$ is
induced by the Barr-Zee type loop diagrams, while $\mu\rightarrow 3e$ and
$\mu\rightarrow e$ conversion in nuclei are Higgs-boson exchange processes
effectively at tree-level. The Higgs boson exchange contributions to the cLFV
processes of tau lepton are also discussed in Ref. higgstau .
Figure 2: Branching ratios for cLFV processes of muon induced by the Higgs
boson exchange in the SUSY SM. This figure comes from Ref. Paradisi:2006jp .
If the alignment between lepton and slepton masses is not complete in the
third hypothesis, the cLFV processes are predicted. Let us show an example
with $U(1)\times U(1)$ flavor symmetries, which is given in Ref. Feng:2007ke .
The charge assignments of the right- and left-handed leptons are
$\displaystyle\bar{E}_{1}(1,0),~{}\bar{E}_{2}(1,-2),~{}\bar{E}_{3}(0,-3),$
$\displaystyle L_{1}(4,0),~{}L_{2}(2,2),~{}L_{3}(0,4),$
and the flavor symmetries are broken by VEVs of flavon fields,
$\phi_{1}(-1,0)$ and $\phi_{2}(0,-1)$. In this case, the charged lepton mass
matrix is
$\displaystyle(m_{l})\sim\lambda\left(\begin{array}[]{ccc}\lambda^{4}&0&0\\\
\lambda^{4}&\lambda^{2}&0\\\ \lambda^{4}&\lambda^{2}&1\end{array}\right),$
(15)
and the hierarchical structure of lepton masses is explained with $\lambda\sim
0.1-0.2$. In this setup, the slepton mass matrices are aligned, and the left-
handed and right-handed slepton mixings are suppressed as
$\theta_{\tilde{\mu}_{L}\tilde{e}_{L}}\sim\lambda^{4}$ and
$\theta_{\tilde{\mu}_{R}\tilde{e}_{R}}\sim\lambda^{2}$. However, these mixings
are marginal to the bounds from $\mu\rightarrow e\gamma$.
The SUSY flavor problem is one of the guidelines to construct realistic SUSY
SMs. As shown above, in the proposed ideas and models to suppress the FCNC
processes, the cLFV processes are not necessarily suppressed, and the on-going
and planed experiments cover the predictions.
The cLFV processes also have good sensitivities to non-SUSY models at TeV
scale as expected. The BSMs at TeV scale are severely constrained from the
cLFV processes, unless some mechanism works to suppress them. In the
following, some concrete examples are reviewed.
First is the little Higgs model ArkaniHamed:2001ca with T parity Cheng:2003ju
. In the little Higgs model, the Higgs boson is a pseudo Nambu-Goldstone boson
for symmetry breaking, and quadratic divergence in the radiative correction to
the Higgs boson mass term is cancelled by heavy particles at one-loop level.
The T parity is introduced for the heavy particles not to contribute to the
electroweak observables at tree level. This extension has a bonus. The
lightest T-odd particle is stable and a candidate of the dark matter in the
universe.
In the little Higgs model with T parity, SU(2)L doublet mirror leptons with T
parity odd are introduced. The mirror leptons have coupling with leptons and
the heavy gauge bosons, which is lepton-flavor violating. The cLFV processes
are generated at one-loop level, similar to the SUSY SM. In Fig. 3, ${\rm
Br}(\mu\rightarrow e\gamma)$ and ${\rm Br}(\mu\rightarrow 3e)$ are shown in
this model. It is shown in Ref. Blanke:2007db that the accidental
cancellation reduces ${\rm Br}(\mu\rightarrow e\gamma)$ and two processes have
comparable branching ratios. Then, this model still is viable even the mirror
leptons are around 1 TeV.
Figure 3: ${\rm Br}(\mu\rightarrow e\gamma)$ and ${\rm Br}(\mu\rightarrow 3e)$
in the little Higgs model with T parity. Here, the mirror leptons masses are
taken from 300 GeV and 1.5 TeV. This figure comes from Ref. Blanke:2007db .
Next is the SM on the Randall-Sundrum (RS) background. The RS geometry is
known as a solution of the hierarchy problem Randall:1999ee . In addition,
when the SM fermions and gauge bosons propagate in the full five-dimensional
space, the fermion mass hierarchy is also explained even from the “anarchic”
structure Gherghetta:2000qt . In this model the Kaluza-Klein (KK) particles
have the LFV interaction. In Fig. 4 ${\rm Br}(\mu\rightarrow e\gamma)$ and
${\rm R}(\mu\rightarrow e;{\rm Ti})$ are shown for the KK scale 10 TeV. The
current experimental bounds give constraints on this model. While
$\mu\rightarrow e\gamma$ is a one-loop process, the $\mu\rightarrow e$
conversion is generated at tree level.
Figure 4: ${\rm Br}(\mu\rightarrow e\gamma)$ and ${\rm R}(\mu\rightarrow
e;{\rm Ti})$ in the SM on the Randall-Sundrum background. Here, the Kaluza-
Klein scale is 10 TeV. This figure comes from Ref. Agashe:2006iy .
Finally, we discuss the relation of the cLFV processes and the neutrino
masses. The seesaw mechanism is the most promising in the candidates of the
neutrino masses. In many models the mechanism is realized at much higher
energy scale than the TeV scale. In such models we cannot expect some direct
relations between the cLFV processes and the origin of the neutrino masses.
Exceptions are the SUSY seesaw models susyseesaw1 ; susyseesaw2 ; however, the
relations are indirect. On the other hand, there are attempts to construct
models of the neutrino mass origin at TeV scale.
One of the models is the triplet Higgs model (type-II seesaw model)
Schechter:1980gr . In this model, the triplet Higgs boson with mass at TeV
scale has sizable lepton-flavor violating coupling, and the coupling is
directly linked to the neutrino mass matrix. In Fig. 5 the branching ratios
for the cLFV processes of muon in this model are shown. The pattern depends on
the neutrino mass structure Kakizaki:2003jk .
Figure 5: Branching ratios for cLFV processes of muon in the triplet Higgs
model. Here, neutrino mass spectrum is inverted-hierarchical (hierarchical) in
the upper (lower) figure. These figures come from Ref. Kakizaki:2003jk
## III Correlations
When the cLFV processes are discovered, it would be important to take
correlations among various processes in order to unveil the origin of the
lepton-flavor violation and the realistic BSM.
The first correlation is among the cLFV processes with common flavor
transition. The correlation discriminates BSMs at TeV scale since the pattern
of correlation depends on models. In Table 1, the ratios of the cLFV rates of
muon are shown in three models; the SUSY SM in which the dipole moment
contribution to the cLFV processes is dominant, the SUSY SM in which the SUSY
particles are so heavy that the Higgs boson exchange dominates, and the little
Higgs model with T parity. While the cLFV processes are radiatively generated
in those models, the pattern of the ratio of the cLFV rates are quite
different. Correlations among the cLFV processes of $\tau$ and among the
$\mu$-$e$ conversions in various nuclei Kitano:2002mt are also useful to
discriminate models.
Table 1: Ratios of the cLFV rates of muon in three models. This table comes from Blanke:2007db . | $\frac{{\rm Br}(\mu\rightarrow 3e)}{{\rm Br}(\mu\rightarrow e\gamma)}$ | $\frac{{\rm R}(\mu\rightarrow e;{\rm Ti})}{{\rm Br}(\mu\rightarrow e\gamma)}$
---|---|---
SUSY SM (dipole) | $\sim 6\times 10^{-3}$ | $\sim 5\times 10^{-3}$
SUSY SM (Higgs) | $\sim 6\times 10^{-3}$ | $0.08$-$0.15$
Little Higgs with T parity | 0.4-2.5 | $10^{-2}$-$10^{2}$
The second correlation is among the cLFV processes with different flavor
transitions. In some models the cLFV processes are related to the neutrino
masses. In the triplet Higgs model, the interaction of the triplet Higgs boson
with leptons is directly linked to the neutrino mass matrix $(m_{\nu})$, as
mentioned above. The model is one of realizations of the minimal flavor
violation hypothesis in the lepton sector Cirigliano:2005ck . As the result,
the ratios of the cLFV rates of tau lepton and muon are given as
$\displaystyle\frac{{\rm Br}(\tau\rightarrow\mu\gamma)}{{\rm
Br}(\mu\rightarrow e\gamma)}\simeq
0.17\times\left(\frac{(m_{\nu}m_{\nu}^{\dagger})_{\tau\mu}}{(m_{\nu}m_{\nu}^{\dagger})_{\tau
e}}\right)^{2}\sim 300$ $\displaystyle\frac{{\rm Br}(\tau\rightarrow
e\gamma)}{{\rm Br}(\mu\rightarrow e\gamma)}\simeq
0.17\times\left(\frac{(m_{\nu}m_{\nu}^{\dagger})_{\tau
e}}{(m_{\nu}m_{\nu}^{\dagger})_{\tau e}}\right)^{2}\sim 0.2.$ (16)
In the SUSY seesaw models, the cLFV processes are indirectly related to the
origin of the neutrino masses even if the energy scale is high. Under the
universal scalar mass hypothesis, the cLFV processes give information of the
type-I SUSY seesaw model, which is independent of the neutrino mass from a
viewpoint of the reconstruction of the model susyseesaw2 . In the type-II SUSY
seesaw model, the relation in Eq. 16 is predicted again susyseesaw2 .
The third correlation is between the hadronic and leptonic FCNC processes. In
GUTs, quarks and leptons are unified so that the cLFV processes are correlated
with hadronic FCNC processes Hisano:2003bd . Let us consider the SUSY SU(5)
GUT with right-handed neutrinos. In this model, lepton doublets and right-
handed down-type quarks are embedded in common SU(5) multiplets. Then,
neutrino Yukawa coupling affects both the right-handed sdown and left-handed
slepton mass matrices Moroi:2000mr .
In Fig. 6 we show the correlation between CP phase of the $B_{s}$ mixing
amplitude, $\phi_{B_{s}}$, and ${\rm Br}(\tau\rightarrow\mu\gamma)$ in the
SUSY SU(5) GUT with right-handed neutrinos. The CP violation in the $B_{s}$
mixing is suppressed in the SM, and then it is also sensitive to the BSMs. The
phase $\phi_{B_{s}}$ is defined to be zero in the SM. It is recently announced
by the Utfit collaboration Bona:2008jn that $\phi_{B_{s}}$ deviates more than
$3\sigma$ from the SM prediction. In the figure, we show the region for
$\phi_{B_{s}}$. The deviation of $\phi_{B_{s}}$ is constrained from null
result of $\tau\rightarrow\mu\gamma$ search in this model, and the $95\%$
probability region derived by the Utfit collaboration is marginal in this
model. When the deviation of $\phi_{B_{s}}$ is established, search for
$\tau\rightarrow\mu\gamma$ would be an important test of the SUSY SU(5) GUTs.
Figure 6: Correlation between CP phase of the $B_{s}$ mixing amplitude,
$\phi_{B_{s}}$, and ${\rm Br}(\tau\rightarrow\mu\gamma)$ in the SUSY SU(5) GUT
with right-handed neutrinos. Here, $\tan\beta=10$ and 30, and we show the
region for $\phi_{B_{s}}$, derived by the Utfit collaboration. These figures
come from Ref. Hisano:2008df
.
When the flavor-violating mass terms for the right-handed squarks are non-
vanishing, the hadronic EDMs are generated at one-loop Dimopoulos:1994gj and
two-loop levels Hisano:2007cz . The non-zero $(m_{\tilde{d}_{R}}^{2})_{23}$
generates the strange-quark chromoelectric dipole moment, which contributes to
the hadronic EDMs Hisano:2003iw . Thus, the correlation between the hadronic
EDMs and the cLFV processes would be the tests of the SUSY GUTs. But, this
program still has difficulties in precisions of the hadronic EDM evaluation
Narison:2008jp , and the further improvements would be required for it.
## IV Summary
In this talk, I discussed charged lepton-flavor violation in physics beyond
the standard model and reviewed topics related to it. The cLFV processes are
accidentally suppressed by the GIM mechanism in the SM even after tiny
neutrino masses are introduced. On the other hand, the suppression is not
necessarily automatic in physics beyond the SM. Studies of cLFVs probe BSMs,
hidden flavor symmetries, and underlying flavor structures.
Current bounds on cLFVs give constraints on physics even around
O($10^{(5-6)}$) GeV. In practical BSMs, cLFVs are suppressed by loop-factors
or small flavor-mixing, or accidental cancellation. Thus, coming MEG
experiment, and planed experiments, Mu2e, COMET and PRISM/PRIME, will cover
interesting regions in various BSMs. Since the cLFVs are pieces of puzzles in
the BSMs, it is important to stress that taking various correlations are
useful to solve the puzzles.
###### Acknowledgements.
The work of JH was also supported in part by the Grant-in-Aid for Science
Research, Japan Society for the Promotion of Science (No. 20244037 and No.
2054252).
## References
* (1) M. Raidal et al., arXiv:0801.1826 [hep-ph].
* (2) M. L. Brooks et al. [MEGA Collaboration], Phys. Rev. Lett. 83 (1999) 1521 [arXiv:hep-ex/9905013].
* (3) U. Bellgardt et al. [SINDRUM Collaboration], Nucl. Phys. B 299 (1988) 1.
* (4) C. Dohmen et al. [SINDRUM II Collaboration.], Phys. Lett. B 317 (1993) 631.
* (5) W. Bertl et al. [SINDRUM II Collaboration], Eur. Phys. J. C 47 (2006) 337.
* (6) S. Banerjee, Nucl. Phys. Proc. Suppl. 169 (2007) 199 [arXiv:hep-ex/0702017].
* (7) K. Abe et al. [Belle Collaboration], arXiv:hep-ex/0609049.
* (8) B. Aubert et al. [BABAR Collaboration], Phys. Rev. Lett. 95 (2005) 041802 [arXiv:hep-ex/0502032].
* (9) B. Aubert et al. [BABAR Collaboration], Phys. Rev. Lett. 96 (2006) 041801 [arXiv:hep-ex/0508012].
* (10) Homepage of MEG experiment, http://meg.web.psi.ch/.
* (11) T. Mori, talk given in the 4th International Workshop on Nuclear and Particle Physics at J-PARC (NP08), Mito, Ibaraki, Japan, March, 2008 (http://j-parc.jp/NP08/).
* (12) R. Bernstein, talk given in the 4th International Workshop on Nuclear and Particle Physics at J-PARC (NP08), Mito, Ibaraki, Japan, March, 2008 (http://j-parc.jp/NP08/).
* (13) A. Sato, talk given in the 4th International Workshop on Nuclear and Particle Physics at J-PARC (NP08), Mito, Ibaraki, Japan, March, 2008 (http://j-parc.jp/NP08/).
* (14) A. G. Akeroyd et al. [SuperKEKB Physics Working Group], arXiv:hep-ex/0406071.
* (15) M. A. Giorgi et al. [SuperB group], INFN Roadmap Report, March 2006.
* (16) A. de Gouvea, S. Lola and K. Tobe, Phys. Rev. D 63 (2001) 035004 [arXiv:hep-ph/0008085].
* (17) As a review, H. P. Nilles, Phys. Rept. 110 (1984) 1.
* (18) M. Dine, A. E. Nelson and Y. Shirman, Phys. Rev. D 51, 1362 (1995); M. Dine, A. E. Nelson, Y. Nir and Y. Shirman, Phys. Rev. D 53, 2658 (1996).
* (19) D. E. Kaplan, G. D. Kribs and M. Schmaltz, Phys. Rev. D 62 (2000) 035010; Z. Chacko, M. A. Luty, A. E. Nelson and E. Ponton, JHEP 0001 (2000) 003; M. Schmaltz and W. Skiba, Phys. Rev. D 62 (2000) 095005.
* (20) L. J. Randall and R. Sundrum, Nucl. Phys. B 557 (1999) 79; G. F. Giudice, M. A. Luty, H. Murayama and R. Rattazzi, JHEP 9812 (1998) 027.
* (21) Y. Nir and N. Seiberg, Phys. Lett. B 309 (1993) 337.
* (22) M. Dine, A. Kagan and S. Samuel, Phys. Lett. B 243, (1990) 250; S. Dimopoulos and G.F. Giudice, Phys. Lett. B 357 (1995) 573; A. Pomarol and D. Tommasini, Nucl. Phys. B 466 (1996) 3; A.G. Cohen, D.B. Kaplan and A.E. Nelson, Phys. Lett. B 388 (1996) 588 ; J. Hisano, K. Kurosawa and Y. Nomura, Phys. Lett. B 445 (1999) 316; Nucl. Phys. B 584 (2000) 3.
* (23) L. J. Hall, V. A. Kostelecky and S. Raby, Nucl. Phys. B 267, 415 (1986).
* (24) R. Barbieri and L. J. Hall, Phys. Lett. B 338 (1994) 212; R. Barbieri, L. J. Hall and A. Strumia, Nucl. Phys. B 445 (1995) 219; J. Hisano, T. Moroi, K. Tobe and M. Yamaguchi, Phys. Lett. B 391 (1997) 341 [Erratum-ibid. B 397 (1997) 357].
* (25) F. Borzumati and A. Masiero, Phys. Rev. Lett. 57 (1986) 961; J. Hisano, T. Moroi, K. Tobe, M. Yamaguchi and T. Yanagida, Phys. Lett. B 357 (1995) 579; J. Hisano, T. Moroi, K. Tobe and M. Yamaguchi, Phys. Rev. D 53 (1996) 2442.
* (26) J. Hisano and D. Nomura, Phys. Rev. D 59 (1999) 116005 [arXiv:hep-ph/9810479].
* (27) J. Hisano, D. Nomura and T. Yanagida, Phys. Lett. B 437 (1998) 351 [arXiv:hep-ph/9711348].
* (28) K. S. Babu and C. Kolda, Phys. Rev. Lett. 89 (2002) 241802.
* (29) P. Paradisi, JHEP 0608 (2006) 047 [arXiv:hep-ph/0601100].
* (30) M. Sher, Phys. Rev. D 66 (2002) 057301 [arXiv:hep-ph/0207136]; A. Brignole and A. Rossi, Nucl. Phys. B 701 (2004) 3 [arXiv:hep-ph/0404211]; A. Brignole and A. Rossi, Phys. Lett. B 566 (2003) 217 [arXiv:hep-ph/0304081]; P. Paradisi, JHEP 0602 (2006) 050 [arXiv:hep-ph/0508054].
* (31) J. L. Feng, C. G. Lester, Y. Nir and Y. Shadmi, arXiv:0712.0674 [hep-ph].
* (32) N. Arkani-Hamed, A. G. Cohen and H. Georgi, Phys. Rev. Lett. 86 (2001) 4757 [arXiv:hep-th/0104005]; Phys. Lett. B 513 (2001) 232 [arXiv:hep-ph/0105239].
* (33) H. C. Cheng and I. Low, JHEP 0309 (2003) 051 [arXiv:hep-ph/0308199]; JHEP 0408 (2004) 061 [arXiv:hep-ph/0405243].
* (34) M. Blanke, A. J. Buras, B. Duling, A. Poschenrieder and C. Tarantino, JHEP 0705 (2007) 013 [arXiv:hep-ph/0702136].
* (35) L. Randall and R. Sundrum, Phys. Rev. Lett. 83 (1999) 3370 [arXiv:hep-ph/9905221].
* (36) T. Gherghetta and A. Pomarol, Nucl. Phys. B 586 (2000) 141 [arXiv:hep-ph/0003129].
* (37) K. Agashe, A. E. Blechman and F. Petriello, Phys. Rev. D 74 (2006) 053011 [arXiv:hep-ph/0606021].
* (38) J. A. Casas and A. Ibarra, Nucl. Phys. B 618 (2001) 171 [arXiv:hep-ph/0103065]; J. R. Ellis, J. Hisano, M. Raidal and Y. Shimizu, Phys. Rev. D 66 (2002) 115013 [arXiv:hep-ph/0206110].
* (39) A. Rossi, Phys. Rev. D 66 (2002) 075003 [arXiv:hep-ph/0207006].
* (40) J. Schechter and J. W. F. Valle, Phys. Rev. D 22 (1980) 2227.
* (41) M. Kakizaki, Y. Ogura and F. Shima, Phys. Lett. B 566 (2003) 210 [arXiv:hep-ph/0304254].
* (42) R. Kitano, M. Koike and Y. Okada, Phys. Rev. D 66 (2002) 096002 [Erratum-ibid. D 76 (2007) 059902] [arXiv:hep-ph/0203110].
* (43) V. Cirigliano, B. Grinstein, G. Isidori and M. B. Wise, Nucl. Phys. B 728 (2005) 121 [arXiv:hep-ph/0507001].
* (44) J. Hisano and Y. Shimizu, Phys. Lett. B 565 (2003) 183 [arXiv:hep-ph/0303071]; M. Ciuchini, A. Masiero, L. Silvestrini, S. K. Vempati and O. Vives, Phys. Rev. Lett. 92 (2004) 071801 [arXiv:hep-ph/0307191]; M. Ciuchini, A. Masiero, P. Paradisi, L. Silvestrini, S. K. Vempati and O. Vives, Nucl. Phys. B 783 (2007) 112 [arXiv:hep-ph/0702144].
* (45) T. Moroi, JHEP 0003 (2000) 019 [arXiv:hep-ph/0002208].
* (46) M. Bona et al. [UTfit Collaboration], arXiv:0803.0659 [hep-ph].
* (47) J. Hisano and Y. Shimizu, arXiv:0805.3327 [hep-ph].
* (48) S. Dimopoulos and L. J. Hall, Phys. Lett. B 344 (1995) 185 [arXiv:hep-ph/9411273]; R. Barbieri, A. Romanino and A. Strumia, Phys. Lett. B 369 (1996) 283 [arXiv:hep-ph/9511305]; A. Romanino and A. Strumia, Nucl. Phys. B 490 (1997) 3 [arXiv:hep-ph/9610485].
* (49) J. Hisano, M. Nagai and P. Paradisi, Phys. Lett. B 642 (2006) 510 [arXiv:hep-ph/0606322]; arXiv:0712.1285 [hep-ph].
* (50) J. Hisano and Y. Shimizu, Phys. Lett. B 581 (2004) 224 [arXiv:hep-ph/0308255]; Phys. Rev. D 70 (2004) 093001 [arXiv:hep-ph/0406091]; J. Hisano, M. Kakizaki, M. Nagai and Y. Shimizu, Phys. Lett. B 604 (2004) 216 [arXiv:hep-ph/0407169].
* (51) S. Narison, arXiv:0806.2618 [hep-ph].
| arxiv-papers | 2008-07-01T14:55:40 | 2024-09-04T02:48:56.501915 | {
"license": "Public Domain",
"authors": "Junji Hisano",
"submitter": "Junji Hisano",
"url": "https://arxiv.org/abs/0807.0149"
} |
0807.0348 | 11institutetext: International school for advanced studies SISSA, via Beirut
2-4, 34014 Trieste, Italy, Istituto nazionale di fisica nucleare, sezione di
Trieste 22institutetext: The Abdus Salam ICTP, Strada costiera 11, 34014,
Trieste, Italy
# On the role of volatility in the evolution of social networks
Daniele De Martino 11 Matteo Marsili 22
(Received: date / Revised version: date)
###### Abstract
We study how the volatility, node- or link-based, affects the evolution of
social networks in simple models. The model describes the competition between
order – promoted by the efforts of agents to coordinate – and disorder induced
by volatility in the underlying social network. We find that when volatility
affects mostly the decay of links, the model exhibit a sharp transition
between an ordered phase with a dense network and a disordered phase with a
sparse network. When volatility is mostly node-based, instead, only the
symmetric (disordered) phase exists These two regimes are separated by a
second order phase transition of unusual type, characterized by an order
parameter critical exponent $\beta=0^{+}$. We argue that node volatility has
the same effect in a broader class of models, and provide numerical evidence
in this direction.
###### pacs:
89.65.-sSocial and economic systems and 05.70.FhPhase transitions: general
studies and 64.60.aqNetworks
## 1 Introduction
The competition between order and disorder is by no means restricted to
physics. Also economies and societies – as systems of many interacting
individuals – organize themselves in different (macroscopic) states, with
different degrees of order – informally interpreted as coordination on social
norms, compliance with laws or conventions PYoung . Besides all its inherent
complexity, one important element of additional richness is that the relation
between the degree of order in a society and the cohesion of the underlying
social network is not unidirectional as in physics, where the topology of
interactions is fixed 5 . Rather the degree of order in a society influences
in important ways the density and topology of interactions. The interplay
between network’s dynamics and collective behavior is important in many
phenomena, ranging from informal contacts in labour market Topa and peer
effects in promoting (anti-)social behaviors 6 to inter-firm agreement for
R$\&$D 8 . The structure of the networks involved in these phenomena is
dynamically shaped by incentives of agents (nodes), be they individuals or
organizations, who establish bilateral interactions (links) when profitable.
In addition, this interplay typically takes place in a volatile environment.
That is, the favourable circumstances that led at same point to the formation
of a particular link may later on deteriorate, causing the removal or rewiring
of that link. This combination of factors raises a number of interesting
issues in statistical physics, as the collective behavior – of e.g. processes
of ordering 14 , opinion spreading, Voter and reaction diffusion Blasius –
may radically change when they are coupled to the dynamics of the network they
are defined on.
We shall here focus on the stylized mathematical description of this generic
phenomena given in Ref. 14 : Here the feedback between nodes and networks
dynamics, arises from assuming that the formation of a link requires some sort
of similarity or proximity of the two parties. This captures different
situation: For example, in cases where trust is essential in the establishment
of new relationships (e.g. in crime or trade networks), linking may be
facilitated by common acquaintances or by the existence of a chain of
acquaintances joining the two parties. In other cases (e.g. in R$\&$D or
scientific networks) a common language, methodology, or comparable level of
technical competence may be required for the link to be feasible or fruitful
to both parties.
This class of models reveals a generic behavior characterized by a
discontinuous transition from an uncoordinated state characterized by a sparse
network, to a coordinated state on a dense network. As discussed in Ref. 14 ,
this agrees with anecdotical evidence which can be summarized as follows:
(i)_Sharp transitions._ Observation on the spread of social pathologies6 , the
growth of research collaborations, both scientific7 and industrial8 , suggest
that networks can shift from a sparse to connected state in short time spans.
(ii)_Resilience._ Once a transition to a highly connected network has taken
place, this setup can survive even to a reversion to unfavorable conditions,
e.g. the thriving performance of Silicon Valley during the computer industy
crisis of the 1980s9 , or the recent development of open-sourcesoftware,
sustained against large odds, thanks to a dense web of collaboration and
trust10 .
(iii)_Equilibrium coexistence._ Under apparently similar environmental
conditions, social networks can be found both in a dense or sparse state. A
good illustration is provided, e.g. by the dual experience of poor
neighborhoods in large cities, where neither poverty nor other socioeconomic
conditions alone can explain wheter there is a degradation in a ghetto with
rampant social problems6 .
This paper focuses on analyzing the effect of node volatility in these simple
models. Indeed, the effect of volatility is limited to link removal in Ref. 14
, but the turnover of agents (i.e. node removal and arrival) may be an
important factor in many real systems. Our focus here will be mostly on the
statistical phenomenon, than on its interpretation in socio-economic terms.
Indeed we find that the introduction of node volatility brings in a
qualitative change, which can be described as a continuous phase transition
with unusual critical properties. In order to show this, we concentrate on the
simplest model for which a full analytic treatment is possible. In the
concluding section, we argue that this qualitative change is expected in a
wider class of model, and it can have much stronger effects.
## 2 model
Our model is a variant of one in the general class of14 . It reproduces in a
stylized manner the mechanisms of co-evolution in social networks
aforementioned, and shows some common elements of the observed phenomenology
(i.e. sharp transitions, resilience, equilibrium coexistence); in particular,
it also shows how the alternative assumptions of link or node based volatility
have profound effects on the dynamics of network formation. The model
describes $N$ agents sitting on the nodes of a network, each of which is
characterized by a variable $\sigma_{i}$ which represents the social norm
(convention or technological standard) adopted by agent $i$. There are $q$
possible social norms, i.e. $\sigma_{i}\in\\{1,\ldots,q\\}$. In terms of
statistical physics, the model can be thought of as a $q$ state Potts model
defined on a graph of $N$ nodes, that evolves in a coupled fashion to the
dynamics of the system. The rules of the dynamics are the followings:
* •
each node attempts to establish a new link with a randomly chosen node at rate
$\eta/2$. The link is established only if the two nodes have the same color
* •
links are destroyed at rate $1$
* •
all the links of a node are destroyed with rate $\alpha$
* •
the color of a node is updated with rate $\nu$ to the color of any of its
neighbors, unless the node is isolated. In the latter case the nodes takes a
random color.
So, the parameter $\alpha$ interpolates between two kinds of volatility. For
$\alpha=0$ volatility only affects links and for $\alpha\gg 1$ it mostly
affects nodes. With respect to the parameters introduced in Ref. 14 , we
observe that the link decay rate has been set to $\lambda=1$ and that, for
notation convenience, we scaled by a factor $2$ the link creation rate $\eta$.
As observed in Ref. 14 , the color update rule is effective only for isolated
nodes, in the long run, and in that case the color is drawn at random. Since
only links between same type agents are created, after a transient all nodes
are either isolated, or connected to nodes of the same color. Therefore the
particular way in which the neighbor is chosen is immaterial. For example,
both a majority rule (most frequent color) or a voter-type rule (random
neighbor) would give the same dynamics. The model can be generalized to a
probabilistic update rule for the colors introducing a finite temperature $T$
(see 14 ). Results do not change considerably as long as $T$ is small enough,
so we shall confine ourselves to the $T=0$ case.
Figure 1: Our model: links are formed with rate $\eta$ between nodes having
the same color, links are destroyed with rate $1$, all the links of a node
desappear with rate $\alpha$ and colors of isolated nodes are randomly updated
with rate $\nu$.
Ref. 14 has shown that for $\alpha=0$, the system shows an hysteretic
transition in $\eta$ from a symmetric to an asymmetric state. The symmetric
state is characterized by a sparse network, with average degree $\langle
k\rangle<1$, with a symmetric distribution of colors. In the asymmetric state,
instead, a dense network with $\langle k\rangle>1$ arises, along with a
dominant color, which is adopted by agents more frequently than the others. In
this sense, the model shows how order and disorder are intimately related with
the dynamics of the social network in a volatile environment.
In what follows, we solve the model in the stationary state for $N\to\infty$,
for all the values of $\alpha$. We find that the $\alpha=0$ behavior is
generic for all $\alpha<1$, but the transition is softened as $\alpha$
increases. For $\alpha>1$ instead we show that the system is always in the
symmetric phase. Hence, in terms of statistical mechanics, $\alpha=1$ is a
second order critical point separating a phase with spontaneously broken
symmetry from a symmetric phase.
## 3 Theory
If we call $n_{k,\sigma}$ the density of nodes with $k$ links and color
$\sigma=1,\dots,q$ we have the following rate equations:
$\displaystyle\dot{n}_{k,\sigma}$ $\displaystyle=$
$\displaystyle(k+1)n_{k+1,\sigma}-kn_{k,\sigma}-\alpha n_{k,\sigma}+$ (1)
$\displaystyle{}+x_{\sigma}(n_{k-1,\sigma}-n_{k,\sigma})$
$\displaystyle\dot{n}_{0,\sigma}$ $\displaystyle=$
$\displaystyle\alpha(n_{\sigma}-n_{0,\sigma})+n_{1,\sigma}-x_{\sigma}n_{k,\sigma}+$
(2)
$\displaystyle{}+\frac{\nu}{q}\sum_{\sigma^{\prime}=1}^{q}(n_{0,\sigma^{\prime}}-n_{0,\sigma})$
where, for future convenience, we have introduced the dynamical variables
$x_{\sigma}=\eta\sum_{k=0}^{\infty}n_{k,\sigma}.$ (3)
aking the sum over all $k$ of these equations and multiplying by $\eta$ we
find
$\dot{x}_{\sigma}=\frac{\eta\nu}{q}\sum_{\sigma^{\prime}=1}^{q}(n_{0,\sigma^{\prime}}-n_{0,\sigma})$
(4)
which implies that, in the stationary state, each component has the same
fraction $n_{0,\sigma}=n_{0}/q$ of disconnected ($k=0$) nodes. This is
equivalent to a detailed balance condition for the density of the different
components.
It is straightforward to derive an equations for the characteristic functions
$\pi_{\sigma}(s)$ of the degree distribution
$p_{\sigma}(k)=n_{k,\sigma}/\sum_{q}n_{q,\sigma}$ of the component $\sigma$.
In the stationary state this reads:
$(1-s)\frac{d\pi_{\sigma}}{ds}=[\alpha+x_{\sigma}(1-s)]\pi_{\sigma}(s)-\alpha.$
(5)
The stationary solution is found by direct integration:
$\pi_{\sigma}(s)=\alpha\int_{0}^{1}\\!dzz^{\alpha-1}e^{-x_{\sigma}(1-s)(1-z)}$
(6)
It is easy to see that this interpolates between a Poisson distribution,
$\pi_{\sigma}(s)=e^{x_{\sigma}(s-1)/\lambda}$ for $\alpha\to 0$, which
coincides with the result of Ref. 14 , and an exponential distribution
$\pi_{\sigma}(s)=\lambda/[\lambda+x_{\sigma}(s-1)]$ for $\alpha\to\infty$.
Notice also that the average degree in component $\sigma$ is $\langle
k\rangle_{\sigma}=\pi_{\sigma}^{\prime}(1)=x_{\sigma}/(1+\alpha)$. This is
precisely what one expects from balance of link creation and destruction of
links in component $\sigma$.
Observing that $\pi_{\sigma}(0)=\eta\frac{n_{0,\sigma}}{x_{\sigma}}=\frac{\eta
n_{0}}{qx_{\sigma}}$ we find an equation for $x_{\sigma}$ in the stationary
state, which reads
$G_{\alpha}(x_{\sigma})\equiv\alpha
x_{\sigma}\int_{0}^{1}\\!duu^{\alpha-1}e^{x_{\sigma}(u-1)}=\frac{\eta
n_{0}}{q}.$ (7)
Notice that the r.h.s. of Eq. (7) is independent of $\sigma$. The variables
$x_{\sigma}$ are determined by Eq. (7) and the normalization condition, which
takes the form
$\sum_{\sigma=1}^{q}x_{\sigma}=\eta.$ (8)
The properties of the solutions of Eqs. (7,8) depend on the behavior of the
function $G_{\alpha}(x)$, which are discussed in the appendix, and can be
classified in symmetric and asymmetric solutions.
### 3.1 $\alpha>1$: The symmetric solution
For $\alpha>1$ the function $G_{\alpha}(x)$ is a monotone increasing function
(see appendix), Hence Eq. (7) has a single solution and Eq. (8) implies that
$x_{\sigma}=\eta/q$ for all components $\sigma$. Hence Eq. (7) yields the
total fraction of disconnected nodes
$n_{0}=\frac{q}{\eta}G_{\alpha}\left(\eta/q\right)$
as a function of the parameters $q,\alpha$ and $\eta$. We can analyze the
stability of the symmetric solution recalling that $n_{0,\sigma}=\eta
G_{\alpha}(x_{\sigma})$. Then Eq. (4) becomes a dynamical equation for
$x_{\sigma}$
$\dot{x}_{\sigma}=\frac{\nu}{q}\sum_{\sigma^{\prime}=1}^{q}\left[G_{\alpha}(x_{\sigma^{\prime}})-G_{\alpha}(x_{\sigma})\right].$
(9)
Linear stability of the symmetric solution is addressed by setting
$x_{\sigma}=\eta/q+\epsilon_{\sigma}$, with
$\sum_{\sigma}\epsilon_{\sigma}=0$. Then to linear order
$\dot{\epsilon}_{\sigma}=\frac{\nu}{q}G_{\alpha}^{\prime}(\eta/q)\sum_{\sigma^{\prime}=1}^{q}[\epsilon_{\sigma^{\prime}}-\epsilon_{\sigma}]=-\nu
G_{\alpha}^{\prime}(\eta/q)\epsilon_{\sigma}.$ (10)
Hence, as long as $G_{\alpha}(x)$ is an increasing function of $x$, the
symmetric solution is stable. This is always the case for $\alpha>1$, as we
shall see, it fails to hold for $\alpha<1$.
### 3.2 $\alpha<1$: The asymmetric solution
For $\alpha<1$ the symmetric solution still exists. However the function
$G_{\alpha}(x)$ now has a maximum for some $x_{0}(\alpha)$ (see the appendix)
and $G_{\alpha}(x)\to\alpha$ from above as $x\to\infty$. Therefore the
symmetric solution becomes unstable when $\eta>\eta_{+}$ where
$\eta_{+}\equiv qx_{0}(\alpha)$ (11)
because beyond that point $G_{\alpha}^{\prime}(\eta/q)<0$.
The occurrence of a maximum in $G_{\alpha}$ also implies that Eq. (7) admits
solutions with $x_{\sigma}=x_{-}<x_{0}(\alpha)$ for some $\sigma$’s and
$x_{\sigma}=x_{+}>x_{0}(\alpha)$ for the other components. Since $x_{\sigma}$
is related to the density of a component $\sigma$, we shall call a component
dense if $x_{\sigma}=x_{+}$ and diluted if $x_{\sigma}=x_{-}$. As in Ref. 14 ,
all solutions with more than one dense component are unstable. Indeed, by the
same argument used to analyze the stability of the symmetric solution, a
perturbation with $\epsilon_{\sigma}=0$ for all diluted components would grow
as $\dot{\epsilon}_{\sigma}=-\nu G_{\alpha}^{\prime}(x_{+})\epsilon_{\sigma}$
on all dense components. These unstable modes correspond to density
fluctuations across dense components and are clearly absent in the solution
with one only dense component. These are the asymmetric solutions we shall
focus on in what follows. There are $q$ of them, depending on which color is
associated with the dense component. The variables $x_{\pm}$ are determined by
the system of equations
$\displaystyle G_{\alpha}(x_{+})$ $\displaystyle=$ $\displaystyle
G_{\alpha}(x_{-})$ (12) $\displaystyle x_{+}+(q-1)x_{-}$ $\displaystyle=$
$\displaystyle\eta$ (13)
This solution is shown in Fig. 2.
Figure 2: Solutions $x_{\pm}$ as a function of $\eta$ for $q=10$ and
$\alpha=0.4$. The dashed line $x=\eta/q$ separating the two curves is the
symmetric solution.
Actually, of the two asymmetric solutions the one with $x_{+}$ decreasing with
$\eta$ is clearly unphysical as this would have a connected component with an
average degree $\langle k\rangle_{\sigma}=x_{+}/(1+\alpha)$ which decreases
with the rate $\eta$ with which links are formed. As in ref. 14 , it is easy
to see that only solutions with $x_{+}$ increasing in $\eta$ are stable.
Indeed, regarding $\eta$ and $x_{-}$ as functions of $x_{+}$ in Eq. (13) we
find
$\frac{d\eta}{dx_{+}}=1+(q-1)\frac{dx_{-}}{dx_{+}}=\frac{G_{\alpha}^{\prime}(x_{-})+(q-1)G_{\alpha}^{\prime}(x_{+})}{G_{\alpha}^{\prime}(x_{-})}.$
Consider perturbations of the form $x_{\sigma}=x_{+}+\epsilon$ for the dense
component and $x_{\sigma}=x_{-}-\epsilon/(q-1)$ for the others. Then by a
derivation analogous to that leading to Eq. (10), we find
$\dot{\epsilon}=-\frac{\nu}{q}\left[G_{\alpha}^{\prime}(x_{-})+(q-1)G_{\alpha}^{\prime}(x_{+})\right]\epsilon=-\frac{\nu}{q}G_{\alpha}^{\prime}(x_{-})\frac{d\eta}{dx_{+}}\epsilon.$
Given that $G_{\alpha}^{\prime}(x_{-})>0$, this implies that on solutions with
$x_{+}$ decreasing with $\eta$, the perturbation $\epsilon$ grows unboundedly.
The asymmetric solution ceases to exist for $\eta<\eta_{-}$111We note, in
passing, that the condition $d\eta/dx_{+}=0$ provides an equation which allows
to determine $\eta_{-}$.. In the region $\eta\in[\eta_{-},\eta_{+}]$ both the
symmetric and the asymmetric solutions co-exist. The coexistence region, in
the $\alpha,\eta$ plane is reported in Fig. 4.
The practical relevance of the results derived so far is best discussed
introducing an order parameter
$m=\frac{x_{+}-x_{-}}{\eta}$ (14)
which is the difference in the density of the dense and diluted components.
This vanishes in the symmetric phase and is non-zero in the asymmetric one. In
Fig. 3 where we report the behavior of the average degree of the network
$\langle
k\rangle=\sum_{k,\sigma}n_{k,\sigma}k=\frac{\eta}{1+\alpha}\frac{1+(q-1)m^{2}}{q}.$
(15)
Fig. 3 shows that as $\eta$ sweeps through the coexistence region the system
undergoes an hysteresis loop: the degree jumps from low to high values at
$\eta_{+}$ as $\eta$ is increased whereas when $\eta$ decreases from large
values, the network collapses back to the symmetric phase when $\eta_{-}$ is
crossed. In the case $\alpha=0$ 14 , the symmetric phase is characterized by
sparse networks, with a vanishing giant component. This is no more true when
$0<\alpha<1$, specially close to $\eta_{+}$ 222Indeed the condition for the
presence of a giant component is $\langle k(k-1)\rangle_{\sigma}>\langle
k\rangle_{\sigma}$ which, by a straightforward calculation, reads $\eta\geq
q(1+\alpha/2)$. At the critical point $\eta_{+}=qx_{0}(\alpha)$ this reads
$x_{0}(\alpha)\geq 1+\alpha/2$ which holds true for all $\alpha>0$.. Numerical
simulations fully confirm this picture, even though for finite systems the
symmetric (asymmetric) phase is meta-stable close to $\eta_{+}$ ($\eta_{-}$)
and therefore the transition occurs for lower (larger) values of $\eta$.
Figure 3: Mean degree $<k>$ as a function of $\eta/\lambda$ for a system with
$q=10$ colors, for $\alpha=0$ and $0.2$, simulations are for systems of $1000$
nodes. Figure 4: Phase diagram for $q=10$. The symmetric phase extends below
and to the right of the (full) line $\eta_{-}(\alpha)$ whereas above the
(dashed) line $\eta_{+}(\alpha)$ only the asymmetric phase is stable. The
coexistence region, where both phases are stable, is delimited by the two
curves.
### 3.3 The critical region: $\alpha\approx 1$
The behavior of the order parameter $m$ on the critical lines which confine
the coexistence region is shown in Fig. 5. This shows that the transition is
continuous but with a peculiar critical behavior. In order to shed light on
this, the appendix shows that, asymptotically for $\alpha\simeq 1$
$\eta_{+}(\alpha)=qx_{0}(\alpha)\simeq-q\log(1-\alpha)+c\log|\log(1-\alpha)|$
(16)
with $c>0$ a constant. A detailed asymptotic analysis of the limit $\alpha\to
1$ (see appendix) also shows that
$m\sim 1/x_{0}(\alpha)\sim|\log(1-\alpha)|^{-1}.$ (17)
In terms of the usual description of critical phenomena, where
$m\sim|1-\alpha|^{\beta}$, this model is consistent with an exponent
$\beta=0^{+}$. Indeed, the singular behavior of $m$ is very close to that of a
first order phase transition.
Figure 5: Order parameter $m$ on the boundary of the coexistence region
$\eta_{-}$ (full line) and $\eta_{+}$ (dashed line) as a function of $\alpha$
for $q=10$.
## 4 Conclusions
The introduction of node volatility, in the simple model discussed here, makes
the transition from a symmetric (disordered) diluted network to an asymmetric
(ordered) dense network less sharp. Indeed when node volatility dominates,
i.e. when the number of links lost per unit time by node decay outnumber those
lost from link decay ($\alpha>1$), the transition disappears altogether, and
the symmetric (disordered) state prevails. The phenomenology is strongly
reminiscent of that of first order phase transitions (e.g. liquid-gas or
paramagnet-ferromagnet) though the critical behavior is highly non-trivial.
The virtue of the particular model studied is that it allows a detailed
analytic approach which allows one to gain insight on all aspects of its
behavior. This model belongs to a general class of models which embody a
generic feedback mechanism between the nodes and the network they are embedded
in, which can be expressed in the following way: while the network promotes
similarity or proximity between nodes, proximity or similarity enhances link
formation. This feedback allows the system to cope with environmental
volatility, which acts removing links at a constant rate. Interestingly, the
emergence of an “ordered” state plays a key role in this evolutionary
struggle.
We believe the general findings discussed here will extend to the general
class of models of Ref. 14 . In particular, we expect the phase transition to
be blurred by the effect of node volatility and to disappear when the latter
exceed a particular threshold.
Actually, Fig. 6 shows that this is the case even for the model of Ref. 13 .
This is a model where link creation occurs either by long distance search at
rate $\eta$ (as in the model discussed here) or through local search (on
second neighbors) at rate $\xi$. Again links decay at unit rate. We refer the
interested reader to Ref. 13 for further details, for the present discussion
let it suffice to say that the effects of (link) volatility are contrasted by
the creation of a dense network with small-world features (a somewhat similar
model with node volatility has been considered in Ref. 12 ). Fig. 6 shows that
the effects of node volatility are very strong. Indeed, even a very small
$\alpha$ reduces considerably the size of the coexistence region and the value
$\alpha_{c}$ at which the latter disappears is also relatively small.
Figure 6: Mean degree as a function of the rate $\xi$ of formation of links
with neighbours of neighbours, for $N=1000$ ($\lambda=1$). Right: $\eta=0.01$,
Left: $\eta=0.1$.
These results suggest that node volatility is indeed a relevant effect in the
co-evolution of socio-economic networks, as it may affect in dramatic ways the
ability of the system to reach a dense and/or coordinated state.
## Appendix
The function $G_{\alpha}(x)$ can be written as
$G_{\alpha}(x)=\alpha\int_{0}^{x}\\!dz\left(1-\frac{z}{x}\right)^{\alpha-1}e^{-z}$
For $\alpha>1$ we have
$\frac{dG_{\alpha}}{dx}=\frac{\alpha(\alpha-1)}{x^{2}}\int_{0}^{x}\\!dzz\left(1-\frac{z}{x}\right)^{\alpha-1}e^{-z}>0$
Notice also that
$\frac{d}{dx}\frac{G_{\alpha}(x)}{x}=-\alpha\int_{0}^{1}\\!duu\left(1-u\right)^{\alpha-1}e^{-ux}$
i.e. $n_{0}=G_{\alpha}(\eta/q)/(\eta/q)$ in the symmetric solution is a
decreasing function of $\eta/q$. In addition $G_{\alpha}(x)\simeq x$ for $x\ll
1$, i.e. $n_{0}\to 1$.
For $\alpha=1-\epsilon$ we can approximate
$G_{\alpha}(x)\simeq(1-\epsilon)\int_{0}^{x}du\left[1-\epsilon\log\left(1-\frac{u}{x}\right)\right]e^{-u}$
$=(1-\epsilon)(1-e^{-x}+\epsilon(E_{i}(x)-\gamma))$
where $E_{i}(x)$ is the exponential integral function, and for $\epsilon\to 0$
we have $x_{\pm}\to\infty$ and $E_{i}(x)\simeq\frac{e^{x}}{x}$ so
$G_{\alpha}\simeq(1-\epsilon)(1-e^{-x}+\epsilon/x)$
We have at the critical point $x_{+}+(q-1)x_{-}=qx_{0}$ where $x_{0}$ is such
that $G^{\prime}_{\alpha}(x_{0})=0$ so $\epsilon\frac{e^{x_{0}}}{x_{0}^{2}}=1$
and $x_{0}\simeq-\log\epsilon+2\log|\log\epsilon|$ by definition
$m=\frac{x_{0}-x_{-}}{x_{0}}$ and $x_{-}=x_{0}(1-m)$ $x_{+}=x_{0}(1+(q-1)m)$
so from $G_{\alpha}(x_{+})=G_{\alpha}(x_{-})$ we have
$e^{-x_{+}}-\epsilon/x_{+}\simeq e^{-x_{-}}-\epsilon/x_{-}$
and then
$\frac{qmx_{0}}{(1-m)(1+(q-1)m)}=e^{mx_{0}}(1-e^{-qmx_{0}})$
we have $mx_{0}\to c$ where $c$ is given by
$c=e^{c}(1-e^{-qc})/q$
## References
* (1) H. P. Young The Journal of Economic Perspectives, 10, 105-122, (Spring, 1996)
* (2) Topa, G., Rev. Ec. Studies 68, 261 (2001).
* (3) M.Granovetter, Am. J. Sociol. 91, 481 (1985)
* (4) C. Nardini, B. Kozma, A. Barrat, Rev. Lett. 100, 158701 (2008); F. Vazquez, V.M. Eguiluz, M. San Miguel, Phys. Rev. Lett. 100, 108702 (2008).
* (5) T. Gross and B. Blasius, J. R. Soc. Interface 5, 259 (2008).
* (6) J.Crane, Am. J. Sociol. 96, 1226 (1991)
* (7) S.Goyal, M.J.van der Leij, J. L. Moraga-Gonzales, J. Polit. Econ. 114, 403 (2006)
* (8) J.Hagedoorn, Research policy 31, 477 (2002)
* (9) A.Saxenian, _Regional Advantage: culture and competition in Silicon Valley and route 128_ (Harvard university press, Cambridge,MA, 1994)
* (10) Y.Benkler, Yale Law Journal 112 369 (2002)
* (11) J.Davidsen, H.Ebel and S.Bornholdt, Phy.Rev.Lett. 88, 12, 128701 (2002)
* (12) M.Marsili, F.Vega-Redondo and F.Slanina, Proc.Natl.Acad.Sci. U.S.A. 101, 1439 (2004)
* (13) G.Ehrardt, M.Marsili and F.Vega-Redondo, Phy.Rev. E 74, 036106 (2006)
| arxiv-papers | 2008-07-02T13:53:13 | 2024-09-04T02:48:56.510104 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Daniele De Martino and Matteo Marsili",
"submitter": "Daniele De Martino",
"url": "https://arxiv.org/abs/0807.0348"
} |
0807.0421 | # Navier–Stokes equations and forward-backward SDEs on the group of
diffeomorphisms of a torus
Ana Bela Cruzeiro1,2 and Evelina Shamarova2,3
###### Abstract
We establish a connection between the strong solution to the spatially
periodic Navier–Stokes equations and a solution to a system of forward-
backward stochastic differential equations (FBSDEs) on the group of volume-
preserving diffeomorphisms of a flat torus. We construct representations of
the strong solution to the Navier–Stokes equations in terms of diffusion
processes.
1 Dep. de Matemática, IST-UTL.
---
2 Grupo de Física Matemática da Universidade de Lisboa.
3 Centro de Matemática da Universidade do Porto.
E-mail addresses: | abcruz@math.ist.utl.pt (A.B. Cruzeiro)
---|---
| evelinas@fc.up.pt (E. Shamarova)
Keywords: Navier–Stokes equations, forward-backward SDEs, diffeomorphims group
## 1\. Introduction
The classical Navier–Stokes equations read as follows:
$\displaystyle\begin{split}&\frac{\partial}{\partial
t}u(t,x)=-(u,\nabla)u(t,x)+\nu\Delta u(t,x)-\nabla p(t,x),\\\
&\mathrm{div\,}u=0,\\\ &u(0,x)=-u_{0}(x),\end{split}$ (1)
where $u_{0}(x)$ is a divergence-free smooth vector field. We fix a time
interval $[0,T]$, and rewrite equations (1) with respect to the function
$\tilde{u}(t,x)=-u(T-t,x).$
Problem (1) is equivalent to the following:
$\displaystyle\begin{split}&\frac{\partial}{\partial
t}\tilde{u}(t,x)=-(\tilde{u},\nabla)\tilde{u}(t,x)-\nu\Delta\tilde{u}(t,x)-\nabla\tilde{p}(t,x),\\\
&\mathrm{div\,}\tilde{u}=0,\\\ &\tilde{u}(T,x)=u_{0}(x),\end{split}$ (2)
where $\tilde{p}(t,x)=p(T-t,x)$.
In what follows, system (2) will be referred to as the backward Navier–Stokes
equations. To this system we associate a certain system of forward-backward
stochastic differential equations on the group of volume-preserving
diffeomorphisms of a flat torus. For simplicity, we work in two dimensions.
However, the generalization of most of the results to the case of $n$
dimensions is straightforward. The necessary constructions and non-
straightforward generalizations related to the $n$-dimensional case are
considered in the appendix.
Assuming the existence of a solution of (2) with the final data in the Sobolev
space $H^{\alpha}$ for sufficiently large ${\alpha}$, we construct a solution
of the associated system of FBSDEs. Conversely, if we assume that a solution
of the system of FBSDEs exists, then the solution of the Navier–Stokes
equations can be obtained from the solution of the FBSDEs. In fact, the
constructed FBSDEs on the group of volume-preserving diffeomorphisms can be
regarded as an alternative object to the Navier–Stokes equations for studying
the properties of the latter.
The connection between forward-backward SDEs and quasi-linear PDEs in finite
dimensions has been studied by many authors, for example in [9], [18], and
[21].
Our construction uses the approach originating in the work of Arnold [1] which
states that the motion of a perfect fluid can be described in terms of
geodesics on the group of volume-preserving diffeomorphisms of a compact
manifold. The necessary differential-geometric structures were developed in
later work by Ebin and Marsden [10]. We note here that [1] and [10] deal only
with differential geometry on the group of maps without involving probability.
The associated system of FBSDEs is solved using the existence of a solution to
(2), and by applying results from the works of Gliklikh ([13], [14], [15],
[16]). The latter works use, in turn, the approach to stochastic differential
equations on Banach manifolds developed by Dalecky and Belopolskaya [4], and
started by McKean [19]. Conversely, a solution of (2) is obtained using the
existence of a solution to the associated FBSDEs as well as some ideas and
constructions from [9]. However, unlike [9], we work in an infinite-
dimensional setting.
Representations of the Navier–Stokes velocity field as a drift of a diffusion
process were initiated in [23] and [24]. A different system of stochastic
equations (but not a system of two SDEs) associated to the Navier–Stokes
system was introduced and studied in [5]. This system also includes an SDE on
the group of volume-preserving diffeomorphisms, but is not a system of
forward-backward SDEs. Also, we mention here the works [2] and [3] discussing
probabilistic representations of solutions to the Navier–Stokes equations, and
the work [6] establishing a stochastic variational principle for the
Navier–Stokes equations. Different probabilistic representations of the
solution to the Navier–Stokes equations were studied for example in [17] and
[7]. We note that the list of literature on probabilistic approaches to the
Navier–Stokes equations as well as connections between finite-dimensional
FBSDEs and PDEs cited in this paper is by no means complete.
The method of applying infinite-dimensional forward-backward SDEs in
connection to the Navier–Stokes equations is employed, to the authors’
knowledge, for the first time.
## 2\. Geometry of the diffeomorphism group of the 2D torus
Let ${\mathbb{T}}^{2}=S^{1}\times S^{1}$ be the two-dimensional torus, and let
$H^{\alpha}({\mathbb{T}}^{2})$, ${\alpha}>2$, be the space of
$H^{\alpha}$-Sobolev maps ${\mathbb{T}}^{2}\to{\mathbb{T}}^{2}$. By
$G^{{\alpha}}$ we denote the subset of $H^{\alpha}({\mathbb{T}}^{2})$ whose
elements are ${\rm C}^{1}$-diffeomorphisms. Let
$G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}$ be the subgroup of
$G^{{\alpha}}$ consisting of diffeomorphisms preserving the volume measure on
${\mathbb{T}}^{2}$.
###### Lemma 1.
Let $g$ be an $H^{\alpha}$-map and a local diffeomorphism of a finite-
dimensional compact manifold $M$, $F$ be an $H^{\alpha}$-section of the
tangent bundle $TM$. Then, $F\circ g$ is an $H^{\alpha}$-map.
###### Proof.
See [15] (p. 139) or [10] (p. 108). ∎
Let $R_{g}$ denote the right translation on $G^{\alpha}$, i.e.
$R_{g}(\eta)=\eta\circ g$.
###### Lemma 2.
The map $R_{g}$ is $C^{\infty}$-smooth for every $g\in G^{\alpha}$.
Furthermore, for every $\eta\in G^{\alpha}$, the tangent map $TR_{g}$
restricted to the tangent space $T_{\eta}G^{\alpha}$ is defined by the
formula:
$\displaystyle TR_{g}:\;T_{\eta}G^{\alpha}\to T_{\eta\circ
g}G^{\alpha},\;X\mapsto X\circ g.$
###### Proof.
The proof easily follows from the $\alpha$-lemma (see [10], [15], [16]). ∎
###### Lemma 3.
The groups $G^{\alpha}$ and ${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$
are infinite-dimensional Hilbert manifolds. The group
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ is a subgroup and a smooth
submanifold of $G^{\alpha}$.
###### Lemma 4.
The tangent space $T_{e}G^{\alpha}$ is formed by all $H^{\alpha}$-vector
fields on ${\mathbb{T}}^{2}$. The tangent space
$T_{e}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ is formed by all
divergence-free $H^{\alpha}$-vector fields on ${\mathbb{T}}^{2}$.
The proof of Lemmas 3 and 4 can be found for example in [10], [15], [16].
###### Lemma 5.
Let $X\in T_{e}G^{\alpha}$ be an $H^{\alpha}$-vector field on
${\mathbb{T}}^{2}$. Then the vector field $\hat{X}$ on $G^{\alpha}$ defined by
$\hat{X}(g)=X\circ g$ is right-invariant. Furthermore, $\hat{X}$ is ${\rm
C}^{k}$-smooth if and only if $X\in H^{{\alpha}+k}$.
###### Proof.
The first statement follows from Lemma 2. The proof of the second statement
can be found in [10]. ∎
The vector field $\hat{X}$ on $G^{\alpha}$ defined in Lemma 5 will be referred
to below as the right-invariant vector field generated by $X\in
T_{e}G^{\alpha}$.
Let $g\in G^{\,{\alpha}}$, $X,Y\in T_{e}G^{{\alpha}}$. Consider the weak
$(\,\cdot\,,\,\cdot\,)_{0}$ and the strong $(\,\cdot\,,\,\cdot\,)_{\alpha}$
Riemannian metrics on $G^{\alpha}$ (see [16]):
$\displaystyle(\hat{X}(g),\hat{Y}(g))_{0}=$
$\displaystyle\int_{{\mathbb{T}}^{2}}(X\circ g(\theta),Y\circ
g(\theta))d\theta,$ (3) $\displaystyle(\hat{X}(g),\hat{Y}(g))_{\alpha}=$
$\displaystyle\int_{{\mathbb{T}}^{2}}(X\circ g(\theta),Y\circ
g(\theta))d\theta$ $\displaystyle+$
$\displaystyle\int_{{\mathbb{T}}^{2}}((d+\delta)^{\alpha}X\circ
g(\theta),(d+\delta)^{\alpha}Y\circ g(\theta))d\theta$ (4)
where $d$ is the differential, $\delta$ is the codifferential, $\hat{X}$ and
$\hat{Y}$ are the right-invariant vector fields on $G^{\alpha}$ generated by
the $H^{\alpha}$-vector fields $X$ and $Y$. Metric (3) gives rise to the
$L_{2}$-topology on the tangent spaces of $G^{\alpha}$, and metric (4) gives
rise to the $H^{\alpha}$-topology on the tangent spaces of $G^{{\alpha}}$ (see
[16]). If $g\in{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$, then scalar
products (3) and (4) do not depend on $g$. Moreover, for the strong metric on
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$, we have the following
formula:
$\displaystyle(\hat{X}(g),\hat{Y}(g))_{\alpha}=\int_{{\mathbb{T}}^{2}}(X\circ
g(\theta),(1+\Delta)^{\alpha}Y\circ g(\theta))d\theta$
where $\Delta=(d\delta+\delta d)$ is the Laplace-de Rham operator (see [22]).
Let us introduce the notation:
$\displaystyle{\mathbb{Z}}_{2}^{+}=\\{(k_{1},k_{2})\in{\mathbb{Z}}^{2}:k_{1}>0\;\text{or}\;k_{1}=0,k_{2}>0\\};$
$\displaystyle
k=(k_{1},k_{2})\in{\mathbb{Z}}_{2}^{+},\;\bar{k}=(k_{2},-k_{1}),\;|k|=\sqrt{k_{1}^{2}+k^{2}_{2}},\;k\cdot\theta=k_{1}\theta_{1}+k_{2}\theta_{2},$
$\displaystyle\theta=(\theta_{1},\theta_{2})\in{\mathbb{T}}^{2},\;\nabla=\Bigl{(}\frac{\partial}{\partial\theta_{1}},\frac{\partial}{\partial\theta_{2}}\Bigr{)},\;(\bar{k},\nabla)=k_{2}\frac{\partial}{\partial\theta_{1}}-k_{1}\frac{\partial}{\partial\theta_{2}}\,,$
and the vectors
$\displaystyle\begin{split}&\bar{A}_{k}(\theta)=\frac{1}{|k|^{{\alpha}+1}}\cos(k\cdot\theta)\left(\begin{matrix}k_{2}\\\
-k_{1}\end{matrix}\right),\quad\bar{B}_{k}(\theta)=\frac{1}{|k|^{{\alpha}+1}}\sin(k\cdot\theta)\left(\begin{matrix}k_{2}\\\
-k_{1}\end{matrix}\right),\\\ &\bar{A}_{0}=\left(\begin{matrix}1\\\
0\end{matrix}\right),\quad\bar{B}_{0}=\left(\begin{matrix}0\\\
1\end{matrix}\right).\end{split}$
Let $\\{A_{k}(g),B_{k}(g)\\}_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}$ be the
right-invariant vector fields on $G^{\alpha}$ generated by
$\\{\bar{A}_{k},\bar{B}_{k}\\}_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}$, i.e.
$\displaystyle\begin{split}&A_{k}(g)=\bar{A}_{k}\circ g,\quad
B_{k}(g)=\bar{B}_{k}\circ g,\quad g\in G^{\alpha},\\\ &A_{0}=\bar{A}_{0},\quad
B_{0}=\bar{B}_{0}.\end{split}$
By $\omega$-lemma (see [15]), $A_{k}$ and $B_{k}$ are ${\rm
C}^{\infty}$-smooth vector fields on $G^{\alpha}$.
###### Lemma 6.
The vectors $A_{k}(g)$, $B_{k}(g)$, $k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}$,
$g\in{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$, form an orthogonal
basis of the tangent space
$T_{g}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ with respect to both
the weak and the strong inner products in
$T_{g}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. In particular, the
vectors $\bar{A}_{k}$, $\bar{B}_{k}$, $k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}$,
form an orthogonal basis of the tangent space
$T_{e}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. Moreover, the weak and
the strong norms of the basis vectors are bounded by the same constant.
###### Proof.
It suffices to prove the lemma for the strong norm. Let us compute
$\Delta^{\alpha}\bar{A}_{k}$. Note that the vectors $\frac{k}{|k|}$ and
$\frac{\bar{k}}{|k|}$ form an orthonormal basis of ${\mathbb{R}}^{2}$. Let us
observe that by the identity $(\bar{k},\nabla)\cos(k\cdot\theta)=0$,
$\delta\bar{A}_{k}=0$. Hence $d\delta\,\bar{A}_{k}=0$ which implies
$\Delta\,\bar{A}_{k}=\delta d\,\bar{A}_{k}$. We obtain:
$\displaystyle\bar{A}_{k}=\frac{1}{|k|^{\alpha}}\cos(k\cdot\theta)\,\frac{\bar{k}}{|k|},$
$\displaystyle
d\,\bar{A}_{k}=-\frac{1}{|k|^{{\alpha}-1}}\sin(k\cdot\theta)\,\frac{k}{|k|}\wedge\frac{\bar{k}}{|k|},$
$\displaystyle\Delta\bar{A}_{k}=\delta
d\,\bar{A}_{k}=\frac{1}{|k|^{{\alpha}-2}}\cos(k\cdot\theta)\,\frac{\bar{k}}{|k|}=|k|^{2}\bar{A}_{k},$
$\displaystyle\Delta^{\alpha}\,\bar{A}_{k}=|k|^{\alpha}\cos(k\cdot\theta)\,\frac{\bar{k}}{|k|}=|k|^{2{\alpha}}\bar{A}_{k}.$
This and the volume-preserving property of
$g\in{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ imply that
$\displaystyle(B_{m}(g),A_{k}(g))_{\alpha}=(\bar{B}_{m},\bar{A}_{k})_{\alpha}=(1+|k|^{2{\alpha}})(\bar{B}_{m},\bar{A}_{k})_{L_{2}}=0,$
$\displaystyle\|A_{k}(g)\|^{2}_{\alpha}=\|\bar{A}_{k}\|^{2}_{\alpha}=\bigl{(}1+|k|^{2{\alpha}}\bigr{)}\|\bar{A}_{k}\|^{2}_{L_{2}}=2\pi^{2}\left(|k|^{-2{\alpha}}+1\right)$
where $\|\,\cdot\,\|_{\alpha}$ is the norm corresponding to the scalar product
$(\,\cdot\,,\,\cdot\,)_{\alpha}$. Thus,
$2\pi^{2}\leqslant\|A_{k}(g)\|^{2}_{\alpha}\leqslant 4\pi^{2}$. Clearly, for
the $\|B_{k}(g)\|^{2}_{\alpha}$ we obtain the same. ∎
It has been shown, for example, in [10] and [16] that the weak Riemannian
metric has the Levi-Civita connection, geodesics, the exponential map, and the
spray. Let $\bar{\nabla}$ and $\tilde{\nabla}$ denote the covariant
derivatives of the Levi-Civita connection of the weak Riemannian metric (3) on
$G^{\alpha}$ and ${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$,
respectively. In [10] (see also [16], [15]), it has been shown that
$\displaystyle\tilde{\nabla}={\rm P\,}\circ\bar{\nabla}$
where ${\rm P\,}:TG^{\alpha}\to
T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ is defined in the following
way: on each tangent space $T_{g}G^{\alpha}$, ${\rm P\,}=P_{g}$ where
$P_{g}=TR_{g}\circ P_{e}\circ TR_{g^{-1}}$, $TR_{g}$ and $TR_{g^{-1}}$ are
tangent maps, and $P_{e}:T_{e}G^{\alpha}\to
T_{e}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ is the projector defined
by the Hodge decomposition.
###### Lemma 7.
Let $\hat{U}$ be the right-invariant vector field on $G^{\alpha}$ generated by
an $H^{{\alpha}+1}$-vector field $U$ on ${\mathbb{T}}^{2}$, and let $\hat{V}$
be the right-invariant vector field on $G^{\alpha}$ generated by an
$H^{\alpha}$-vector field $V$ on ${\mathbb{T}}^{2}$. Then
$\bar{\nabla}_{\hat{V}}\hat{U}$ is the right-invariant vector field on
$G^{\alpha}$ generated by the $H^{\alpha}$-vector field $\nabla_{V}U$ on
${\mathbb{T}}^{2}$.
###### Lemma 8.
Let $\hat{U}$ be the right-invariant vector field on
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ generated by a divergence-
free $H^{{\alpha}+1}$-vector field $U$ on ${\mathbb{T}}^{2}$, and let
$\hat{V}$ be the right-invariant vector field on $G^{\alpha}$ generated by a
divergence-free $H^{\alpha}$-vector field $V$ on ${\mathbb{T}}^{2}$. Then
$\tilde{\nabla}_{\hat{V}}\hat{U}$ is the right-invariant vector field on
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ generated by the divergence-
free $H^{\alpha}$-vector field $P_{e}\nabla_{V}U$ on ${\mathbb{T}}^{2}$.
The proofs of Lemmas 7 and 8 follow from the right-invariance of covariant
derivatives on $G^{\alpha}$ and
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ (see [16]).
###### Remark 1.
The basis
$\\{\bar{A}_{k},\bar{B}_{k}\\}_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}$ of
$T_{e}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ can be extended to a
basis of the entire tangent space $T_{e}G^{\alpha}$. Indeed, let us introduce
the vectors:
$\displaystyle\bar{\mathcal{A}_{k}}(\theta)=\frac{1}{|k|^{{\alpha}+1}}\cos(k\cdot\theta)\left(\begin{matrix}k_{1}\\\
k_{2}\end{matrix}\right),\;\bar{\mathcal{B}_{k}}(\theta)=\frac{1}{|k|^{{\alpha}+1}}\sin(k\cdot\theta)\left(\begin{matrix}k_{1}\\\
k_{2}\end{matrix}\right),\;k\in{\mathbb{Z}}^{+}_{2}.$
The system $\bar{A}_{k}$, $\bar{B}_{k}$,
$k\in{\mathbb{Z}}^{+}_{2}\cup\\{0\\}$, $\bar{\mathcal{A}_{k}}$,
$\bar{\mathcal{B}_{k}}$, $k\in{\mathbb{Z}}^{+}_{2}$, form an orthogonal basis
of $T_{e}G^{\alpha}$. Further let $\mathcal{A}_{k}$ and $\mathcal{B}_{k}$
denote the right-invariant vector fields on $G^{\alpha}$ generated by
$\bar{\mathcal{A}_{k}}$ and $\bar{\mathcal{B}_{k}}$.
## 3\. The FBSDEs on the group of diffeomorphisms of the 2D torus
Let $h:{\mathbb{T}}^{2}\to{\mathbb{R}}^{2}$ be a divergence-free
$H^{{\alpha}+1}$-vector field on ${\mathbb{T}}^{2}$, and let $\hat{h}$ be the
right-invariant vector field on $G^{\alpha}$ generated by $h$. Further let the
function $V(s,\,\cdot\,)$ be such that there exists a function $p:[t,T]\to
H^{{\alpha}+1}({\mathbb{T}}^{2},{\mathbb{R}})$ satisfying
$V(s,\,\cdot\,)=\nabla p(s,\,\cdot\,)$ for all $s\in[t,T]$. For each
$s\in[t,T]$, $\hat{V}(s,\,\cdot\,)$ denotes the right-invariant vector field
on $G^{\alpha}$ generated by $V(s,\,\cdot\,)\in
H^{{\alpha}}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$.
Let $E$ be a Euclidean space spanned on an orthonormal, relative to the scalar
product in $E$, system of vectors
$\\{e_{k}^{A},e_{k}^{B},e_{0}^{A},e_{0}^{B}\\}_{k\in{\mathbb{Z}}_{2}^{+},|k|\leqslant
N}$. Consider the map
$\displaystyle\sigma(g)=\sum_{\begin{subarray}{c}k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},\\\
|k|\leqslant N\end{subarray}}A_{k}(g)\otimes e_{k}^{A}+B_{k}(g)\otimes
e_{k}^{B},\;g\in G^{\alpha},$
i.e. $\sigma(g)$ is a linear operator $E\to T_{g}G^{\alpha}$ for each $g\in
G^{\alpha}$.
Let $(\Omega,\mathcal{F},\mathbb{P})$ be a probability space, and $W_{s}$,
$s\in[t,T]$, be an $E$-valued Brownian motion:
$\displaystyle
W_{s}=\sum_{\begin{subarray}{c}k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},\\\
|k|\leqslant
N\end{subarray}}(\beta^{A}_{k}(s)e^{A}_{k}+\beta^{B}_{k}(s)e^{B}_{k})$
where
$\\{\beta^{A}_{k},\beta^{B}_{k}\\}_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},|k|\leqslant
N}$ is a sequence of independent Brownian motions. We consider the following
system of forward and backward SDEs:
$\begin{cases}dZ_{s}^{t,e}=Y_{s}^{t,e}ds+\epsilon\,\sigma(Z_{s}^{t,e})dW_{s},\\\
dY_{s}^{t,e}=-\hat{V}(s,Z_{s}^{t,e})ds+X_{s}^{t,e}dW_{s},\\\
Z^{t,e}_{t}=e;\;Y_{T}^{t,e}=\hat{h}(Z_{T}^{t,e}).\end{cases}$ (5)
The forward SDE of (5) is an SDE on
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ where
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ is considered as a Hilbert
manifold. Stochastic differentials and stochastic differential equations on
Hilbert manifolds are understood in the sense of Dalecky and Belopolskaya’s
approach (see [4]). More precisely, we use the results from [15] which
interprets the latter approach for the particular case of SDEs on Hilbert
manifolds. The stochastic integral in the forward SDE can be explicitly
written as follows:
$\displaystyle\int_{t}^{s}\sigma(Z^{t,e}_{r})dW_{r}=\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},|k|\leqslant
N}\int_{t}^{s}A_{k}(Z^{t,e}_{r})d\beta^{A}_{k}(r)+B_{k}(Z^{t,e}_{r})d\beta^{B}_{k}(r).$
(6)
Let us consider the backward SDE:
$\displaystyle
Y_{s}^{t,e}=\hat{h}(Z_{T}^{t,e})+\int_{s}^{T}\hat{V}(r,Z^{t,e}_{r})dr-\int_{s}^{T}X_{r}^{t,e}dW_{r}.$
(7)
Note that the processes $\hat{V}(s,Z^{t,e}_{s})=V(s,\,\cdot\,)\circ
Z^{t,e}_{s}$ and $\hat{h}(Z^{t,e}_{T})=h\circ Z^{t,e}_{T}$ are
$H^{\alpha}$-maps by Lemma 1. Therefore, it makes sense to understand SDE (7)
as an SDE in the Hilbert space
$H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$. Let
$\mathcal{F}_{s}=\sigma(W_{r},r\in[0,s])$. We would like to find an
$\mathcal{F}_{s}$-adapted triple of stochastic processes
$(Z_{s}^{t,e},Y_{s}^{t,e},X_{s}^{t,e})$ solving FBSDEs (5) in the following
sense: at each time $s$, the process $(Z_{s}^{t,e},Y_{s}^{t,e})$ takes values
in an $H^{\alpha}$-section of the tangent bundle
$T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. Namely, for each
$s\in[t,T]$ and $\omega\in\Omega$,
$Z_{s}^{t,e}\in{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$,
$Y_{s}^{t,e}\in
T_{Z_{s}^{t,e}}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. Therefore,
the forward SDE is well-posed on both $G^{\alpha}$ and
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$, and can be written in the
Dalecky–Belopolskaya form:
$\displaystyle dZ_{s}^{t,e}=\exp_{Z_{s}^{t,e}}$
$\displaystyle\\{Y_{s}^{t,e}ds+\epsilon\,\sigma(Z_{s}^{t,e})dW_{s}\\}$ or
$\displaystyle dZ_{s}^{t,e}=\tilde{\exp}_{Z_{s}^{t,e}}$
$\displaystyle\\{Y_{s}^{t,e}ds+\epsilon\,\sigma(Z_{s}^{t,e})dW_{s}\\}$
where $\exp$ and $\tilde{\exp}$ are the exponential maps of the Levi-Civita
connection of the weak Riemannian metrics (3) on $G^{\alpha}$ and resp.
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. Below, we will show that
using either of these representations leads to the same solution of FBSDEs
(5).
Finally, the process $X_{s}^{t,e}$ takes values in the space of linear
operators $\mathcal{L}(E,H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2}))$, i.e.
$\displaystyle
X_{s}^{t,e}=\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},|k|\leqslant
N}X^{kA}_{s}\otimes e^{A}_{k}+X^{kB}_{s}\otimes e^{B}_{k}$ (8)
where the processes $X^{kA}_{s}$ and $X^{kB}_{s}$ take values in
$H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$.
###### Remark 2.
The results obtained below also work in the situation when the Brownian motion
$W_{s}$ is infinite dimensional (as in [8]). Namely, when
$W_{s}=\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}a_{k}\beta_{k}^{A}\otimes
e^{A}_{k}+b_{k}\beta_{k}^{B}\otimes e^{B}_{k}$ where $a_{k}$, $b_{k}$,
$k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}$, are real numbers satisfying
$\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}|a_{k}|^{2}+|b_{k}|^{2}<\infty$.
However, this requires an additional analysis on the solvability of the
forward SDE based on the approach of Dalecky and Belopolskaya [4] since the
results of Gliklikh ([13], [15], [16]) applied below are obtained for the case
of a finite-dimensional Brownian motion.
## 4\. Constructing a solution of the FBSDEs
### 4.1 The forward SDE
Let us consider the backward Navier–Stokes equations in ${\mathbb{R}}^{2}$:
$\displaystyle\begin{split}&y(s,\theta)=h(\theta)+\int_{s}^{T}\bigl{[}\nabla
p(r,\theta)+\bigl{(}y(r,\theta),\nabla\bigr{)}y(r,\theta)+\nu\Delta
y(r,\theta)\bigr{]}dr,\\\ &\mathrm{div}\,y(s,\theta)=0\end{split}$ (9)
where $s\in[t,T]$, $\theta\in{\mathbb{T}}^{2}$, $\Delta$ and $\nabla$ are the
Laplacian and the gradient.
###### Assumption 1.
Let us assume that on the interval $[t,T]$ there exists a solution
$\bigl{(}y(s,\,\cdot\,),p(s,\,\cdot\,)\bigr{)}$ to (9) such that the functions
$p:[t,T]\to H^{{\alpha}+1}({\mathbb{T}}^{2},{\mathbb{R}})$ and $y:[t,T]\to
H^{{\alpha}+1}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$ are continuous.
Clearly, $y(s,\,\cdot\,)\in
T_{e}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. Let
$\\{Y^{t;kA}_{s},Y^{t;kB}_{s}\\}_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}$ be the
coordinates of $y(s,\,\cdot\,)$ with respect to the basis
$\\{\bar{A}_{k},\bar{B}_{k}\\}_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}$, i.e.
$y(s,\theta)=\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}Y^{t;kA}_{s}\bar{A}_{k}(\theta)+Y^{t;kB}_{s}\bar{B}_{k}(\theta).$
Let $\hat{Y}_{s}(\,\cdot\,)$ denote the right-invariant vector field on
$G^{\alpha}$ generated by the solution $y(s,\,\cdot\,)$, i.e.
$\hat{Y}_{s}(g)=y(s,\,\cdot\,)\circ g$. On each tangent space
$T_{g}G^{\alpha}$, the vector $\hat{Y}_{s}(g)$ can be represented by a series
converging in the $H^{\alpha}$-topology:
$\displaystyle\hat{Y}_{s}(g)=\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}Y^{t;kA}_{s}A_{k}(g)+Y^{t;kB}_{s}B_{k}(g).$
(10)
In this paragraph we will study the SDE:
$\displaystyle
dZ^{t,e}_{s}=\hat{Y}_{s}(Z^{t,e}_{s})ds+\epsilon\,\sigma(Z^{t,e}_{s})dW_{s}.$
(11)
Later, in Theorem 6, we will show that the solution $Z^{t,e}_{s}$ to (11) and
the process $Y^{t,e}_{s}=\hat{Y}_{s}(Z^{t,e}_{s})$ are the first two processes
in the triple $(Z^{t,e}_{s},Y^{t,e}_{s},X^{t,e}_{s})$ that solves FBSDEs (5).
###### Theorem 1.
There exists a unique strong solution $Z^{t,e}_{s}$, $s\in[t,T]$, to (11) on
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$, with the initial condition
$Z^{t,e}_{t}=e$.
###### Proof.
Below, we verify the assumptions of Theorem 13.5 of [16]. The latter theorem
will imply the existence and uniqueness of the strong solution to (11). Note
that, if sum (6) representing the stochastic integral
$\int_{t}^{s}\sigma(Z^{t,e}_{s})\,dW_{s}$ contains only the terms
$A_{0}(\beta^{A}_{0}(s)-\beta^{A}_{0}(t))$ and
$B_{0}(\beta^{B}_{0}(s)-\beta^{B}_{0}(t))$, i.e., informally speaking, if the
Brownian motion runs only along the constant vectors $A_{0}$ and $B_{0}$, then
the statement of the theorem follows from Theorem 28.3 of [16]. If sum (6)
contains also terms with $A_{k}$ and $B_{k}$, $k\in{\mathbb{Z}}_{2}^{+}$, or,
informally, when the Brownian motion runs also along non-constant vectors
$A_{k}$ and $B_{k}$, $k\in{\mathbb{Z}}_{2}^{+}$, then the assumptions of
Theorem 13.5 of [16] require the boundedness of $A_{k}$ and $B_{k}$ with
respect to the strong norm. The latter fact holds by Lemma 6.
Hence, all the assumptions of Theorem 13.5 of [16] are satisfied. Indeed, the
proof of Theorem 28.3 of [16] shows that the Levi-Civita connection of the
weak Riemannian metric (3) on
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ is compatible (see
Definition 13.7 of [16]) with the strong Riemannian metric (4). The function
$\sigma(g)=\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},|k|\leqslant
N}A_{k}(g)\otimes e_{k}^{A}+B_{k}(g)\otimes e_{k}^{B}$ is ${\rm
C}^{\infty}$-smooth since $A_{k}$ and $B_{k}$ are ${\rm C}^{\infty}$-smooth.
Moreover, by Lemma 6, $\sigma(g)$ is bounded on
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. Next, since $y:[t,T]\to
H^{{\alpha}+1}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$ is continuous, then it is
also bounded with respect to (at least) the $H^{\alpha}$-norm. Hence, the
generated right-invariant vector field $\hat{Y}_{s}(g)$ is bounded in $s$ with
respect to the strong metric (4), and it is at least ${\rm C}^{1}$-smooth in
$g$. The boundedness of $\hat{Y}_{s}$ in $g$ follows from the volume-
preserving property of $g$. ∎
###### Theorem 2.
There exists a unique strong solution $Z^{t,e}_{s}$, $s\in[t,T]$, to (11) on
$G^{\alpha}$, with the initial condition $Z^{t,e}_{t}=e$. This solution
coincides with the solution to SDE (11) on
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$.
###### Proof.
Consider the identical embedding
$\imath:{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}\to G^{\alpha}$. By
results of [4] (Proposition 1.3, p. 146; see also [16], p. 64), the stochastic
process $\imath(Z^{t,e}_{s})=Z^{t,e}_{s}$, $s\in[t,T]$, is a solution to SDE
(11) on $G^{\alpha}$, i.e. with respect to the exponential map $\exp$. This
easily follows from the fact that
$T\imath:T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}\to TG^{\alpha}$,
where $T$ is the tangent map, is the identical imdedding, and that
$\imath\bigl{(}\exp(X)\bigr{)}=\tilde{\exp}(T\imath\circ X)$. The solution
$Z^{t,e}_{s}$ to (11) on $G^{\alpha}$ is unique. This follows from the
uniqueness theorem for SDE (11) considered on the manifold $G^{\alpha}$
equipped with the weak Riemannian metric. Indeed, $\sigma(g)$ and
$\hat{Y}_{s}(g)$ are bounded with respect to the weak metric (3) since the
functions $\bar{A}_{k}$, $\bar{B}_{k}$, $k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}$,
are bounded on ${\mathbb{T}}^{2}$, and $y(\,\cdot\,,\,\cdot\,)$ is bounded on
$[t,T]\times{\mathbb{T}}^{2}$. Moreover $\sigma(g)$ is ${\rm
C}^{\infty}$-smooth and $\hat{Y}_{s}$ is at least ${\rm C}^{1}$-smooth on
$G^{\alpha}$. ∎
One can also consider (11) as an SDE with values in the Hilbert space
$H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$.
###### Theorem 3.
There exists a unique strong solution $Z^{t,e}_{s}$ to the
$H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$-valued SDE (11) on $[t,T]$,
with the initial condition $Z^{t,e}_{t}=e$ where $e$ is the identity of
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. This solution coincides
with the solution to SDE (11) on
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ or $G^{\alpha}$.
###### Proof.
By Theorem 1, SDE (11) on ${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$
has a unique strong solution $Z^{t,e}_{s}$ on $[t,T]$. Let us prove that the
solution $Z^{t,e}_{s}$ to (11) solves this SDE considered as an SDE in
$H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$. Consider the identical
imbedding $\imath_{\scriptscriptstyle
V}:\,{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}\to
H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$, $g\mapsto g$. Applying Itô’s
formula to $\imath_{\scriptscriptstyle V}$, and taking into account that
$A_{k}(g)\imath_{\scriptscriptstyle V}(g)=\nabla_{\bar{A}_{k}}\theta\circ
g=A_{k}(g)$ and that $A_{k}(g)A_{k}(g)\imath_{\scriptscriptstyle
V}(g)=A_{k}(g)A_{k}(g)=0$, we obtain that the solution $Z^{t,e}_{s}$ to (11)
on ${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ solves the
$H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$-valued SDE (11). Note that by
the uniqueness theorem for SDEs in Hilbert spaces, SDE (11) can have only one
solution in $L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$. This proves the
uniqueness of its solution in $H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$
as well. Thus the solutions to (11) on $G^{\alpha}$,
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$, and in
$H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$ coincide. ∎
Let us find the representations of SDE (11) in normal coordinates on
$G^{\alpha}$ and ${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. First, we
prove the following lemma.
###### Lemma 9.
The following equality holds:
$\displaystyle\int_{t}^{s}\sigma(Z^{t,e}_{r})\circ
dW_{r}=\int_{t}^{s}\sigma(Z^{t,e}_{r})dW_{r},$
i.e. instead of the Itô stochastic integral in (11) we can write the
Stratonovich stochastic integral $\int_{t}^{s}\sigma(Z^{t,e}_{r})\circ
dW_{r}$.
###### Proof.
We have:
$\displaystyle\sigma(Z^{t,e}_{r})\circ
dW_{r}=\sigma(Z^{t,e}_{r})dW_{r}+\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},|k|\leqslant
N}dA_{k}(Z^{t,e}_{r})d\beta^{A}_{k}(r)+dB_{k}(Z^{t,e}_{r})d\beta^{B}_{k}(r).$
Hence, we have to prove that $dA_{k}(Z^{t,e}_{r})d\beta^{A}_{k}(r)=0$ and
$dB_{k}(Z^{t,e}_{r})d\beta^{B}_{k}(r)=0$. For simplicity of notation we use
the notation $A_{\nu}$ for both of the vector fields $A_{k}$ and $B_{k}$ and
the notation $\bar{A}_{\nu}$ for $\bar{A}_{k}$ and $\bar{B}_{k}$,
$k\in{\mathbb{Z}}^{+}_{2}\cup\\{0\\}$. Also, we use the notation
$\beta_{\nu}(s)$ for the Brownian motions
$\\{\beta^{A}_{k}(s),\beta^{B}_{k}(s)\\}_{k\in{\mathbb{Z}}^{+}_{2}\cup\\{0\\},|k|\leqslant
N}$. We obtain:
$\displaystyle d(\bar{A}_{\nu}\circ
Z^{t,e}_{s})=\sum_{\gamma}A_{\gamma}(Z^{t,e}_{s})\bigl{(}\bar{A}_{\nu}\circ
Z^{t,e}_{s}\bigr{)}\circ
d\beta_{\gamma}(s)+Y^{t,e}_{s}\bigl{(}\bar{A}_{\nu}\circ
Z^{t,e}_{s}\bigr{)}dt.$
This implies
$\displaystyle d(\bar{A}_{\nu}\circ Z^{t,e}_{s})\cdot
d\beta_{\nu}=A_{\nu}(Z^{t,e}_{s})\bigl{(}\bar{A}_{\nu}\circ
Z^{t,e}_{s}\bigr{)}ds=0$
which holds by the identity
$(\bar{k},\nabla)\cos(k\cdot\theta)=(\bar{k},\nabla)\sin(k\cdot\theta)=0$ or
by differentiating of constant vector fields. ∎
Let
$\bar{Z}^{t}_{s}=\\{Z_{s}^{t;kA},Z_{s}^{t;kB}\\}_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}$
be the vector of local coordinates of the solution $Z^{t,e}_{s}$ to (11) on
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$, i.e. the vector of normal
coordinates provided by the exponential map
$\tilde{\exp}:T_{e}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}\to{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$.
Let $U_{e}$ be the canonical chart of the map $\tilde{\exp}$.
###### Theorem 4 (SDE (11) in local coordinates).
Let
$\displaystyle\tau=\inf\\{s\in[t,T]:Z_{s}^{t,e}\notin U_{e}\\}.$ (12)
On the interval $[t,\tau]$, SDE (11) has the following representation in local
coordinates:
$\displaystyle\begin{split}&Z^{t,kA}_{s\wedge\tau}=\int_{t}^{{s}\wedge\tau}Y^{t;kA}_{r}dr+\delta_{k}\epsilon\,(\beta_{k}^{A}(s\wedge\tau)-\beta_{k}^{A}(t)),\\\
&Z^{t,kB}_{s\wedge\tau}=\int_{t}^{{s}\wedge\tau}Y^{t;kB}_{r}dr+\delta_{k}\epsilon\,(\beta_{k}^{B}(s\wedge\tau)-\beta_{k}^{B}(t)).\end{split}$
(13)
where $\delta_{k}=1$ if $|k|\leqslant N$, and $\delta_{k}=0$ if $|k|>N$.
###### Proof.
Let $\bar{g}=\\{g^{kA},g^{kB}\\}_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}$ be
local coordinates in the neighborhood $U_{e}$ provided by the map
$\tilde{\exp}$. Let $f\in{\rm
C}^{\infty}(G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}})$, and let
$\tilde{f}:T_{e}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}\to{\mathbb{R}}$
be such that $\tilde{f}=f\circ\tilde{\exp}$. Since $\tilde{\exp}$ is a
$C^{\infty}$-map (see [10]), then $\tilde{f}\in{\rm C}^{\infty}(U_{0})$, where
$U_{0}=\tilde{\exp}^{-1}U_{e}$. Note that $\frac{\partial}{\partial
g^{kA}}\tilde{f}(\bar{g})=A_{k}(g)f(g)$ and $\frac{\partial}{\partial
g^{kB}}\tilde{f}(\bar{g})=B_{k}(g)f(g)$. By Itô’s formula, we obtain:
$f(Z^{t,e}_{s\wedge\tau})-f(e)=\tilde{f}(\bar{Z}^{t,0}_{s\wedge\tau})-\tilde{f}(0)\\\
=\int_{t}^{s\wedge\tau}dr\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}\frac{\partial\tilde{f}}{\partial
g^{kA}}(\bar{Z}^{t}_{r})Y^{t;kA}_{r}+\int_{t}^{s\wedge\tau}\epsilon\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}\delta_{k}\,\frac{\partial\tilde{f}}{\partial
g^{kA}}(\bar{Z}^{t}_{r})Y^{t;kA}_{r}\circ d\beta_{k}^{A}(r)\\\
+\int_{t}^{s\wedge\tau}dr\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}\frac{\partial\tilde{f}}{\partial
g^{kB}}(\bar{Z}^{t}_{r})Y^{t;kB}_{r}+\int_{t}^{s\wedge\tau}\epsilon\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}\delta_{k}\,\frac{\partial\tilde{f}}{\partial
g^{kB}}(\bar{Z}^{t}_{r})Y^{t;kB}_{r}\circ d\beta_{k}^{B}(r)\\\
=\int_{t}^{s\wedge\tau}dr\,\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}\bigl{(}Y^{t;kA}_{r}A_{k}(Z^{t,e}_{r})f(Z^{t,e}_{r})+Y^{t;kB}_{r}B_{k}(Z^{t,e}_{r}\bigr{)}f(Z^{t,e}_{r})\bigr{)}\\\
+\int_{t}^{s\wedge\tau}\epsilon\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}\delta_{k}\,\bigl{(}A_{k}(Z^{t,e}_{r})f(Z^{t,e}_{r})\circ
d\beta^{A}_{k}(r)+B_{k}(Z^{t,e}_{r})f(Z^{t,e}_{r})\circ
d\beta_{k}^{B}(r)\bigr{)}.$
Using representations (10) and (6) we obtain:
$\displaystyle
f(Z^{t,e}_{s\wedge\tau})-f(e)=\int_{t}^{s\wedge\tau}\hat{Y}_{r}(Z^{t,e}_{r})f(Z^{t,e}_{r})dr+\int_{t}^{s\wedge\tau}\epsilon\,\sigma(Z^{t,e}_{r})f(Z^{t,e}_{r})\circ
dW_{r}.$
This shows that the process
$\displaystyle\exp\Bigl{\\{}\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}Z^{t,kA}_{s\wedge\tau}\bar{A}_{k}+Z^{t,kB}_{s\wedge\tau}\bar{B}_{k}\Bigr{\\}}$
solves SDE (11) on the interval $[t,\tau]$. ∎
Let
$\displaystyle\check{\bar{Z}}_{s}^{t}=\\{\check{Z}_{s}^{t;kA},\check{Z}_{s}^{t;kB},\check{Z}_{s}^{t;k\mathcal{A}},\check{Z}_{s}^{t;k\mathcal{B}},\check{Z}_{s}^{t;0A},\check{Z}_{s}^{t;0B}\\}_{k\in{\mathbb{Z}}_{2}^{+}}$
be the vector of local coordinates of the solution $Z^{t,e}_{s}$ to (11) on
$G^{\alpha}$, i.e. the vector of normal coordinates provided by the
exponential map $\exp:T_{e}G^{\alpha}\to G^{\alpha}$. Further let
$\check{U}_{e}$ be the canonical chart of the map $\exp$.
###### Theorem 5.
Let
$\displaystyle\check{\tau}=\inf\\{s\in[t,T]:Z_{s}^{t,e}\notin\check{U}_{e}\\}.$
Then, a.s. $\check{\tau}=\tau$, where the stopping time $\tau$ is defined by
(12), and on $[t,\tau]$, $\check{Z}_{s}^{t;kA}=Z_{s}^{t;kA}$,
$\check{Z}_{s}^{t;kB}=Z_{s}^{t;kB}$, $k\in{\mathbb{Z}}^{2}_{+}\cup\\{0\\}$,
$\check{Z}_{s}^{t;k\mathcal{A}}=\check{Z}_{s}^{t;k\mathcal{B}}=0$,
$k\in{\mathbb{Z}}_{2}^{+}$, a.s.
###### Proof.
Let us introduce additional local coordinates
$g^{k\mathcal{A}},g^{k\mathcal{B}}$, $k\in{\mathbb{Z}}_{2}^{+}$, and perform
the same computation as in the proof of Theorem 4. We have to take into
account that $Y^{k\mathcal{A}}_{s}=Y^{k\mathcal{B}}_{s}=0$,
$k\in{\mathbb{Z}}_{2}^{+}$, and that the components of the Brownian motion are
non-zero only along divergence-free and constant vector fields. We obtain that
the coordinate process $\check{\bar{Z}}_{s}^{t}$ verifies SDEs (13) and the
equations $\check{Z}_{s}^{t;k\mathcal{A}}=\check{Z}_{s}^{t;k\mathcal{B}}=0$,
$k\in{\mathbb{Z}}_{2}^{+}$. ∎
### 4.2 The backward SDE and the solution of the FBSDEs
We have the following result:
###### Theorem 6.
Let $\hat{Y}_{s}$ be the right-invariant vector field generated by the
solution $y(s,\,\cdot\,)$ to the backward Navier–Stokes equations (9). Further
let $Z^{t,e}_{s}$ be the solution to SDE (11) on
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. Then there exists an
$\epsilon>0$ such that the triple of stochastic processes
$\displaystyle
Z^{t,e}_{s},\,Y^{t,e}_{s}=\hat{Y}_{s}(Z^{t,e}_{s}),\,X^{t,e}_{s}=\epsilon\,\sigma(Z^{t,e}_{s})\hat{Y}_{s}(Z^{t,e}_{s})$
solves FBSDEs (5) on the interval $[t,T]$.
###### Remark 3.
The expression $\sigma(Z^{t,e}_{s})\hat{Y}_{s}(Z_{s}^{t,e})$ means the
following:
$\displaystyle\sigma(Z^{t,e}_{s})\hat{Y}_{s}(Z_{s}^{t,e})=\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},|k|\leqslant
N}A_{k}(Z^{t,e}_{s})\hat{Y}_{s}(Z_{s}^{t,e})\otimes
e^{A}_{k}+B_{k}(Z^{t,e}_{s})\hat{Y}_{s}(Z_{s}^{t,e})\otimes e^{B}_{k}$
where $\hat{Y}_{s}(\,\cdot\,)$ is regarded as a function
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}\to
H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$, and $A_{k}(g)\hat{Y}_{s}(g)$
means differentiation of
$\hat{Y}_{s}:\,{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}\to
H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$ along the vector field $A_{k}$
at the point $g\in{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. Let
$\gamma_{\xi}$ be the geodesic in
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ such that $\gamma_{0}=e$ and
$\gamma^{\prime}_{0}=\bar{A}_{k}$. We obtain:
$A_{k}(g)\hat{Y}_{s}(g)(\theta)=\frac{d}{d\xi}\,\hat{Y}_{s}(\gamma_{\xi}\circ
g)(\theta)|_{\xi=0}=R_{g}\,\frac{d}{d\xi}\,y(s,\gamma_{\xi}\theta)|_{\xi=0}\\\
=R_{g}\,\nabla_{\bar{A}_{k}}y(s,\theta)=\bar{\nabla}_{A_{k}}\hat{Y}_{s}(g)(\theta).$
(14)
Thus,
$\displaystyle
X^{t,e}_{s}=\epsilon\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},|k|\leqslant
N}[\nabla_{\bar{A}_{k}}y(s,\,\cdot\,)\otimes
e^{A}_{k}+\nabla_{\bar{B}_{k}}y(s,\,\cdot\,)\otimes e^{B}_{k}]\circ
Z_{s}^{t,e},$ (15)
and the stochastic integral in (7) can be represented as
$\int_{s}^{T}X^{t,e}_{r}dW_{r}\\\
=\epsilon\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},|k|\leqslant
N}\int_{s}^{T}\nabla_{\bar{A}_{k}}y(r,\,\cdot\,)\circ
Z_{r}^{t,e}d\beta^{A}_{k}(r)+\int_{s}^{T}\nabla_{\bar{B}_{k}}y(r,\,\cdot\,)\circ
Z_{r}^{t,e}d\beta^{B}_{k}(r).$
In particular, if $N=0$,
$\displaystyle\int_{s}^{T}X^{t,e}_{r}dW_{r}=\epsilon\left(\int_{s}^{T}\frac{\partial}{\partial\theta_{1}}y(r,\,\cdot\,)\circ
Z_{r}^{t,e}d\beta^{A}_{0}(r)+\int_{s}^{T}\frac{\partial}{\partial\theta_{2}}y(r,\,\cdot\,)\circ
Z_{r}^{t,e}d\beta^{B}_{0}(r)\right).$
A result similar to Lemma 10 below was obtained in [6].
###### Lemma 10 (The Laplacian of a right-invariant vector field).
Let $\hat{V}$ be the right-invariant vector field on $G^{\tilde{\alpha}}$
generated by an $H^{\tilde{\alpha}+2}$-vector field $V$ on ${\mathbb{T}}^{2}$.
Further let $\epsilon>0$ be such that
$\frac{\epsilon^{2}}{2}\Bigl{(}1+\frac{1}{2}\sum_{k\in{\mathbb{Z}}_{2}^{+},|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}}}\Bigr{)}=\nu$. Then for all $g\in
G^{\tilde{\alpha}}$,
$\displaystyle\frac{\epsilon^{2}}{2}\sum_{\begin{subarray}{c}k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},\\\
|k|\leqslant
N\end{subarray}}\bigl{(}\bar{\nabla}_{A_{k}}\bar{\nabla}_{A_{k}}+\bar{\nabla}_{B_{k}}\bar{\nabla}_{B_{k}}\bigr{)}\,\hat{V}(g)=\nu\,\Delta
V\circ g.$ (16)
Here $\tilde{\alpha}$ is an integer which is not necessary equal to
${\alpha}$.
###### Proof.
By the right-invariance of the vector fields
$\bar{\nabla}_{A_{k}}\bar{\nabla}_{A_{k}}\hat{V}$ and
$\bar{\nabla}_{B_{k}}\bar{\nabla}_{B_{k}}\hat{V}$ (Lemma 7), it suffices to
show (16) for $g=e$. We observe that
$\displaystyle(\bar{k},\nabla)\cos(k\cdot\theta)=(\bar{k},\nabla)\sin(k\cdot\theta)=0.$
Then, for $k\in{\mathbb{Z}}_{2}^{+}$, $\theta\in{\mathbb{T}}^{2}$,
$\displaystyle\bar{\nabla}_{A_{k}}\bar{\nabla}_{A_{k}}\hat{V}(e)(\theta)=\frac{1}{|k|^{2{\alpha}+2}}\cos(k\cdot\theta)(\bar{k},\nabla)\bigl{[}\cos(k\cdot\theta)(\bar{k},\nabla)V(\theta)\bigr{]}$
$\displaystyle=\frac{1}{|k|^{2{\alpha}+2}}\cos(k\cdot\theta)^{2}(\bar{k},\nabla)^{2}V(\theta).$
Similarly,
$\bar{\nabla}_{B_{k}}\bar{\nabla}_{B_{k}}\hat{V}(e)(\theta)=\frac{1}{|k|^{2{\alpha}+2}}\sin(k\cdot\theta)^{2}(\bar{k},\nabla)^{2}V(\theta)$.
Hence, for each $k\in{\mathbb{Z}}_{2}^{+}$,
$\displaystyle(\bar{\nabla}_{A_{k}}\bar{\nabla}_{A_{k}}+\bar{\nabla}_{B_{k}}\bar{\nabla}_{B_{k}})\,\hat{V}(e)(\theta)=\frac{1}{|k|^{2{\alpha}+2}}(\bar{k},\nabla)^{2}V(\theta).$
(17)
Note that for each $k\in{\mathbb{Z}}_{2}^{+}$, either $\bar{k}$ or $-\bar{k}$
is in ${\mathbb{Z}}_{2}^{+}$, and
$\displaystyle(\bar{k},\nabla)^{2}+(k,\nabla)^{2}=|k|^{2}\Delta.$
Summation over $k\in{\mathbb{Z}}_{2}^{+}$, $|k|\leqslant N$, in (17), and
coupling the terms numbered by $k$ and $\bar{k}$ (or $-\bar{k}$) gives:
$\displaystyle\sum_{k\in{\mathbb{Z}}_{2}^{+},|k|\leqslant
N}(\bar{\nabla}_{A_{k}}\bar{\nabla}_{A_{k}}+\bar{\nabla}_{B_{k}}\bar{\nabla}_{B_{k}})\hat{V}(e)(\theta)=\frac{1}{2}\sum_{k\in{\mathbb{Z}}_{2}^{+},|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}}}\,\Delta V(\theta).$
Note that
$(\bar{\nabla}_{A_{0}}\bar{\nabla}_{A_{0}}+\bar{\nabla}_{B_{0}}\bar{\nabla}_{B_{0}})\hat{V}(e)(\theta)=\Delta
V(\theta)$. Finally, we obtain:
$\displaystyle\sum_{\begin{subarray}{c}k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},\\\
|k|\leqslant
N\end{subarray}}(\bar{\nabla}_{A_{k}}\bar{\nabla}_{A_{k}}+\bar{\nabla}_{B_{k}}\bar{\nabla}_{B_{k}})\hat{V}(e)(\theta)=\Bigl{(}1+\frac{1}{2}\sum_{k\in{\mathbb{Z}}_{2}^{+},|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}}}\Bigr{)}\Delta V(\theta).$
The lemma is proved. ∎
###### Corollary 1.
Let the function $\varphi:{\mathbb{T}}^{2}\to{\mathbb{R}}^{2}$ be ${\rm
C}^{2}$-smooth. Further let $A_{k}(g)[\varphi\circ g]$ and
$B_{k}(g)[\varphi\circ g]$, $k\in{\mathbb{Z}}^{+}_{2}$, mean the
differentiation of the function $G^{\tilde{\alpha}}\to
L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$, $g\mapsto\varphi\circ g$ along
$A_{k}$ and resp. $B_{k}$. Then for all $g\in G^{\tilde{\alpha}}$,
$\displaystyle\frac{\epsilon^{2}}{2}\sum_{\begin{subarray}{c}k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},\\\
|k|\leqslant
N\end{subarray}}\bigl{(}A_{k}(g)A_{k}(g)+B_{k}(g)B_{k}(g)\bigr{)}\,[\varphi\circ
g]=\nu\,\Delta\varphi\circ g.$ (18)
###### Proof.
The computation that we made in (14) but applied to $\varphi\circ g$ implies
that
$\displaystyle A_{k}(g)[\varphi\circ
g]=\Bigl{[}\frac{1}{|k|^{\alpha+1}}\cos(k\cdot\theta)(\bar{k},\nabla)\varphi(\theta)\Bigr{]}\circ
g.$
Similarly, we compute $B_{k}(g)[\varphi\circ g]$. Now we just have to repeat
the proof of Lemma 10 to come to (18). ∎
###### Lemma 11.
Let $\Phi_{r}$, $r\in[t,T]$, $t\in[0,T)$, be an
$H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$-valued stochastic process
whose trajectories are integrable, and let $\phi_{T}$ be an
$H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$-valued random element so that
both $\Phi_{r}$ and $\phi_{T}$ possess finite expectations. Then there exists
an $\mathcal{F}_{s}$-adapted
$H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})\times\mathcal{L}\bigl{(}E,H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})\bigr{)}$-valued
pair of stochastic processes $(Y_{s},X_{s})$ solving the BSDE
$\displaystyle
Y_{s}=\phi_{T}+\int_{s}^{T}\Phi_{r}\,dr-\int_{s}^{T}X_{r}\,dW_{r}$ (19)
on $[t,T]$. The $Y_{s}$-part of the solution has the representation
$\displaystyle
Y_{s}=\mathbb{E}\,[\,\phi_{T}+\int_{s}^{T}\Phi_{r}\,dr\,|\,\mathcal{F}_{s}]\,,$
(20)
and therefore is unique. The $X_{s}$-part of the solution is unique with
respect to the norm
$\|X_{s}\|^{2}=\int_{t}^{T}\|X_{s}\|_{\mathcal{L}(E,H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2}))}^{2}\,ds$.
The proof of the lemma uses some ideas from [20].
###### Proof.
Representation (20) follows from (19). Let us extend the process $Y_{s}$ to
the entire interval $[0,T]$ by setting $Y_{s}=Y_{t}$ for $s\in[0,t]$, and note
that the extended process $Y_{s}$ is a solution of the SDE
$\displaystyle
Y_{s}=\phi_{T}+\int_{s}^{T}\operatorname{\mathbb{I}}_{[t,T]}\Phi_{r}\,dr-\int_{s}^{T}X_{r}dW_{r}$
on $[0,T]$. Let
$X_{s}\in\mathcal{L}\bigl{(}E,H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})\bigr{)}$,
$s\in[0,T]$, be such that
$\displaystyle\mathbb{E}\,\bigl{[}\,\phi_{T}+\int_{0}^{T}\operatorname{\mathbb{I}}_{[t,T]}\,\Phi_{r}\,dr-
Y_{0}\,|\,\mathcal{F}_{s}\bigr{]}=\int_{0}^{s}X_{r}\,dW_{r}.$ (21)
The process $X_{s}$ exists by the martingale representation theorem. Indeed,
on the right-hand side of (21) we have a Hilbert space valued martingale.
By Theorem 6.6 of [12], each component of the
$H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$-valued martingale on the
right-hand side of (21) can be represented as a sum of real-valued stochastic
integrals with respect to the Brownian motions
$\\{\beta^{A}_{k}(s),\beta^{B}_{k}(s)\\}_{k\in{\mathbb{Z}}^{+}_{2}\cup\\{0\\},|k|\leqslant
N}$. Hence, there exist $\mathcal{F}_{s}$-adapted stochastic processes
$\\{X_{s}^{kA},X_{s}^{kB}\\}_{k\in{\mathbb{Z}}^{+}_{2}\cup\\{0\\},|k|\leqslant
N}$ such that
$\displaystyle\mathbb{E}\bigl{[}\phi_{T}+\int_{0}^{T}\operatorname{\mathbb{I}}_{[t,T]}\Phi_{r}\,dr-
Y_{0}\,|\,\mathcal{F}_{s}\bigr{]}=\sum_{\begin{subarray}{c}k\in{\mathbb{Z}}^{+}_{2}\cup\\{0\\},\\\
|k|\leqslant
N\end{subarray}}\int_{0}^{s}X_{r}^{kA}\,d\beta^{A}_{k}(r)+\int_{0}^{s}X_{r}^{kB}\,d\beta^{B}_{k}(r).$
Let the process $X_{s}$ be defined by (8) via the processes $X_{s}^{kA}$ and
$X_{s}^{kB}$, $k\in{\mathbb{Z}}^{+}_{2}\cup\\{0\\},|k|\leqslant N$. Itô’s
isometry shows that
$\mathbb{E}\int_{0}^{T}\|X_{r}\|^{2}_{\mathcal{L}(E,H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2}))}<\infty$.
Note that for all $s\in[0,t]$,
$\int_{0}^{s}X_{r}dW_{r}=\int_{0}^{t}X_{r}dW_{r}$. This shows that $X_{s}=0$
for almost all $\omega\in\Omega$ and almost all $s\in[0,t]$, and therefore can
be chosen equal to zero on $[0,t]$. Thus, (21) takes the form:
$\displaystyle\mathbb{E}\,\bigl{[}\,\phi_{T}+\int_{t}^{T}\Phi_{r}\,dr-
Y_{t}\,|\,\mathcal{F}_{s}\bigr{]}=\int_{t}^{s}X_{r}\,dW_{r}.$ (22)
It is easy to verify that the pair $(Y_{s},X_{s})$ defined by (20) and (22)
solves BSDE (19). To prove the uniqueness, note that any
$\mathcal{F}_{s}$-adapted solution to (19) takes the form (20), (22).
Moreover, if the processes $X_{s}$ and $X^{\prime}_{s}$ satisfy (22), then
$\displaystyle\int_{t}^{T}\|X_{s}-X^{\prime}_{s}\|_{\mathcal{L}(E,H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2}))}^{2}\,dr=\Bigl{\|}\int_{t}^{T}(X_{s}-X^{\prime}_{s})\,dW_{r}\Bigr{\|}_{H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}^{2}=0.$
∎
###### Proof of Theorem 6.
Let us consider BSDE (7) as an
$L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$-valued SDE, and $\hat{Y}_{s}$ as a
function ${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}\to
L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$. Since for each $s\in[t,T]$,
$y(s,\,\cdot\,)\in H^{{\alpha}+1}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$ and
${\alpha}>2$ by assumption, then
$\hat{Y}_{s}:{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}\to
L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$ is at least ${\rm C}^{2}$-smooth.
Equations (2) show that the function
$\partial_{s}y(\,\cdot\,,\,\cdot\,):\,[t,T]\to
L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$ is continuous since $\nabla p$,
$\Delta y$, and $(y,\nabla y)$ are continuous functions $[t,T]\to
L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$ by Assumption 1. Taking into account
that the diffeomorphisms of ${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$
are volume-preserving, we conclude that for each fixed
$g\in{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$,
$\partial_{s}\hat{Y}_{s}(g):[t,T]\to L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$
is a continuous function. Hence,
$\hat{Y}_{\bullet}:[t,T]\times{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}\to
L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$ is ${\rm C}^{1}$-smooth in
$s\in[t,T]$ and ${\rm C}^{2}$-smooth in
$g\in{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. Itô’s formula is
therefore applicable to $\hat{Y}_{s}(Z^{t,e}_{s})$. Below we use the fact that
$Z^{t,e}_{s}$ is a solution to forward SDE (11) and the identity
$\frac{\partial\hat{Y}_{s}}{\partial s}(Z^{t,e}_{s})=\frac{\partial
y(s,\,\cdot\,)}{\partial s}\circ Z^{t,e}_{s}$. For the latter derivative we
substitute the right-hand side of the first equation of (2). The notation
$\hat{X}(g)[\hat{Y}_{s}(g)]$ (sometimes without square brackets) means
differentiation of the function
$\hat{Y}_{s}:\,{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}\to
L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$ along the right-invariant vector
field $\hat{X}$ on ${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ at the
point $g\in{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. The same argument
as in Remark 3 implies that
$\hat{X}(g)[\hat{Y}_{s}(g)]=\bar{\nabla}_{\hat{X}}\hat{Y}_{s}(g)$. Taking into
account this argument, we obtain:
$\hat{Y}_{s}(Z^{t,e}_{s})-\hat{h}(Z^{t,e}_{T})=-\int_{s}^{T}\partial_{r}\hat{Y}_{r}(Z^{t,e}_{r})\,dr-\int_{s}^{T}dr\,\hat{Y}_{r}(Z_{r}^{t,e})[\hat{Y}_{r}(Z_{r}^{t,e})]\\\
-\int_{s}^{T}dr\,{\textstyle\frac{\epsilon^{2}}{2}}\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},|k|\leqslant
N}\bigl{[}A_{k}(Z_{r}^{t,e})A_{k}(Z_{r}^{t,e})\hat{Y}_{r}(Z_{r}^{t,e})+B_{k}(Z_{r}^{t,e})B_{k}(Z_{r}^{t,e})\hat{Y}_{r}(Z_{r}^{t,e})\bigr{]}\\\
-\int_{s}^{T}\epsilon\,\sigma(Z^{t,e}_{r})\hat{Y}_{r}(Z_{r}^{t,e})\,dW_{r}.$
(23)
Note that
$\displaystyle\hat{Y}_{r}(Z_{r}^{t,e})[\hat{Y}_{r}(Z_{r}^{t,e})]=[(y(r,\,\cdot\,),\nabla)y(r,\,\cdot\,)]\circ
Z^{t,e}_{r}.$
Also, let us observe that
$\displaystyle\frac{\epsilon^{2}}{2}\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},|k|\leqslant
N}\bigl{[}A_{k}(Z_{r}^{t,e})A_{k}(Z_{r}^{t,e})\hat{Y}_{r}(Z_{r}^{t,e})+B_{k}(Z_{r}^{t,e})B_{k}(Z_{r}^{t,e})\hat{Y}_{r}(Z_{r}^{t,e})\bigr{]}$
$\displaystyle=\frac{\epsilon^{2}}{2}\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},|k|\leqslant
N}\bigl{[}\bar{\nabla}_{A_{k}}\bar{\nabla}_{A_{k}}\hat{Y}_{r}(Z^{t,e}_{r})+\bar{\nabla}_{B_{k}}\bar{\nabla}_{B_{k}}\hat{Y}_{r}(Z^{t,e}_{r})\bigr{]}$
$\displaystyle\phantom{\frac{\epsilon^{2}}{2}\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},|k|\leqslant
N}\bigl{[}A_{k}(Z_{r}^{t,e})A_{k}(Z_{r}^{t,e})\hat{Y}_{r}(Z_{r}^{t,e})}=\nu[\Delta
y(s,\,\cdot\,)]\circ Z^{t,e}_{r}$
where the latter equality holds by Lemma 10, and $\epsilon>0$ is chosen so
that
$\frac{\epsilon^{2}}{2}\Bigl{(}1+\frac{1}{2}\sum_{k\in{\mathbb{Z}}_{2}^{+},|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}}}\Bigr{)}=\nu$. Note that the terms
$\bar{\nabla}_{A_{k}}\bar{\nabla}_{A_{k}}\hat{Y}_{r}(Z^{t,e}_{r})$ and
$\bar{\nabla}_{B_{k}}\bar{\nabla}_{B_{k}}\hat{Y}_{r}(Z^{t,e}_{r})$ are
elements of $TG^{{\alpha}-1}$, and therefore are well defined in
$L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$. Continuing (23), we obtain:
$\hat{Y}_{s}^{t}(Z^{t,e}_{s})-\hat{h}(Z^{t,e}_{T})\\\
=\int_{s}^{T}dr\Bigl{[}\hat{V}(r,Z^{t,e}_{r})+[(y(r,\,\cdot\,),\nabla)y(r,\,\cdot\,)]\circ
Z^{t,e}_{r}+\nu[\Delta y(r,\,\cdot\,)]\circ Z^{t,e}_{r}\Bigr{]}\\\
-\int_{s}^{T}[(y(r,\,\cdot\,),\nabla)y(r,\,\cdot\,)]\circ
Z^{t,e}_{r}\,dr-\int_{s}^{T}\nu[\Delta y(r,\,\cdot\,)]\circ Z^{t,e}_{r}\,dr\\\
-\int_{s}^{T}\epsilon\,\sigma(Z^{t,e}_{r})\hat{Y}_{r}(Z_{r}^{t,e})\,dW_{r}=\int_{s}^{T}\hat{V}(r,Z^{t,e}_{r})\,dr-\int_{s}^{T}\epsilon\,\sigma(Z^{t,e}_{r})\hat{Y}_{r}(Z_{r}^{t,e})\,dW_{r}.$
(24)
Thus the pair of stochastic processes
$(\hat{Y}_{s}(Z^{t,e}_{s}),\epsilon\,\sigma(Z^{t,e}_{s})\hat{Y}_{s}(Z_{s}^{t,e}))$
is a solution to BSDE (7) in $L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$. It is
$\mathcal{F}_{s}$-adapted since $Z^{t,e}_{s}$ is $\mathcal{F}_{s}$-adapted. By
Lemma 11, we know that there exists a unique $\mathcal{F}_{s}$-adapted
solution $(Y^{t,e}_{s},X^{t,e}_{s})$ to (7) in
$H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$. Clearly,
$(Y^{t,e}_{s},X^{t,e}_{s})$ is also a unique $\mathcal{F}_{s}$-adapted
solution to (7) in $L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$. Hence,
$Y^{t,e}_{s}=\hat{Y}_{s}(Z^{t,e}_{s})$ and
$\int_{t}^{T}\|X^{t,e}_{s}-\epsilon\,\sigma(Z^{t,e}_{s})\hat{Y}_{s}(Z_{s}^{t,e}))\|^{2}_{\mathcal{L}(E,H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2}))}\,ds=0$,
and therefore the pair of stochastic processes
$\bigl{(}\hat{Y}_{s}(Z^{t,e}_{s}),\epsilon\,\sigma(Z^{t,e}_{s})\hat{Y}_{s}(Z_{s}^{t,e})\bigr{)}$
is a unique $\mathcal{F}_{s}$-adapted solution to BSDE (7) in
$H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$. The theorem is proved. ∎
## 5\. Some identities involving the Navier–Stokes solution
The backward SDE allows us to obtain the representation below for the
Navier–Stokes solution. Also, it easily implies the well-known energy identity
for the Navier–Stokes equations.
### 5.1 Representation of the Navier–Stokes solution
###### Theorem 7.
Let $t\in[0,T]$, and let $Z^{t,e}_{s}$ be the solution to SDE (11) on $[t,T]$
with the initial condition $Z^{t,e}_{t}=e$. Then the following representation
holds for the solution $y(t,\,\cdot\,)$ to (9).
$\displaystyle
y(t,\,\cdot\,)=\mathbb{E}\Bigl{[}\hat{h}(Z_{T}^{t,e})+\int_{t}^{T}\nabla
p(s,\,\cdot\,)\circ Z^{t,e}_{s}ds\Bigr{]}.$
###### Proof.
Note that $\hat{Y}_{t}(Z^{t,e}_{t})=y(t,\,\cdot\,)$, and
$\mathbb{E}[\int_{t}^{T}X_{r}^{t,e}dW_{r}]=0$. Taking the expectation from the
both parts of (7) at time $s=t$ we obtain the above representation. ∎
### 5.2 A simple derivation of the energy identity
Itô’s formula applied to the squared
$L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$-norm of $Y^{t,e}_{s}$ gives:
$\|Y^{t,e}_{s}\|_{L_{2}}^{2}=\|\hat{h}(Z^{t,e}_{T})\|^{2}_{L_{2}}\\\
+2\int_{s}^{T}(Y^{t,e}_{r},\hat{V}(Z^{t,e}_{r}))_{L_{2}}dr-2\int_{s}^{T}(Y^{t,e}_{r},X^{t,e}_{r}dW_{r})_{L_{2}}-\int_{s}^{T}\|X^{t,e}_{s}\|_{L_{2}}^{2}dr.$
(25)
Using representation (15) for the process $X^{t,e}_{s}$ we obtain:
$\displaystyle\|X^{t,e}_{s}\|_{L_{2}}^{2}=\epsilon^{2}\Bigl{[}\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},\,|k|\leqslant
N}\|\nabla_{\bar{A}_{k}}y(s,\,\cdot\,)\|_{L_{2}}^{2}+\|\nabla_{\bar{B}_{k}}y(s,\,\cdot\,)\|^{2}_{L_{2}}\,\Bigr{]}$
$\displaystyle=\epsilon^{2}\Bigl{[}\,\sum_{k\in{\mathbb{Z}}_{2}^{+},\,|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}+2}}\,\|(\bar{k},\nabla
y(s,\,\cdot\,))\|^{2}_{L_{2}}+\|\nabla y(s,\,\cdot\,)\|^{2}_{L_{2}}\,\Bigr{]}$
$\displaystyle=\epsilon^{2}\Bigl{[}\,{\textstyle\frac{1}{2}}\sum_{k\in{\mathbb{Z}}_{2}^{+},\,|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}+2}}\bigl{(}\|(\bar{k},\nabla
y(s,\,\cdot\,))\|^{2}_{L_{2}}+\|(k,\nabla
y(s,\,\cdot\,))\|^{2}_{L_{2}}\bigr{)}+\|\nabla
y(s,\,\cdot\,)\|^{2}_{L_{2}}\Bigr{]}$
$\displaystyle=\epsilon^{2}\Bigl{(}1+\frac{1}{2}\sum_{k\in{\mathbb{Z}}_{2}^{+},\,|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}}}\Bigr{)}\,\|\nabla
y(s,\,\cdot\,)\|^{2}_{L_{2}}=2\,\nu\,\|\nabla y(s,\,\cdot\,)\|^{2}_{L_{2}}.$
Taking the expectation in (25) and using the volume-preserving property of
$Z^{t,e}_{s}$, we obtain:
$\displaystyle\|y(s,\,\cdot\,)\|_{L_{2}}^{2}+2\nu\int_{s}^{T}\|\nabla
y(r,\,\cdot\,)\|_{L_{2}}^{2}dr=\|h\|_{L_{2}}^{2}.$
## 6\. Constructing the solution to the Navier–Stokes equations from a
solution to the FBSDEs
Let us prove now a result which is, in some sense, a converse of Theorem 6. In
this section we consider (5) as a system of forward and backward SDEs in the
Hilbert space $H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$, where
$\alpha\geqslant 3$. As before, let $\hat{V}(s,Z^{t,e}_{s})$ denote $\nabla
p(s,\,\cdot\,)\circ Z^{t,e}_{s}$, and let $\mathcal{F}_{s}$ denote the
filtration $\sigma\\{W_{r},r\in[0,s]\\}$.
###### Theorem 8.
Assume, for an $H^{\alpha+1}$-smooth function $p(s,\,\cdot\,)$, $s\in[0,T]$,
and for any $t\in(0,T)$, the existence of an $\mathcal{F}_{s}$-adapted
solution $(Z^{t,e}_{s},Y^{t,e}_{s},X^{t,e}_{s})$ to (5) on $[t,T]$ such that
the processes $Z^{t,e}_{s}$ and $Y^{t,e}_{s}$ have a.s. continuous
trajectories and such that $Z^{t,e}_{s}$ take values in
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. Then there exists $T_{0}>0$
such that for all $T<T_{0}$ there exists a deterministic function
$y(s,\,\cdot\,)\in T_{e}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ on
$[0,T]$, such that a.s. on $[t,T]$ the relation
$Y^{t,e}_{s}=y(s,\,\cdot\,)\circ Z^{t,e}_{s}$ holds. Moreover, the pair of
functions $(y,p)$ solves the backward Navier–Stokes equations (9) on $[0,T]$.
Lemmas 12–18 below are the steps in the proof of Theorem 8.
###### Lemma 12.
For all $t\in[0,T)$ and for any $\mathcal{F}_{t}$-measurable
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$-valued random variable
$\xi$, the triple of stochastic processes
$\displaystyle(Z_{s}^{t,\xi},Y_{s}^{t,\xi},X_{s}^{t,\xi})=(Z^{t,e}_{s}\circ\xi,Y^{t,e}_{s}\circ\xi,X^{t,e}_{s}\circ\xi)$
(26)
is $\mathcal{F}_{s}$-adapted and solves the FBSDEs
$\begin{cases}Z_{s}^{t,\xi}=\xi+\int_{t}^{s}Y_{r}^{t,\xi}\,dr+\int_{t}^{s}\sigma(Z_{r}^{t,\xi})\,dW_{r}\\\
Y_{s}^{t,\xi}=h(Z_{T}^{t,\xi})+\int_{s}^{T}\hat{V}(r,Z_{r}^{t,\xi})dr-\int_{s}^{T}X_{r}^{t,\xi}\,dW_{r}\end{cases}$
(27)
on the interval $[t,T]$ in the space
$H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$.
###### Proof.
Let us apply the operator $R_{\xi}$ of the right translation to the both sides
of FBSDEs (5). We only have to prove that we are allowed to write $R_{\xi}$
under the signs of both stochastic integrals in (5). Let us prove that it is
true for an $\mathcal{F}_{t}$-measurable stepwise function
$\xi=\sum_{i=1}^{\infty}g_{i}\operatorname{\mathbb{I}}_{A_{i}}$, where
$g_{i}\in{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ and the sets $A_{i}$
are $\mathcal{F}_{t}$-measurable. Indeed, let $s$ and $S$ be such that
$t\leqslant s<S\leqslant T$, and let $\Phi_{r}$ be an
$\mathcal{F}_{r}$-adapted stochastically integrable process. We obtain:
$\int_{s}^{S}\Phi_{r}\,dW_{r}\circ\sum_{i=1}^{\infty}g_{i}\operatorname{\mathbb{I}}_{A_{i}}=\sum_{i=1}^{\infty}\operatorname{\mathbb{I}}_{A_{i}}\int_{s}^{S}\Phi_{r}\circ
g_{i}\,dW_{r}=\sum_{i=1}^{\infty}\int_{s}^{S}\operatorname{\mathbb{I}}_{A_{i}}\Phi_{r}\circ
g_{i}\,dW_{r}\\\
=\int_{s}^{S}\Phi_{r}\circ\sum_{i=1}^{\infty}g_{i}\operatorname{\mathbb{I}}_{A_{i}}\,dW_{r}.$
Next, we find a sequence of $\mathcal{F}_{t}$-measurable stepwise functions
converging to $\xi$ in the space of continuous functions ${\rm
C}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$. This is possible due to the
separability of ${\rm C}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$. Indeed, let us
consider a countable number of disjoint Borel sets $O^{n}_{i}$ covering ${\rm
C}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$, and such that their diameter in the
norm of ${\rm C}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$ is smaller than
$\frac{1}{n}$. Let $A_{i}^{n}=\xi^{-1}(O_{i}^{n})$ and $g_{i}^{n}\in
O_{i}^{n}\cap{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. Define
$\xi_{n}=\sum_{i=1}^{\infty}g_{i}^{n}\operatorname{\mathbb{I}}_{A_{i}^{n}}$.
Then it holds that for all $\omega\in\Omega$, $\|\xi-\xi_{n}\|_{{\rm
C}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}<\frac{1}{n}$. Let $I(\Phi)$ and
$I(\Phi\circ\xi)$ denote $\int_{s}^{S}\Phi_{r}\,dW_{r}$ and resp.
$\int_{s}^{S}\Phi_{r}\circ\xi\,dW_{r}$. We have to prove that a.s.
$I(\Phi)\circ\xi=I(\Phi\circ\xi)$. For this it suffices to prove that
$\displaystyle\lim_{n\to\infty}\mathbb{E}\|I(\Phi)\circ\xi_{n}-I(\Phi)\circ\xi\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}$
$\displaystyle=0,$ (28)
$\displaystyle\lim_{n\to\infty}\mathbb{E}\|I(\Phi\circ\xi_{n})-I(\Phi\circ\xi)\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}$
$\displaystyle=0.$ (29)
Due to the volume-preserving property of $\xi$ and $\xi_{n}$,
$\|I(\Phi)\circ\xi_{n}\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}=\|I(\Phi)\circ\xi\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}=\|I(\Phi)\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}$.
Hence, by Lebesgue’s theorem, in (28) we can pass to the limit under the
expectation sign. Relation (28) holds then by the continuity of $I(\Phi)$ in
$\theta\in{\mathbb{T}}^{2}$. To prove (29) we observe that by Itô’s isometry,
the limit in (29) equals to
$\lim_{n\to\infty}\mathbb{E}\int_{s}^{S}\|\Phi_{r}\circ\xi_{n}-\Phi_{r}\circ\xi\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}dr$.
The same argument that we used to prove (28) implies that we can pass to the
limit under the expectation and the integral signs. Relation (29) follows from
the continuity of $\Phi_{r}$ in $\theta\in{\mathbb{T}}^{2}$.
Hence, $(Z^{t,e}_{s}\circ\xi,Y^{t,e}_{s}\circ\xi,X^{t,e}_{s}\circ\xi)$ is a
solution to (27). This solution is clearly $\mathcal{F}_{s}$-adapted. ∎
Lemmas 13–17 below use some ideas and constructions from [9].
###### Lemma 13.
The map $[0,T]\times{\mathbb{T}}^{2}\to{\mathbb{R}}^{2}$, $(t,\theta)\mapsto
Y^{t,e}_{t}(\theta)$ is deterministic.
###### Proof.
Let us extend the solution $(Z_{s}^{t,e},Y_{s}^{t,e},X_{s}^{t,e})$ to the
interval $[0,t]$ by setting $Z^{t,e}_{s}=e$, $Y^{t,e}_{s}=Y^{t,e}_{t}$,
$X^{t,e}_{s}=0$ for all $s\in[0,t]$. The extended process solves the problem:
$\displaystyle\begin{cases}Z_{s}^{t,e}=e+\int_{0}^{s}\operatorname{\mathbb{I}}_{[t,T]}(r)Y_{r}^{t,e}\,dr+\int_{0}^{s}\operatorname{\mathbb{I}}_{[t,T]}(r)\sigma(Z_{r}^{t,e})\,dW_{r}\\\
Y_{s}^{t,e}=h(Z_{T}^{t,e})+\int_{s}^{T}\operatorname{\mathbb{I}}_{[t,T]}(r)\hat{V}(r,Z_{r}^{t,e})\,dr-\int_{s}^{T}X_{r}^{t,e}\,dW_{r}.\end{cases}$
(30)
The random vector $Y_{0}^{t,e}$ is $\mathcal{F}_{0}$-measurable, and hence is
deterministic by Blumenthal’s zero-one law. Since $Y_{t}^{t,e}=Y_{0}^{t,e}$,
the result follows. ∎
###### Lemma 14.
There exists a constant $T_{0}>0$ such that for $T<T_{0}$ the function
$[0,T]\to H^{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$, $t\mapsto Y^{t,e}_{t}$ is
continuous.
###### Proof.
Let $(Z^{t,e}_{s},Y^{t,e}_{s},X^{t,e}_{s})$ and
$(Z^{t^{\prime}\\!,e}_{s},Y^{t^{\prime}\\!,e}_{s},X^{t^{\prime}\\!,e}_{s})$ be
solutions to (27) which start at the identity $e$ at times $t$ and resp.
$t^{\prime}$, and let $t<t^{\prime}$. These solutions can be regarded as
solutions of (30) if we extend them to the entire interval $[0,T]$ as it was
described in Lemma 13. The application of Itô’s formula to
$\|Y^{t,e}_{s}\|_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}^{2}$ and the
backward SDE of (27) imply that the expectation
$\mathbb{E}\|Y^{t,e}_{s}\|_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}^{2}$ is
bounded. The forward SDE of (30), Gronwall’s lemma, and usual stochastic
integral estimates imply that there exists a constant $K_{1}>0$ such that
$\displaystyle\mathbb{E}\|Z^{t,e}_{s}-Z^{t^{\prime}\\!,e}_{s}\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}<K_{1}\Bigl{[}\int_{0}^{s}\operatorname{\mathbb{I}}_{[t,T]}\mathbb{E}\|Y^{t,e}_{r}-Y^{t^{\prime}\\!,e}_{r}\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}dr+(t^{\prime}-t)\Bigr{]}.$
Let us apply Itô’s formula to
$\|Y^{t,e}_{s}-Y^{t^{\prime}\\!,e}_{s}\|_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}^{2}$
when using the backward SDE of (30). Again, Gronwall’s lemma, usual stochastic
integral estimates and the above estimate for
$\mathbb{E}\|Z^{t,e}_{s}-Z^{t^{\prime}\\!,e}_{s}\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}$
imply that there exists a constant $K_{2}>0$ such that
$\displaystyle\mathbb{E}\|Y^{t,e}_{s}-Y^{t^{\prime}\\!,e}_{s}\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}<K_{2}\Bigl{[}\int_{0}^{T}\mathbb{E}\|Y^{t,e}_{r}-Y^{t^{\prime}\\!,e}_{r}\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}dr+(t^{\prime}-t)\Bigr{]}.$
We take $T_{0}$ smaller than $\frac{1}{K_{2}}$. Then there exists a constant
$K>0$ such that
$\displaystyle\sup_{s\in[0,T]}\mathbb{E}\|Y^{t,e}_{s}-Y^{t^{\prime}\\!,e}_{s}\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}<K(t^{\prime}-t).$
(31)
Evaluating the right-hand side at the point $s=t$, and taking into account
that $Y^{t^{\prime}\\!,e}_{t}=Y^{t^{\prime}\\!,e}_{t^{\prime}}$ we obtain that
$\displaystyle\|Y^{t,e}_{t}-Y^{t^{\prime}\\!,e}_{t^{\prime}}\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}<K(t^{\prime}-t).$
(32)
Differentiating (30) with respect to $\theta$ we obtain the following system
of forward and backward SDEs:
$\displaystyle\begin{cases}\nabla
Z_{s}^{t,e}=I+\int_{0}^{s}\operatorname{\mathbb{I}}_{[t,T]}(r)\nabla
Y_{r}^{t,e}\,dr+\int_{0}^{s}\operatorname{\mathbb{I}}_{[t,T]}(r)\nabla\sigma(Z_{r}^{t,e})\nabla
Z_{r}^{t,e}\,dW_{r}\\\ \nabla Y_{s}^{t,g}=\nabla h(Z_{T}^{t,e})\nabla
Z_{T}^{t,e}+\int_{s}^{T}\operatorname{\mathbb{I}}_{[t,T]}(r)\nabla\hat{V}(r,Z_{r}^{t,g})\nabla
Z_{r}^{t,e}\,dr\\\ \phantom{\nabla Y_{s}^{t,g}=\nabla h(Z_{T}^{t,e})\nabla
Z_{T}^{t,e}}-\int_{s}^{T}\nabla X_{r}^{t,e}\,dW_{r}.\end{cases}$
Again, standard estimates imply the boundedness of $\mathbb{E}\|\nabla
Z^{t,e}_{s}\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}$ and
$\mathbb{E}\|\nabla
Y^{t,e}_{s}\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}$. The same
argument that we used to obtain (32) as well as the estimate for the
$\sup_{s\in[0,T]}\mathbb{E}\|Z^{t,e}_{s}-Z^{t^{\prime}\\!,e}_{s}\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}$,
which easily follows from (31), and the forward SDE imply that there exists a
constant $L>0$ such that for all $t$ and $t^{\prime}$ from the interval
$[0,T]$,
$\displaystyle\|\nabla Y^{t,e}_{t}-\nabla
Y^{t^{\prime}\\!,e}_{t^{\prime}}\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}<L|t^{\prime}-t|.$
(33)
Differentiating (30) the second time and using the same argument once again we
obtain that there exist a constant $M>0$ such that for all $t$ and
$t^{\prime}$ belonging to $[0,T]$,
$\displaystyle\|\nabla\nabla Y^{t,e}_{t}-\nabla\nabla
Y^{t^{\prime}\\!,e}_{t^{\prime}}\|^{2}_{L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})}<M|t^{\prime}-t|.$
(34)
Now (32), (33), and (34) imply the continuity of the map $t\mapsto
Y^{t,e}_{t}$ with respect to the
$H^{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$-topology. ∎
Everywhere below we assume that $T<T_{0}$ where $T_{0}$ is the constant
defined in Lemma 14.
###### Lemma 15.
For every $t\in[0,T)$ and for every $\mathcal{F}_{t}$-measurable random
variable $\xi$, the solution $(Z^{t,\xi}_{s},Y^{t,\xi}_{s},X^{t,\xi}_{s})$ to
(27) is unique on $[t,T]$.
###### Proof.
Let us assume that there exists another solution
$(\tilde{Z}^{t,\xi}_{s},\tilde{Y}^{t,\xi}_{s},\tilde{X}^{t,\xi}_{s})$ to (27)
on $[t,T]$. The same argument as in the proof of Lemma 14 implies the
uniqueness of solution to (27). Specifically, the argument that we applied to
the pair of solutions $(Z^{t,e}_{s},Y^{t,e}_{s},X^{t,e}_{s})$ and
$(Z^{t^{\prime}\\!,e}_{s},Y^{t^{\prime}\\!,e}_{s},X^{t^{\prime}\\!,e}_{s})$
has to be applied to $(Z^{t,\xi}_{s},Y^{t,\xi}_{s},X^{t,\xi}_{s})$ and
$(\tilde{Z}^{t,\xi}_{s},\tilde{Y}^{t,\xi}_{s},\tilde{X}^{t,\xi}_{s})$, and it
has to be taken into account that $t=t^{\prime}$. ∎
###### Lemma 16.
Let the function $y:[0,T]\times{\mathbb{T}}^{2}\to{\mathbb{R}}^{2}$ be defined
by the formula:
$\displaystyle y(t,\theta)=Y^{t,e}_{t}(\theta).$ (35)
Then, for every $t\in[0,T]$, $y(t,\,\cdot\,)$ is $H^{\alpha}$-smooth, and a.s.
$\displaystyle Y_{u}^{t,e}=y(u,\,\cdot\,)\circ Z_{u}^{t,e}.$ (36)
###### Proof.
Note that (26) implies that if $\xi$ is $\mathcal{F}_{t}$-measurable then
$\displaystyle Y^{t,\xi}_{t}=y(t,\,\cdot\,)\circ\xi.$ (37)
Further, for each fixed $u\in[t,T]$, $(Z_{s}^{t,e},Y_{s}^{t,e},X_{s}^{t,e})$
is a solution of the following problem on $[u,T]$:
$\begin{cases}Z_{s}^{t,e}=Z_{u}^{t,e}+\int_{u}^{s}Y^{t,e}_{r}dr+\int_{u}^{s}\sigma(Z^{t,e}_{r})dW_{r}\\\
Y_{s}^{t,e}=h(Z_{T}^{t,e})+\int_{s}^{T}\hat{V}(r,Z_{r}^{t,e})dr-\int_{s}^{T}X^{t,e}_{r}dW_{r}.\end{cases}$
By uniqueness of solution, it holds that $Y_{s}^{t,e}=Y_{s}^{u,Z_{u}^{t,e}}$
a.s. on $[u,T]$. Next, by (37), we obtain that
$Y_{u}^{u,Z_{u}^{t,e}}=y(u,\,\cdot\,)\circ Z_{u}^{t,e}$. This implies that
there exists a set $\Omega_{u}$ (which depends on $u$) of full
$\mathbb{P}$-measure such that (36) holds everywhere on $\Omega_{u}$. Clearly,
one can find a set $\Omega_{\mathbb{Q}}$, $\mathbb{P}(\Omega_{\mathbb{Q}})=1$,
such that (36) holds on $\Omega_{\mathbb{Q}}$ for all rational $u\in[t,T]$.
But the trajectories of $Z^{t,e}_{s}$ and $Y^{t,e}_{s}$ are a.s. continuous.
Furthermore, Lemma 14 implies the continuity of $y(t,\,\cdot\,)$ in $t$ with
respect to (at least) the $L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$-topology.
Therefore, (36) holds a.s. with respect to the
$L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$-topology. Since both sides of (36)
are continuous in $\theta\in{\mathbb{T}}^{2}$ it also holds a.s. for all
$\theta\in{\mathbb{T}}^{2}$. ∎
###### Lemma 17.
The function $y$ defined by formula (35) is $C^{1}$-smooth in $t\in[0,T]$.
###### Proof.
Let $\delta>0$. We obtain:
$\displaystyle
y(t+\delta,\,\cdot\,)-y(t,\,\cdot\,)=Y^{t+\delta,e}_{t+\delta}-Y^{t,e}_{t}=Y^{t+\delta,e}_{t+\delta}-Y^{t,e}_{t+\delta}+Y^{t,e}_{t+\delta}-Y^{t,e}_{t}.$
Let $\hat{Y}_{s}$ be the right-invariant vector field on $G^{\alpha}$
generated by $y(s,\,\cdot\,)$. Lemma 16 implies that a.s.
$\displaystyle Y^{t,e}_{t+\delta}=\hat{Y}_{t+\delta}(Z^{t,e}_{t+\delta}).$
Thus we obtain that a.s.
$\displaystyle
y(t+\delta,\,\cdot\,)-y(t,\,\cdot\,)=\bigl{(}\hat{Y}_{t+\delta}(e)-\hat{Y}_{t+\delta}(Z^{t,e}_{t+\delta})\bigr{)}+(Y^{t,e}_{t+\delta}-Y^{t,e}_{t}).$
We use the backward SDE for the second difference and apply Itô’s formula to
the first difference when considering $\hat{Y}_{t+\delta}$ as a $C^{2}$-smooth
function ${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}\to
L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$. We obtain:
$\displaystyle\hat{Y}_{t+\delta}(Z^{t,e}_{t+\delta})-\hat{Y}_{t+\delta}(e)$
$\displaystyle=\int_{t}^{t+\delta}dr\,\hat{Y}^{t,e}_{r}(Z^{t,e}_{r})[\hat{Y}_{t+\delta}(Z^{t,e}_{r})]+\int_{t}^{t+\delta}\epsilon\,\sigma(Z^{t,e}_{r})\,\hat{Y}_{t+\delta}(Z^{t,e}_{r})\,dW_{r}$
$\displaystyle+\int_{t}^{t+\delta}dr\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}[A_{k}(Z^{t,e}_{r})A_{k}(Z^{t,e}_{r})+B_{k}(Z^{t,e}_{r})B_{k}(Z^{t,e}_{r})]\,\hat{Y}_{t+\delta}(Z^{t,e}_{r}).$
The same argument as in Theorem 6 implies:
$\hat{Y}_{t+\delta}(Z^{t,e}_{t+\delta})-\hat{Y}_{t+\delta}(e)=\int_{t}^{t+\delta}dr\,\nabla_{y(r,\,\cdot\,)}\,y(t+\delta,\,\cdot\,)\circ
Z^{t,e}_{r}\\\ +\int_{t}^{t+\delta}dr\,\nu\,\Delta\,y(t+\delta,\,\cdot\,)\circ
Z^{t,e}_{r}+\int_{t}^{t+\delta}\epsilon\,\sigma(Z^{t,e}_{r})\,\hat{Y}_{t+\delta}(Z^{t,e}_{r})\,dW_{r}.$
Further we have:
$\displaystyle Y^{t,e}_{t}-Y^{t,e}_{t+\delta}=\int_{t}^{t+\delta}dr\,\nabla
p(r,\,\cdot\,)\circ Z^{t,e}_{r}-\int_{t}^{t+\delta}X^{t,e}_{r}\,dW_{r}.$
Finally we obtain that
$\frac{1}{\delta}\,\bigl{(}y(t+\delta,\,\cdot\,)-y(t,\,\cdot\,)\bigr{)}=-\frac{1}{\delta}\,\mathbb{E}\Bigl{[}\int_{t}^{t+\delta}dr\,[\,(y(r,\,\cdot\,),\nabla)\,y(t+\delta,\,\cdot\,)\\\
+\nu\,\Delta\,y(t+\delta,\,\cdot\,)+\nabla p(r,\,\cdot\,)]\circ
Z^{t,e}_{r}\Bigr{]}.$ (38)
Note that $Z^{t,e}_{r}$, $\nabla p(r,\,\cdot\,)$, and
$(y(r,\,\cdot\,),\nabla)\,y(t+\delta,\,\cdot\,)\circ Z^{t,e}_{r}$ are
continuous in $r$ a.s. with respect to the
$L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$-topology. By Lemma 14,
$\nabla\,y(t,\,\cdot\,)$ and $\Delta\,y(t,\,\cdot\,)$ are continuous in $t$
with respect to the $L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$-topology.
Formula (38) and the fact that $Z^{t,e}_{t}=e$ imply that in the
$L_{2}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$-topology
$\displaystyle\partial_{t}y(t,\,\cdot\,)=-[\nabla_{y(t,\,\cdot\,)}\,y(t,\,\cdot\,)+\nu\,\Delta\,y(t,\,\cdot\,)+\nabla
p(t,\,\cdot\,)].$ (39)
Since the right-hand side of (39) is an $H^{\alpha-2}$-map, so is the left-
hand side. This implies that $\partial_{t}y(t,\,\cdot\,)$ is continuous in
$\theta\in{\mathbb{T}}^{2}$. Relation (39) is obtained so far for the right
derivative of $y(t,\theta)$ with respect to $t$. Note that the right-hand side
of (39) is continuous in $t$ which implies that the right derivative
$\partial_{t}y(t,\theta)$ is continuous in $t$ on $[0,T)$. Hence, it is
uniformly continuous on every compact subinterval of $[0,T)$. This implies the
existence of the left derivative of $y(t,\theta)$ in $t$, and therefore, the
existence of the continuous derivative $\partial_{t}y(t,\theta)$ everywhere on
$[0,T]$. ∎
###### Lemma 18.
For every $t\in[0,T]$, the function
$y(t,\,\cdot\,):{\mathbb{T}}^{2}\to{\mathbb{R}}^{2}$ is divergence-free.
Moreover, the pair $(y,p)$ verifies the backward Navier–Stokes equations.
###### Proof.
Fix a $t>0$, and consider the
$T_{e}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$-valued curve
$\gamma_{\zeta}=\mathbb{E}\,[{\exp}^{-1}Z^{t,e}_{\zeta}]$, $\zeta\geqslant t$,
in a neigborhood of the origin of
$T_{e}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. The forward SDE of
(27) can be represented as an SDE on $G^{\alpha}$:
$\displaystyle\begin{cases}dZ^{t,e}_{s}=\exp\\{\hat{Y}_{s}(Z^{t,e}_{s})\,ds+\sigma(Z^{t,e}_{s})\,dW_{s}\\},\\\
Z^{t,e}_{t}=e,\end{cases}$
where $\hat{Y}_{s}$ is the right-invariant vector field on $G^{\alpha}$
generated by $y(s,\,\cdot\,)$. This implies that
$\displaystyle\frac{\partial}{\partial\zeta}\gamma_{\zeta}\Bigl{|}_{\zeta=t}=y(t,\,\cdot\,),$
and therefore $y(t,\,\cdot\,)\in
T_{e}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. Next, the backward SDE
of (27) implies that $Y^{t,e}_{T}=h(Z^{t,e}_{T})$. This and relation (36)
imply that $y(T,\,\cdot\,)=h$. Since we already obtained (39) in Lemma 17 the
proof of the lemma is now complete. ∎
## 7\. The backward SDE as an SDE on a tangent bundle
Let $(Z^{t,e}_{s},Y^{t,e}_{s},X^{t,e}_{s})$ be a solution to FBSDEs (5). We
will show that the backward SDE can be represented as an SDE on the tangent
bundle $T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ as well as an SDE on
$TG^{\alpha}$. We will construct a backward SDE in the Dalecky–Belopolskaya
form (see [4]) and show that the process $Y^{t,e}_{s}$ is its unique solution.
### 7.1 The representation of the backward SDE on
$T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$
Let $y(s,\,\cdot\,)$, $s\in[t,T]$, be the solution to the backward
Navier–Stokes equations (9). Let $\hat{Y}_{s}$ be the right-invariant vector
field on ${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ generated by
$y(s,\,\cdot\,)$. The connection map on the manifold
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ generates the connection map
on the manifold $T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ as it was
shown in [4], p. 58 (see also [11]). As before, we consider the Levi-Civita
connection of the weak Riemannian metric (3) on
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. Let $\overline{\exp}$
denote the exponential map of the generated connection on
$T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. More precisely,
$\overline{\exp}$ is given as follows:
$\displaystyle\overline{\exp}_{\left(\begin{subarray}{c}x\\\
a\end{subarray}\right)}\left(\begin{matrix}\alpha\\\
\beta\end{matrix}\right)=\left(\begin{matrix}\gamma_{\alpha}(1)\\\
\eta_{\beta}(1)\end{matrix}\right)$
where $\left(\begin{matrix}\gamma_{\alpha}(t)\\\
\eta_{\beta}(t)\end{matrix}\right)$ is the geodesic curve on
$T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ with the initial data
$\gamma^{\prime}_{\alpha}(0)=\alpha$, $\eta_{\beta}^{\prime}(0)=\beta$,
$\gamma_{\alpha}(0)=x$, $\eta_{\beta}(0)=a$. Let the vector fields $A_{k}^{\sc
H}$ and $B_{k}^{\sc H}$ be the horizontal lifts of $A_{k}$ and $B_{k}$ onto
$TT{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. Further let
$\partial_{s}\hat{Y}_{s}^{\ell}$ be the vertical lift of
$\partial_{s}\hat{Y}_{s}$ onto
$TT{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. Let us consider the
backward SDE on $T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$:
$\displaystyle\begin{split}dY^{t,e}_{s}&=\overline{\exp}_{Y^{t,e}_{s}}\Bigl{\\{}\partial_{s}\hat{Y}_{s}^{\ell}(Y^{t,e}_{s})ds+{\mathrm{S}}(Y^{t,e}_{s})ds\\\
&+\epsilon\,\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},|k|\leqslant
N}\bigl{[}A_{k}^{\sc H}(Y^{t,e}_{s})\otimes e^{A}_{k}+B_{k}^{\sc
H}(Y^{t,e}_{s})\otimes e^{A}_{k}\bigr{]}\,dW_{s}\Bigr{\\}},\\\
Y^{t,e}_{T}&=\hat{h}(Z^{t,e}_{T})\end{split}$ (40)
where $\mathrm{S}$ is the geodesic spray of the Levi-Civita connection of the
weak Riemannian metric (3) on
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ (see [15] or [16]), and
$Z^{t,e}_{s}$, $s\in[t,T]$, is the solution to (11) on
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ with the initial condition
$Z^{t,e}_{t}=e$.
###### Theorem 9.
There exists a solution to (40) on $[t,T]$. Moreover, if
$\partial_{s}y(s,\,\cdot\,)\in H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$,
then this solution is unique and coincides with the $Y^{t,e}_{s}$-part of the
unique $\mathcal{F}_{s}$-adapted solution $(Y^{t,e}_{s},X^{t,e}_{s})$ to (7).
###### Proof.
From the proof of Theorem 6 we know that the pair of stochastic processes
$(\hat{Y}_{s}(Z^{t,e}_{s}),\epsilon\,\sigma(Z^{t,e}_{s})\hat{Y}_{s}(Z^{t,e}_{s}))$
is the unique $\mathcal{F}_{s}$-adapted solution to (7) in
$H^{\alpha}({\mathbb{T}}^{2},{\mathbb{R}}^{2})$. Let us prove that
$\hat{Y}_{s}(Z^{t,e}_{s})$ is a strong solution to (40). First we describe a
system of local coordinates
$(g^{kA},X^{kA},g^{kB},X^{kB})_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}$ in a
neighborhood $U_{e}g\times
T_{e}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ of the point
$\hat{X}(g)\in T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ where
$U_{e}\subset{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ is the canonical
chart. The vector
$\bar{g}=(g^{kA},g^{kB})_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}$ is the vector
of normal coordinates in the neighborhood $U_{e}g$,
$g\in{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. The vector
$\bar{X}=(X^{kA},X^{kB})_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}$ represents the
coordinates of the decomposition of the vector $\hat{X}(g)\in
T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ in the basis
$\\{A_{k},B_{k}\\}_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}$:
$\hat{X}(g)=\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}(X^{kA}A_{k}(g)+X^{kB}B_{k}(g))$.
Let $f$ be a smooth function on
$T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$, and let
$\tilde{f}(\bar{X},\bar{g})=f(\hat{X}(g))$, where $\hat{X}(g)\in
T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. Let $\tau$ be the exit time
of the process $Z^{t,e}_{r}$ from the neighborhood $U_{e}Z^{t,e}_{s}$. We will
compute the difference $f(Y^{t,e}_{s})-f(Y^{t,e}_{\tau})$ using Itô’s formula.
Let
$(\bar{Z}_{r},\bar{Y}_{r})=(Z^{kA}_{r},Z^{kB}_{r},Y^{kA}_{r},Y^{kB}_{r})_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\}}$
be the vector of local coordinates of the process $\hat{Y}_{r}(Z^{t,e}_{r})$
on $[s,\tau]$. Using SDE (40), we obtain:
$f(Y^{t,e}_{s})-f(Y^{t,e}_{\tau})=-\sum_{k\in{\mathbb{Z}}^{+}_{2}\cup\\{0\\}}\int_{s}^{\tau}\Bigl{[}(Y^{kA}_{r})^{\prime}\frac{\partial\tilde{f}(\bar{Y}_{r},\bar{Z}_{r})}{\partial
Y^{kA}_{r}}+(Y^{kB}_{r})^{\prime}\frac{\partial\tilde{f}(\bar{Y}_{r},\bar{Z}_{r})}{\partial
Y^{kB}_{r}}\\\
+Y^{kA}_{r}\frac{\partial\tilde{f}(\bar{Y}_{r},\bar{Z}_{r})}{\partial
Z^{kA}_{r}}+Y^{kB}_{r}\frac{\partial\tilde{f}(\bar{Y}_{r},\bar{Z}_{r})}{\partial
Z^{kB}_{r}}+\frac{\epsilon^{2}}{2}\delta_{k}\Bigl{(}\frac{\partial^{2}}{\partial(Z^{kA}_{r})^{2}}+\frac{\partial^{2}}{\partial(Z^{kB}_{r})^{2}}\Bigr{)}\tilde{f}(\bar{Y}_{r},\bar{Z}_{r})\Bigr{]}dr\\\
-\epsilon\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},|k|\leqslant
N}\int_{s}^{\tau}\Bigl{[}\frac{\partial\tilde{f}(\bar{Y}_{r},\bar{Z}_{r})}{\partial
Z^{kA}_{r}}\otimes
e^{A}_{k}+\frac{\partial\tilde{f}(\bar{Y}_{r},\bar{Z}_{r})}{\partial
Z^{kB}_{r}}\otimes e^{A}_{k}\Bigr{]}\,dW_{r}$ (41)
where $\delta_{k}=1$ if $|k|\leqslant N$, and $\delta_{k}=0$ otherwise. Since
$f$ is a smooth function on
$T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$, all its restrictions to
the tangent spaces of ${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ are
smooth. Hence, one can talk about derivatives of $f$ restricted to a tangent
space along the vectors of this tangent space. Namely, the following relation
holds:
$\displaystyle\frac{\partial\tilde{f}(\bar{Y}_{r},\bar{Z}_{r})}{\partial
Y^{kA}_{r}}=f^{\prime}(\hat{Y}_{r}(Z^{t,e}_{r}))A_{k}(Z^{t,e}_{r}).$
Note that the differentiation of $\tilde{f}$ with respect to $Z^{kA}_{r}$ and
$Z^{kB}_{r}$ can be regarded as the differentiation of the composite function
$f\circ\hat{Y}_{r}$ along the vectors $A_{k}$ and $B_{k}$. Namely,
$\frac{\partial\tilde{f}(\bar{Y}_{r},\bar{Z}_{r})}{\partial
Z^{kA}_{r}}=A_{k}(Z^{t,e}_{r})[(f\circ\hat{Y}_{r})(Z^{t,e}_{r})]$. This
implies:
$f(Y^{t,e}_{s})-f(\hat{h}(Z^{t,e}_{T}))=-\int_{s}^{T}\,dr\,\bigl{[}\,\partial_{r}(f\circ\hat{Y}_{r})(Z^{t,e}_{r})+\hat{Y}_{r}(Z^{t,e}_{r})(f\circ\hat{Y}_{r})(Z^{t,e}_{r})\\\
+\frac{\epsilon^{2}}{2}\sum_{k\in{\mathbb{Z}}^{+}_{2}\cup\\{0\\},|k|\leqslant
N}\bigl{(}A_{k}(Z^{t,e}_{r})A_{k}(Z^{t,e}_{r})+B_{k}(Z^{t,e}_{r})B_{k}(Z^{t,e}_{r})\bigr{)}(f\circ\hat{Y}_{r})(Z^{t,e}_{r})\,\bigr{]}\\\
-\epsilon\sum_{k\in{\mathbb{Z}}^{+}_{2}\cup\\{0\\},|k|\leqslant
N}\int_{s}^{T}[A_{k}(Z^{t,e}_{r})(f\circ\hat{Y}_{r})(Z^{t,e}_{r})\otimes
e^{A}_{k}+B_{k}(Z^{t,e}_{r})(f\circ\hat{Y}_{r})(Z^{t,e}_{r})\otimes
e^{B}_{k}]dW_{r}.$ (42)
We extended the integration to the entire interval $[s,T]$ since the local
coordinates no longer appear under the integral signs. This is also possible
since (41) holds also with respect to the local coordinates in the
neighborhood $U_{e}Z^{t,e}_{\tau}$ and a new exit time $\tau_{1}$. The same
argument can be repeated with respect to the local coordinates in the
neighborhood $U_{e}Z^{t,e}_{\tau_{1}}$, etc. Let us consider now
$f\circ\hat{Y}_{s}$ as a time-dependent function of
$g\in{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. Applying Itô’s formula
to $(f\circ\hat{Y}_{s})(Z^{t,e}_{s})$ on the interval $[s,T]$ and using SDE
(11) on ${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$, we obtain exactly
the above identity. This proves that $Y^{t,e}_{s}=\hat{Y}_{s}(Z^{t,e}_{s})$ is
a strong solution to (40) on
$T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. By results of [15],
$\partial_{s}\hat{Y}_{s}^{\ell}$ is ${\rm C}^{1}$-smooth. Moreover
$\mathrm{S}$, $A_{k}^{\sc H}$ and $B_{k}^{\sc H}$, $k\in{\mathbb{Z}}_{2}^{+}$,
are ${\rm C}^{\infty}$-smooth. Again, by results of [15], the solution of BSDE
(40) on $T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ is unique. ∎
### 7.2 The representation of the backward SDE on $TG^{\alpha}$
Applying Proposition 1.3 (p. 146) of [4] (see also [16], p. 64) to the
manifolds $T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ and $TG^{\alpha}$
and the identical imbedding $\imath_{\scriptscriptstyle
V}:\,T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}\to TG^{\alpha}$, we
obtain that the process $\imath_{\scriptscriptstyle
V}\bigl{(}\hat{Y}_{s}(Z^{t,e}_{s})\bigr{)}=\hat{Y}_{s}(Z^{t,e}_{s})$ solves
the following backward SDE on $TG^{\alpha}$:
$\displaystyle\begin{split}dY^{t,e}_{s}&=\bar{\exp}_{Y^{t,e}_{s}}\Bigl{\\{}\partial_{s}\hat{Y}_{s}^{\bar{\ell}}(Y^{t,e}_{s})ds+\bar{\mathrm{S}}(Y^{t,e}_{s})ds\\\
&+\epsilon\,\sum_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},|k|\leqslant
N}\bigl{[}A_{k}^{\bar{\sc H}}(Y^{t,e}_{s})\otimes e^{A}_{k}+B_{k}^{\bar{\sc
H}}(Y^{t,e}_{s})\otimes e^{A}_{k}\bigr{]}\,dW_{s}\Bigr{\\}},\\\
Y^{t,e}_{T}&=\hat{h}(Z^{t,e}_{T})\end{split}$ (43)
where $\bar{\mathrm{S}}$ is the geodesic spray of the Levi-Civita connection
of the weak Riemannian metric on $G^{\alpha}$,
$\partial_{s}\hat{Y}_{s}^{\bar{\ell}}$ denotes the vertical lift of
$\partial_{s}\hat{Y}_{s}$ onto $TTG^{\alpha}$, $A_{k}^{\bar{\sc H}}$ and
$B_{k}^{\bar{\sc H}}$ denote the horizontal lifts of $A_{k}$ and $B_{k}$ onto
$TTG^{\alpha}$, the process $Z^{t,e}_{s}$, $s\in[t,T]$, is the solution to
(11) on $G^{\alpha}$ with the initial condition $Z^{t,e}_{t}=e$. The
exponential map $\bar{\exp}$ on $TTG^{\alpha}$ is defined similarly to the map
$\overline{\exp}$ on $TT{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$.
Namely, the Levi-Civita connection of the weak Riemannian metric on
$G^{\alpha}$ generates a connection on $TG^{\alpha}$. The latter gives rise to
the exponential map $\bar{\exp}$ on $TTG^{\alpha}$ as it was described in
Paragraph 7.1. We actually have obtained the following theorem.
###### Theorem 10.
Backward SDE (43) has a unique strong solution. Moreover, this solution
coincides with the unique strong solution to BSDE (40) on
$T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$, and with the
$Y^{t,e}_{s}$-part of the unique $\mathcal{F}_{s}$-adapted solution
$(Y^{t,e}_{s},X^{t,e}_{s})$ to (7).
###### Proof.
We have already shown that the process $\hat{Y}_{s}(Z^{t,e}_{s})$ solves BSDE
(43). The uniqueness of solution can be proved in exactly the same way as the
uniqueness of solution to (40) on
$T{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ (see the proof of Theorem
9). ∎
## 8\. Appendix
### 8.1 Geometry of the group of volume-preserving diffeomorphisms of the
$n$-dimensional torus
Let $\mathbb{T}^{n}=\underbrace{S^{1}\times\dots\times S^{1}}_{n}$ denote the
$n$-dimensional torus. Let us describe a basis of the tangent space
$T_{e}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ of the group
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ of volume-preserving
diffeomorphisms of $\mathbb{T}^{n}$. We introduce the following notation:
$\displaystyle{\mathbb{Z}}_{n}^{+}=\\{(k_{1},k_{2},\ldots,k_{n})\in{\mathbb{Z}}^{n}:k_{1}>0\;\text{or}\;k_{1}=\cdots=k_{i-1}=0,\,k_{i}>0,$
$\displaystyle\phantom{{\mathbb{Z}}_{n}^{+}=(k_{1},k_{2},\ldots,k_{n})\in{\mathbb{Z}}^{n}:k_{1}>0k_{1}=\cdots=k_{i-1}=0,}i=2,\dots,n\\};$
$\displaystyle
k=(k_{1},\ldots,k_{n})\in{\mathbb{Z}}_{n}^{+},\quad|k|=\sqrt{\sum_{i=1}^{n}k_{i}^{2}},\quad
k\cdot\theta=\sum_{i=1}^{n}k_{i}\theta_{i}\,,$
$\displaystyle\theta=(\theta_{1},\ldots,\theta_{n})\in\mathbb{T}^{n},\;\nabla=\Bigl{(}\frac{\partial}{\partial\theta_{1}},\frac{\partial}{\partial\theta_{2}},\dots,\frac{\partial}{\partial\theta_{n}}\Bigr{)}.$
For every $k\in{\mathbb{Z}}_{n}^{+}$, $(\bar{k}^{1},\ldots,\bar{k}^{n-1})$
denotes an orthogonal system of vectors of length $|k|$ which is also
orthogonal to $k$. Introduce the vector fields on $\mathbb{T}^{n}$:
$\displaystyle\bar{A}^{i}_{k}=\frac{1}{|k|^{{\alpha}+1}}\cos(k\cdot\theta)\,\bar{k}^{i},\;\bar{B}^{i}_{k}=\frac{1}{|k|^{{\alpha}+1}}\sin(k\cdot\theta)\,\bar{k}^{i},\;i=1,\ldots,n-1,\;k\in{\mathbb{Z}}_{n}^{+},$
and the constant vector fields $\bar{A}^{i}_{0}$, $i=1,\ldots,n$, whose $i$th
coordinate is $1$ and the other coordinates are $0$. Let
$A^{i}_{k},B^{i}_{k}$, $i=1,\ldots,n-1$, $k\in{\mathbb{Z}}_{n}^{+}$, denote
the right-invariant vector fields on
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ generated by
$\bar{A}^{i}_{k},\bar{B}^{i}_{k}$, $i=1,\ldots,n-1$,
$k\in{\mathbb{Z}}_{n}^{+}$, respectively, and let $A^{i}_{0}=\bar{A}^{i}_{0}$,
$i=1,\ldots,n$, stand for constant vector fields on
${G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. The following lemma is an
analog of Lemma 6.
###### Lemma 19.
The vectors $A^{i}_{k}(g)$, $B^{i}_{k}(g)$, $k\in{\mathbb{Z}}_{n}^{+}$,
$i=1,\ldots,n-1$, $g\in{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$,
$A^{i}_{0}$, $i=1,\ldots,n$, form an orthogonal basis of the tangent space
$T_{g}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$ with respect to both
the weak and the strong inner products in
$T_{g}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$. In particular, the
vectors $\bar{A}^{i}_{k}$, $\bar{B}^{i}_{k}$, $k\in{\mathbb{Z}}_{n}^{+}$,
$i=1,\ldots,n-1$, $\bar{A}^{i}_{0}$, $i=1,\ldots,n$, form an orthogonal basis
of the tangent space $T_{e}{G^{\,{\alpha}}_{\scriptscriptstyle\mathrm{V}}}$.
Moreover, the weak and the strong norms of the basis vectors are bounded by
the same constant.
The other lemmas of Section 2 hold in the $n$-dimensional case, with respect
to the system $A^{i}_{k}$, $B^{i}_{k}$, $k\in{\mathbb{Z}}_{n}^{+}$,
$i=1,\ldots,n-1$, $A^{i}_{0}$, $i=1,\ldots,n$, without changes. The index
${\alpha}$ of the Sobolev space $H^{\alpha}$ is an integer bigger than
$\frac{n}{2}+1$.
### 8.2 The Laplacian of a right-invariant vector field on
$G^{\alpha}(\mathbb{T}^{n})$
One of the most important steps in the proof of Theorems 6 and 8 is Lemma 10,
i.e. the computation of the Laplacian of a right-invariant vector field on
$G^{\alpha}$ with respect to the subsystem
$\\{A_{k},B_{k}\\}_{k\in{\mathbb{Z}}_{2}^{+}\cup\\{0\\},|k|\leqslant N}$ where
$N$ can be fixed arbitrary. Below we prove an $n$-dimensional analog of this
lemma.
###### Lemma 20.
Let $\hat{V}$ be the right-invariant vector field on
$G^{\tilde{\alpha}}(\mathbb{T}^{n})$ generated by an
$H^{\tilde{\alpha}+2}$-vector field $V$ on $\mathbb{T}^{n}$. Further let
$\epsilon>0$ be such that
$\displaystyle\frac{\epsilon^{2}}{2}\Bigl{(}1+\frac{n-1}{n}\sum_{k\in{\mathbb{Z}}^{+}_{n},|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}}}\Bigr{)}=\nu.$
Then for all $g\in G^{\tilde{\alpha}}$,
$\displaystyle\frac{\epsilon^{2}}{2}\,\Bigl{[}\sum_{k\in{\mathbb{Z}}_{n}^{+},|k|\leqslant
N}\sum_{i=1}^{n-1}\bigl{(}\bar{\nabla}_{A_{k}^{i}}\bar{\nabla}_{A_{k}^{i}}+\bar{\nabla}_{B_{k}^{i}}\bar{\nabla}_{B_{k}^{i}}\bigr{)}+\sum_{i=1}^{n}\bar{\nabla}_{A_{0}^{i}}\bar{\nabla}_{A_{0}^{i}}\Bigr{]}\,\hat{V}(g)=\nu\,\Delta
V\circ g.$
###### Proof.
As it was mentioned in the proof of Lemma 7, it suffices to consider the case
$g=e$. We observe that for all $i=1,\ldots,n-1$,
$\displaystyle(\bar{k}^{i},\nabla)\cos(k\cdot\theta)=-\sin(k\cdot\theta)(\bar{k}^{i},k)=0.$
Similarly, $(\bar{k}^{i},\nabla)\sin(k\cdot\theta)=0$. Then, for
$k\in{\mathbb{Z}}_{n}^{+}$, $\theta\in\mathbb{T}^{n}$,
$\displaystyle\sum_{i=1}^{n-1}\bar{\nabla}_{A_{k}^{i}}\bar{\nabla}_{A_{k}^{i}}\hat{V}(e)(\theta)=\frac{1}{|k|^{2{\alpha}+2}}\sum_{i=1}^{n-1}\cos(k\cdot\theta)(\bar{k}^{i},\nabla)\bigl{[}\cos(k\cdot\theta)(\bar{k}^{i},\nabla)V(\theta)\bigr{]}$
$\displaystyle=\frac{1}{|k|^{2{\alpha}+2}}\cos(k\cdot\theta)^{2}\sum_{i=1}^{n-1}(\bar{k}^{i},\nabla)^{2}V(\theta)=\frac{1}{|k|^{2{\alpha}+2}}\cos(k\cdot\theta)^{2}(|k|^{2}\Delta-(k,\nabla)^{2})V(\theta).$
The latter equality holds by the identity
$\sum_{i=1}^{n-1}(\bar{k}^{i},\nabla)^{2}+(k,\nabla)^{2}=|k|^{2}\Delta$ that
follows, in turn, from the fact that the system
$\bigl{\\{}\frac{\bar{k}^{i}}{|k|},\frac{k}{|k|}\bigr{\\}}$, $i=1,\ldots,n-1$,
forms an orthonormal basis of ${\mathbb{R}}^{n}$. Similarly,
$\displaystyle\sum_{i=1}^{n-1}\bar{\nabla}_{B_{k}^{i}}\bar{\nabla}_{B_{k}^{i}}\hat{V}(e)(\theta)=\frac{1}{|k|^{2{\alpha}+2}}\sin(k\cdot\theta)^{2}(|k|^{2}\Delta-(k,\nabla)^{2})V(\theta).$
Hence, for each $k\in{\mathbb{Z}}_{n}^{+}$,
$\displaystyle\sum_{i=1}^{n-1}(\bar{\nabla}_{A_{k}^{i}}\bar{\nabla}_{A_{k}^{i}}+\bar{\nabla}_{B_{k}^{i}}\bar{\nabla}_{B_{k}^{i}})\hat{V}(e)(\theta)=\frac{1}{|k|^{2{\alpha}+2}}(|k|^{2}\Delta-(k,\nabla)^{2})V(\theta).$
(44)
Further we have:
$\sum_{k\in{\mathbb{Z}}_{n}^{+},|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}+2}}(k,\nabla)^{2}=\frac{1}{2}\sum_{k\in{\mathbb{Z}}_{n},|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}+2}}(k,\nabla)^{2}\\\
=\frac{1}{2}\sum_{k\in{\mathbb{Z}}_{n},|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}+2}}\sum_{i=1}^{n}k_{i}^{2}\partial_{i}^{2}+\sum_{k\in{\mathbb{Z}}_{n},|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}+2}}\sum_{i\neq j}k_{i}k_{j}\partial_{i}\partial_{j}$
where $\partial_{i}=\frac{\partial}{\partial\theta_{i}}$, and due to the
factor $\frac{1}{2}$ we perform the summation over all $k\in{\mathbb{Z}}_{n}$.
Clearly, the second sum is zero. To show this, we have to specify the way of
summation. Let us collect in a group the terms
$k_{i}k_{j}\partial_{i}\partial_{j}$ attributed to those
$k\in{\mathbb{Z}}_{n}$ whose coordinates except the $i$th and the $j$th
coincide, while the $i$th and the $j$th coordinates satisfy the following
rules: they are obtained from $k_{i}$ and $k_{j}$ attributed to one of the
vectors of the group by means of an arbitrary assignment of a sign. This
operation specifies four vectors. The other four vectors are obtained from the
first four vectors of the group by means of the permutation of the $i$th and
the $j$th coordinates. In total, we get eight vectors in the group. Clearly,
the summands $k_{i}k_{j}\partial_{i}\partial_{j}$ attributed to these vectors
cancel each other. Let us compute the first sum.
$\displaystyle\sum_{k\in{\mathbb{Z}}_{n},|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}+2}}\sum_{i=1}^{n}k_{i}^{2}\partial_{i}^{2}=\sum_{i=1}^{n}\,\Bigl{[}\sum_{k\in{\mathbb{Z}}_{n},|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}+2}}\,k_{i}^{2}\Bigr{]}\,\partial_{i}^{2}.$
Note that
$\displaystyle\sum_{k\in{\mathbb{Z}}_{n},|k|=const}\,k_{1}^{2}=\cdots=\sum_{k\in{\mathbb{Z}}_{n},|k|=const}\,k_{n}^{2}=\frac{1}{n}\sum_{k\in{\mathbb{Z}}_{n},|k|=const}|k|^{2}.$
This implies:
$\displaystyle\sum_{k\in{\mathbb{Z}}_{n},|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}+2}}\sum_{i=1}^{n}k_{i}^{2}\partial_{i}^{2}=\frac{1}{n}\sum_{k\in{\mathbb{Z}}_{n},|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}}}\,\Delta=\frac{2}{n}\sum_{k\in{\mathbb{Z}}_{n}^{+},|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}}}\,\Delta.$
Together with (44) it gives:
$\sum_{k\in{\mathbb{Z}}_{n}^{+},|k|\leqslant
N}\sum_{i=1}^{n-1}(\bar{\nabla}_{A_{k}^{i}}\bar{\nabla}_{A_{k}^{i}}+\bar{\nabla}_{B_{k}^{i}}\bar{\nabla}_{B_{k}^{i}})\hat{V}(e)(\theta)=\frac{n-1}{n}\sum_{k\in{\mathbb{Z}}_{n}^{+},|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}}}\,\Delta V(\theta).$
We also have to take into consideration the term
$\displaystyle\sum_{i=1}^{n}\bar{\nabla}_{A_{0}^{i}}\bar{\nabla}_{A_{0}^{i}}\hat{V}(e)(\theta)=\Delta
V(\theta).$
Finally, we obtain:
$\displaystyle\Bigl{[}\sum_{k\in{\mathbb{Z}}_{n}^{+},|k|\leqslant
N}\sum_{i=1}^{n-1}(\bar{\nabla}_{A_{k}^{i}}\bar{\nabla}_{A_{k}^{i}}+\bar{\nabla}_{B_{k}^{i}}\bar{\nabla}_{B_{k}^{i}})+\sum_{i=1}^{n}\bar{\nabla}_{A_{0}^{i}}\bar{\nabla}_{A_{0}^{i}}\Bigr{]}\,\hat{V}(e)(\theta)$
$\displaystyle=\Bigl{(}1+\frac{n-1}{n}\sum_{k\in{\mathbb{Z}}_{n}^{+},|k|\leqslant
N}\frac{1}{|k|^{2{\alpha}}}\Bigr{)}\,\Delta V(\theta).$
The lemma is proved. ∎
## Acknowledgements
We would like to thank the referee for meaningful questions. This work was
supported by the Portuguese Foundation for Science and Technology (FCT) under
the projects PTDC/MAT/69635/06 and SFRH/BPD/48714/2008.
## References
* [1] V. I. Arnold, Sur la géométrie différentielle des groupes de lie de dimension infinie et ses applications a l’hidrodynamique des fluides parfaits. Ann. Inst. Fourier 16, 316–361, 1966.
* [2] S. Albeverio, Ya. Belopolskaya, Probabilistic approach to hydrodynamic equations. In the book “Probabilistic Methods in Fluids.” World Scientific, p. 1-21, 2003.
* [3] S. Albeverio, Ya. Belopolskaya, Generalized solutions of the Cauchy problem for the Navier–Stokes system and diffusion processes, arXiv:0709.1008v2 [math.PR].
* [4] Ya. I. Belopolskaya, Yu. L. Dalecky, Stochastic equations and differential geometry, Series: Mathematics and its Applications, Kluwer Academic Publishers, Netherlands, (1989), 260 p.
* [5] Ya. I. Belopolskaya, Probabilistic Representation of Solutions of Hydrodynamic Equations, Journal of Mathematical Sciences (New York), 101, no. 5, (2000), 3422–3436.
* [6] F. Cipriano, A. B. Cruzeiro, Navier–Stokes Equation and Diffusions on the Group of Homeomorphisms of the Torus, Commun. Math. Phys. 275, (2007), 255–269.
* [7] P. Constantin, G. Iyer, A stochastic Lagrangian representation of the three-dimensional incompressible Navier–Stokes equations, Comm. Pure Appl. Math., 61, no. 3, (2008), 330–345.
* [8] A. B. Cruzeiro, P. Malliavin, Nonergodicity of Euler fluid dynamics on tori versus positivity of the Arnold–Ricci tensor, J. Funct. Anal. 254, (2008), 1903–1925.
* [9] F. Delarue, On the existence and uniqueness of solutions to FBSDEs in a non-generate case, Stochastic Processes and their Applications, 99, (2002), 209–286.
* [10] D. Ebin, J. Marsden, Groups of diffeomorphisms and the motion of an incompressible fluid, Annals Math, 92, no. 1, (1970), 102–163.
* [11] H. Eliasson, Geometry of manifolds of maps, J. Differential geometry, 1, (1967), 169–194.
* [12] N. Ikeda, S. Watanabe, Stochastic differential equations and diffusion processes. 2nd ed. (English) North-Holland Mathematical Library, 24. Amsterdam etc.: North-Holland; Tokyo: Kodansha Ltd. xvi, (1981), 555 p.
* [13] Yu. E. Gliklikh, Calculus on Riemannian manifolds and Problems of Mathematical Physics, Voronezh university press, (1989), 192 p. (in Russian)
* [14] Yu. E. Gliklikh, New version of the Lagrange approach to the dynamics of a viscous incompressible fluid, Mathematical Notes, 55, No. 4, (1994), 344–350.
* [15] Yu. E. Gliklikh, Ordinary and Stochastic Differential Geometry as a Tool for Mathematical Physics, Springer (1996), 189 p.
* [16] Yu. E. Gliklikh, Global Analysis in Mathematical Physics: Geometric and Stochastic Methods, Springer (1997), 213 p.
* [17] Y. Le Jan, A.-S. Sznitman, Stochastic cascades and 3-dimensional Navier–Stokes equations. Probab. Theory Rel. Fields 109, (1997), 343–366.
* [18] J. Ma, P. Protter, J. Yong, Solving forward-backward stochastic differential equations explicitly - a four step scheme, Probab. Theory Related Fields, 98, (1994), 339–359.
* [19] H. P. McKean, Stochastic integrals, Academic Press, N. Y., (1969).
* [20] E. Pardoux, S. G. Peng, Adapted solution of a backward stochastic differential equation, Systems and Control Letters, 14, No. 1, (1990), 55–61.
* [21] E. Pardoux, S. Tang, Forward-backward stochastic differential equations and quasilinear parabolic PDEs, Probab. Theory Relat. Fields, 114, (1999), 123–150.
* [22] S. Shkoller, Geometry and curvature of diffeomorphisms groups with $H^{1}$ metric and mean hydrodynamics, J. Funct. Anal. 160,(1998), 337–365.
* [23] K. Yasue, A variational principle for the Navier-Stokes equation, J. Funct. Anal., 51(2), (1983), 133–141
* [24] T. Nakagomi, K. Yasue, J.-C. Zambrini, Stochastic variational derivations of the Navier-Stokes equation, Lett. Math. Phys., 160, (1981), 337–365.
| arxiv-papers | 2008-07-02T19:02:09 | 2024-09-04T02:48:56.516966 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Ana Bela Cruzeiro and Evelina Shamarova",
"submitter": "Evelina Shamarova",
"url": "https://arxiv.org/abs/0807.0421"
} |
0807.0478 | # Low-lying Proton Intruder State in 13B
S. Ota S. Shimoura H. Iwasaki M. Kurokawa S. Michimasa N. Aoi H. Baba
K. Demichi Z. Elekes T. Fukuchi T. Gomi S. Kanno S. Kubono K. Kurita H.
Hasegawa E. Ideguchi N. Iwasa Y.U. Matsuyama K.L. Yurkewicz T. Minemura
T. Motobayashi T. Murakami M. Notani A. Odahara A. Saito H. Sakurai E.
Takeshita S. Takeuchi M. Tamaki T. Teranishi Y. Yanagisawa K. Yamada M.
Ishihara Center for Nuclear Study, University of Tokyo, Saitama 351-0198,
Japan Department of Physics, University of Tokyo, Tokyo 113-0033, Japan
RIKEN Nishina Center for Accelerator-Based Science, Saitama 351-0198, Japan
Department of Physics, Rikkyo University, Tokyo 171-8501, Japan Hungarian
Acad. Sci., Inst. Nucl. Res., H-4001 Debrecen, Hungary Department of Physics,
Tohoku University, Miyagi 980-8578, Japan NSCL, Michigan State University,
East Lansing, MI 48824, USA Department of Physics, Kyoto University, Kyoto
606-8502, Japan Argonne National Laboratory, Argonne, IL 60439, USA
Department of Physics, Osaka University 560-0043, Japan Department of
Physics, Kyushu University, 812-8581
###### Abstract
The neturon rich nucleus 13B was studied via the proton transfer reaction
4He(12Be,13B$\gamma$) at 50$A$ MeV. The known 4.83-MeV excited state was
strongly populated and its spin and parity were assigned to 1/2+ by comparing
the angular differential cross section data with DWBA calculations. This low-
lying 1/2+ state is interpreted as a proton intruder state and indicates a
deformation of the nucleus.
###### keywords:
Intruder state , Proton single particle state , Proton transfer reaction ,
4He(12Be,13B$\gamma$) ,
###### PACS:
21.10.Pc , 25.55.Hp , 27.20.+n , 29.30.Kv
††journal: Physics Letters B
The existence of intruder states in light neutron-rich unstable nuclei is
often considered to be evidence for one or more $\hbar\omega$ configurations
in the low-lying states around the $psd$ shell. The ground state of 11Be is
$1/2^{+}$ which is lower in energy by 0.3 MeV than the $1/2^{-}$ state [1]. In
12Be, there is a $1^{-}$ intruder state at 2.7-MeV excitation energy [2]. The
energies of these low-lying, non-normal parity states indicate $1\hbar\omega$
configurations. Furthermore, the presence of low-lying 2${}_{1}^{+}$ [3] and
0${}_{2}^{+}$ [4, 5] states in 12Be suggests a $2\hbar\omega$ configuration.
Three theoretical interpretations have been proposed for these one or more
$\hbar\omega$ configurations in the low-lying states of neutron-rich unstable
nuclei: (1) the monopole interaction of the tensor force [6], (2) the loosely
bound nature of some orbitals [7], and (3) nuclear deformation [5, 8, 9]. In
Ref. [6],the effective interaction was determined so that the model reproduces
the energy levels in light, neutron-rich nuclei including intruder states, and
the importance of monopole interaction due to the tensor force was pointed
out. Reference [7] discusses the fact that in the $psd$ shell, the $2s_{1/2}$
orbital gains its energy relative to the other orbitals due to its loosely
bound nature. The non-zero $\hbar\omega$ configurations can also be
intuitively explained by the deformed mean field picture. As seen in the
Nilsson diagram, the gap becomes smaller with increasing deformation [5, 8,
9]. Since the combination of these effects, which are provided by their
corresponding theoretical models, changes the neutron shell structure in
neutron-rich nuclei, such as 11,12Be, the relative importance of the three
theoretical approaches has not been clarified.
For the proton shell in light neutron-rich nuclei, the effects due to the
tensor force and the loosely bound nature of the orbitals are expected to be
small since the $\nu p_{1/2,3/2}$ orbitals are fully filled and the proton(s)
are deeply bound. However, deformation is still presumed to affect the proton
shell structure. Proton intruder states are, therefore, signatures for the
importance of deformation. In the present study, we focus on the proton shell
structure in the $N=8$ nucleus 13B by investigating the proton single-particle
states.
In order to investigate the proton single-particle states in 13B, we used the
$(\alpha,t)$ reaction on 12Be in inverse kinematics. This process at an
intermediate energy has a relatively large cross section since the proton to
be transferred is deeply bound in the $\alpha$ particle and, thus, has high-
momentum components, which reduce the effect of the momentum mismatch of the
($\alpha$,$t$) reaction [10].
The experiment was performed at the RIKEN Accelerator Research Facility. A
12Be beam was produced by the fragmentation reaction of a 100$A$ MeV 18O beam
on a 9Be target with a thickness of 1.85-g/cm2. The 12Be beam was separated by
the RIKEN Projectile-fragment Separator (RIPS) [12]. The incident particles
were identified on an event-by-event basis using the measured time-of-flight
and the energy deposited. The time-of-flight, over a path length of 5.4 m, and
the energy deposited were measured using two plastic scintillators located at
the last two foci of the RIPS. The intensity and the purity of the 12Be beam
respectively were typically 2 $\times$ 105 counts per second and 90%.
The 50$A$ MeV 12Be beam bombarded a secondary target of liquid helium[11]
located at the final focus of the RIPS. A liquid helium target was chosen
because of its statistical advantage in terms of the experimental yields. The
helium was condensed by a cryogenic refrigerator and kept below 4 K during the
experiment. A target thickness of 143$\pm$5 mg/cm2 was estimated from the
velocity difference between outgoing particles measured with and without the
liquid helium. The positions and directions of the incident particles on the
secondary target were deduced from the position information of two parallel
plate avalanche counters [13] located 30-cm apart from each other around the
final focus. The outgoing particles were detected by a plastic scintillator
hodoscope 3.5-m downstream from the secondary target with a 1 $\times$ 1 m2
active area and an angular acceptance of up to 8 degrees in the laboratory
frame. The plastic scintillator hodoscope consisted of 5-mm thick $\Delta E$
and 60-mm thick $E$ layers. The $\Delta E$ layer was divided into 13 plastic
scintillator bars vertically and the $E$ layer into 16 plastic scintillator
bars horizontally. The outgoing particles were identified on an event-by-event
basis using the measured time-of-flight and the energy deposited in the
$\Delta E$ and $E$ layers. The mass distribution for the $Z=5$ particles is
shown in Fig. 1; the mass resolution $(\delta A)$ was determined to be
$\sim$0.25. The time-of-flight between the secondary target and the hodoscope
was deduced from the time information of the plastic scintillators located
upstream from the secondary target and the plastic scintillator hodoscope.
Position information for the outgoing particles was deduced from the time
difference between the output signals from the two photomultiplier tubes
attached to both ends of each scintillator bar, and was used to determine the
scattering angle. The angular resolution of the scattering angle in the
laboratory frame was 0.5 degrees in one standard deviation.
The de-excitation $\gamma$ rays were detected by an array of germanium
detectors: Gamma-Ray detector Array with Position and Energy sensitivities
(GRAPE)[14]. This consisted of 6 germanium detectors located at 140∘ with
respect to the beam axis. Each detector contains two cylindrical crystals 6 cm
in diameter and 2 cm thick, with a common anode between them. The each cathode
attached to each crystal is segmented into a $3\times 3$ matrix. The GRAPE
provides position information of the $\gamma$-ray interaction point, which is
extracted from a pulse shape analysis of the signal from the cathode [14, 15].
The intrinsic energy resolution and the full energy peak efficiency were
typically 2.7 keV (FWHM) and 0.4%, respectively, for 1332-keV $\gamma$ rays
from a 60Co standard source. The energy resolution after correcting for the
Doppler shift was deduced to be 1.3% (FWHM) for 2.1-MeV $\gamma$ rays,
corresponding to the decay of the first 2+ state of 12Be moving with 30% the
light velocity. The excited states of 13B populated in the reaction were
identified by measuring the energy of the de-excitation $\gamma$ rays.
Figure 2 shows the energy spectrum of $\gamma$ rays measured in coincidence
with 13B after correcting for the Doppler shift. There are three peaks
corresponding to the transitions from (3.68-, 3.73-), 4.13- and 4.83-MeV
excited states to the ground state. The peaks seen in the low energy region is
considered to originate from the reaction of the beam and the window of the
target cell. Other significant transitions including those between the excited
states were not observed in the present measurement. In the figure, hatched
areas show the response functions of the GRAPE for the de-excitation $\gamma$
rays obtained by means of a Monte Carlo simulation using the GEANT4 code [16,
17] and for the background $\gamma$ rays. The background was assumed to
consist of two components, the natural background $\gamma$ rays, estimated by
putting the gate in the non-prompt region of the time spectrum of the GRAPE,
and the $\gamma$ rays from the isomer state of 12Be, which were simulated by
assuming a life time of 331 ns [5]. The ratio of the isomer state to the
ground state in the secondary beam was less than 2 percent at the secondary
target. In the simulations, all the resolutions of the detectors associated
with the correction for the Doppler shift were taken into account. The
intensity of each decay to the ground state was deduced by fitting the
experimental spectrum with the sum of the response functions and the
background. Assuming no cascade decay, the derived relative intensities of the
observed $\gamma$ rays from the known excited states are shown in Table 1,
together with the previously reported data including two neutron transfer
[18], neutron knockout [19], $\beta$-decay followed by neutron decay [20], and
multi-nucleon transfer [21]. In the present reaction, the excited state at
4.83 MeV is strongly populated relative to the other excited states, while it
is less excited in the other reactions except for the multi-nucleon transfer
reaction which is expected to populate proton excited states. Considering the
selectivity of the proton transfer reaction, it is conceivable that the
4.83-MeV excited state is of proton single-particle nature.
In order to determine the angular differential cross section of the
4He(12Be,13B∗) reaction, individual $\gamma$-ray spectra with 0.5-degree-
scattering-angle cuts were fitted with the same response functions and
background described above, and they were analyzed to deduce the populations.
Figure 3 shows the experimental angular distribution for the 4.83-MeV
excitation with filled circles. The transferred angular momentum ($\Delta L$)
in the reaction is determined by comparing the obtained angular distributions
with predictions of the distorted-wave Born approximations (DWBA), calculated
with the DWUCK5 code [22]. The optical potentials for the entrance and exit
channels were obtained by adopting a single folding model used in Ref. [23].
The density distribution of 12Be for the folding is calculated by using the
mean field calculation code, TIMORA[24, 25]. The density distribution of 13B
is assumed to be the same as that of 12Be with $R(^{13}{\rm B})/R(^{12}{\rm
Be})=(13/12)^{1/3}$. For the entrance channel, the depth of the imaginary
potential is adjusted so as to represent the inelastic scattering data of
12Be, excited to the $2^{+}_{1}$ state. For the exit channel, the depths of
the real and imaginary potentials are varied in order to estimate the
statistical error for the spectroscopic factor. The DWBA predictions with
$\Delta L=0,1,2$ are shown as curves in Fig. 3. The absolute magnitudes of the
predictions are normalized so as to minimize the $\chi^{2}$ values. The
reduced $\chi^{2}$ value is 0.74 for the $L=0$ calculation. The forward angle
peak in the experimental angular distribution is well described by the $\Delta
L$ = 0 DWBA calculation. Therefore, we assigned a spin and parity of 1/2+ to
the 4.83-MeV excited state. Its low energy indicates that it is a proton
intruder state from the $sd$ shell. The spectroscopic strength $C^{2}S$ was
deduced to be $0.20\pm 0.02$, where $S$ is a spectroscopic factor and $C^{2}$
is an isospin Clebsch-Gordan coefficient. The systematic errors of the
$C^{2}S$ due to the choice of the optical potentials in the DWBA calculation
and to the geometrical uncertainty of the GRAPE in the simulation were
evaluated to be 50% and 10%, respectively. The former one is obtained from the
difference between the folding model and a separate calculation using a global
optical potential for 12C in Ref. [26, 27]. In the absolute magnitude of the
differential cross section, there is no ambiguity originated from cascade
decays to or from the 4.83-MeV excited state. No excited state was found above
4.83 MeV up to the neutron threshold of 4.87 MeV in 13B, therefore, the
cascade decays to the 4.83-MeV excited state are unlikely to occur. On the
other hand, the cascade decay from the 4.83-MeV excited state is also small as
explained below. The excited state (1/2+) is expected to decay to the ground
state (3/2-) via an $E1$ transition, whose decay rate is proportional to the
cube of the transition energy, i.e., $E_{\gamma}^{3}$ [8]. The maximum
possible transition energy to a known excited state is 1.4 MeV corresponding
to the decay to the 3.48-MeV excited state. Even if the decay to this state is
by $E1$ transition, its decay rate is about 40 times smaller than to the
ground state. Decay rates other than $E1$ are also expected to be much
smaller. In fact, no cascade decay of the 4.83-MeV excited state was observed
in the present and the previous experiments.
There are few theoretical predictions for 13B. In Ref. [19], the result of a
shell model calculation for 13B has been shown, however, the calculation is
performed from the viewpoint of the neutron structure. Focusing on the proton
shell structure, we performed a shell model calculation for 13B(12Be) with the
calculation code, OXBASH [28], wherein the interaction including the effect of
the tensor force and reproducing intruder states in the neutron-rich nuclei
such as 11,12Be [6], is used. The model space consisting of the $psd$ shell
with maximum 3(2)$\hbar\omega$ excitation was considered for 13B(12Be). The
spectroscopic factor was calculated as the overlap between each excited state
of 13B and the ground state of 12Be. The shell model calculation predicts four
1/2+ states below 10 MeV excitation energy. Three lower states are predicted
at 3.7, 6.5 and 9.1 MeV with spectroscopic strengths of 0.01, 0.03 and 0.003,
respectively. These strengths are too small to explain the experimental
result. The fourth 1/2+ state with a spectroscopic strength of 0.30 is close
to the experimental spectroscopic strength; however its excitation energy is
9.5 MeV. These results of the shell model calculation may indicate that the
theoretical approach based on the effect of the tensor force does not explain
the experimental result, or the model space consisting of the $psd$ shell and
the excitation of 3$\hbar\omega$ are not large enough.
A simple expression for the observed $1/2^{+}$ state with proton single
particle nature is $(\pi sd)^{1}\otimes^{12}$Be. Concerning the possible large
deformation of 12Be, we examine the qualitative consequence of a deformed
core. The $1/2^{+}$ state arises with a configuration of
$(\pi[220\frac{1}{2}])^{1}$ to the 12Be(g.s.) core. As seen in the Nilsson
diagram, at large deformation, the $\pi[220\frac{1}{2}]$ orbital gains energy.
The energy gain of the total system compared to the spherical $s_{1/2}$ is
estimated to be around 7 MeV from the diagram and the semi-empirical mass
formula with a deformation parameter of 0.5. The discrepancy between the shell
model prediction mentioned above and the present result for the $1/2^{+}$
state is thus ascribed to a core deformation effect. In the present reaction,
the observed $1/2^{+}$ state is considered to be populated by the transfer of
a proton to the $\pi[220\frac{1}{2}]$ orbital. Hence, the low excitation
energy of the 1/2+ state suggests a deformed mean field. Since the proton is
transfered to the $[220\frac{1}{2}]$ orbital by an $s$-wave ($\Delta L=0$),
the spectroscopic factor of the spherical $2s_{1/2}$ component in the
$[220\frac{1}{2}]$ orbital, e.g. $S=0.273$ at $\delta=0.4$ [8], may also
reproduce the experimental result.
Based on the discussion above, the $1/2^{+}$ intruder state at 4.83 MeV is a
signature of a deformed field and indicates the importance of a deformation in
this nuclei. On the other hand, the ground state of 13B is supposed to be
spherical with a normal $p$-shell configuration [29]. The excitation of only
one proton thus changes the structure of the nucleus.
Recently an AMD calculation related to the experimentally assigned 1/2+ state
of 13B was carried out by Kanada-En’yo et al [30]. The calculation predicts a
largely deformed 1/2${}_{1}^{+}$ state with proton $1\hbar\omega$ excitation
and neutron $2\hbar\omega$ excitation. In the excited state, the last proton
occupies a molecular $\sigma$ orbital which has a density distribution similar
to the $[220\frac{1}{2}]$ Nilsson orbital. The 1/2${}_{1}^{+}$ state predicted
by AMD may correspond to the experimental one although the calculated
excitation energy is 8 MeV.
In summary, we have studied the proton transfer reaction on 12Be in inverse
kinematics. A spin and parity of $1/2^{+}$ for the 4.83-MeV excited state in
13B were assigned for the first time. A spectroscopic strength of $0.20\pm
0.02$ was also obtained with 60% systematic error. This state is interpreted
to be a proton intruder state from the $sd$-shell because of its non-normal
parity and small excitation energy. The nuclear deformation provides a simple
mechanism for the existence of such a low-lying proton intruder state. The
present study shows the importance of deformation for proton shell structure
in neutron-rich nuclei and the change of the proton shell structure deu to the
excitation of one proton.
We would like to thank the RIKEN Ring Cyclotron staff members for their
operation during the experiment. The present work was partially supported by
the Grant-in-Aid for Scientific Research (No.15204018) by the Ministry of
Education, Culture, Sports, Science and Technology, of Japan.
## References
* [1] D. Wilkinson, et al., Phys. Rev. 113 (1959) 563.
* [2] H. Iwasaki, et al., Phys. Lett. B 491 (2000) 8.
* [3] H. Iwasaki, et al., Phys. Lett. B 481 (2000) 7.
* [4] S. Shimoura, et al., Phys. Lett. B 560 (2003) 31.
* [5] S. Shimoura, et al., Phys. Lett. B 654 (2007) 87.
* [6] T. Suzuki, et al., Phys. Rev. C 67 (2003) 044302.
* [7] I. Hamamoto, Nucl. Phys. A 731 (2004) 211.
* [8] A. Bohr, B. R. Mottelson, Nuclear Structure, Vol. II, World Scientific (1998).
* [9] I. Hamamoto, S. Shimoura J. Phys. G 34 (2007) 2715
* [10] S. Michimasa, et al., Phys. Lett. B 638 (2006) 146.
* [11] H. Ryuto, et al., Nucl. Instr. and Meth. A 555 (2005) 1.
* [12] T. Kubo, et al., Nucl. Instr. and Meth. B 70 (1992) 309.
* [13] H. Kumagai, et al., Nucl. Instr. and Meth. A 470 (2001) 562.
* [14] S. Shimoura, Nucl. Instr. and Meth. A 525 (2004) 188.
* [15] M. Kurokawa, et al., IEEE TRANSACTIONS ON NUCLEAR SCIENCE 50 (2003) 1309.
* [16] S. Agostinelli, et al., Nucl. Instr. and Meth. A 506 (2003) 250.
* [17] Geant4 a toolkit for the simulation of the passage of particles through matter, http://geant4.web.cern.ch/geant4/index.shtml.
* [18] F. Ajzenberg-Selove, et al., Phys. Rev. C 17 (1978) 1283.
* [19] V. Guimarães, et al., Phys. Rev. C 61 (2000) 064609.
* [20] N. Aoi, et al., Phys. Rev. C 66 (2002) 014301.
* [21] R. Kalpakchieva, et al., Eur. Phys. J. A 7 (2000) 451.
* [22] P. D. Kunz, computer code DWUCK5, unpublished.
* [23] G. R. Satchler, Phys. Rev. C 55 (1997) 285.
* [24] C. J. Horowitz, Computational Nuclear Physics, Vol. I, Springer-Verlag (1991).
* [25] C. J. Horowitz, Nucl. Phys. A 368 (1981) 503.
* [26] A. Ingemarsson, et al., Nucl. Phys. A 676 (2000) 3.
* [27] A. Ingemarsson, et al., Nucl. Phys. A 696 (2001) 3.
* [28] B. A. Brown, A. Etchegoyen, W. D. M. Rae, The computer code OXBASH, MSU-NSCL report number 524 (1986).
* [29] T. Nagatomo, et al., Nucl. Phys. A 746 (2004) 509c
* [30] Y. Kanada-En’yo, et al., submitted to Phys. Rev C (arXiv:0706.3503v1 [nucl–th]).
Figure 1: Mass distribution of the outgoing particles for $Z=5$ extracted from the correlation between TOF and $\Delta E$ measured with the plastic scintillator hodoscope. Figure 2: Doppler corrected $\gamma$-ray spectrum in coincidence with 13B particles. The spectrum is decomposed with the sum of response functions of the GRAPE for the de-excitation and background $\gamma$ rays. The resulting responce functions are shown with hatched areas. The response functions for two doublets around 3.5 and 3.7 MeV are summed. The response function for the background includes the natrural background and the decay of the 12Be isomer (see the text). Figure 3: Angular distribution of the 4.83-MeV excited 13B. Experimental data is shown by the filled circles with statistical errors. The solid, dotted and dashed curves show the DWBA predictions with assumptions of $\Delta L$ = 0, 1, and 2, respectively. See the text for more detail including the optical potentials. Table 1: List of the relative populations of the excited states. The intensities are normalized by the most intense population. The errors are statistical only. In the last column, the spins and parities which have been assigned in the previous and present studies are listed. a:Present reaction of 4He(12Be,13B$\gamma$). b:11B (t,p)13B. c:9Be (14B,13B)X. d:14Be $\beta$ delayed n. e:16O(14C,17F)13B. For the reaction b the relative intensity is at $\theta_{\rm lab}=10^{\circ}$ and for the reaction e at $\theta_{\rm c.m.}=5.4^{\circ}$. | Relative intensities |
---|---|---
Ex (MeV) | a | b | c | d | e | $J^{\pi}$
3.48 | 0.19 $\pm$ 0.05 | 0.06 | 0.60 $\pm$ 0.14 | | |
3.53 | 0.20 $\pm$ 0.05 | 0.19 | | 1 | |
3.68 | 0.74 $\pm$ 0.07 | 0.38 | 1 | | |
3.71 | 0.68 $\pm$ 0.07 | 0.25 | | | |
4.13 | 0.49 $\pm$ 0.04 | 1 | 0.04 $\pm$ 0.04 | | |
4.83 | 1 | 0.03 | | | 1 | 1/2+
| arxiv-papers | 2008-07-03T02:49:57 | 2024-09-04T02:48:56.525471 | {
"license": "Public Domain",
"authors": "S. Ota, S. Shimoura, H. Iwasaki, M. Kurokawa, S. Michimasa, N. Aoi, H.\n Baba, K. Demichi, Z. Elekes, T. Fukuchi, T. Gomi, S. Kanno, S. Kubono, K.\n Kurita, H. Hasegawa, E. Ideguchi, N. Iwasa, Y.U. Matsuyama, K.L. Yurkewicz,\n T. Minemura, T. Motobayashi, T. Murakami, M. Notani, A. Odahara, A. Saito, H.\n Sakurai, E. Takeshita, S. Takeuchi, M. Tamaki, T. Teranishi, Y. Yanagisawa,\n K. Yamada and M. Ishihara",
"submitter": "Shinsuke Ota",
"url": "https://arxiv.org/abs/0807.0478"
} |
0807.0479 | # Parametric interaction and intensification of nonlinear Kelvin waves
###### Abstract
Observational evidence is presented for nonlinear interaction between
mesoscale internal Kelvin waves at the tidal – $\omega_{t}$ or the inertial –
$\omega_{i}$ frequency and oscillations of synoptic – $\Omega$ frequency of
the background coastal current of Japan/East Sea. Enhanced coastal currents at
the sum – $\omega_{+}$ and dif – $\omega_{-}$ frequencies:
$\omega_{\pm}=\omega_{t,i}\pm\Omega$ have properties of propagating Kelvin
waves suggesting permanent energy exchange from the synoptic band to the
mesoscale $\omega_{\pm}$ band. The interaction may be responsible for the
greater than predicted intensification, steepen and break of boundary trapped
and equatorially trapped Kelvin waves, which can affect El Niño. The problem
on the parametric interaction of the nonlinear Kelvin wave at the frequency
$\omega$ and the low-frequency narrow-band nose with representative frequency
$\Omega\ll\omega$ is investigated with the theory of nonlinear week dispersion
waves.
Gindraft=false NOVOTRYASOV AND STEPANOV INTENSIFICATION OF INTERNAL KELVIN
WAVES Vadim Novotryasov, V.I.Il’ichev Pacific Oceanological Institute of the
Russian Academy of Sciences, 43 Baltiiskaya Street, 690041 Vladivostok,
Russia. (vadimnov@poi.dvo.ru) Dmitriy Stepanov, V.I.Il’ichev Pacific
Oceanological Institute of the Russian Academy of Sciences, 43 Baltiiskaya
Street, 690041 Vladivostok, Russia. (step-nov@poi.dvo.ru)
## 1 Introduction
Internal Kelvin waves play a significant role in the dynamics of the oceans.
There are two basic types of the waves: equatorially trapped and boundary
trapped. Kelvin waves propagating on the equatorial thermocline participate in
the adjustment of the tropical ocean to wind stress forcing [Philander,1990].
Extensive data on Kelvin waves have been obtained recently, motivated in part
by possible connection between the initial stages of El Niño and equatorial
nonlinear Kelvin waves which may precede this event. Any relaxation or
reversal of the steady trade winds (the easterlies) results in the excitation
of a Kelvin wave, which can affect El Niño. In other words Kelvin waves play a
critical role in generating and sustaining of the Southern Oscillation
[e.g.,Fedorov, 2000].
In the dynamics of the coastal oceans Kelvin waves are used to interpret such
phenomena as the instability of alongshore currents, the generation and
variability of wind currents on the shelf, and upwelling [Brink, 1991].
Fedorov and Melville [1995, 1996] considered Kelvin waves trapped boundaries
and showed that waves could manifest nonlinear properties that is steepen and
overturn or break. A wave of depression deepens the thermocline and breaks on
the forward face of the wave. A broken wave of depression can form a jump
[also called shocks or fronts: Lighthill, 1978; Philander, 1990]. In turn, the
breaking of Kelvin waves may be important in mixing, momentum and energy
transfer in coastal oceans. Other example of nonlinear Kelvin wave dynamics
includes the problem of nonlinear geostrophic adjustment in the presence of a
boundary [Helfrich et al., 1999; Reznik and Grimshaw, 2002].
The spectrum represents one of the major characteristics of the waves. It is
used as a representative statistical description of the wave field in studies
of nonlinear interaction [e.g., Hibiya et al., 1998], acoustic propagation
[e.g., Colosi et al., 1998], and mixing parametrization [Polzin, 1995].
Filonov and Novotryasov [2005, 2007] studded the wave band of temperature
fluctuation spectra in the coastal zone of Pacific Ocean and observed that in
the high-frequency band of temperature spectra the spectral exponent tends to
$\sim\omega^{-1}$ at the time of spring tide and on the western shelf of the
Japan/East Sea, in the $\omega_{i}\ll\omega\ll N_{\ast}$ range, where
$N_{\ast}$ is the representative buoyancy frequency and $\omega_{i}$ is the
inertial frequency, the spectral exponent tends to $\sim\omega^{-3}$. These
features Filonov and Novotryasov [2007] simulated by the model spectrum of
nonlinear internal waves in the shallow water. They considered interaction of
high-frequency waves with the wave at the tidal frequency and shown that the
spectrum of high-frequency internal waves take the universal form and the
spectral exponent tends to $\sim\omega^{-1}$.
In this paper, we present observational evidence for nonlinear interaction
between synoptic oscillations of the background current at the representative
frequency $\Omega$ and nonlinear Kelvin waves at the tidal – $\omega_{t}$ or
the inertial – $\omega_{i}$ frequencies. Findings are based on a well defined
spectral peaks at the sum – $\omega_{+}$ and dif – $\omega_{-}$ frequencies of
inertial, semidiurnal and synoptic motions measured by current meters and
temperature records collected in the coastal Japan/East Sea. We made rotary
spectral analyses of these records and found that the clockwise rotary
spectrum for coastal currents measured at 35–m depth in the 1999 year has
well-defined spectral peaks at the frequencies $\omega_{t}\sim 1/12.4$(cph),
$\Omega_{1,2}\sim 1/64,1/102$(cph), as well as sym and difference frequencies
$\omega_{\pm}=\omega_{s}\pm\Omega_{1,2}$ and their overtones
$\omega_{n}=n\omega_{\pm}(n=1,2,3)$. Analyses of coastal temperature records
collected in the 2004 year showed that the spectrum of temperature variations
has an analog form with spectral peaks at the inertial frequency
$\omega_{i}\sim 1/17.8$ (cph) and $\Omega_{1,2}\sim 1/80,1/160$ (cph) synoptic
frequencies, as well as sym and difference frequencies
$\omega_{\pm}=\omega_{i}\pm\Omega_{1,2}$ and overtones
$\omega_{n}=n\omega_{\pm}(n=1,2,3)$.
With the theory of nonlinear interaction among week dispersion waves [Gurbatov
et al., 1990] we consider the problem on the parametric interaction of the
nonlinear Kelvin wave at the frequency $\omega$ and the low-frequency narrow-
band nose with representative frequency $\Omega\ll\omega$. We show that
nonlinear interaction between Kelvin wave and no stationary low-frequency
component of the background coastal current excited by the atmospheric forcing
leading to the intensification of tidal and inertial currents on the sub-
surfers and near-bottom layers of the coastal zone. Analogy phenomena of the
Kelvin wave intensification can occur in the equatorial ocean.
## 2 Observations
In the first experiment, observations of internal waves were performed at the
coast of the Gamov Peninsula area of the Japan/East Sea in 1999 year. The
records of current meter collected from a mooring deployed water depth 40 m,
during 2 weeks in September (Figure 1a). To gauge temperature, speed and
direction of currents, the buoy was instrumented with the Russian-made POTOK
integral instrument at a depth of 35 meters. The POTOK had a temperature
measurement resolution of 0.05∘C. The sampling rate was 15 minutes. The time
series of the meridional (light curve) and zonal (heavy curve) current
components from this mooring are shown on Figure 1c. The Canadian Guideline
CTD profiler was used for the vertical profiling of water. The hydrostatic
pressure, temperature and electric conductivity of seawater were measured
during profiling. The profiling errors were no greater than 0.01∘C in
temperature and 0.02 psu in salinity. Figure 1b shows the mean temperature and
buoyancy frequency profiles for the study site. Details of this experiment
(results, methods of their processing, etc.) were reported by Novotraysov, et.
al [2003].
Figure 1: (a) Study area on the Japan/East Sea shelf, September 1999, 2004\.
The mooring location in 2004 is shown in Arabic. The circle indicates the
mooring location in 1999 and the location of the mooring vessel, from which
hourly casts were conducted on September, 1999, 2004. (b) Daily mean vertical
profiles of temperature $T$ and buoyancy frequency $N$. (c) Records of the
meridional (light curve) and zonal (heavy curve) components of the current at
the 35-m depth in autumn 1999. (d) Record of the temperature variations at the
mooring buoy 1 in autumn 2004.
In the second experiment, measurements of internal wave band temperature
fluctuations were performed during 18 days starting 3 September 2004. The time
records of temperature were collected from two moorings deployed along the
coastline at a distance of 800m from it, approximately at 40m depth and
separated by a distance of 5.5 km from each other. The first of them was
deployed at 28m depth and the second one, at 35m depth (below the surface).
The moorings were equipped with digital thermographs made by the Russian
manufacturer. The devices had a measuring precision of 0.05∘C for temperature.
The sampling rate was 1 minutes. The temperature and salinity vertical
profiles were performed on 20–21 September, from a vessel anchored between the
moorings with the Canadian Guideline CTD profiler. In total, 25 hourly casts
were made.
The obtained data were analyzed by unified standard spectral techniques [Emery
and Thomson, 1997]. The approach involved (i) the elimination of low-frequency
components with a Tukey’s cosine filter, (ii) the splitting of the resulting
series into nine 37.2-hour segments (three semidiurnal tidal periods each),
(iii) the calculation and averaging of spectral densities by segments, and
(iv) the smoothing of the averaged spectral densities by a five-point Tukey’s
filter. The number of degrees of freedom in the processing amounted to
approximately 10 for the first experiment and 20 for the second, providing a
reasonable reliability of the results of spectral analysis.
## 3 The model
A spectrum model of nonlinear interactions among internal Kelvin waves is
developed. It is assumed that $H/\lambda\ll 1$, and $A/H\ll 1$, where $H$ is
the water depth, $\lambda$ is a characteristic wave-length, $A$ is a
representative wave amplitude. The basic component of this model is the simple
wave equation. For the first most powerful mode of the small-amplitude Kelvin
waves the equation is written as
${\partial u}/{\partial x}-\alpha u{\partial u}/{\partial\tau}=0,$ (1)
where $u(t,x)$ is the alongshore current, $x$ is a horizontal coordinate,
$\tau=t-x/c$, $t$ is time. The parameters $\alpha$ and $c$ are the coefficient
of nonlinearity and the phase speed of long internal waves, respectively.
Parameter of nonlinearity is determined by the background density and is
related in the Boussinesq approximation as:
$\alpha=\left({3H\int\limits_{-H}^{0}{\phi_{z}^{3}dz}}\right)\times\left({2c^{2}\int\limits_{-H}^{0}{\phi_{z}^{2}dz}}\right)^{-1},$
(2)
where $z$ is a vertical coordinate, positive upward. The phase speed $c$ and
the amplitude function of the wave mode $\phi(z)$ are determined from the
solution of the eigenvalue problem
$d^{2}\phi{\kern 1.0pt}/dz^{2}+c^{-2}N^{2}(z){\kern 1.0pt}\,\phi=0,$ (3)
with boundary condition
$\phi(-H)=\phi(0)=0,$ (4)
and with the normalization
$\phi_{\max}=1.$ (5)
Equation (1) is a basic model for study of Kelvin waves interaction. Let us
consider interaction nonlinear Kelvin wave with frequency $\omega_{0}$ (tidal
$\omega_{t}$ or inertial $\omega_{i})$ and narrow-band synoptic noise (later
SN) with frequency $\Omega\ll\omega_{0}$ using the spectrum model of nonlinear
internal waves described in terms of the equation (1). Let the alongshore
current $u(x,t)$ at the boundary of the coastal area $x=0$ be the
superposition of internal wave with frequency $\omega_{0}$, amplitude $A_{0}$
and the noise $\vartheta(t)$ with typical frequency $\gamma<<\omega_{0}$
$u(t,x=0)=u_{0}(t)=A_{{\kern 1.0pt}0}\cos{\kern 1.0pt}\,(\omega_{0}{\kern
1.0pt}t+\varphi)+\vartheta(t),$ (6)
where $\varphi$ – is a random phase with uniform distribution in the interval
$[-\pi,+\pi]$.
We confine our analysis to the wave evolution stage, which is characterized by
condition $x<x_{\ast}$, where $x_{\ast}=\,(\alpha A_{0}\omega_{0})^{-1}$. In
this stage the front Kelvin wave appears and it is not accompanied by
generation of internal solitons. We introduce parameter $d=x/x_{\ast}$, which
determine the similarity between a Kelvin shock-wave and the wave and consider
the case $d<1$. For this case the spectral density of the wave field $u$ is
$\displaystyle a(\omega;x)=-\frac{J_{0}(\omega d/\omega_{0})}{2\pi
i\omega\alpha x}\int\limits_{-\infty}^{+\infty}\\{\exp[-i\omega\alpha
x\vartheta(t)]-1\\}e^{i\omega t}dt-$
$\displaystyle-\sum\limits_{n=-\infty\atop n\neq
0}^{\infty}\frac{i^{n}J_{n}(\omega d/\omega_{0})}{2\pi i\omega\alpha
x}\int\limits_{-\infty}^{+\infty}{\\{\exp[in\varphi-i\omega\alpha
x\vartheta(t)]-1\\}}\times$
$\displaystyle\times\exp[i(\omega-n\omega_{0})t]dt\,\,\,\,$ (7)
From equation (7) we can affirm that the first term is the spectral density of
the synoptic oscillations and the second term is the spectral density of the
Kelvin waves distorted by the synoptic oscillations.
We consider the spectral composition of the Kelvin wave near of the harmonics
with the number $n$ of the carrier frequency $\omega_{0}$. Given that
$\gamma\ll\omega_{0}$ we make replacement $\omega$ at $n\omega_{0}$ in the
exponential rate of the equation (7). Then we obtain the equation for the
harmonic with number $n$
$u_{n}\left({t,x}\right)=A_{n}(x)\cos[n\omega_{0}t+n(\varphi(t)-\omega_{0}\beta
x\vartheta(t))]$ (8)
and her amplitude depending on the parameter $d=A_{0}\alpha\omega_{0}x$ is
$A_{n}(x)=2J_{n}(nd)A_{0}/nd.$ (9)
From the equation (9), we can affirm that the amplitude of the harmonic is
decreased with increasing of harmonic number $\sim 1/n$.
Figure 2: (a) Normalized clockwise rotary spectra of currents versus for
current measured at 35–m depth in autumn 2004 near the Gamov peninsula. The
numerical digits over peaks are their periods (h). (b) The increased fragment
of the spectrum in the surrounding of the semidiurnal frequency. Dark numerals
12.2; 6.2; 4.3 are carrier periods (h).
Perform the more detailed spectral analysis of the nonlinear Kelvin waves,
when the synoptic noise consists of regular and stochastic components:
$\vartheta(t)=a\cos(\Omega t)+\vartheta_{ns}(t)$. Let the amplitude of the
phase modulation related with regular component $M=\alpha n\Omega x$ is small,
that is $M\ll 1$. Let the Kelvin waves have narrow-band spectrum, and the
synoptic noise width $\gamma\ll\omega_{0}$, besides the noise, amplitude and
phase of Kelvin waves have Gaussian distribution. In this case the correlation
function of the harmonic with number $n$ taking into account the equation (8)
is
$\displaystyle B_{n}(\tau,x)=\\{\frac{A_{n}^{2}}{2}\\{\cos
n\omega_{0}\tau+\frac{(a\alpha
n\omega_{0}x)^{2}}{2}[\cos(n\omega_{0}+\Omega\tau)+$
$\displaystyle+\cos(n\omega_{0}-\Omega)\tau]\\}\\}\exp[-n^{2}D_{\psi}\tau/2],\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,$
(10)
where
$D_{\psi}(\tau)=D_{\varphi}(\tau)+(\alpha\omega_{0}x)^{2}(\sigma_{ns}^{2}-B_{ns}(\tau))$.
Here $D_{\phi}$ is the structure function of the Kelvin wave phase,
$B_{ns}(\tau)=<\vartheta_{ns}(t+\tau)\vartheta_{ns}(t)>$ is the correlation
function of the synoptic noise. We perform the Fourier transform of the
correlation function $B_{n}\left({\tau,x}\right)$ and obtain the formula for
the spectrum of the harmonic
$\displaystyle
S_{n}\left({\omega;x}\right)=\frac{A_{n}^{2}}{2}\tilde{S}_{n}\left({\omega-n{\kern
1.0pt}\omega_{0}}\right)+\frac{(A_{n}a\alpha n\omega_{{\kern
1.0pt}0}x)^{2}}{4}\times$
$\displaystyle\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\times\\{\tilde{S}_{n}\left({\omega-(n{\kern
1.0pt}\omega_{0}+\Omega}\right))+\tilde{S}_{n}(\omega-(n{\kern
1.0pt}\omega_{{\kern 1.0pt}0}-\Omega))\\},$ (11)
where
$\tilde{S}_{n}\left({\omega\,;x}\right)=\frac{1}{2\pi}\int\limits_{-\infty}^{+\infty}{\exp[-n^{2}D_{\psi}(\tau)/2]\cos}{\kern
1.0pt}(\omega\tau)d\tau.$ (12)
From equation (3), we can affirm that spectrum of the wave have harmonics with
frequencies $n\omega_{0}$ as well as harmonics with sum and dif frequencies
$\omega_{\pm}=n\omega_{0}\pm\Omega$ and amplitudes increasing with growth of
distance traversed by the wave. This means that at the coastal zone between
Kelvin waves and synoptic noise takes place nonlinear interaction, which is
accompanied by intensification Kelvin waves with side frequencies
$\omega_{\pm}=n\omega_{0}\pm\Omega$.
## 4 Discussion and Conclusion
Table 1: The periods of the peaks of the temperature variations spectrum (digit numerals) and the calculated periods (light numerals) $1/\Omega(h)$ | $1/\omega_{+}(h)$ | $1/\omega_{s}(h)$ | $1/\omega_{-}(h)$ | $1/\omega_{+}(h)$ | $1/\omega_{i}(h)$ | $1/\omega_{-}(h)$ | $1/\omega_{+}(h)$ | $2/\omega_{i}(h)$ | $1/\omega_{-}(h)$
---|---|---|---|---|---|---|---|---|---
84 | 10.8 | 12.4 | 14.5 | 14.7 | 17.8 | 22.6 | 8.0 | 8.9 | 10.0
84 | 11.0 | 12.4 | 14.5 | 15.0 | 17.5 | 22.5 | 8.0 | 8.7 | 10.0
168 | 11.5 | 12.4 | 13.4 | 16.1 | 17.8 | 19.9 | 8.5 | 8.9 | 9.4
168 | 11.5 | 12.4 | 13.0 | 16.2 | 17.5 | – | 8.5 | 8.7 | 9.5
Let us turn to the records of the current velocity. Figure 1c presents records
smoothed by Tukey window with the width 1 h of the meridional (light curve)
and zonal (heavy curve) components of the current in the coastal zone of
Japan/East Sea in autumn 1999 year. It is of interest to note a low-frequency
component of the variation in the record of the current velocity.
Figure 3: Normalized spectrum versus for record of the temperature variations
at the mooring buoy 1 in autumn 2004 near the Gamov peninsula. The numerical
digits over peaks are their periods (h).
Let us analyze circular rotary current component with the rotary spectral
analysis [Emery and Thomson, 1997]. The spectrum of the clockwise rotary (CWR)
component of the velocity is
$S_{-}(\omega)=0,125\cdot(S_{uu}+S_{vv}-2Q_{uv}),$
where $S_{uu}$ and $S_{vv}$ are the one-sided autospectra of the $u$ and $v$
Cartesian components of the velocity and $Q_{uv}$ is the quadrature spectrum
between the two components. These spectra were determined by unified
statistical spectral techniques using the algorithms presented in [Emery and
Thomson, 1997]. The spectra have N degrees of freedom, where N $\geq$ 10\.
Figure 2a shows the CWR spectrum of currents versus $\log{\kern
1.0pt}\,\omega$ normalized by maximal value. It is of interest to note the
groups of the well defined significant spectral peaks in the surrounding of
the semidiurnal frequency and her $1,2,3$ and $4$ harmonics. Our estimations
show that the magnitudes of these spectral peaks interrelated with number
harmonics by the approximate relation
$A_{12}:A_{6}:A_{4}:A_{3}:A_{2,4}\approx 1:2:3:4:5$
(with error $<$ 20%).
There are two peaks with light and heavy tops at the representative
frequencies $\Omega_{1}\sim 1/64$ and $\Omega_{2}\sim 1/102$ (cph) in the low-
frequency band of the spectra.
Figure 2b presents the increased fragment of the spectrum in the surrounding
of the semidiurnal frequency. It is of interest to note a significant well
defined spectral peak at the $\omega_{s}=1/12.2$ (cph) frequency surrounded by
side peaks with dark end light tops. Not hard to make sure that the
frequencies of these picks are determined by formula
$\omega_{\pm}=\omega_{{\kern 1.0pt}s}\pm\Omega_{1,2}$ with error equals of
experimental error uncertainty. In particularly the strong spectral peak at
the difference frequency $\omega_{-}=1/12.2-1/64$ (cph) suggests that the
motion in study region is dominated by not only the tide movements, but and
the inertial movements. The pick at the frequency $\omega_{i}=1/17.7$ (cph) is
formed by the inertial movements. Thus the spectrum has fine structure in the
band 1/14 - 1/16.5 (cph), which don’t segregate with help our dates and
methods of analyses.
Let us consider the structure of the spectrum in surroundings of the harmonics
of semidiurnal frequency $\omega_{s}$ (Figure 2b). As in the past picks at the
carrier frequencies: $\omega_{s1},\omega_{s2}$ are surrounded by side picks at
the frequencies, whish are determined by formula
$\omega_{\pm}=\omega_{s1,2}\pm\Omega_{1,2}$.
Thus spectral analysis of the clockwise rotary velocity shows that her
spectrum is determined by oscillations with the tidal frequency
$\omega_{s}\approx 1/12.2$ (cph) and the frequencies $\Omega_{1,2}\approx
1/72,106$ (cph) from synoptic band and also that spectrum has the fine
structure in the neighborhood of the frequency $\omega_{s}$ with side peaks at
the frequencies, which are determined by formula
$\omega_{\pm}=\omega_{s}\pm\Omega_{1,2}$ The similar structure has spectrum in
the neighborhood of the first and the second harmonics of the semidiurnal
frequency $\omega_{s}$.
Then turn to the records of the temperature of the coastal water of Japan/East
sea collected in autumn 2004 year. Figure 3 shows the averaged spectrum of the
temperature records collected from two moorings and normalized by the maximum
value of the spectrum. Spectra are calculated by unified standard spectral
techniques [Emery and Thomson, 1997]. The number of freedom degrees amounted
to approximately 20.
Of particular interest are the groups of the significant spectral peaks at the
inertial frequency and her first harmonic and also the peaks in the
surrounding of the semidiurnal frequency $\omega_{s}$.
Arabic numerals over spectral peaks equal the periods (hour) of these peaks.
These periods (h) are placed in the Table 1. Digit numerals equal the periods
of the spectral peaks and light numerals equal the periods calculated by the
formula $1/\omega_{\pm}=1/\omega_{s,i}\pm 1/\Omega_{1,2}$.
As indicated by Table 1 the periods of the spectral peaks and the periods
calculated by the formula have approximately equal values. Thus the spectral
analysis of current and temperature meter records of the Japan/East Sea
coastal water shows that the spectrum is determined by oscillations with
frequencies of near inertial $\omega_{i}\approx 1/17.8$ (cph) and tidal
$\omega_{s}\approx 1/12.2$ (cph) frequencies as well as the frequencies
$\Omega_{1,2}\approx 1/72,106$ (cph) from synoptic band. The analysis shows
that the spectrum has the fine structure in the neighborhood of the
frequencies $\omega_{i,s}$ with side picks at the frequencies, which are
determined by formula $\omega_{\pm}=\omega_{i,s}\pm\Omega_{1,2}$. The similar
structure has spectrum in the neighborhood of the first harmonics of the
inertial frequency $\omega_{i}$.
As a conclusion, we showed that among nonlinear internal Kelvin wave with
frequency $\omega$ and the low-frequency narrow-band nose with representative
frequency $\Omega\ll\omega$ occur the parametric interaction, i.e. among the
low-frequency band and high-frequency band of internal Kelvin waves occur the
energy exchange. Hence, as a result of nonlinear interaction the energy of
internal Kelvin waves in the coastal and tropical oceans can grow at the
expense of the low-frequency narrow-band nose energy.
###### Acknowledgements.
This work was supported by the Presidential Grant No. MK–1364.2008.5 by the
Prezidium of the Russian Academy of Sciences (Program “Mathematical methods in
nonlinear dynamics.” Project No. 06-I-13-048).
## References
* [1] Emery, W. J., and R. E. Thomson (1997), Data Analysis Methods in Physical Oceanography, 634 pp., Pergamon, New York.
* [2] Brink, K. H. (1991), Coastal-trapped waves and wind-driven currents over the continental shelf Ann. Rev. Fluid Mech., 23, 389–412.
* [3] Colosi, J. A., and the ATOS Group (1999), A review of recent results on ocean acoustic wave propagation in random media: Basin scales, IEEE J. Oceanic Eng., 24, 138-155.
* [4] Fedorov, A. V., and W. K. Melville (1995), Propagation and breaking of nonlinear Kelvin waves, J. Phys. Oceanogr., 25, 2518–31.
* [5] Fedorov, A. V., W. K. Melville (1996), Hydraulic jumps at boundaries in rotating fluids, J. Fluid Mech., 324, 55–82.
* [6] Fedorov, A. V., W. K. Melville (2000), Kelvin fronts on the equatorial thermocline, J. Phys. Oceanogr., 30, 1692–705.
* [7] Filonov, A., and V. Novotryasov (2005), Features of the nonlinear wave spectrum in the coastal zone, Geophiys. Res. Lett., 32, L15602, doi: 10.1029/2005GL023046.
* [8] Filonov, A., and V. Novotryasov (2007), On a spectrum of nonlinear internal waves in the oceanic coastal zone, Nonlin. Processes Geophys., 14, 757–762.
* [9] Gurbatov, S.N., A.N. Malakhov, and A.I. Saichev (1990), Nonlinear Random Waves in Media with Zero Dispersion, 215 pp., Nauka, Moscow.
* [10] Helfrich, K.R., A.C. Kuo, L.J. Pratt (1999), Nonlinear Rossby adjustment in a channel, J. Fluid Mech., 390, 187–222.
* [11] Hibiya, T., Y. Niwa and K. Fujiwara (1998), Numerical experiments of nonlinear energy transfer within the oceanic internal wave spectrum, J. Geophys. Res., 103, 18715–17722.
* [12] Lighthill, M. J. (1978), it Waves in Fluids, 504 pp., Cambridge University Press, UK.
* [13] Novotryasov, V. V., N. S. Vanin, A. A. Karnaukhov (2005), Manifestation of nonlinear properties of Kelvin internal waves in the coastal zone of the sea of Japan, Izv. Atm. and Ocean. Phys-ics., 41, 611–619.
* [14] Philander S.G. (1990), El Niño, La Niña and the Southern Oscillation, 293 pp., Academic Press, New York.
* [15] Polzin, K. L., J. M. Toole and R. W. Schmitt (1995), Finescale parametrization of turbulent disspation, J. Phys. Oceanogr., 25, 306–328.
* [16] Reznik, G.M. and R. Grimshaw (2002), Nonlinear geostrophic adjustment in the presence of a boundary, J. Fluid Mechanics, 471, 257–283.
| arxiv-papers | 2008-07-03T04:24:55 | 2024-09-04T02:48:56.529556 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Vadim Novotryasov and Dmitriy Stepanov",
"submitter": "Stepanov Dmitriy",
"url": "https://arxiv.org/abs/0807.0479"
} |
0807.0645 | # Searches for Large Extra Dimensions at the Tevatron
Vyacheslav Krutelyov1 for the CDF and the DØ Collaborations
1- University of California at Santa Barbara - Dept of Physics
Santa Barbara CA 93106-9530 - U.S.A.
Work supported by the U.S. Department of Energy under contract No. DE-
AC02-07CH11359.
###### Abstract
The presence of extra dimensions can be probed in high energy collisions via
the production or exchange of gravitons. The former corresponds to signatures
with missing energy while the latter corresponds to modifications of the final
state spectra. Here I review results of analyses performed by the CDF and DØ
Collaborations on $p\overline{p}$ collisions at
$\sqrt{s}=1.96~{}\mathrm{Te\kern-1.00006ptV}$ in signatures sensitive to large
extra dimensions. These include analyses of
$\gamma+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ and
$\mathrm{jet}+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ as signatures of graviton
production as well as analyses of dilepton and diboson final states sensitive
to graviton exchange.
## 1 Introduction
The standard model of particle physics (SM) does not account for gravitational
interactions. A number of theoretical models beyond the SM naturally include
gravity and require the existence of extra dimensions where only gravity
propagates freely. In particular, a model proposed in Ref. [1] assumes there
are large extra dimensions (LED) compactified on a sub-millimeter scale $R$.
In this model gravity can become strong at the $\mathrm{Te\kern-1.00006ptV}$
scale, which effectively solves the hierarchy problem present in the SM. For
$n$ extra dimensions the four-dimensional Planck mass $M_{\mathrm{Pl}}$ is
related to the $4+n$-dimensional fundamental scale $M_{D}$ by
$M_{\mathrm{Pl}}^{2}=8\pi M_{D}^{2}(M_{D}R)^{n}$. For $M_{D}$ in the
$\mathrm{Te\kern-1.00006ptV}$ range $R$ is of the order of 0.1 $\mathrm{mm}$
(10 $\mathrm{fm}$) for $n=2$ ($6$). To solve the hierarchy problem $M_{D}$
should not be too large, which rules out $n=1$. Tests of the Newton’s law
constrain $R$ to be below $37~{}\mathrm{\mu m}$ for $n=2$ [2]. Other
constraints come from astrophysics (up to $n=3$) and from high energy
colliders [2].
The cross sections of the LED processes depend on $M_{D}$ and can be at levels
detectable at the Tevatron [3]. In $p\overline{p}$ collisions the graviton can
be produced in $q\overline{q}\to gG$, $qg\to qG$, or $gg\to gG$ corresponding
to a $\mathrm{jet}+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ final state, and
$q\overline{q}\to\gamma G$ corresponding to a
$\gamma+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ detector signature [4]. Graviton
exchange can be studied in a range of $2\to 2$ processes with the best
sensitivity in final states with two leptons, photons, or $Z$ bosons.
In this review [5] I first describe searches for graviton production in the
$\gamma+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ and
$\mathrm{jet}+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ signatures reported by CDF and in
the $\gamma+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ signature reported by DØ. Then I
briefly describe searches sensitive to graviton exchange in dimuon, dielectron
and diphoton final states reported by DØ and in $ZZ$ final states by both CDF
and DØ.
## 2 CDF and DØ detectors
A detailed description of the CDF and the DØ detectors can be found in Ref.
[6, 7]. The vertexing and the tracking detectors surrounding the interaction
region are used to reconstruct charged particle trajectories. Further out are
the calorimeters with electromagnetic and hadronic longitudinal segmentation
used to identify and measure photons, electrons, and jets. Outside the
calorimeter are the muon detectors used to identify muons from the collisions
and cosmic rays.
In analyses of $\gamma+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ events one of the major
backgrounds is from cosmic ray muons misidentified as photons. The detector
features used to reject this background are: photon pointing available at DØ
allowing the reconstruction of direction of the photon from the longitudinal
segmentation of the electromagnetic calorimeter [8]; photon timing available
at CDF from the EMTiming system [9] providing the measurement of the time of
the energy deposit and allowing to suppress cosmic-ray background by a factor
of 20 or more.
## 3 Searches for graviton production
The production of gravitons in hadron collisions is probed in events where the
final-state hadronic jet or photon recoils against the graviton, which is not
detected and results in ${E\\!\\!\\!\\!/_{T}}$.
Background Events $Z\to\nu\nu$ $388\pm 30$ $W\to\ell\nu$ $362\pm 17$ Multi-jet
$23\pm 20$ $\gamma+$jet $17\pm 5$ Non-collision $10\pm 10$ Total predicted
$808\pm 62$ Data observed $809$
Table 1: The numbers of expected background and observed events in the
$\mathrm{jet}+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ sample used by CDF.
Searches in both the $\mathrm{jet}+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ and the
$\gamma+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ require exclusive signatures: only a
jet or a photon is in the event with a veto on the presence of other objects.
These searches have similar backgrounds. The dominant and only irreducible
background is production of $Z\to\nu\nu$ in association with a photon or a
jet.
A search for LED in $\mathrm{jet}+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ events has
been performed by CDF using 1.1 $\mathrm{fb}^{-1}$ of data, updating the
analysis in Ref. [10]. Events with high ${E\\!\\!\\!\\!/_{T}}$, a high-$E_{T}$
jet and no second jet with $E_{T}>60~{}\mathrm{Ge\kern-1.00006ptV}$ are
selected. All major backgrounds are estimated in a data-driven way. The
contributions from $Z\to\nu\nu$ and from $W\to\ell\nu$, where the lepton is
lost, are estimated using measured $Z\to\ell\ell$ and $W\to\ell\nu$ events.
After an a priori optimization for the best sensitivity to LED, events with
the leading jet $E_{T}>150~{}\mathrm{Ge\kern-1.00006ptV}$ and
$\mbox{${E\\!\\!\\!\\!/_{T}}$}>120~{}\mathrm{Ge\kern-1.00006ptV}$ are
selected. The summary of predicted and observed events is given in Table 1.
The constraints on LED are given at the end of this section.
Background CDF DØ $Z\gamma\to\nu\nu\gamma$ $24.8\pm 2.8$ $12.1\pm 1.3$
$W\to(e\to\gamma)\nu$ $2.6\pm 0.4$ $3.8\pm 0.3$ $W\to(\mu/\tau\to\gamma)\nu$
$1.0\pm 0.2$ – $W\gamma\to\ell\nu\gamma$ $5.0\pm 1.4$ $1.5\pm 0.2$
jet$\to\gamma$ – $2.2\pm 1.5$ $\gamma\gamma\to\gamma$ $2.3\pm 0.6$ - Non-
collision $9.8\pm 1.3$ $2.8\pm 1.4$ Total predicted $46.3\pm 3.0$ $22.4\pm
2.5$ Observed data $40$ $29$
Table 2: The numbers of expected background and observed events in the
$\gamma+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ samples analyzed by CDF and DØ. The
contribution from jet$\to\gamma$ in CDF is included in $(W/\gamma)\gamma$ by
virtue of the background estimation method.
Results for the $\gamma+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ signature in the
Tevatron Run II have been recently reported by CDF in 2 $\mathrm{fb}^{-1}$ of
data and by DØ in 1.05 $\mathrm{fb}^{-1}$ of data [8]. Both analyses use
events with high ${E\\!\\!\\!\\!/_{T}}$ and a high-$E_{T}$ photon in the
central detector region with $|\eta|<1.1$ [4]. No jets with
$E_{T}>15~{}\mathrm{Ge\kern-1.00006ptV}$ are allowed in the event in order to
suppress QCD backgrounds. Also, no high-$p_{T}$ track is allowed, suppressing
contributions from leptonic $W$ and $Z$ decays: tracks with
$p_{T}>10~{}\mathrm{Ge\kern-1.00006ptV}/c$ (isolated with
$p_{T}>6.5~{}\mathrm{Ge\kern-1.00006ptV}/c$) are vetoed by CDF (DØ). The
contribution from cosmic rays is suppressed based on presence of hits in the
muon detectors as well as using the photon pointing or the photon timing
mentioned in Section 2.
The background contributions for both analyses are summarized in Table 2. The
number of events from $Z\to\nu\nu$ is estimated from simulation. The
contribution from $W\to e\nu$, where the electron is identified as the photon,
is extracted from the number of events with electrons passing all other
requirements applied to the photon signal sample, scaled by the
misidentification rate. Contributions from cosmic rays and jets misidentified
as photons are estimated at DØ using the photon pointing method. The remaining
backgrounds in the analysis by DØ are estimated using simulation. The
contribution from cosmics at CDF is estimated using events with the photon
time significantly different from the collision time. The remaining
backgrounds at CDF are from processes where an object is lost. Except for
$W\to\tau\nu$, which is estimated from simulation, each of these backgrounds
is given by the number of events in data with this object identified, scaled
by the simulated rate for it to be lost. This approach implicitly includes
contributions from a jet identified as a photon.
Source LEP DØ CDF $n$ $\gamma+\mbox{${E\\!\\!\\!\\!/_{T}}$}$
$\gamma+\mbox{${E\\!\\!\\!\\!/_{T}}$}$
$\mathrm{jet}+\mbox{${E\\!\\!\\!\\!/_{T}}$}$
$\gamma+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ combined 2 1600 921 1310 1080 1400 3
1200 877 1080 1000 1150 4 940 848 980 970 1040 5 770 821 910 930 980 6 660 810
880 900 940
Table 3: Lower limits on $M_{D}$ in $\mathrm{Ge\kern-1.00006ptV}/{c^{2}}$ at
$95\%$ C.L. for $n$ from 2 to 6 observed in the
$\mathrm{jet}+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ and
$\gamma+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ signatures and the combination of the
two at CDF, and in the $\gamma+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ signature at DØ,
along with the constraints from the LEP experiments [11]. Constraints provided
by DØ for $n$ above 6 can be found in [8].
After an a priori optimization for the best sensitivity, events with a photon
with $E_{T}>90~{}\mathrm{Ge\kern-1.00006ptV}$ and
$\mbox{${E\\!\\!\\!\\!/_{T}}$}>50(70)~{}\mathrm{Ge\kern-1.00006ptV}$ are
selected in the CDF (DØ) signal sample. Event counts observed in the data are
consistent with expectations from the backgrounds. The constraints on the LED
model are summarized in Table 3. Since the sensitivity in the
$\gamma+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ is similar to that in the
$\mathrm{jet}+\mbox{${E\\!\\!\\!\\!/_{T}}$}$ mode, the constraints can be
improved in a combination. The result of this combination using the CDF data
is shown in Table 3. Depending on the number of extra dimensions, the
sensitivity to LED at the Tevatron is comparable to or better than that from
the LEP experiments [11].
## 4 Searches for virtual graviton exchange
In hadron collisions the processes most sensitive to virtual graviton exchange
in LED are those with two leptons (electrons or muons), two photons, or two
$Z$ bosons in the final state, where it is possible to reconstruct the full
kinematics of the final state with high precision. The value of the graviton
exchange amplitude depends on the cutoff energy $\Lambda$, presumably of order
$M_{D}$ [3]. The cross section is proportional to $\Lambda^{-8}$.
A search for LED in the dimuon final state has been reported by DØ using 200
$\mathrm{pb}^{-1}$ of data [12]. The sample is comprised of events with two
opposite charge muons with $p_{T}$ above $15~{}\mathrm{Ge\kern-1.00006ptV}/c$
and both muons passing additional quality and isolation requirements. The
dilepton invariant mass and the scattering angle in the dilepton center of
mass are analyzed simultaneously to improve sensitivity to the exchange of a
graviton (spin 2 particle). A similar analysis using a di-em final state where
no discrimination is made between electrons and photons has been performed at
DØ with 240 $\mathrm{pb}^{-1}$ of data [13]. In this case events were selected
if they had two em objects with $E_{T}>25~{}\mathrm{Ge\kern-1.00006ptV}$
passing quality and isolation requirements. No excess over the expected
backgrounds is observed. The constraints on the parameter $\Lambda$ are in the
range of 1 to 2 $\mathrm{Te\kern-1.00006ptV}$ depending on the theoretical
parameters.
Both CDF and DØ have recently reported on searches for $p\overline{p}\to ZZ$
production in a four-lepton final state [14, 15]. The observed number of
events is consistent with the SM expectations. Although no constraint has been
provided on LED by either experiment, based on Ref. [16] the constraint on
$\Lambda$ is expected to be in the range of 1.5 to 2.5
$\mathrm{Te\kern-1.00006ptV}$.
## 5 Summary
Searches for LED in signatures with direct graviton production and those with
virtual graviton exchange are explored by the CDF and the DØ experiments. No
evidence of LED has been seen. Additional $p\overline{p}$ collision data
currently produced at the Tevatron is expected to give more insight about the
possible presence of large extra spacial dimensions.
## Acknowledgements
I would like to acknowledge the funding institutions supporting the CDF and DØ
Collaborations. The full list of agencies can be found in, e.g, [6, 7].
## References
* [1] N. Arkani-Hamed, S. Dimopoulos and G. R. Dvali, Phys. Lett. B 429, 263 (1998) [arXiv:hep-ph/9803315].
* [2] W. M. Yao et al. [Particle Data Group], J. Phys. G 33, 1 (2006).
* [3] G. F. Giudice, R. Rattazzi and J. D. Wells, Nucl. Phys. B 544, 3 (1999) [arXiv:hep-ph/9811291].
* [4] The polar coordinate system is used with an origin at the center of a detector and the $z$-axis ($\theta=0$) along the proton beam. The pseudorapidity $\eta$ is defined as $\eta=-\ln[\tan(\theta/2)]$. The transverse momentum $p_{T}$ is defined as $p_{T}=p\cdot\sin{\theta}$, where $p$ is the particle’s momentum. The transverse energy is defined as $E_{T}=E\cdot\sin{\theta}$, where $E$ is the energy measured by the calorimeter. The missing $E_{T}$ (${E\\!\\!\\!\\!/_{T}}$) is defined as $\vec{\mbox{${E\\!\\!\\!\\!/_{T}}$}}=-\sum_{i}E_{T}^{i}\hat{n}_{i}$, where $i$ is the index of a calorimeter tower (a segmentation unit in $\eta$-$\phi$) and $\hat{n}_{i}$ is a unit vector perpendicular to the beam axis pointing to the tower from the origin.
* [5] Slides:
`http://indico.cern.ch/contributionDisplay.py?contribId=104&sessionId=15&confId=24657`
* [6] A. Abulencia et al. [CDF Collaboration], J. Phys. G 34, 2457 (2007) [arXiv:hep-ex/0508029].
* [7] V. M. Abazov et al. [D0 Collaboration], Nucl. Instrum. Meth. A 565, 463 (2006) [arXiv:physics/0507191].
* [8] V. M. Abazov et al. [D0 Collaboration], arXiv:0803.2137 [hep-ex], accepted by Phys. Rev. Lett.
* [9] M. Goncharov et al., Nucl. Instrum. Meth. A 565, 543 (2006) [arXiv:physics/0512171].
* [10] A. Abulencia et al. [CDF Collaboration], Phys. Rev. Lett. 97, 171802 (2006) [arXiv:hep-ex/0605101].
* [11] J. Abdallah et al. [DELPHI Collaboration], Eur. Phys. J. C 38, 395 (2005) [arXiv:hep-ex/0406019]; P. Achard et al. [L3 Collaboration], Phys. Lett. B 587, 16 (2004) [arXiv:hep-ex/0402002].
* [12] V. M. Abazov et al. [D0 Collaboration], Phys. Rev. Lett. 95, 161602 (2005) [arXiv:hep-ex/0506063].
* [13] V. M. Abazov et al. [D0 Collaboration], DØ note 4336 (2004), unpublished.
* [14] T. Aaltonen et al. [CDF Collaboration], arXiv:0801.4806 [hep-ex].
* [15] V. M. Abazov et al. [D0 Collaboration], Phys. Rev. Lett. 100, 131801 (2008) [arXiv:0712.0599 [hep-ex]].
* [16] M. Kober, B. Koch and M. Bleicher, Phys. Rev. D 76, 125001 (2007) [arXiv:0708.2368 [hep-ph]] and references therein.
| arxiv-papers | 2008-07-03T20:58:12 | 2024-09-04T02:48:56.536809 | {
"license": "Public Domain",
"authors": "Vyacheslav Krutelyov (for the CDF and D0 Collaborations)",
"submitter": "Vyacheslav Krutelyov",
"url": "https://arxiv.org/abs/0807.0645"
} |
0807.0647 | New constraints on $H_{0}$ and $\Omega_{m}$ from SZE/X-RAY data and Baryon
Acoustic Oscillations
R. F. L. HOLANDA111holanda@astro.iag.usp.br, J. V.
CUNHA222cunhajv@astro.iag.usp.br and J. A. S. LIMA333limajas@astro.iag.usp.br
Departamento de Astronomia, Universidade de São Paulo
Rua do Matão, 1226 - 05508-900, São Paulo, SP, Brazil
###### Abstract
The Hubble constant, $H_{0}$, sets the scale of the size and age of the
Universe and its determination from independent methods is still worthwhile to
be investigated. In this article, by using the Sunyaev-Zel‘dovich effect and
X-ray surface brightness data from 38 galaxy clusters observed by Bonamente et
al. (2006), we obtain a new estimate of $H_{0}$ in the context of a flat
$\Lambda$CDM model. There is a degeneracy on the mass density parameter
($\Omega_{m}$) which is broken by applying a joint analysis involving the
baryon acoustic oscillations (BAO) as given by Sloan Digital Sky Survey
(SDSS). This happens because the BAO signature does not depend on $H_{0}$. Our
basic finding is that a joint analysis involving these tests yield
$H_{0}=0.765^{+0.035}_{-0.033}$ km s-1 Mpc-1 and
$\Omega_{m}=0.27^{+0.03}_{-0.02}$. Since the hypothesis of spherical geometry
assumed by Bonamente et al. is questionable, we have also compared the above
results to a recent work where a sample of triaxial galaxy clusters has been
considered.
## 1 Introduction
Ten years ago, two different teams using SNe type Ia as standard candles
announced that the expansion of the Universe is speeding up and not slowing as
believed by the cosmology community[1]. The authors interpreted their results
as implying that the cosmological constant $\Lambda$ is greater than zero, an
assumption that is not only surviving but has been strengthened in the last
decade by many independent astronomical observations[2, 3, 4, 5, 6, 7] (see
also Peebles and Ratra[8], and Padmanabhan[9] for reviews).
On the other hand, until now there is no consensus about the present value of
the Hubble parameter even when the estimates of $H_{0}$ are limited to the
Cepheid period-luminosity relation and the correlation of the luminosity at
maximum light for SNe type Ia. For instance, Riess et al.[10] using a hybrid
sample composed by Cepheids in galaxies and recent type Ia SNe concluded that
$H_{0}=73\pm 6$ $km.s^{-1}.Mpc^{-1}$ while Sandage[11] and coworkers based on
data from the Hubble Space Telescope (HST) key project are claiming that
$H_{0}=62\pm 5$ $km.s^{-1}.Mpc^{-1}$. Therefore, is still worthwhile to
improve the direct local estimates of $H_{0}$, and, probably, more important,
to investigate their concordance with independent methods. An interesting one
is provided by the combination of Sunyaev-Zel’dovich Effect (SZE) and
measurements of X-ray surface brightness from galaxy clusters.
The so-called SZE is a small distortion on the Cosmic Microwave Background
(CMB) spectrum provoked by the inverse Compton scattering of the CMB photons
passing through a population of hot electrons[12, 13]. Observing the
temperature decrement of the CMB in the direction of galaxy clusters and the
X-ray spectrum emitted by the hot electron gas pervading the cluster, it is
possible to break the degeneracy between concentration and temperature[14],
and, therefore, to calculate the angular diameter distance (ADD)[15, 16, 17].
This technique for measuring distances is completely independent of other
methods (as the one provided by the luminosity distance), and it can be used
to measure distances at high redshifts directly. Review papers on this subject
have been published by Birkinshaw[18] and Carlstrom, Holder and Reese[19].
More recently, such a technique has been applied for a fairly large number of
clusters[20, 21, 22, 23, 24].
In 2006, Bonamente and collaborators[23] determined the ADD distance to 38
clusters of galaxies in the redshift range $0.14\leq z\leq 0.89$ using X-ray
data from Chandra and Sunyaev-Zeldovich effect data data from the Owens Valley
Radio Observatory and the Berkeley-Illinois-Maryland Association
interferometric arrays. Assuming spherical symmetry, the cluster plasma and
dark matter distributions were analyzed by using a hydrostatic equilibrium
model accounting for radial variations in density, temperature and abundances.
In addition, by fixing the earlier cosmic concordance $\Lambda$CDM model
($\Omega_{m}=0.3$, $\Omega_{\Lambda}=0.7$), they obtained for the Hubble
constant $H_{0}=76.9^{+3.5+10.0}_{-3.4-8.0}\,km.s^{-1}.Mpc^{-1}$ (statistical
followed by systematic uncertainty at 1$\sigma$ c. l.). The main advantage of
this method for estimating $H_{0}$ is that it does not rely on the
extragalactic distance ladder being fully independent of any local calibrator.
In a point of fact, once the angular diameter distance has been measured a
large amount of astrophysical and cosmological tests can be performed in order
to constrain the relevant free parameters[25, 26, 27].
It should be stressed, however, that in the above quoted determination of the
Hubble constant, a specific cosmology was fixed from the very beginning. This
is needed because the determination of the Hubble parameter from SZE/X-ray
technique is endowed with a strong degeneracy with respect to the matter
density parameter which can only be broken by considering additional
cosmological tests. An interesting possibility is given by the signature of
the baryon acoustic oscillations (BAO) from the last scattering surface (LSS).
In 2005, Eisenstein et al.[28] presented the large scale correlation function
from the Sloan Digital Sky Survey (SDSS) showing clear evidence for the baryon
acoustic peak at $100h^{-1}$ Mpc scale, a result in good agreement with the
analyses from CMB data[3]. The breaking on the degeneracy between $H_{0}$ and
$\Omega_{m}$ by applying this BAO signature is possible because it does not
depend on $H_{0}$ and is highly sensitive to the matter density parameter. The
combination SZE/X-ray with BAO has been recently discussed in the literature
by Cunha, Marassi and Lima[29].
In this article, we discuss the determinations of $H_{0}$ and $\Omega_{m}$ by
considering the same 38 clusters from Bonamente et al. sample where spherical
symmetry was assumed. Our basic findings follow from a joint analysis
involving the data from SZE and X-ray surface brightness with the recent SDSS
measurements of the baryon acoustic peak. The influence of the intrinsic
cluster geometry on $H_{0}$ it will be quantified by comparing the results
derived here with a recent determination of both parameters[29] based on De
Filippis et al. sample formed by 25 triaxial clusters[22].
## 2 Basic Equation and Statistical Analysis
Let us now consider that the Universe is described by a flat Friedmann-
Robertson-Walker (FRW) geometry
$ds^{2}=dt^{2}-a^{2}(t)\left[dr^{2}+r^{2}(d\theta^{2}+sin^{2}\theta
d\phi)\right],$ (1)
driven by cold dark matter plus a cosmological constant ($\Lambda$CDM).
In this background, the angular diameter distance, ${\cal{D}}_{A}$, can be
written as[29, 30, 31]
${\cal{D}}_{A}(z;h,\Omega_{m})=\frac{3000h^{-1}}{(1+z)}\int_{o}^{z}\frac{dz^{\prime}}{{\cal{H}}(z^{\prime};\Omega_{m})}\quad\mbox{Mpc},$
(2)
where $h=H_{0}/100$ km s-1 Mpc-1 and the dimensionless function
${\cal{H}}(z^{\prime};\Omega_{m})$ is given by
${\cal{H}}=\left[\Omega_{m}(1+z^{\prime})^{3}+(1-\Omega_{m})\right]^{1/2}.$ As
it appears, the above expression has only two free parameters
($h,\Omega_{m}$). In this way, we perform a statistical fit over the
$h-\Omega_{m}$ plane in light of Bonamente et al. data sample[23]. In our
analysis we use a maximum likelihood that can be determined by a $\chi^{2}$
statistics,
$\chi^{2}(z|\mathbf{p})=\sum_{i}{({\cal{D}}_{A}(z_{i};\mathbf{p})-{\cal{D}}_{Ao,i})^{2}\over\sigma_{{\cal{D}}_{Ao,i}}^{2}+\sigma_{stat}^{2}},$
(3)
where ${\cal{D}}_{Ao,i}$ is the observational ADD, $\sigma_{{\cal{D}}_{Ao,i}}$
is the uncertainty in the individual distance, $\sigma_{stat}$ is the
contribution of the statistical errors (see table 3 in Bonamente et al.[23]
(2006)) added in quadrature ($\approx 20$%) and the complete set of parameters
is given by $\mathbf{p}\equiv(h,\Omega_{m})$.
In what follows, we first consider the SZE/X-ray distances separately, and,
further, we present a joint analysis including the BAO signature from the SDSS
catalog. Note that a specific flat $\Lambda$CDM cosmology has not been fixed a
priori in the analysis below.
Figure 1: a) Angular diameter distance as a function of redshift for
$\Omega_{m}=0.3$, $\Omega_{\Lambda}=0.7$ and some selected values of the $h$
parameter. The data points correspond to the the SZE/X-ray distances for 38
clusters from Bonamente et al. sample. b) Confidence regions ($68.3$%, $95.4$%
and $99.7$%) in the $(\Omega_{m},h)$ plane provided by the SZE/X-ray data. The
best fit values are $h=0.743$ and $\Omega_{m}=0.37$. As remarked in the text,
the possible values of $H_{0}$ are heavily dependent on the allowed values of
$\Omega_{m}$, and, therefore, such a degeneracy need to be broken by adding a
new cosmological test.
### 2.1 Limits from SZE/X-ray
To begin with, let us consider the influence of the Hubble parameter on the
ADD once the cosmology is fixed.
In Fig. 1(a), we display the residual Hubble diagram and the galaxy cluster
data sample by considering a flat cosmic concordance model
($\Omega_{m}=0.3,\Omega_{\Lambda}=0.7$). As expected, for a given redshift,
the distances increase for smaller values of $H_{0}$.
In Fig. 1(b) we show the contours of constant likelihood (68.3%, 95.4% and
99.7%) in the space parameter $h-\Omega_{m}$ for the SZ/X-ray data above
discussed. Note that a large range for the $h$ parameter is allowed,
($0.65\leq h\leq 0.86$), at $1\sigma$ of confidence level. In particular, we
found $h=0.755^{+0.065}_{-0.065}$ with $\chi^{2}_{min}=35.2$ at $68.3$%
confidence level (c.l.). Naturally, such bounds on $h$ are reasonably
dependent on the cosmological model adopted. For example, if we fix
$\Omega_{m}=0.3$ we have $h=0.76$, for $\Omega_{m}=1.0$ we have $h=0.65$, and
both cases are permitted with high degree of confidence. This means that using
only the Bonamente et al. sample (spherical cluster hypothesis) we cannot
constrain severely the energetic components of the $\Lambda$CDM model with
basis on the SZE/X-ray data alone. As one may conclude, one additional
cosmological test (fixing $\Omega_{m}$) is necessary in order to break the
degeneracy on the $(\Omega_{m},h)$ plane.
### 2.2 Joint analysis for SZE/X-ray and BAO
As remarked in the introduction, more stringent constraints on the space
parameter ($h,\Omega_{m}$) can be obtained by combining the SZE/X-ray with the
BAO signature[28]. The peak detected (from a sample of 46748 luminous red
galaxies selected from the SDSS Main Sample) is predicted to arise precisely
at the measured scale of 100 $h^{-1}$ Mpc. Let us now consider it as an
additional cosmological test over the spherical cluster sample. Such a
measurement is characterized by
Figure 2: a) Contours in the $\Omega_{m}-h$ plane using the SZE/X-ray and BAO
joint analysis. The contours correspond to $68.3$%, $95.4$% and $99.7$%
confidence levels. The best-fit model converges to $h=0.765$ and
$\Omega_{m}=0.27$. b) Likelihood function for the $h$ parameter in a flat
$\Lambda$CDM universe from SZE/X-ray emission and BAO. The horizontal lines
are cuts in the regions of $68.3$% probability and $95.4$%.
$\displaystyle{\cal{A}}\equiv{\Omega_{\rm{m}}^{1/2}\over{{\cal{H}}(z_{\rm{*}})}^{1/3}}\left[\frac{1}{z_{\rm{*}}}\Gamma(z_{*})\right]^{2/3}=0.469\pm
0.017,$ (4)
where $z_{\rm{*}}=0.35$ is the redshift at which the acoustic scale has been
measured and $\Gamma(z_{*})$ is the dimensionless comoving distance to
$z_{*}$. Note that the above quantity is independent of the Hubble constant,
and, as such, the BAO signature alone constrains only the $\Omega_{m}$
parameter.
Table 1: Limits to $h$ using SZ/X-ray method from galaxy clusters ($\Lambda$CDM) Reference (data) | $\Omega_{m}$ | $h$ ($1\sigma$) | $\chi^{2}$
---|---|---|---
Mason et al. 2001 (7 clusters).. | $0.3$ | $0.66^{+0.14}_{-0.11}$ | $\simeq 2$
Reese et al. 2002 (18 cluster).. | $0.3$ | $0.60^{+0.04}_{-0.04}$ | $16.5$
Reese 2004 (41 clusters).. | $0.3$ | $0.61^{+0.03}_{-0.03}$ | –
Jones et al. 2005 (5 clusters).. | $0.3$ | $0.66^{+0.11}_{-0.10}$ | –
Bonamente et al. 2006 (38 clusters).. | $0.3$ | $0.77^{+0.04}_{-0.03}$ | $31.6$
Cunha et al. 2007 (24 triaxial clusters) + BAO.. | $0.273^{+0.03}_{-0.02}$ | $0.738^{+0.042}_{-0.033}$ | $24.5$
This paper (38 clusters)+BAO.. | $0.273^{+0.03}_{-0.02}$ | $0.765^{+0.035}_{-0.033}$ | $35.3$
In Fig. 2(a), we show the confidence regions for the SZE/X-ray cluster
distance and BAO joint analysis. By comparing with Fig. 1(b), one may see how
the BAO signature breaks the degeneracy in the $(\Omega_{\rm{m}},h)$ plane. As
it appears, the BAO test presents a striking orthogonality centered at
$\Omega_{m}=0.274^{+0.036}_{-0.026}$ with respect to the angular diameter
distance data as determined from SZE/X-ray processes. We find
$h=0.765^{+0.035}_{-0.033}$ at $68.3$% c.l. and
$\Omega_{m}=0.273^{+0.03}_{-0.02}$ at $68.3$% c.l. for $1$ free parameter
($\chi^{2}_{min}=35.3$). In light of these results, the important lesson here
is that the combination of SZE/X-ray with BAO provides an interesting approach
to constrain the Hubble constant.
In Fig. 2(b), we have plotted the likelihood function for the $h$ parameter in
a flat $\Lambda$CDM universe for the SZE/X-ray + BAO data set. The dotted
lines are cuts in the regions of $68.3$% probability and $95.4$% (1 free
parameter).
At this point, it is interesting to compare our results with others recent
works. In table 1, a list of recent Hubble parameter determinations based on
cluster data is displayed. Note that in the first five works the $h$ values
were obtained by fixing the cosmology ($\Omega_{m}=0.3$,
$\Omega_{\Lambda}=0.7$), while in our estimate the SZE/X-ray + BAO technique
was used (no one specific cosmology has been fixed).
On the other hand, the importance of the intrinsic geometry of the cluster has
been emphasized by many authors[32, 33, 34, 35]. As a consequence, the
standard spherical geometry has been severely questioned, since Chandra and
XMM-Newton observations have shown that clusters usually exhibit an elliptical
surface brightness. In this concern, a previous determination of $H_{0}$ from
SZE/X-ray + BAO by Cunha et al.[29] was based in a smaller sample (25 triaxial
clusters) observed by De Fillipis and collaborators[22]. Their results
suggested that 15 clusters are in fact more elongated along the line of sight,
while the remaining 10 clusters are compressed. Actually, the assumed cluster
shape seems to affect considerably the SZE/X-ray distances, and, therefore,
the $H_{0}$ estimates.
As shown in table 1, the central value of $H_{0}$ in this case (triaxial
geometry) is exactly the same determined by Riess et al.[10] by using a hybrid
sample (Cepheids + Supernovae). It is also closer to the Hubble Space
Telescope (HST) key project determination of the Hubble parameter announced by
Friedman et al.[36] based only on Cepheids as distance calibrators.
## 3 Comments and Conclusions
In this paper we have discussed a new determination of the Hubble constant
based on the SZE/X-ray distance technique for a sample of 38 clusters as
compiled by Bonamente et al. assuming spherical symmetry[23]. The degeneracy
on the $\Omega_{m}$ parameter was broken trough a joint analysis applying the
baryon acoustic oscillation signature from the SDSS catalog. The Hubble
constant was constrained to be $h=0.765^{+0.035}_{-0.033}$ for $1\sigma$ and
$\Omega_{m}=0.273^{+0.03}_{-0.02}$. These limits were derived assuming the
flat $\Lambda$CDM scenario and a spherical $\beta$-model. The central $h$
value derived here is in perfect agreement with the Bonamente et al. value,
$h=0.769$ (assuming a hydrostatic equilibrium and fixing the concordance
model), as well as with others recent estimates coming from WMAP and Hubble
Space telescope Key Project, where $h=0.73$. However, it differs slightly of a
recent estimation by Cunha, Marassi & Lima, $h=0.74$, where the SZE/X-Ray +
BAO technique was applied for 24 triaxial galaxies clusters. In general
ground, such results are suggesting that the combination of these three
independent phenomena (SZE/X-ray and BAO) provides an interesting method to
constrain the Hubble constant.
## 4 Acknowledgements
The authors are very grateful to the organizers of the IWARA (International
Workshop on Relativistic Astrophysics, João Pessoa, Brazil, 2007). RFLH and
JVC are supported by FAPESP No. 07/5291-2 and 05/02809-5, respectively. JASL
is supported by CNPq and FAPESP No. 04/13668-0 (Brazilian Research Agencies).
## References
* [1] S. Perlmutter et al., Nature, 391, 51 (1998); S. Perlmutter et al., ApJ. 517, 565 (1999); A. G. Riess et al., Astron. J. 116, 1009 (1998).
* [2] P. Astier et al., Astron. Astrophys. 447, 31 (2006); A. G. Riess et al., Astrophys. J. 659, 98 (2007); T. M. Davis et al., Astrophys. J. 666, 716 (2007).
* [3] D. N. Spergel et al., Astrop. J. Suppl. 148, 175 (2003); D. N. Spergel et al., Astrophys. J. Supl. 170, 377 (2007); J. Dunkley et al., arXiv:0803.0586 [astro-ph]; E. Komatsu et al., [ArXiv: 0803.0547v1].
* [4] R. G. Carlberg et al., ApJ. 462, 32 (1996); A. Dekel, D. Burstein and S. D. M. White, In Critical Dialogues in Cosmology, edited by N. Turok World Scientific, Singapore (1997); P. J. E. Peebles, in Formation of Structure in the Universe, edited by A. Dekel and J. P. Ostriker, Cambridge UP, Cambridge (1999).
* [5] D. J. Eisenstein et al., Astrophys. J. 633, 560 (2005).
* [6] S. W. Allen, R. W. Schmidt and A. C. Fabian, MNRAS 334, L11 (2002); S. Ettori, P. Tozzi and P. Rosati, A&A 398, 879 (2003); J. A. S. Lima, J. V. Cunha and J. S. Alcaniz, Phys. Rev. D 68, 023510 (2003); J. V. Cunha, L. Marassi and R. C. Santos, IJMPD 16, 403 (2007).
* [7] J. Dunlop et al., Nature 381, 581 (1996); Y. Yoshii, T. Tsujimoto and K. Kawara, ApJ 507, L133 (1998); J. S. Alcaniz and J. A. S. Lima, ApJ 521, L87 (1999)[astro-ph/9902298]; J. S. Alcaniz, J. A. S. Lima and J. V. Cunha, MNRAS 340, L39 (2003)[astro-ph/0301226]; J. V. Cunha and R. C. Santos, IJMP D 13, 1321 (2004)[astro-ph/0402169]; A. Friaca, J. S. Alcaniz and J. A. S. Lima, Mon. Not. R. Astron. Soc. 362, 1295 (2005), [astro-ph/0504031]; J. F. Jesus, Gen. Rel. Grav. 40, 791 (2008), [astro-ph/0603142]; J. A. S. Lima, J. F. Jesus and J. V. Vunha, arXiv:0709.2195 [astro-ph].
* [8] P. J. E. Peebles and B. Ratra Rev. Mod. Phys. 75, 559 (2003).
* [9] T. Padmanabhan, Phys. Rept. 380, 235 (2003).
* [10] A. G. Riess et al., Astrophys. J., 627, 579 (2005).
* [11] A. Sandage et al., Astrophys. J., 653, 843 (2006).
* [12] R. A. Sunyaev and Ya. B. Zel’dovich, Comments Astrophys. Space Phys. 4 (1972) 173.
* [13] N. Itoh, Y. Kohyama and S. Nozawa, Astrophys. J. 502, 7 (1998).
* [14] The SZE and X-ray surface brightness are sensible to different combinations of the cluster electron gas density $n_{e}$ and its temperature $T_{e}$.
* [15] J. Silk and S. D. M. White, Astrophys. J. 226, L103 (1978).
* [16] J. G. Bartlett and J. Silk, Astrophys. J. 423, 12 (1994); J. G. Bartlett, Astrop. Space Sci. 290, 105 (2004).
* [17] A. Cavaliere, L. Danese and G. De Zotti, Astron. and Astrophys. 75, 322 (1979).
* [18] M. Birkinshaw, Phys. Rep. 310 (1999) 97.
* [19] J. E. Carlstrom, G. P. Holder and E. D. Reese, ARA&A 40, 643 (2002).
* [20] E. D. Reese et al., Astrophys. J. 581, 53 (2002).
* [21] M. E. Jones et al., MNRAS 357, 518 (2002).
* [22] E. De Filippis, M. Sereno, M.W. Bautz and G. Longo, Astrophys. J. 625, 108 (2005).
* [23] M. Bonamente et al., Astrophys. J. 647, 25 (2006).
* [24] S. LaRoque et al. Astrophys. J. 652, 917 (2006).
* [25] J. G. Bartlett and J. Silk, Astrophys. J. 423, 12 (1994) .
* [26] Y. Rephaeli, ARA&A, 33, 541 (1995).
* [27] S. Kobayashy, S. Sasaki and Y. Suto, PASJ 48 L107 (1996).
* [28] D. J. Eisenstein, Astrophys. J. 633, 560 (2005).
* [29] J. V. Cunha, L. Marassi and J. A. S. Lima, MNRAS 379 (2007) L1[astro-ph/0611934].
* [30] Note that we have replaced $H_{0}^{-1}=3000h^{-1}$ Mpc. See for instance, E. W. Kolb and M. S. Turner, The Early Universe, Addison-Wesley, New York (1990).
* [31] J. A. S. Lima and J. S. Alcaniz, Astron. Astrophys. 357, 393 (2000)[astro-ph/0003189]; Astrophys. J. 566, 15 (2002)[astro-ph/0201056]; Astrophys. J. 618, 16 (2005)[astro-ph/0308465]; Z-H. Zhu and M.-K. Fujimoto, Astrophys. J., 581, 1 (2002); J. S. Alcaniz, Phys. Rev. D 69, 083521 (2004)[astro-ph/0312424]; J. V. Cunha, J. S. Alcaniz and J. A. S. Lima, Phys. Rev. D 69, 083501 (2004)[astro-ph/0306319].
* [32] D. C. Fox and U.-L. Pen, Astrophys. J. 574, 38 (2002).
* [33] Y. P. Jing and Y. Suto, Astrophys. J. 574, 538 (2002).
* [34] M. Plionis, S. Basilakos and C. Ragone-Figueroa, Astrophys. J. 650, 770 (2006).
* [35] M. Sereno, E. De Filippis, G. Longo and M. W. Bautz, Astrophys. J. 645, 170 (2006).
* [36] W. L. Freedman et al., Astrophys. J., 553, 47 (2001).
| arxiv-papers | 2008-07-03T21:09:17 | 2024-09-04T02:48:56.540561 | {
"license": "Public Domain",
"authors": "R. F. L. Holanda, J. V. Cunha and J. A. S. Lima",
"submitter": "Rodrigo Holanda",
"url": "https://arxiv.org/abs/0807.0647"
} |
0807.0669 | # Complete one-loop electroweak corrections to $ZZZ$ production at the ILC
Su Ji-Juan, Ma Wen-Gan, Zhang Ren-You, Wang Shao-Ming, and Guo Lei
Department of Modern Physics, University of Science and Technology
of China (USTC), Hefei, Anhui 230027, P.R.China
###### Abstract
We study the complete ${\cal O}(\alpha_{ew})$ electroweak (EW) corrections to
the production of three $Z^{0}$-bosons in the the framework of the standard
model(SM) at the ILC. The leading order and the EW next-to-leading order
corrected cross sections are presented, and their dependence on the colliding
energy $\sqrt{s}$ and Higgs-boson mass $m_{H}$ is analyzed. We investigate
also the LO and one-loop EW corrected distributions of the transverse momentum
of final $Z^{0}$ boson, and the invariant mass of $Z^{0}Z^{0}$-pair. Our
numerical results show that the EW one-loop correction generally suppresses
the tree-level cross section, and the relative correction with
$m_{H}=120~{}GeV(150~{}GeV)$ varies between $-15.8\%(-13.9\%)$ and
$-7.5\%(-6.2\%)$ when $\sqrt{s}$ goes up from $350~{}GeV$ to $1~{}TeV$.
PACS: 11.15.Ex ,12.15.Lk ,13.66.Jn ,14.70.Hp
## I. Introduction
To discover the signature of new physics beyond the standard model(SM)[1, 2]
is one of the main goals for the forthcoming collider experiments. The
precision measurements of the trilinear gauge-boson couplings are helpful for
verification of non-abelian gauge structure, and the investigation of the
quartic gauge-boson couplings can either confirm the symmetry breaking
mechanism or present the direct test on the new physics beyond the SM[3]. The
direct study of quartic gauge-boson couplings requires the investigations of
the processes involving at least three external gauge-bosons. In Refs.[4, 5,
6, 7, 8], the precise predictions for the $VVV$ productions at hadron
colliders were provided. It shows that the QCD corrections increase the $WWZ$
cross section at the LHC by more than $70\%$, and the QCD corrections to $ZZZ$
production at the LHC increase the LO cross section by about $50\%$ [6]. Thus,
any quantitative measurement of the concerned gauge couplings will have to
take QCD corrections into account,
Due to heavy backgrounds, the precise measurement at a hadron collider is more
difficult than at linear collider. The proposed International Linear Collider
(ILC) by the particle physics community will be built with the entire
colliding energy in the range of $200~{}GeV<\sqrt{s}<500~{}GeV$ and an
integrated luminosity of around $500~{}(fb)^{-1}$ in four years. The machine
should be upgradeable to $\sqrt{s}\sim 1~{}TeV$ with an integrated luminosity
of $1~{}(ab)^{-1}$ in three years[9]. Among all the ILC physics goals, the
verification of gauge theory in the SM and finding the evidence of new physics
via experimental measurement of the electroweak gauge boson couplings are
crucial tasks too. The measurement will be able to be improved considerably at
ILC compared with at Fermilab Tevatron and CERN LHC, and therefore a precise
understand of the SM phenomenology at ILC to at least one-loop order is
necessary[10]. Without the accurate theoretical predictions and reliable error
estimates for important observables at ILC, it is impossible to interpret
experimental data properly.
The process of $ZZZ$ production with the subsequential leptonic decays of
vector bosons at ILC is not only an important process as a background for
various new physics processes, but also possible to provide further tests for
the quadrilinear gauge boson couplings, including the four gauge boson
coupling, such as between $ZZZZ$, which does not exist at tree-level in the
SM, because this kind of quadrilinear couplings would induce deviations from
the SM predicted observables[11].
In this paper we present the calculations of the cross sections for the
process $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ at the leading order(LO) and involving
complete electroweak (EW) one-loop (${\cal O}(\alpha_{ew})$) corrections. The
paper is organized as follows: In the next section we present the calculation
descriptions for the tree-level process $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ . The
calculation of the electroweak corrections at one-loop level is reported in
section III. The numerical results and discussions are given in section IV. In
the last section we give a short summary.
## II. Leading-order $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ process
In the calculations of the tree-level and one-loop level cross sections for
the $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ process, we use the ’t Hooft-Feynman
gauge. The analytically calculation of the leading order cross section for
$e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ process is presented in this section. We
describe the lowest order $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ process adopted for
evaluating the cross section as
$e^{+}(p_{1})+e^{-}(p_{2})\to Z^{0}(p_{3})+Z^{0}(p_{4})+Z^{0}(p_{5}),$ (2.1)
where $p_{i}~{}(i=1-5)$ label the four-momenta of incoming positron, electron
and outgoing $Z^{0}$-bosons, respectively. Since the mass of electron/positron
is negligible comparing with the colliding energy and the Yukawa coupling
strength between Higgs/Goldstone and fermions is proportional to the fermion
mass, in our work we ignore the contributions of the Feynman diagrams
involving the couplings of $H^{0}-e^{+}-e^{-}$ and $G^{0}-e^{+}-e^{-}$. We
depict the tree-level Feynman diagrams contributing to the cross section of
the production process of three $Z^{0}$-bosons at the ILC in Fig.1. There we
have 9 generic tree-level diagrams for the process $e^{+}e^{-}\to
Z^{0}Z^{0}Z^{0}$ in the framework of the SM. All the Born-level diagrams can
be grouped in two different topologies. Figs.1(a-c) belong to s-channel,
Figs.1(d-i) are grouped in t(u)-channel.
Figure 1: The generic tree-level Feynman diagrams for $e^{+}e^{-}\to
Z^{0}Z^{0}Z^{0}$ process.
The differential cross section for the process $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$
at the tree-level is then obtained as
$\displaystyle d\sigma_{LO}=\frac{1}{3!}\frac{1}{4}\sum_{spin}|{\cal
M}_{LO}|^{2}d\Phi_{3},$ (2.2)
where ${\cal M}_{LO}$ is the amplitude of all the tree-level diagrams in
Fig.1. The factors $\frac{1}{3!}$ and $\frac{1}{4}$ are due to three identical
final $Z^{0}$-bosons and spin-averaging of the initial particles,
respectively. The summation in Eq.(2.2) is taken over the spins of the initial
and final particles, and $d\Phi_{3}$ is the three-particle phase space element
defined as
$\displaystyle
d\Phi_{3}=\delta^{(4)}\left(p_{1}+p_{2}-\sum_{i=3}^{5}p_{i}\right)\prod_{j=3}^{5}\frac{d^{3}\textbf{{p}}_{j}}{(2\pi)^{3}2E_{j}}.$
(2.3)
## III. Electroweak (${\cal O}(\alpha_{ew})$) corrections
The ${\cal O}(\alpha_{ew})$ order electroweak corrections to the Born-level
$e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ process consist of two parts, i.e.,
* •
The virtual contributions to the leading order process $e^{+}e^{-}\to
Z^{0}Z^{0}Z^{0}$ from the electroweak one-loop and their corresponding
counterterm diagrams;
* •
The contribution from the real photon emission process $e^{+}e^{-}\to
Z^{0}Z^{0}Z^{0}$ $\gamma$. The soft photon emission process $e^{+}e^{-}\to
Z^{0}Z^{0}Z^{0}$ ($\gamma$) consists IR singularities, which will be cancelled
by the IR singularities in the contributions of the one-loop diagrams. There
is no collinear IR singularity since we keep the nonzero mass of
electron(positron);
In the following subsections, we describe in detail the calculation procedure
and discuss the calculation of each contribution part.
### III..1 Virtual corrections
There are totally 2313 electroweak one-loop and corresponding counterterm
Feynman diagrams being taken into account in our calculation, and they can be
classified into self-energy, triangle, box, pentagon and counterterm diagrams.
We depict some representative samples among 66 pentagon diagrams in Fig.2. In
our calculation of the electroweak (${\cal O}(\alpha_{ew})$) corrections to
$e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ process, all the one-loop Feynman diagrams and
their relevant amplitudes are created by using $FeynArts~{}3.3$[12], and the
Feynman amplitudes are subsequently implemented by applying FormCalc5.3
programs[13] and our in-house routines. The electroweak one-loop amplitude
involves five point tensor integrals up to rank 4\. The numerical calculation
of the integral functions($n\leq 4$) are implemented by using the expressions
presented in Refs.[14, 15]. We use our independent Fortran subroutines
following the expressions for the scalar and tensor five-point integrals in
Ref.[16], and find agreement with LoopTools2.2[13]. The Grace2.2.1 program[17]
is used to accomplish five-body phase-space integration for hard photon
radiation process $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}\gamma$ .
Figure 2: Some representative pentagon Feynman diagrams for the process
$e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ .
The total unrenormalized amplitude corresponding to all the one-loop Feynman
diagrams contains both ultraviolet (UV) and infrared (IR) singularities. We
regulate all singularities adopting dimensional regularization(DR) scheme [21]
where the dimensions of spinor and space-time manifolds are extended to
$D=4-2\epsilon$. The relevant fields are renormalized by taking the on-mass-
shell (OMS) scheme [22, 23]. The IR singularity is regularized by introducing
a infinitesimal fictitious mass $m_{\gamma}$. All the tensor coefficients of
the one-loop integrals can be calculated by using the reduction formulae
presented in Refs.[16, 24]. As we expect, the UV divergence contributed by
virtual one-loop diagrams can be cancelled by that contributed from the
counterterms exactly both analytically and numerically.
### III..2 Real photon emission process $e^{+}e^{-}\to
Z^{0}Z^{0}Z^{0}\gamma$
In our calculation for one-loop diagrams, there exists soft IR divergence. In
order to get an IR finite cross section for $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ up
to the order of ${\cal O}(\alpha_{ew}^{4})$, we should consider the ${\cal
O}(\alpha_{ew})$ corrections to $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ process due to
real photon emission. The soft IR divergence in virtual photonic corrections
for the process $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ can be exactly cancelled by
adding the real photonic bremsstrahlung corrections to this process in the
soft photon limit. In the real photon emission process
$\displaystyle e^{+}(p_{1})+e^{-}(p_{2})\to
Z^{0}(p_{3})+Z^{0}(p_{4})+Z^{0}(p_{5})+\gamma(p_{6}),$ (3.1)
a real photon radiates from the electron/positron, and can be soft or hard.
The general phase-space-slicing (PSS) method [25] is adopted to isolate the
soft photon emission singularity part in the real photon emission process
$e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}\gamma$ , and the cross section of the real
photon emission process (3.1) is decomposed into soft and hard terms
$\Delta\sigma_{{real}}=\Delta\sigma_{{S}}+\Delta\sigma_{{H}}=\sigma_{LO}(\delta_{{S}}+\delta_{{H}}).$
(3.2)
where the ’soft’ and ’hard’ describe the energy nature of the radiated photon.
The energy $E_{6}$ of the radiated photon in the center of mass system(c.m.s.)
frame is considered soft if $E_{6}\leq\Delta E$, and hard if $E_{6}>\Delta E$,
respectively. Then both $\sigma_{{S}}$ and $\sigma_{{H}}$ should depend on the
arbitrary soft cutoff $\delta_{s}\equiv\Delta E/E_{b}$, where $E_{b}$ is the
electron beam energy in the c.m.s. frame and equals to $\sqrt{s}/2$, but the
total cross section of the real photon emission process $\sigma_{{real}}$ is
cutoff $\Delta E/E_{b}$ independent. Since the soft cutoff $\Delta E/E_{b}$ is
taken to be a small value in our calculations, the terms of order $\Delta
E/E_{b}$ can be neglected and the soft contribution can be evaluated by using
the soft photon approximation analytically [22, 23, 26]
$\displaystyle{d}\Delta\sigma_{S}=-d\sigma_{{LO}}\frac{\alpha_{ew}}{2\pi^{2}}\int_{|\vec{p}_{6}|\leq\Delta
E}\frac{d^{3}p_{6}}{2E_{6}}\left(\frac{p_{1}}{p_{1}\cdot
p_{6}}-\frac{p_{2}}{p_{2}\cdot p_{6}}\right)^{2}.$ (3.3)
Our calculation demonstrates that the IR singularity in the soft contribution
from Eq.(3.3) is cancelled exactly with that from the virtual photonic
corrections. Therefore, $\Delta\sigma_{v}+\Delta\sigma_{real}$, the sum of the
${\cal O}(\alpha_{ew}^{4})$ cross section corrections from virtual, soft and
hard photon emission contribution parts, is independent of the cutoff value
$\delta_{s}$. The hard contribution, which is UV and IR finite, is computed by
using the Monte Carlo technique. Finally, the electroweak corrected cross
section for the $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ process up to the order of
${\cal O}(\alpha^{4}_{ew})$ can be obtained by
$\displaystyle\sigma_{{tot}}=\sigma_{LO}+\Delta\sigma_{v}+\Delta\sigma_{real}=\sigma_{LO}\left(1+\delta_{tot}\right).$
(3.4)
### III..3 QED and total ${\cal O}(\alpha_{ew})$ order corrections
In analyzing the originations of the electroweak corrections, we split the
full ${\cal O}(\alpha_{ew})$ corrections to the process $e^{+}e^{-}\to
Z^{0}Z^{0}Z^{0}$ into two parts, the QED correction part and the weak
correction part. Correspondingly we define the total relative correction as
$\delta_{tot}=\delta_{QED}+\delta_{weak}$. The QED correction part is
contributed by the diagrams with virtual photon in loop, and real photon
emission process $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}\gamma$ . For the counterterms
involved in the QED contribution, the electron/positron wave function
renormalization constants include only photonic contribution. The remainders
of the total virtual electroweak corrections belong to weak correction part.
With above definitions we can express the full one-loop electroweak corrected
total cross section as a summation of several parts.
$\displaystyle\sigma_{tot}$ $\displaystyle=$
$\displaystyle\sigma_{LO}+\Delta\sigma_{v,QED}+\Delta\sigma_{s}+\Delta\sigma_{h}+\Delta\sigma_{v,weak}$
(3.5) $\displaystyle=$
$\displaystyle\sigma_{LO}\left(1+\delta_{QED}+\delta_{weak}\right)=\sigma_{LO}\left(1+\delta_{tot}\right),$
where $\Delta\sigma_{v}$ and $\Delta\sigma_{s}$ are the cross section
corrections contributed by the virtual electroweak one-loop diagrams and the
soft photon emission process respectively, $\Delta\sigma_{v,QED}$,
$\Delta\sigma_{s}$, $\Delta\sigma_{h}$ and $\Delta\sigma_{v,weak}$ are the
corrections from the virtual QED contribution, the soft photon emission
process, the hard photon emission process and the virtual weak contribution,
separately. $\delta_{QED}$, $\delta_{weak}$ and $\delta_{tot}$ are the
relative corrections contributed by the QED correction part, the weak
correction part and the total electroweak correction, respectively.
As we mentioned above, there exist both ultraviolet(UV) divergency and
infrared(IR) soft singularity in the contributions of the electroweak one-loop
diagrams for $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ process, but no collinear IR
singularity since we keep the nonzero mass of electron/positron in our
calculation of one-loop order corrections. After doing the renormalization
procedure, we verified that the UV singularity is vanished, and the IR soft
divergency appeared in the virtual correction is cancelled by the the soft
photon emission process $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}\gamma$ .
In order to discuss the origin of the large correction when the colliding
energy is close to the threshold of the production of three $Z^{0}$-bosons, we
discuss the photonic (QED) corrections and the genuine total electroweak
corrections separately. The QED corrections comprise two parts: the QED
virtual corrections $\Delta\sigma_{v,QED}$ which contributed by the loop
diagrams with virtual photon exchange in loop and the corresponding QED parts
of the counterterms, and the real photon emission corrections
$\Delta\sigma_{real}$. Therefore, the QED relative correction $\delta_{QED}$
can be expressed as
$\displaystyle\delta_{QED}=\delta_{v,QED}+\delta_{real},$ (3.6)
where $\delta_{v,QED}=\Delta\sigma_{v,QED}/\sigma_{LO}$, and the genuine weak
relative correction $\delta_{w}$ can be got from
$\displaystyle\delta_{w}=\delta_{tot}-\delta_{QED}.$ (3.7)
## IV. Numerical results and discussion
For the numerical calculation, we take the input parameters as follows[27]:
$\displaystyle m_{e}$ $\displaystyle=$ $\displaystyle 0.51099892~{}{\rm
MeV},~{}m_{\mu}~{}=~{}105.658369~{}{\rm MeV},~{}m_{\tau}~{}=~{}1776.99~{}{\rm
MeV},$ $\displaystyle m_{u}$ $\displaystyle=$ $\displaystyle 66~{}{\rm
MeV},~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}m_{c}~{}=~{}1.25~{}{\rm
GeV},~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}m_{t}~{}=~{}174.2~{}{\rm GeV},$
$\displaystyle m_{d}$ $\displaystyle=$ $\displaystyle 66~{}{\rm
MeV},~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}m_{s}~{}=~{}95~{}{\rm
MeV},~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}m_{b}~{}=~{}4.7~{}{\rm GeV},$
$\displaystyle m_{W}$ $\displaystyle=$ $\displaystyle 80.403~{}{\rm
GeV},~{}~{}~{}~{}~{}~{}~{}~{}~{}m_{Z}~{}=~{}91.1876~{}{\rm GeV}.$ (4.1)
There we use the experimental value of W-boson mass as input parameter, but
not the $m_{W}$ evaluated from $G_{\mu}$ as in $\alpha_{ew}$ scheme[28]. We
take the electric charge defined in the Thomson limit
$\alpha_{ew}(0)=1/137.036$ and the effective values of the light quark masses
($m_{u}$ and $m_{d}$) which can reproduce the hadronic contribution to the
shift in the fine structure constant $\alpha_{ew}(m_{Z}^{2})$ [29]. As we know
that the LEP II experiments provide the lower limit on the SM Higgs mass as
$114.4~{}GeV$ at the $95\%$ confidence level from the results of direct
searches for $e^{+}e^{-}\to Z^{0}H^{0}$ production[30, 31], and the
electroweak precision measurements indicate indirectly the upper bound as
$m_{H}\lesssim 182~{}GeV$ at the $95\%$ C.L., when the lower limit on $m_{H}$
is used in determination of this upper limit[31]. Therefore, in our numerical
evaluation it is reasonable to take the mass of Higgs-boson being in the range
of $115~{}GeV<m_{H}<170~{}GeV$. Then we shall not encounter the resonance
problem of Higgs-boson during our calculation.
We checked the correctness of the numerical results of the LO cross section
for process $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ , by using Grace2.2.1[17] and
FeynArts3.3/FormCalc5.3 [12, 13] packages separately. In adopting Grace2.2.1
and FeynArts3.3/FormCalc5.3 programs, we used both ’t Hooft-Feynman and
unitary gauges separately in the calculations of the LO cross section to check
the gauge invariance, and we got coincident numerical results. The numerical
results of the LO cross section for the process $e^{+}e^{-}\to
Z^{0}Z^{0}Z^{0}$ , by using ’t Hooft-Feynman gauge and taking
$\sqrt{s}=500~{}GeV$ and the other input parameters shown in Eqs.(IV.), are
listed in Table 1. There it is shown that there is a good agreement between
the numerical results by adopting different packages.
$m_{H}(GeV)$ | $\sigma_{LO}(fb)$(Grace) | $\sigma_{LO}(fb)$ (FeynArts)
---|---|---
115 | 1.0056(4) | 1.0055(2)
120 | 1.0139(4) | 1.0138(2)
150 | 1.0975(4) | 1.0975(2)
170 | 1.2565(4) | 1.2564(2)
Table 1: The numerical results of the LO cross sections for the process
$e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ by using Grace2.2.1 and
FeynArts3.3/FormCalc5.3 packages separately, and taking the input parameters
as shown in Eqs.(IV.) and $\sqrt{s}=500~{}GeV$.
In the one-loop calculation, we must set the values of the IR regulator
$m_{\gamma}$, the fictitious photon mass, and soft cutoff $\delta_{s}=\Delta
E/E_{b}$ besides the parameters mentioned in Eqs.(IV.). As we know, the total
cross section should have no relation with these two parameters if the IR
divergency does really vanish. Our numerical results show that the cross
section correction at ${\cal O}(\alpha^{4}_{ew})$ order
$\Delta\sigma_{tot}=\Delta\sigma_{real}+\Delta\sigma_{v}$ is invariable within
the calculation errors when $m_{H}=120~{}GeV$, $\sqrt{s}=500~{}GeV$,
$\delta_{s}=10^{-3}$ and the fictitious photon mass $m_{\gamma}$ varies from
$10^{-15}~{}GeV$ to $10^{-1}~{}GeV$.
Fig.3(a) presents a verification of the correctness of our calculation for
process $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ including ${\cal O}(\alpha_{ew})$
order corrections. The amplified curve for $\Delta\sigma_{tot}$ including
calculation errors is depicted in Fig.3(b). Both figures are to show the
independence of the total ${\cal O}(\alpha_{{\rm ew}})$ electroweak correction
on the soft cutoff $\delta_{s}$, when we take $m_{H}=120~{}{{\rm GeV}}$ and
$\sqrt{s}=500~{}{\rm GeV}$. From Fig.3(b) we can say that the total EW
relative correction $\Delta\sigma_{tot}$ has no relation to the value of
$\delta_{s}$ within the calculation error range. In the further calculations,
we set $m_{\gamma}=10^{-2}~{}GeV$ and $\delta_{s}=10^{-3}$.
Figure 3: (a) The dependence of the ${\cal O}(\alpha_{{\rm ew}})$ correction
to cross section of $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ on the soft cutoff
$\delta_{s}$ with $m_{H}=120~{}GeV$ and $\sqrt{s}=500~{}GeV$. (b) the
amplified plot of the curve for the total correction $\Delta\sigma_{tot}$ in
Fig.3(a), where it includes calculation errors.
In Table 2, we list some representative numerical results of the LO and one-
loop EW corrected cross sections($\sigma_{LO}$, $\sigma_{tot}$), the QED and
total EW corrections to the cross sections for $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$
process($\Delta\sigma_{QED}$, $\Delta\sigma_{tot}$), and their corresponding
relative corrections($\delta_{QED}$, $\delta_{tot}$) when $\sqrt{s}=500~{}GeV$
and the values of Higgs-boson mass are taken to be $115~{}GeV$, $150~{}GeV$
and $170~{}GeV$ separately. From these data we can see that the one-loop EW
corrections suppress the LO cross section of the process $e^{+}e^{-}\to
Z^{0}Z^{0}Z^{0}$ , and the relative corrections are about minus few percent.
We can conclude also the LO and EW corrected cross sections increase with the
increment of $m_{H}$, while the absolute total EW correction decreases when
the value of Higgs-boson mass goes up.
$m_{H}(GeV)$ | $\sigma_{LO}(fb)$ | $\sigma_{tot}(fb)$ | $\Delta\sigma_{QED}(fb)$ | $\Delta\sigma_{tot}(fb)$ | $\delta_{QED}(\%)$ | $\delta_{tot}(\%)$
---|---|---|---|---|---|---
115 | 1.0055(2) | 0.9159(7) | -0.0451(7) | -0.0896(7) | -4.49(7) | -8.91(7)
150 | 1.0975(2) | 1.0194(8) | -0.0444(8) | -0.0780(8) | -4.04(7) | -7.11(7)
170 | 1.2564(2) | 1.1989(9) | -0.0393(8) | -0.0575(9) | -3.12(7) | -4.58(7)
Table 2: The numerical results of $\sigma_{LO}$, $\sigma_{tot}$,
$\Delta\sigma_{QED}$, $\Delta\sigma_{tot}$(in femto bar), and their
corresponding relative EW and QED corrections($\delta_{tot}$, $\delta_{QED}$)
for the process $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ , when $\sqrt{s}=500~{}GeV$
and $m_{H}=115~{}GeV,~{}150~{}GeV,~{}170~{}GeV$ respectively.
The numerical results of the LO, ${\cal O}(\alpha_{ew})$ EW, QED corrected
cross sections($\sigma_{LO}$, $\sigma_{tot}$, $\sigma_{QED}$) and the total
relative EW, QED corrections($\delta_{tot}$, $\delta_{QED}$) for the process
$e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ with $m_{H}=120~{}GeV,~{}150~{}GeV$ as the
functions of colliding energy $\sqrt{s}$ are plotted in Figs.4(a) and (b)
respectively. As indicated in Fig.4(a), The curves for the cross sections of
$\sigma_{LO}$, $\sigma_{tot}$ and $\sigma_{QED}$ increase quickly in the
$\sqrt{s}$ region of $[350~{}GeV,~{}550~{}GeV]$ and decrease when
$\sqrt{s}>600~{}GeV$. The two figures show the ${\cal O}(\alpha_{ew})$
corrections always suppress the corresponding LO cross sections of process
$e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ in both cases of $m_{H}=120~{}GeV$ and
$m_{H}=150~{}GeV$ separately, but the QED correction parts can enhance the LO
cross sections when $\sqrt{s}>700~{}GeV$. We can read out from Fig.4(b) that
the corresponding total EW relative corrections for $m_{H}=120~{}GeV$ and
$150~{}GeV$ vary in the ranges of $[-15.8\%,~{}-7.5\%]$ and
$[-13.9\%,~{}-6.2\%]$ respectively, when $\sqrt{s}$ runs from $350~{}GeV$ to
$1~{}TeV$. We can see also from these two plots that in the colliding energy
$\sqrt{s}$ region near the threshold of the production of three
$Z^{0}$-bosons, the main contribution to the total EW relative
correction($\delta_{tot}$) comes from the QED correction part. That is due to
the Coulomb singularity effect coming from the instantaneous photon exchange
in loops which has a small spatial momentum.
Figure 4: (a) The LO($\sigma_{LO}$), ${\cal O}(\alpha_{ew})$ EW, QED
corrected cross sections($\sigma_{tot}$, $\sigma_{QED}$) for the process
$e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ as the functions of colliding energy
$\sqrt{s}$ with $m_{H}=120~{}GeV$, $150~{}GeV$ separately. (b) The
corresponding relative EW, QED relative corrections($\delta_{tot}$,
$\delta_{QED}$) versus $\sqrt{s}$.
We present the distributions of the transverse momenta of final $Z^{0}$-bosons
at leading order and up to one-loop order, $d\sigma_{LO}/dp_{T}^{Z}$,
$d\sigma_{NLO}/dp_{T}^{Z}$, when $m_{H}=120~{}GeV$ and $\sqrt{s}=500~{}GeV$ in
Fig.5. There we pick the $p_{T}^{Z}$ of each of the three $Z^{0}$-bosons as an
entry in this histograms, then the final result of the differential cross
section is obtained by multiplying factor $1/3$. In this figure we can see
that the EW one-loop correction suppresses obviously the LO differential cross
section $d\sigma_{LO}/dp_{T}^{Z}$ when $p_{T}^{Z}>50~{}GeV$, but the EW
correction is small when $p_{T}^{Z}<25~{}GeV$. It also shows that the EW
corrections do not observably change the LO distribution line-shape of
$p_{T}^{Z}$, and both the differential cross sections of
$d\sigma_{LO}/dp_{T}^{Z}$ and $d\sigma_{NLO}/dp_{T}^{Z}$ have their maximal
values at about $p_{T}^{Z}\sim 50~{}GeV$ respectively.
Figure 5: The distributions of the transverse momenta of $Z^{0}$-bosons
($P_{T}^{Z}$) at the LO and up to EW one-loop order with $m_{H}=120~{}GeV$ and
$\sqrt{s}=500~{}GeV$.
We plot the distributions of the invariant mass of $Z^{0}Z^{0}$-pair, denoted
as $M_{ZZ}$, at the LO and up to EW one-loop order with $m_{H}=120~{}GeV$ and
$\sqrt{s}=500~{}GeV$ in Fig.6. Here we can see that the EW correction slightly
enhances the LO differential cross section when $M_{ZZ}<250~{}GeV$, but
suppresses $d\sigma_{LO}/dM_{ZZ}$ obviously when $M_{ZZ}>250~{}GeV$. The
suppression of $d\sigma_{LO}/dM_{ZZ}$ is due to the contribution from the hard
photon emission process $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}\gamma$ at the ${\cal
O}(\alpha_{ew}^{4})$ order, in which the momentum balance between the sum of
the momenta of three $Z^{0}$-bosons and that of the radiated hard photon will
reduce the value of invariant mass $M_{ZZ}$ and change the line-shape in the
range with large $M_{ZZ}$.
Figure 6: The distributions of the invariant mass of
$Z^{0}Z^{0}$-pair($M_{ZZ}$) at the LO and up to EW one-loop order when
$m_{H}=120~{}GeV$ and $\sqrt{s}=500~{}GeV$.
## V. Summary
In this paper we describe the impact of the complete one-loop EW corrections
to the scattering process $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ in the SM. This
channel can be used to measure the quartic vector boson coupling at ILC. We
investigate the dependence of the LO, ${\cal O}(\alpha_{ew})$ EW and QED
corrected cross sections on colliding energy $\sqrt{s}$ and Higgs-boson mass,
and present the LO and EW one-loop corrected distributions of the transverse
momenta of final $Z^{0}$-bosons and the LO and EW corrected differential cross
sections of invariant mass of $Z^{0}Z^{0}$-pair. To see the origin of some of
the large corrections clearly, we calculate the QED and genuine weak
corrections separately. We conclude that both the Born cross section and the
EW corrected cross section for $e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ process are
sensitive to the Higgs boson mass in the range of $115~{}GeV<m_{H}<170~{}GeV$.
We find the ${\cal O}(\alpha_{ew})$ corrections generally suppress the LO
cross section, the LO distribution of the momenta of $Z^{0}$-bosons and the LO
differential cross sections of invariant mass of $Z^{0}Z^{0}$-pair for process
$e^{+}e^{-}\to Z^{0}Z^{0}Z^{0}$ . Our numerical results show that when
$m_{H}=120~{}GeV(150~{}GeV)$ and the colliding energy goes up from $350~{}GeV$
to $1~{}TeV$, the relative EW correction varies from $-15.8\%(-13.9\%)$ to
$-7.5\%(-6.2\%)$.
Acknowledgments: This work was supported in part by the National Natural
Science Foundation of China, Specialized Research Fund for the Doctoral
Program of Higher Education(SRFDP) and a special fund sponsored by Chinese
Academy of Sciences.
## References
* [1] S. L. Glashow, Nucl. Phys. B 22 (1961) 579 ; S. Weinberg, Phys. Rev. Lett. 19 (1967) 1264; A. Salam, in Proceedings of the 8th Nobel Symposium, Stockholm, 1968, edited by N. Svartholm (Almquist and Wiksells, Stockholm, 1968), p.367; H. D. Politzer, Phys. Rep. 14 (1974) 129.
* [2] P. W. Higgs, Phys. Lett 12 (1964) 132, Phys. Rev. Lett. 13 (1964) 508 ; Phys. Rev. 145 (1966) 1156; F. Englert and R. Brout, Phys. Rev. Lett. 13 (1964) 321; G. S. Guralnik, C. R. Hagen, and T. W. B. Kibble, ibid. 13 (1964) 585 ; T. W. B. Kibble, Phys. Rev. 155 (1967) 1554.
* [3] T. Han, H.-J. He, and C.-P. Yuan, Phys. Lett. B422 (1998) 294, arXiv:hep-ph/9711429v3.
* [4] J. Ohnemus, arXiv:hep-ph/9503389; L.J. Dixon, Z. Kunszt and A. Signer, Phys. Rev. D60 (1999) 114037, arXiv:hep-ph/9907305; L.J. Dixon, Z. Kunszt and A. Signer, Nucl. Phys. B531 (1998) 3, arXiv:hep-ph/9803250; J. Ohnemus, Phys. Rev. D50 (1994) 1931, arXiv:hep-ph/9403331.
* [5] J.M. Campbell and R.K. Ellis, Phys. Rev. D60 (1999) 113006, arXiv:hep-ph/9905386.
* [6] A. Lazopoulos, K. Melnikov and F. Petriello, Phys. Rev. D76 (2007) 014001, arXiv:hep-ph/0703273.
* [7] V. Hankele and D. Zeppenfeld, KA-TP-35-2007, SFB/CPP-07-94, arXiv:hep-ph/0712.3544v1; J. Bagger, V. Barger, K. Cheung, J. Gunion, T. Han, G.A. Ladinsky, R. Rosenfeld, and C.-P. Yuan, Phys. Rev. D49 (1994) 1246; D52 (1995) 3878.
* [8] T. Binoth, G. Ossola, C.G. Papadopoulos and R. Pittau, JHEP 0806, 082 (2008), arXiv:hep-ph/0804.0350v1.
* [9] Parameters for Linear Collider, http://www.fnal.gov/directorate/icfa/LC_parameters.pdf
* [10] O.J.P. Eboli, M.C. Gonzalez-Garcia, S.M. Lietti and S.F. Novaes, Phys. Rev. D63 (2001) 075008, arXiv:hep-ph/0009262; D. Green, arXiv:hep-ex/0310004.
* [11] M.Baillargeon, F. Boudjema, F.Cuypers, E. Gabrielli, and B. Mele, Nucl. Phys. B424 (1994) 343; Y.-J. Zhou,W.-G. Ma, R.-Y. Zhang, Y. Jiang, and L. Han, Phys. Rev. D73 (2006)073009.
* [12] T. Hahn, Comput. Phys. Commun. 140 (2001)418.
* [13] T. Hahn, M. Perez-Victoria, Comput. Phys. Commun. 118 (1999)153.
* [14] G.’t Hooft and M. Veltman, Nucl. Phys. B153 (1979) 365.
* [15] A. Denner, U Nierste and R Scharf, Nucl. Phys. B367 (1991) 637\.
* [16] A. Denner and S. Dittmaier, Nucl. Phys. B658 (2003) 175.
* [17] T. Ishikawa,et al., (MINMI-TATEYA collaboration) ”GRACE User’s manual version 2.0”, August 1, 1994.
* [18] G. ’t Hooft and M. Veltman, Nucl. Phys. B44, (1972) 189.
* [19] D. A. Ross and J. C. Taylor, Nucl. Phys. B51, (1979) 25.
* [20] A. Denner, Fortschr. Phys. 41, (1993) 307, arXiv:hep-ph/0709.1075v1.
* [21] G. ’t Hooft and M. Veltman, Nucl. Phys. B44, (1972) 189.
* [22] D. A. Ross and J. C. Taylor, Nucl. Phys. B51, (1979) 25.
* [23] A. Denner, Fortschr. Phys. 41, (1993) 307.
* [24] G. Passarino and M. Veltman, Nucl. Phys. B160, (1979) 151.
* [25] W. T. Giele and E. W. N. Glover, Phys. Rev. D46, (1992) 1980; W. T. Giele, E. W. Glover and D. A. Kosower, Nucl. Phys. B403, 1993) 633.
* [26] G. ’t Hooft and Veltman, Nucl. Phys. B153, (1979) 365.
* [27] W.M. Yao, et al., J. of Phys. G33(2006) 1.
* [28] G. Eilam, P.R. Mendel, R. Migneron, and A. Soni, Phys. Rev. Letts 66,3105(1991).
* [29] F. Jegerlehner, Report No. DESY 01-029, arXiv:hep-ph/0105283v1.
* [30] R.Barate et al., Phys. Lett. B565 (2003) 61.
* [31] The LEP Collaborations ALEPH, DELPHI, L3, OPAL, and the LEP Electroweak Working Group. LEPEWWG/2007-01 and arxiv:0712.0929.
| arxiv-papers | 2008-07-04T03:27:18 | 2024-09-04T02:48:56.546965 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Su Ji-Juan, Ma Wen-Gan, Zhang Ren-You, Wang Shao-Ming, Guo Lei",
"submitter": "Su Jijuan",
"url": "https://arxiv.org/abs/0807.0669"
} |
0807.0918 | # Enhanced Ferromagnetic Stability in Cu Doped Passivated GaN Nanowires
H. J. Xiang National Renewable Energy Laboratory, Golden, Colorado 80401, USA
Su-Huai Wei National Renewable Energy Laboratory, Golden, Colorado 80401, USA
###### Abstract
Density functional calculations are performed to investigate the room
temperature ferromagnetism in GaN:Cu nanowires (NWs). Our results indicate
that two Cu dopants are most stable when they are near each other. Compared to
bulk GaN:Cu, we find that magnetization and ferromagnetism in Cu doped NWs is
strongly enhanced because the band width of the Cu $t_{d}$ band is reduced due
to the 1D nature of the NW. The surface passivation is shown to be crucial to
sustain the ferromagnetism in GaN:Cu NWs. These findings are in good agreement
with experimental observations and indicate that ferromagnetism in this type
of systems can be tuned by controlling the size or shape of the host
materials.
Dilute magnetic semiconductors (DMSs) have attracted wide interest recently
because they possess both semiconducting and magnetic properties at the same
time, thus suitable for spintronic applications. One of the major focus in
this important research field is to produce DMSs with Curie temperatures
($T_{c}$) at or above room temperature. It has been predicted that the use of
wide band-gap semiconductors, e.g., GaN, as the host material for DMSs can
lead to high $T_{c}$ Dietl2000 . However, experimental studies on transition-
metal (TM)-doped bulk GaN have led to conflicting results. For example, some
experiments have shown the existence of ferromagnetism in Ga1-xMnxN but with
$T_{c}$ varying between 8 K Sarigiannidou2006 and 904 K Sonoda2002 at
$x=0.06$. In contrast, a magneto-optical study Ando2003 suggested that
Ga1-xMnxN is paramagnetic. Theoretical studies also show that the magnetic
properties of bulk Ga1-xMnxN is quite unique: depending on the Mn
concentration, carrier density, and pressure, it can change from
ferromagnetism to antiferromagnetism Dalpian2005 ; Dalpian2006 .
Recently, Wu et al. Wu2006 predicted that two Cu ions in a configuration
about 6.2 Å from each other in bulk GaN are coupled ferromagnetically with the
total moment 2 $\mu_{B}$/Cu. Subsequently, room temperature ferromagnetism was
found in Cu-implanted GaN samples, although with a much smaller saturation
magnetization (from 0.01 to 0.27 $\mu_{B}$/Cu), and the results are very
sensitive to the annealing temperature Lee2007 ; Seipel2007 . A more recent
calculation by Rosa and Ahuja Rosa2007 points out that the ferromagnetic
coupling in bulk GaN:Cu is much weak than expected from previous calculation
Wu2006 because the magnetic moment on Cu is very sensitive to the Cu-Cu
distance; it becomes much smaller when Cu-Cu becomes nearest neighbor.
However, a recent experimental study by Seong et al. reported that room
temperature ferromagnetism can be achieved in Ga1-xCuxN nanowires (NWs)
synthesized in a chemical vapor transport system under flow of NH3 and the
saturation magnetic moment is significantly higher than that in bulk GaN
($\sim$1.00 $\mu_{B}$/Cu at 5 K) Seong2007 . The dramatically different
saturation magnetic moment of Cu in bulk GaN and GaN NWs suggest that the
magnetic behavior of Cu doped GaN could be manipulated by tuning the size or
shape of host. However, the mechanism of the enhanced magnetic coupling in
GaN:Cu NWs is not known, including what is the effect of the surface
passivation in the NWs.
In this Letter, we perform a comprehensive first principles study to
understand the structural and magnetic properties of Cu doping in bulk GaN and
GaN NWs. Both bare and passivated GaN NWs are studied. Our first-principles
spin-polarized density functional theory (DFT) calculations were performed on
the basis of the projector augmented wave method PAW encoded in the Vienna ab
initio simulation package VASP using the generalized-gradient approximation
(GGA) Perdew1996 and the plane-wave cutoff energy of 400 eV. For relaxed
structures, the atomic forces are less than 0.02 eV/Å.
For bulk GaN, the optimized wurtzite lattice constants are a$=3.218$ Å and
c$=5.240$ Å with the internal parameter u=$0.3767$ (experimental values:
a$=3.189$ Å, c$=5.186$ Å, and u=$0.377$). The calculated band gap for bulk GaN
is 1.72 eV. The GaN NWs are orientated along the [0001] direction with a
diameter about 1 nm, as shown in Fig. 1(a) and (b). For passivated NWs,
pseudo-hydrogen is used to saturate the dangling bond of the NW surface. In
the simulation, the NW axis is along the $c$ direction, and the lateral
supercell size is chosen so that the closest distance between two neighbor NWs
is larger than 8 Å. After optimization, the diameters of the GaN NWs become
slightly smaller, and consequently, to minimize the strain, the lattice
constant $c$ is increased by 0.018 Å and 0.046 Å for passivated and bare GaN
NWs, respectively, as in the case of bare ZnO NWs Xiang2006 . Both bare and
passivated NWs are found to have a direct bang gap (1.67 eV and 3.25 eV,
respectively) at $\Gamma$. The small band gap of the bare wire is due to the
formation of surface defect levels. In case of Cu doping in NWs, we use a
$1\times 1\times 2$ supercell with a $1\times 1\times 4$ k-mesh. For
comparison, Cu doping in bulk GaN is also studied by using a $3\times 3\times
2$ supercell (shown in Fig. 1(e)) with a $4\times 4\times 4$ k-mesh. The
lattice constants of the doped systems are fixed to those of the undoped hosts
since substitutional Cu changes only marginally the lattice constant Wu2006 ;
Seipel2007 .
To compare the dopability of Cu in bulk GaN and GaN NWs and identify stable
dopant position, we first calculate the formation energy of Cu substitution
for Ga in GaN host as
$\begin{array}[]{ccl}\Delta
H_{f}&=&E(\mathrm{GaN:Cu})-E(\mathrm{host})+\mu(\mathrm{Ga})-\mu(\mathrm{Cu})\\\
&=&\Delta E+\mu(\mathrm{Ga})-\mu(\mathrm{Cu})\end{array}$ (1)
where $E(\mathrm{GaN:Cu})$ is the total energy of the doped system,
$E(\mathrm{host})$ denotes the total energy of the GaN host for the same
supercell in the absence of the defect, $\mu(\mathrm{Ga})$ and
$\mu(\mathrm{Cu})$ are the chemical potential for Ga and Cu, respectively. We
note that the absolute formation energy depends on $\mu(\mathrm{Ga})$ and
$\mu(\mathrm{Cu})$. However, it is sufficient to calculate $\Delta E$ for the
purpose of comparing the relative stability of Cu in GaN host. For a NW as
shown in Fig. 1(a) and (b), there are three inequivalent Ga positions, i.e.,
A, B, and C. In all cases, we confirm that Cu will not form AX center in GaN
due to the delocalization of the Cu 3d related defect bands. Our results are
reported in Table. 1. For bare GaN NW, we find that the formation energy of
CuGa is smaller than that in bulk GaN, because CuGa can relax more easily in
the NW than in the bulk. The lowering of the formation energy is most dramatic
at the surface of the NW, where the formation energy is about 0.96 eV smaller.
For the passivated GaN NW, we find that the formation energy of CuGa is about
0.3 eV larger than that in bulk GaN and is most stable at the A site. This
result follow the general trends observed in nanocrystal quantum dots Li2008
suggesting that the increased formation energy for this non-isovalent dopants
is mainly due to the quantum confinement-induced band gap increase, but not
sensitive to the shape of the quantum structure. The above results show that
the doping ability of Cu in GaN NW depends sensitively on the surface
termination of the NW. We also find that in all cases with an isolated Cu atom
in GaN hosts, the total magnetic moment is 2 $\mu_{B}$ and the local moment on
Cu is about 0.65 $\mu_{B}$ due to the strong Cu 3d and N 2p hybridization
Wu2006 .
Experimentally, the valence state of Cu in Cu doped GaN NW is found to be
close to that of Cu in CuO Seong2007 . To confirm this theoretically, we
calculate the charge tranfer in Cu doped NWs using atoms in molecules (AIM)
theory Bader . We find that the valence state of Cu dopant in bare
(passivated) GaN NW is about 0.84 (0.88). For CuO and Cu2O, the calculated Cu
valence state is 0.99 and 0.55, respectively. So, the valence state of Cu
dopant in GaN NW is indeed close to that in CuO. The calculated valence state
is not close to the nominal value due to the strong covalency in the system.
The results also indicate that Cu occupies Ga site in the experimental sample
Seong2007 .
To investigate the interaction between Cu dopants in GaN NWs, we substitute
two Ga atoms with two Cu atoms. The most stable configuration of two Cu atoms
in GaN NWs is shown in Fig. 1. We can see that two Cu atoms tend to be next to
each other, i.e., they are bonded to the same N atom. This is due to the
strong bonding interaction between Cu 3d $t$ states. In case of passivated NW,
both Cu atoms occupied A positions. In contrast, these Cu atoms occupied C
positions for the bare GaN NW case. This is understandable since an isolated
Cu atom prefers to occupy the C (A) position in case of bare (passivated) GaN
NW.
The spin exchange interaction between Cu 3d moments in GaN NWs is studied
through calculating different spin states, i.e., ferromagnetic (FM) and
antiferromagnetic (AFM) states. In case of Cu doped passivated GaN NW (Fig.
1(d)), the FM state with a total moment 2.8 $\mu_{B}$ is more favorable over
the AFM state by 90 and 31 meV for the unrelaxed and relaxed structures,
respectively. Surprisingly, for the doped bare GaN NW as shown in Fig. 1(c),
it is a non-magnetic semiconductor with a gap of 0.44 eV within the d bands.
To probe the origin of the different nature of magnetism in these two cases,
we plot the density of states (DOS) in Fig. 2. To facilitate the analysis, the
DOSs of the systems with an isolated Cu dopant are also shown. In both cases,
we can see that there is a band gap in the spin up component. In case of the
passivated NW, the spin down $t$ orbitals (In the crystal field of a wurtzite
system, 3d orbitals are splitted into low-lying two-fold $e$ states and three
nearly degenerated $t$ orbitals) are partially occupied. However, when a Cu
atom occupies the surface position [position C in Fig. 1(a)] of the bare GaN
NW, the different crystal field arising from a missing neighbor N atom at the
surface will split the $t$ orbitals into a low-lying one-fold $a$ and a high-
lying two-fold $e^{\prime}$ sets. The presence of the band gap is consistent
with the stability of the surface Cu atom. When the two Cu dopants are next to
each other [Fig. 1(c)], the coupling between two Cu $t$-derived orbitals will
result in a splitting between the bonding and anti-bonding orbitals. Since the
distance between two Cu atoms in bare NW is only 2.93 Å (the Cu-Cu distance in
bulk GaN and passivated GaN NW are 3.13 Å and 3.12 Å respectively), the
splitting will be larger than the spin exchange splitting, resulting in the
low spin non-magnetic state. Similar results are observed in GaN:Mn, where
pressure or surface strain can turn the system from a high-spin to a low-spin
configuration, diminish the ferromagnetism Dalpian2005 ; Wang2004 . Our
calculations reveal that passivation of the surface is crucial in this system
to obtain ferromagnetism for Cu doped GaN NWs. It is noted that previous
theoretical studies focused only on TM (Mn Wang2005A and Cr Wang2005B ) doped
GaN bare NWs. In addition, the coupling between a pair of Cr atoms substituted
in unsaturated GaN nanoholes was found to be FM Wang2007 .
Table 1: Relative formation energy $\Delta E$ of an isolated CuGa defect in bare and passivated GaN NWs. Please refer to Fig. 1 for the definition of positions A, B, and C. $\Delta E$ is related to the absolute formation energy $\Delta H$ by Eq. 1. For bulk GaN, $\Delta E$ of an isolated CuGa defect is 3.33 eV. | site A | site B | site C
---|---|---|---
Bare NW | 3.05 | 3.27 | 2.37
Passivated NW | 3.52 | 3.68 | 3.52
Figure 1: (a) Top view of a bare GaN NW with a diameter about 1.0 nm, (b) top
view of a GaN NW passivated by pseudo H atoms, (c) side view of the most
stable configuration of two substitutional Cu atoms in the bare GaN NW, (d)
side view of the most stable configuration of two substitutional Cu atoms in a
passivated GaN NW, (e) side view of a $3\times 3\times 2$ bulk GaN supercell,
and (f) top view of a configuration of two substitutional Cu atoms in a
passivated GaN NW with a larger diameter 1.6 nm. A, B, and C in (a) and (b)
denote three inequivalent Ga positions. The numbers in (c) and (d) give the
distance (in Å) between two Cu dopants. Number 1, 2 , and 3 in (e) label
different Ga atoms in the supercell. Figure 2: DOS of (a) a bare GaN NW with
an isolated Cu dopant at the C position (Fig. 1(a)), (b) a bare GaN NW with
two Cu dopants (Fig. 1(c)), (c) a passivated GaN NW with an isolated Cu dopant
at the A position (Fig. 1(b)), (d) a passivated GaN NW with two Cu dopants
(Fig. 1(d)), (e) an isolated Cu dopant in a bulk GaN supercell (Fig. 1(e)),
and (f) two Cu dopants at positions 1 and 2 in a bulk GaN supercell (Fig.
1(e)), respectively. Here, high energy conduction bands of GaN hosts are not
shown.
To compare the different magnetic behavior of Cu doping in bulk GaN and GaN
NW, we also study the Cu-Cu interaction in bulk GaN. First, we find that Cu
dopants in bulk GaN also tend to be nearest neighbor. The most favorable
configuration is that two Cu atoms occupy the in-plane nearest neighbor
positions [positions 1 and 2 in Fig. 1(e)] with a magnetic moment of 0.6
$\mu_{B}$/Cu. However, the energy difference between the FM and the AFM state
in this configuration is only $-1.6$ meV (the value for the unrelaxed
structure is 10.5 meV), which is much smaller than that in passivated NW and
in the case when the Cu-Cu distance is about 6.2 Å Wu2006 . In some other
configurations, we find that the AFM state could be even more stable than the
FM state. For instance, if two Cu atoms occupy positions 1 and 3 [Fig. 1(e)],
the FM state has higher energy by 41 meV. These results indicate that FM in Cu
doped bulk GaN is not robust, which is consistent with experimental
observations and the recent calculations of Rosa and Ahuja Rosa2007 .
To unravel the origin of the different behavior of Cu dopants in bulk GaN and
passivated GaN NWs, we illustrate the DOS of the $t$ orbitals near the fermi
level in Fig. 3. For the isolated Cu dopant, the main difference between bulk
GaN:Cu and GaN:Cu NW is that the band width of the $t$ states is smaller in
NWs than that in bulk GaN (0.13 eV v.s. 0.53 eV) due to the quasi-1D nature of
NWs. The DOSs for an isolated Cu in bulk GaN and GaN NWs resemble those shown
in Fig. 3(a) and (d), respectively. When two Cu dopants are close to each
other, the $t$ orbitals in the same spin will couple to each other, resulting
in a low-lying bonding set and a high-lying anti-bonding one in the FM state,
broadening the band. This stabilize the FM state over the AFM state due to the
more occupation of the spin-down bonding orbitals than the anti-bonding state
Dalpian2006 ; Sluiter2005 . However, it also increases the overlapping between
the spin-up and spin-down bands, causing electron transfer from the spin-up
states to spin-down states as shown in [Fig. 3(b) and (e)], reducing the
magnetic moment. The transfer is large in bulk GaN than in GaN NW because the
band width in bulk GaN is wider. The reduced magnetic moment further leads to
a reduced exchange splitting and more charge transfer. The final DOSs of the
FM state for two Cu atoms in bulk GaN is shown in Fig. 3(c). This explains why
Cu-Cu pairs in bulk GaN has a very small magnetic moment when the Cu-Cu
distance is small and why the magnetic interaction is weak in bulk GaN. This
is because for bulk GaN:Cu, when Cu-Cu distance is small, the coupling between
Cu $d$ orbital is large, but the magnetic moment is small. When the Cu-Cu
distance is large, the magnetic moment is recovered, but the $d$ orbital
coupling becomes small. On the other hand, the Cu $t$ orbital band width is
much smaller in GaN NW, thus the charge transfer and reduction of magnetic
moment is small in GaN:Cu NWs even when the Cu-Cu distance is small [Fig.
3(f)]. This explains why magnetization and ferromagnetism is enhanced in
GaN:Cu NWs.
Due to the configurational entropy, Cu dopants might have many different
configurations in GaN NWs under experimental growth condition. Here to examine
the dependence of the magnetic properties on the configuration, we study all
possible inequivalent configurations (102 configurations in total) of two Cu
atoms in passivated GaN NW. To reduce the computational amount, no structural
relaxation is performed. It turns out that in all cases, FM is always
preferred over the AFM state, as shown in Fig. 4. Interestingly, the maximum
magnitude of the energy difference occur at some configurations with the Cu-Cu
distance about 6 Å, instead of the nearest neighbor configurations. This is
because when the Cu-Cu distance is smaller than 6 Å the magnetic moment start
to decrease, thus reducing the magnetic interactions, as discussed above.
However, unlike in bulk GaN:Cu, the energy difference is still negative and
large at nearest neighbor Cu-Cu distance. Thus FM in Cu doped passivated GaN
NW is robust with respect to Cu configurations in GaN NWs.
In the experimental work by Seong et al. Seong2007 , the diameter of the GaN
NW is larger than that of the NW we discussed above. Here to investigate the
magnetic interaction between Cu dopants in larger GaN NWs, we also consider
two Cu dopants in passivated GaN NW with a diameter about 1.6 nm, as shown in
Fig. 1(f). From our calculation, the FM state is more stable by 75 and 54 meV
than the AFM state for the unrelaxed and relaxed structures, respectively. So
the change of exchange interaction between Cu atoms in GaN NW is not very
significant when the diameter of the NW changes.
Figure 3: Schematic illustration of the origin of the different magnetic
behavior between Cu doped passivated GaN NW and Cu doped bulk GaN. (a) shows
the DOS for the $t$ orbitals of an isolated Cu atom in bulk GaN, (b) and (c)
are the DOSs of two in-plane neighbor Cu atoms in bulk GaN in the FM state
with the exchange splitting (EX) fixed and varied, respectively. (d), (e), and
(f) are the corresponding DOSs for Cu in a passivated GaN NW (Fig. 1(d)). “W”
denotes the band width. Figure 4: Energy difference between the AFM and FM
state for all possible inequivalent configurations of two Cu dopants in a
passivated GaN nanowire as shown in Fig. 1(b).
In summary, we explained why the ferromagnetism of GaN:Cu is enhanced in
passivated GaN NWs. We show that due to the 1D nature of the NW, the band
width of the Cu $t_{d}$ band is reduced, thus increases the magnetization. It
is found that passivation of the nanowire surface has a significant impact on
the substitutional position, and thus on the magnetic properties: For the bare
GaN NWs, Cu tends to substitute Ga on the surface, and the strong direct
interaction between two neighboring surface substitutional Cu atoms results in
a non-magnetic semiconducting state. However, when the surface is passivated,
Cu tends to substitute Ga inside the NW, and leads to ferromagnetism arising
from the $d$-$d$ exchange interactions. Our findings, thus, indicate that
surface passivation of the NW is crucial to sustain the FM state.
This work is supported by the U.S. Department of Energy, under Contract No.
DE-AC36-99GO10337. We thank Dr. Juarez L. F. Da Silva and Dr. Sukit
Limpijumnong for useful discussions.
## References
* (1) Dietl, T.; Ohno, H.; Matsukura, F.; Cibert, J.; Ferrand, D. Science 2000, 287, 1019.
* (2) Sarigiannidou, E.; Wilhelm, F.; Monroy, E.; Galera, R. M.; Bellet-Amalric, E.; Rogalev, A.; Goulon, J.; Cibert, J.; Mariette, H. Phys. Rev. B 2006, 74, 041306(R).
* (3) Sonoda, S.; Shimizu, S.; Sasaki, T.; Yamamoto, Y.; Hori, H. J. Cryst. Growth 2002, 237, 1358.
* (4) Ando, K. Appl. Phys. Lett. 2003, 82, 100.
* (5) Dalpian G. M.; Wei, S.-H. J. Appl. Phys. 2005, 98, 083905.
* (6) Dalpian, G. M.; Wei, S.-H.; Gong, X. G.; da Silva, A. J. R.; Fazzio, A. Solid State Commun. 2006, 138, 353.
* (7) Wu, R. Q.; Peng, G. W.; Liu, L.; Feng, Y. P.; Huang, Z. G.; Wu, Q. Y. Appl. Phys. Lett. 2006, 89, 062505.
* (8) Lee, J.-H.; Choi, I.-H.; Shin, S.; Lee, S.; Lee, J.; Whang, C.; Lee, S.-C.; Lee, K.-R.; Baek, J.-H.; Chae, K. H.; Song, J. Appl. Phys. Lett. 2007, 90, 032504.
* (9) Seipel, B.; Erni, R.; Gupta, A.; Li, C.; Owens, F. J.; Rao, K. V.; Browning, N. D.; Moeck, P. J. Mater. Res. 2007, 22, 1396.
* (10) Rosa, A. L.; Ahuja, R. Appl. Phys. Lett. 2007, 91, 232109.
* (11) Seong, H.-K.; Kim, J.-Y.; Kim, J.-J.; Lee, S.-C.; Kim, S.-R.; Kim, U.; Park, T.-E.; Choi, H.-J. Nano Lett. 2007, 7, 3366.
* (12) Blöchl, P. E. Phys. Rev. B 1995, 50, 17953; Kresse, G; Joubert, D. Phys. Rev. B 1999, 59, 1758.
* (13) Kresse, G; Furthmüller, J Comput. Mater. Sci. 1996, 6, 15; Phys. Rev. B 1996, 54, 11169.
* (14) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865.
* (15) Xiang, H. J.; Yang, J. L.; Hou, J. G.; Zhu, Q. S. Appl. Phys. Lett. 2006, 89, 223111.
* (16) Li, J.; Wei, S.-H.; Li, S. S.; Xia, J. B. Phys. Rev. B 2008 77, 113304.
* (17) Bader, R. F. W. Atoms in Molecules-A Quantum Theory (Oxford University Press, Oxford, 1990).
* (18) Wang, Q.; Sun, Q.; Jena, P.; Kawazoe, Y. Phys. Rev. Lett. 2004, 93, 155501.
* (19) Wang, Q.; Sun, Q.; Jena, P. Phys. Rev. Lett. 2005, 95, 167202.
* (20) Wang, Q.; Sun, Q.; Jena, P.; Kawazoe, Y. Nano Lett. 2005, 5, 1587.
* (21) Wang, Q.; Sun, Q.; Jena, P.; Kawazoe, Y. Phys. Rev. B 2007, 75, 075312.
* (22) Sluiter, M. H. F.; Kawazoe, Y.; Sharma, P.; Inoue, A.; Raju, A. R.; Rout, C.; Waghmare, U. V. Phys. Rev. Lett. 2005, 94, 187204.
| arxiv-papers | 2008-07-06T17:24:17 | 2024-09-04T02:48:56.555826 | {
"license": "Public Domain",
"authors": "H. J. Xiang and Su-Huai Wei",
"submitter": "H. J. Xiang",
"url": "https://arxiv.org/abs/0807.0918"
} |
0807.0951 | # Prompt High-Energy Emission from Proton-Dominated Gamma-Ray Bursts
Katsuaki Asano11affiliation: Interactive Research Center of Science, Tokyo
Institute of Technology, 2-12-1 Ookayama, Meguro-ku, Tokyo 152-8550, Japan ,
Susumu Inoue22affiliation: Department of Physics, Kyoto University, Oiwake-
cho, Kitashirakawa, Sakyo-ku, Kyoto 606-8502, Japan , and Peter
Mészáros33affiliation: Department of Astronomy & Astrophysics; Department of
Physics; Center for Particle Astrophysics; Pennsylvania State University,
University Park, PA 16802 asano@phys.titech.ac.jp,
inoue@tap.scphys.kyoto-u.ac.jp, nnp@astro.psu.edu
(Submitted; accepted)
###### Abstract
The prompt emission of gamma-ray bursts (GRBs) is widely thought to be
radiation from accelerated electrons, but an appreciably larger amount of
energy could be carried by accelerated protons, particularly if GRBs are the
sources of ultra-high-energy cosmic rays (UHECRs). We model the expected
photon spectra for such “proton-dominated” GRBs in the internal shock scenario
through Monte Carlo simulations, accounting for various processes related to
high-energy electrons and protons. Besides proton and muon synchrotron
components, emission from photomeson-induced secondary pair cascades becomes
crucial, generally enhancing the GeV-TeV and/or eV-keV photons and offering a
signature of UHE protons. In some cases, it can overwhelm the primary electron
component and result in GRBs peaking in the 10 MeV - 1 GeV range, which may be
relevant to some bursts discussed in a recent re-analysis of EGRET TASC data.
The dependence of the spectra on key quantities such as the bulk Lorentz
factor, magnetic field and proton-to-electron ratio is nontrivial due to the
nonlinear nature of cascading and the interplay of electron- and proton-
induced components. Observations by Fermi, ground-based telescopes and other
facilities should test these expectations and provide critical constraints on
the proton acceleration efficiency.
cosmic rays — gamma rays: bursts — gamma rays: theory — radiation mechanisms:
nonthermal
## 1 Introduction
The prompt emission of gamma-ray bursts (GRBs) is believed to arise from
ultrarelativistic outflows with bulk Lorentz factors $\Gamma\gtrsim 100$ (see,
e.g., reviews by Piran, 2005; Mészáros, 2006). In the popular internal shock
model, collisions among inhomogeneities within the flow lead to formation of
shocks that convert a fraction of the bulk kinetic energy into Fermi-
accelerated relativistic electrons, whose synchrotron emission powers the
observed MeV-band gamma-rays (Rees & Mészáros, 1994). Initially, most of the
kinetic energy as well as the internal energy generated via shock dissipation
are likely carried by protons, so such models entail the operation of a
physical mechanism that transfers energy from protons to electrons on
sufficiently short timescales. This presumably occurs via collective
electromagnetic processes, as simple Coulomb collisions may be too slow. A
general problem in collisionless shock theory and GRB models in particular is
that this mechanism is poorly understood, and one must frequently resort to a
phenomenological parametrization. In view of the large observed energy in MeV
gamma rays, the efficiency of proton-to-electron energy transfer is usually
considered to be high. However, this is by no means physically guaranteed. In
the case of supernova remnant shocks, the total energy in accelerated
electrons is often constrained observationally to be much less than in protons
(e.g. Aharonian et al., 2006). Since we do not yet understand the nature and
total energy budget of the central engine, we cannot readily exclude the
possibility that GRBs actually contain a significantly larger amount of energy
in protons compared to that radiated by the accelerated electrons.
Furthermore, a natural expectation is that the shocked protons are also Fermi-
accelerated. The physical conditions in internal shocks may allow maximum
energies $\gtrsim 10^{20}$ eV, so GRBs are potential sources of the observed
ultra-high-energy cosmic rays (UHECRs; Waxman, 1995; Vietri, 1995; Milgrom &
Usov, 1996). The total energy in accelerated protons that must be supplied per
burst depends on a number of uncertain factors (see also App. B of Murase et
al. (2008)). The required local UHECR emissivity at proton energy
$\varepsilon_{p}\sim 10^{19}$ eV is
$\varepsilon_{p}^{2}d\dot{N}_{p}/d\varepsilon_{p}\simeq 0.8\times 10^{44}\
{\rm erg\ Mpc^{-3}yr^{-1}}$ (Waxman & Bahcall, 1998; Dermer, 2007). Post-SWIFT
estimates of the local rate of long GRBs range from $0.2-1\ {\rm
Gpc^{-3}yr^{-1}}$ if the GRB rate is proportional to the star formation rate,
down to $\sim 0.05\ {\rm Gpc^{-3}yr^{-1}}$ if the GRB rate evolves more
strongly with redshift, which may be observationally favored (e.g. Daigne et
al., 2006; Le & Dermer, 2007; Guetta & Piran, 2007). Assuming a power-law
proton spectrum with index $p_{p}=2$, the necessary isotropic-equivalent
energy per burst in accelerated protons integrated over $\varepsilon_{p}\sim
10^{9}-10^{20}$ eV is $E_{\rm p}\sim 2\times 10^{54}-3\times 10^{55}$ erg,
which is approximately independent of the actual beaming factor. Steeper
spectra and hence even larger $E_{\rm p}$ are called for if GRBs also
contribute significantly to CRs below $10^{19}$ eV (Wick et al., 2004). To be
compared is the corresponding energy in accelerated electrons $E_{\rm e}$,
which can be roughly equated with the observed, isotropic-equivalent MeV
gamma-ray energy $E_{\gamma,{\rm iso}}$, typically $\sim 10^{53}$ erg and up
to $\sim 10^{54}$ erg in the $1-10^{4}$ keV rest-frame band (Kocevski &
Butler, 2008). Thus, in order for GRBs to be viable sources of UHECRs, the
latest observations point to a highly proton-dominated energy budget, $E_{\rm
p}/E_{\rm e}\gtrsim 10-100$. The observed heterogeneity of GRBs also suggests
that not all bursts may be equally efficient UHECR accelerators, in which case
even higher $E_{\rm p}/E_{\rm e}$ may be warranted for a subset of the bursts.
It is therefore of great interest whether such “proton-dominated” GRBs can be
diagnosed observationally. A promising window is GeV-TeV gamma-rays, where
distinctive signatures of UHE proton acceleration may show up, such as
synchrotron emission from protons, muons or secondary particles injected via
photomeson interactions (e.g. Vietri, 1997; Böttcher & Dermer, 1998; Gupta &
Zhang, 2007; Asano & Inoue, 2007, hereafter AI07).
AI07 recently undertook a detailed investigation of such emission processes
utilizing a comprehensive Monte Carlo code. However, having assumed that the
accelerated protons do not carry excessive extra energy, their study was
restricted to $E_{\rm p}/E_{\rm e}=1$. In view of the above possibilities,
here we follow and extend the work of AI07 to $E_{\rm p}/E_{\rm e}>1$. The
results, which are often qualitatively and drastically different from AI07,
are discussed in relation to existing and upcoming observations. Note that
high-energy emission from proton-dominated GRBs has been discussed previously
in different contexts (e.g. Totani, 1998; Asano & Takahara, 2003).
After a recap of our formulation in §2, we discuss the results and their
observational implications in §3 and §4, respectively, and conclude in §5.
## 2 Model and Methods
We briefly summarize the model and methods of AI07, which should be consulted
for more details. In accord with the internal shock paradigm, the emitting
region corresponding to an individual pulse in the prompt light curve is taken
to be a homogeneous shell expanding with $\Gamma$ at radii $R$ from the
central engine. The comoving width of the shell is $l=R/\Gamma$ and the pulse
timescale in the observer frame is $\Delta t=R/\Gamma^{2}c$, as long as $R$
exceeds the shell spreading radius (Mészáros et al.,, 1993), which is always
the case here. Shock dynamics and time variability are not explicitly treated,
so our results should be interpreted as the time-averaged spectra for each
pulse.
With given injection of accelerated electrons and protons in magnetic field
$B$, we solve self-consistently for the distribution of particles and photons
in the shell using Monte Carlo techniques. The time steps are always taken to
be sufficiently shorter than the particle cooling timescales (Asano, 2005). In
addition to synchrotron and inverse Compton (IC) emission from all particles,
our code includes synchrotron self-absorption, cascade processes with photon-
photon ($\gamma\gamma$) production of electron-positron pairs ($e^{\pm}$) and
Klein-Nishina regime Compton scattering, as well as proton-induced processes
such as photomeson ($p\gamma$) interactions and secondary pion, muon and pair
injection. We adopt experimental results for the cross sections of $p\gamma\to
n\pi^{+}$, $p\pi^{0}$, $n\pi^{+}\pi^{0}$ and $p\pi^{+}\pi^{-}$, while
$p\gamma\to p\pi^{0}\pi^{0}$ is neglected in view of its small cross section.
In case the primary proton is converted to a neutron, we assume that it
continues to interact with photons in the shell during the comoving expansion
timescale $t_{\rm exp}=l/c$. We do not acccount for the minor contribution
from neutron-decay electrons (Razzaque & Mészáros, 2006). More details on the
treatment of meson production and their decay products can be found in Asano
(2005) and Asano & Nagataki (2006).
Furthermore, we now account for the Bethe-Heitler (BH) pair production process
($p\gamma\to pe^{+}e^{-}$), whose cross section and inelasticity are taken
from Chodorowski et al. (1992). In the present context, the proton energy loss
is always dominated by photopion production, and the huge compactness of GRBs
implies that the resultant electromagnetic cascade emission is not very
sensitive to the details of particle injection at high energies. Thus,
compared to cases neglecting the BH process, we find that its inclusion here
only leads to modest enhancements of the secondary photon emission, by at most
a few tens of percent.
Primary electrons with total energy density $U_{\rm e}$ are injected with a
power-law distribution $n_{\rm e}(\gamma_{\rm e})\propto\gamma_{\rm
e}^{-p_{\rm e}}$ in the range of Lorentz factors $\gamma_{\rm
e,min}\leq\gamma_{\rm e}\leq\gamma_{\rm e,max}$. The balance of Fermi
acceleration and radiative cooling timescales gives $\gamma_{\rm e,max}$.
Likewise, protons with total energy density $U_{\rm p}$ are injected with a
distribution $n_{\rm p}(\gamma_{\rm p})\propto\gamma_{\rm p}^{-p_{\rm p}}$ in
the range $\gamma_{\rm p,min}\leq\gamma_{\rm p}\leq\gamma_{\rm p,max}$. We
obtain $\gamma_{\rm p,max}$ by equating $t_{\rm acc}=\gamma_{\rm
p}m_{p}c^{2}/eBc$, the Fermi acceleration time in relativistic shocks, to
$\min(t_{\rm exp},t_{\rm loss})$, where $t_{\rm loss}$ is the energy loss time
due to synchrotron, IC and $p\gamma$ cooling (Asano, 2005). In mildly
relativistic internal shocks, $\gamma_{\rm p,min}$ should be of order unity;
here we take $\gamma_{\rm p,min}=10$.
The injection index for electrons is fiducially chosen to be $p_{\rm e}=2.5$,
implying $\beta\simeq 2.25$ for the spectral index above the synchrotron peak
energy. This is consistent with the mean of the $\beta$ values measured by
BATSE, albeit with a considerable dispersion, from $\beta\lesssim 1.5$ to
$\beta\gtrsim 3.0$ (Preece et al., 2000; Kaneko et al., 2006). For protons,
our fiducial index is $p_{\rm p}=2.0$, appropriate when GRBs contribute to
UHECRs only above $10^{19}$ eV (Waxman & Bahcall, 1998); steeper spectra would
increase still the energy demands. Note that the values of $p_{\rm e}$ and
$p_{\rm p}$ relevant to our results each correspond to very different energy
ranges; GeV-TeV for electrons and 10-100 PeV for protons in the comoving
frame. Although the injection spectra for the two species are expected to be
the same at low energies where their gyroradii overlap, $p_{\rm e}>p_{\rm p}$
may be effectively realized if the proton spectrum covering 7-8 decades in
energy deviates from a pure power-law and becomes concave. This may plausibly
occur due to 1) nontrivial geometry and wavelength distribution of magnetic
turbulence at the shock (Niemiec et al., 2006), 2) nonlinear back-reaction of
CR pressure on the shock structure (Baring & Kirk, 1991; Malkov & Drury,
2001), or 3) superposition of pre-existing and newly-injected particles
originating from different regions in the outflow (Bosnjak & Daigne, 2008).
Nevertheless, in view of the observed spread in $\beta$ and the uncertainties
associated with obtaining time-integrated spectra, we also discuss cases with
$p_{\rm e}=p_{\rm p}$ in §3.2.
Some combinations of the remaining parameters are constrained so as to
reproduce typically observed properties of the MeV emission. For given $B$ and
$\Gamma$, $\gamma_{\rm e,min}$ is chosen such that the observed synchrotron
peak energy for nearby bursts is $\varepsilon_{\rm pk}=\Gamma\gamma_{\rm
e,min}^{2}\hbar eB/m_{\rm e}c\simeq 300$ keV. Since the fast-cooling, primary
electrons radiate away most of their energy as MeV photons within $\Delta t$,
$E_{\rm e}=(4\pi\Gamma^{2}R^{2}c\Delta t)U_{\rm e}\simeq(4\pi R^{3})U_{\rm e}$
can be identified with $E_{\rm sh}$, the observable, isotropic-equivalent MeV
pulse energy.
Instead of $U_{\rm e}$, $U_{\rm p}$ and $U_{\rm B}=B^{2}/8\pi$, hereafter we
use $\epsilon_{\rm e}$, $\epsilon_{\rm p}$ and $\epsilon_{\rm B}$, the
conventional parametrization of the corresponding energies as fractions of the
shock-dissipated internal energy (e.g. Mészáros, 2006). Thus $\epsilon_{\rm
B}/\epsilon_{\rm e}=U_{\rm B}/U_{\rm e}$ and $\epsilon_{\rm p}/\epsilon_{\rm
e}=U_{\rm p}/U_{\rm e}=E_{\rm p}/E_{\rm e}$. In place of $R$, we choose the
observable $\Delta t$ as a parameter and set $\Delta t=0.1$ s for simplicity.
Below we only show the spectra corresponding to single pulses. For bursts
composed of $N$ similar pulses, the duration-integrated energy would be simply
$N$ times larger, $E_{\gamma,{\rm iso}}=NE_{\rm sh}$. The set of parameters
are then $\Delta t$, $E_{\rm sh}$, $\Gamma$, $\epsilon_{\rm B}/\epsilon_{\rm
e}$ and $\epsilon_{\rm p}/\epsilon_{\rm e}$. All spectra are plotted as
observed fluence versus photon energy, assuming a GRB redshift $z=0.1$.
Spectral attenuation by intergalactic $\gamma\gamma$ absorption is neglected.
## 3 Results
### 3.1 Fiducial Spectral Indices
First we discuss different cases with our fiducial values of $p_{\rm e}=2.5$
and $p_{\rm p}=2.0$. As mentioned above, $\varepsilon_{\rm pk}$ is chosen to
have the typically observed value of 300 keV. Prompt emission spectra of
single pulses for $E_{\rm sh}=10^{51}$ erg, $\Gamma=300$, $\epsilon_{\rm
B}/\epsilon_{\rm e}=1$ and varying $\epsilon_{\rm p}/\epsilon_{\rm e}=10-100$
are shown in Figure 1. The sharp spectral cutoffs at low and high energies are
due to synchrotron self-absorption and $\gamma\gamma$ absorption,
respectively. This applies to all spectra below when such sharp cutoffs are
seen. Most remarkable is the prominent $e^{\pm}$ cascade component, i.e.
synchrotron and IC emission from secondary $e^{\pm}$ triggered by $p\gamma$
interactions of UHE protons with low energy photons. For the lower range of
$\epsilon_{\rm p}/\epsilon_{\rm e}$, primary synchrotron photons constitute
the main $p\gamma$ target. However, as the proton content increases, the
target photons become dominated by synchrotron emission from the low energy
part of the secondary $e^{\pm}$ themselves. The dependence of the spectra on
$\epsilon_{\rm p}/\epsilon_{\rm e}$ is therefore nonlinear and not simply
proportional, as apparent in Figure 1. The secondary photons also affect the
primary synchrotron component (dashed curves in Figure 1) through enhanced IC
cooling, even though the injection distribution is unchanged.
Figure 1: Single pulse, prompt photon spectra for varying $\epsilon_{\rm
p}/\epsilon_{\rm e}$ as labeled. Other parameters are marked above the figure.
Dashed curves denote the primary contribution only, whose peak flux decreases
with $\epsilon_{\rm p}$. Dot-dashed curves denote separately the electron
synchrotron (labeled eSY) and inverse Compton (eIC) components without
$\gamma\gamma$-absorption effects for $\epsilon_{\rm p}/\epsilon_{\rm e}=100$.
In general, cascade emission significantly hardens the high-energy spectra.
Since secondary $e^{\pm}$ with Lorentz factors $<\gamma_{\rm e,min}$ can be
injected in the cascade, it can also give rise to excess UV-to-X-ray emission
lying above the extrapolation of the sub-MeV spectra, as seen for
$\epsilon_{\rm p}/\epsilon_{\rm e}=10-30$ in Figure 1. The entire spectra thus
tends to become flat in $\varepsilon f(\varepsilon)$.
The case of $\epsilon_{\rm p}/\epsilon_{\rm e}=100$ is drastically different.
Here the proton-induced secondary emission totally overwhelms any primary
electron component, resulting in a hard spectrum peaking at 10-100 MeV.
Although approximately a single power-law between 100 eV and 30 MeV, in fact
it comprises two emission processes by secondary $e^{\pm}$, mainly synchrotron
$\lesssim$ MeV and IC $\gtrsim$ MeV (dot-dashed curves in Figure 1). Despite
$\epsilon_{\rm B}/\epsilon_{\rm e}=1$, IC can dominate over synchrotron since
the energy density of secondary $e^{\pm}$ exceeds both $U_{\rm B}$ and $U_{\rm
e}$.
The comoving photon density $n_{\gamma}$ is decisive for both 1) the
$\gamma\gamma$ optical depth $\tau_{\gamma\gamma}$ and hence the
$\gamma\gamma$ cutoff energy $\varepsilon_{\gamma\gamma}$, and 2) the
efficiency of $p\gamma$ interactions and hence the secondary cascade emission.
Figure 2 displays single pulse spectra for $\Gamma=300$, $\epsilon_{\rm
B}/\epsilon_{\rm e}=1$, $\epsilon_{\rm p}/\epsilon_{\rm e}=10$, and varying
pulse energies $E_{\rm sh}=10^{49}-10^{51}$ erg. Higher $E_{\rm sh}$ implies
higher $n_{\gamma}$, and consequently stronger $p\gamma$ components as well as
lower $\varepsilon_{\gamma\gamma}$. Since $n_{\gamma}\propto\Gamma^{-5}$ with
other parameters fixed, varying $\Gamma$ has larger effects. Shown in Figure 3
are single pulse spectra for $E_{\rm sh}=10^{50}$ erg, $\epsilon_{\rm
B}/\epsilon_{\rm e}=1$, $\epsilon_{\rm p}/\epsilon_{\rm e}=30$ and
$\Gamma=100-1000$. $\Gamma=100$ allows a high $\varepsilon_{\rm pk}$, cascade-
dominated spectrum, even though $\epsilon_{\rm p}/\epsilon_{\rm e}$ is 3 times
less than the analogous case in Figure 1. Increasing $\Gamma$ leads to higher
maximum energies and less cascade contribution. The spectral hardening
$\gtrsim 0.1$ GeV for $\Gamma=300$ and $\gtrsim 10$ GeV for $\Gamma=1000$ is
due to secondary IC.
Figure 2: Single pulse, prompt photon spectra for varying $E_{\rm sh}$ as
labeled. Other parameters are marked above the figure. Figure 3: Single
pulse, prompt photon spectra for varying $\Gamma$ as labeled. Other parameters
are marked above the figure.
Thus high proton-dominance does not always result in conspicuous proton-
induced emission if $\Gamma$ is sufficiently high. Conversely, the absence of
hard, high-energy components does not necessary rule out proton-dominated
GRBs. In fact, the conditions most favorable for contributing to UHECRs is
that they escape the source with minimal $p\gamma$ losses, which corresponds
roughly to the criterion $\Gamma\gtrsim 300(\Delta t/0.1{\rm s})^{-0.3}(E_{\rm
sh}/10^{51}{\rm erg})^{0.2}$ in our model (AI07). On the other hand, $\Gamma$
can be observationally constrained through its strong influence on
$\varepsilon_{\gamma\gamma}$ (e.g. Lithwick & Sari, 2001, AI07). Since the
pulse energy $E_{\rm sh}$ and timescale $\Delta t$ are also measurable, we may
hope to identify bursts where $p\gamma$ losses are likely to be efficient, and
then constrain $\epsilon_{\rm p}/\epsilon_{\rm e}$ from the high-energy
spectra, although some degeneracy with $\epsilon_{\rm B}/\epsilon_{\rm e}$
will remain.
Figure 4 shows single pulse spectra for $E_{\rm sh}=10^{51}$ erg,
$\Gamma=300$, $\epsilon_{\rm p}/\epsilon_{\rm e}=30$ and varying
$\epsilon_{\rm B}/\epsilon_{\rm e}=0.1-10$. The $\epsilon_{\rm
B}/\epsilon_{\rm e}=1$ case is the same as in Figure 1. Higher $B$ causes
steeper spectra with stronger secondary synchrotron relative to secondary IC,
while lower $B$ is vice-versa and produces a 100 MeV peak spectrum. However,
the dependence on $B$ can also be nontrivial. In Figure 5, we show spectra for
$E_{\rm sh}=10^{51}$ erg, $\Gamma=1000$, $\epsilon_{\rm p}/\epsilon_{\rm
e}=100$, and varying $\epsilon_{\rm B}/\epsilon_{\rm e}=0.1-100$ (note that
$\epsilon_{\rm B}/\epsilon_{\rm p}\leq 1$). The higher $\Gamma$ allows spectra
extending into the TeV regime, but renders $p\gamma$ processes inefficient
despite the high proton-dominance. All cases exhibit spectral bumps around
0.1-1 TeV, but their origins are quite different. For $\epsilon_{\rm
B}/\epsilon_{\rm e}\lesssim 1$, this is due to secondary $e^{\pm}$ IC, which
is weaker for higher $B$. However, when $\epsilon_{\rm B}/\epsilon_{\rm
e}\gtrsim 10$, the bump is stronger again, owing to the appearance of
synchrotron emission from protons and muons, their ratio being roughly 2 to 1
for $\epsilon_{\rm B}/\epsilon_{\rm e}=10$ (dot-dashed curves in Figure 5).
For $\epsilon_{\rm B}/\epsilon_{\rm e}=100$, we obtain a pronounced proton
synchrotron TeV peak, as well as enhanced emission at lower energies from
synchrotron radiation by $e^{\pm}$ produced via $\gamma\gamma$ absorption.
Figure 4: Single pulse, prompt photon spectra for varying $\epsilon_{\rm
B}/\epsilon_{\rm e}$ as labeled. Other parameters are marked above the figure.
Dashed curves denote the primary components only, whose peak flux decreases
with $\epsilon_{\rm B}$. Figure 5: Single pulse, prompt photon spectra
varying $\epsilon_{\rm B}/\epsilon_{\rm e}$ as labeled. Other parameters are
marked above the figure. Dot-dashed curves denote separately the electron
synchrotron (eSY), proton synchrotron (pSY) and muon synchrotron ($\mu$SY)
components without $\gamma\gamma$-absorption effects for $\epsilon_{\rm
B}/\epsilon_{\rm e}=10$ and 100.
### 3.2 Equal Proton and Electron Indices
We now consider situations with $p_{\rm e}=p_{\rm p}$, as would occur if the
proton spectrum was a single power-law over its entire energy range. Similar
to the above, Figure 6 testifies that the spectrum for $\epsilon_{\rm
p}/\epsilon_{\rm e}$=30 and $p_{\rm e}=p_{\rm p}$=2.0 can result in a hard GRB
with photon index $\sim 2$ up to 10 GeV. It is interesting to note that in
such cases, the spectral shape around the MeV peak alone may not always reveal
the correct value of $p_{\rm e}$. However, for $p_{\rm e}=p_{\rm p}=2.2$, the
fraction of UHE protons and the associated cascade emission is greatly
diminished, except for a slight distortion of the spectrum above 100 MeV. The
proton contribution becomes totally negligible for $p_{\rm e}=p_{\rm p}=2.5$,
for which neither UHECRs nor neutrinos are significantly generated at any
rate.
Figure 6: Single pulse, prompt photon spectra for varying values of $p_{\rm
e}=p_{\rm p}$ as labeled. Other parameters are marked above the figure. Thick
and thin dashed curves denote the primary components only, for $p_{\rm
e}=p_{\rm p}$=2.0 and 2.2, respectively.
## 4 Observational Implications
A unique property of proton-dominated GRBs is that their photon spectra can
sometimes manifest very high peak energies in the 10 MeV-1 GeV range due to
$p\gamma$ cascade emission (Figures 1,3,4,6). This seems at variance with
commonly observed values of $\varepsilon_{\rm pk}\sim$ 0.1-1 MeV (Kaneko et
al., 2006). However, through a recent re-analysis of EGRET TASC data, Kaneko
et al. (2008) reported a GRB with apparently very high $\varepsilon_{\rm
pk}>170$ MeV, as well as a few others with significant high-energy excess (see
also González et al., 2003). Some studies have also indicated potential
observational biases against BATSE detections of high $\varepsilon_{\rm pk}$
(Lloyd & Petrosian, 1999). At this moment, it is unclear how often such high
$\varepsilon_{\rm pk}$ bursts occur, and whether they are relevant to the
proton-dominated cases discussed here, or simply reflect a primary synchrotron
peak energy that is much higher than average (rather than the values we have
assumed here). In any case, the existence and nature of such bursts will be
definitively probed through ongoing observations by Fermi (Omodei, 2006) and
AGILE (Longo et al., 2007). Note that it is also conceivable that some GRBs
possess conservative proton energies, say $E_{\rm p}\sim 10^{53}$ erg, but
with $E_{\rm p}/E_{\rm e}\gg 1$ so that the MeV emission is relatively weak.
Even if unimportant for UHECRs (§1), new generation satellites should also
probe such MeV-weak bursts.
The $p\gamma$ cascade can also induce excess low-energy emission (Figures
1,2,4), which do not seem typical of known GRBs. However, they may be relevant
for some BATSE bursts with soft excess components (Preece et al., 1996), or
possibly a fraction of the X-ray rich GRBs (Sakamoto et al., 2005). Fermi and
AGILE observations of the accompanying high-energy excess will provide a test.
TeV detections of GRBs have yet to be achieved (e.g. Atkins et al., 2005;
Albert et al., 2007; Horan et al., 2007; Aharonian et al., 2009), but some of
the components discussed above may be eventually observed by current ground-
based facilities such as MAGIC (II), HESS (II), VERITAS, CANGAROO III, or the
future projects CTA, AGIS, HAWC, etc. For example, MAGIC may detect the
luminous proton synchrotron emission for $\epsilon_{\rm B}/\epsilon_{\rm
e}=100$ in Figure 5 at 0.1 TeV beyond $z\sim 1$, assuming $E_{\gamma,{\rm
iso}}=10^{53}$ erg and the latest estimates of intergalactic $\gamma\gamma$
absorption (Albert et al., 2008).
Distinguishing between primary electron IC and proton-induced emission
components may not be easy from the spectral shape alone. However, since the
synchrotron and/or photomeson cooling timescales for UHE protons are
considerably longer than the cooling timescales for GeV-TeV emitting primary
electrons, we can expect important differences in their variability
properties, which should provide further observational clues. Although this
work was limited to time-averaged pulse spectra, a desirable next step is to
perform explicitly time-dependent calculations.
## 5 Conclusions and Outlook
Proton-dominated GRBs are motivated by physical considerations of particle
acceleration in collisionless shocks, as well as their potential to be the
origin of UHECRs. In GRB UHECR scenarios, the spectral index for protons at
UHE must generally be harder than the typical indices for electrons emitting
in the multi-MeV range, which may be possible depending on the physics of
particle acceleration, cooling and/or shock formation, as discussed in §2.
Characteristic emission signatures can then result, such as high peak energy
bursts and/or excess low-energy emission from photomeson-triggered pair
cascades, or luminous spectral bumps from proton synchrotron emission. If the
indices for electrons and protons at the respective energies are equal,
proton-related components may still be visible as long as the index $\lesssim
2.2$, but not for steeper spectra. Through detailed observations of spectra
and variability, we may hope to disentangle the proton-induced components from
the competing emission process of inverse Compton from primary electrons.
Other observable consequences of proton-dominated GRBs may include
contributions to Galactic CRs (e.g. Wick et al., 2004) and the diffuse high-
energy neutrino background (e.g. Murase, 2007).
We note that if some GRBs actually emit stronger GeV-TeV components than
previously expected as discussed here, they could play an increased role in
probing high-$z$ intergalactic radiation fields (Inoue et al., in prep.) as
well as intergalactic magnetic fields (Ichiki et al., 2008, and references
therein).
We thank Chuck Dermer for very informative correspondence, and Kohta Murase
for valuable comments. Support is acknowledged from NSF PHY 0757155, NASA
NNX08AL40G, Grants-in-Aid for Scientific Research Nos. 19047004 and 19540283,
as well as the Global COE Program ”The Next Generation of Physics, Spun from
Universality and Emergence” from the Ministry of E.C.S.S.T. (MEXT) of Japan.
## References
* Aharonian et al. (2006) Aharonian, F. A. et al. 2006, A&A, 449, 223
* Aharonian et al. (2009) Aharonian, F. A. et al. 2009, A&A, submitted, arXiv:0901.2187
* Albert et al. (2007) Albert, J. et al. 2007, ApJ, 667, 358
* Albert et al. (2008) Albert, J. et al. 2008, Science, 320, 1752
* Asano (2005) Asano, K. 2005, ApJ, 623, 967
* Asano & Inoue (2007) Asano, K., & Inoue, S. 2007, ApJ, 671, 645 (AI07)
* Asano & Nagataki (2006) Asano, K., & Nagataki, S. 2006, ApJ, 640, L9
* Asano & Takahara (2003) Asano, K., & Takahara, F. 2003, PASJ, 55, 433
* Atkins et al. (2005) Atkins, R. et al. 2005, ApJ, 630, 996
* Baring & Kirk (1991) Baring, M., & Kirk, J. 1991, A&A, 241, 329
* Bosnjak & Daigne (2008) Bosnjak, Z., & Daigne, F., A&A, submitted, arXiv:0811.2956
* Böttcher & Dermer (1998) Böttcher, M., & Dermer, C. D. 1998, ApJ, 499, L131
* Chodorowski et al. (1992) Chodorowski, M. J., Zdziarski, A. A., & Sikora, M. 1992, ApJ, 400, 181
* Daigne et al. (2006) Daigne, F., Rossi, E., & Mochkovitch, R., MNRAS, 372, 1034
* Dermer (2007) Dermer, C. D. 2007, arXiv:0711.2804
* González et al. (2003) González, M. M. et al. 2003, Nature, 424, 749
* Guetta & Piran (2007) Guetta, D., & Piran, T. S. 2007, JCAP, 07, 003
* Gupta & Zhang (2007) Gupta, N., & Zhang, B., 2007, MNRAS, 380, 78
* Horan et al. (2007) Horan, D. et al. 2007, ApJ, 655, 396
* Ichiki et al. (2008) Ichiki, K., Inoue, S., & Takahashi, K., ApJ, 682, 127
* Kaneko et al. (2006) Kaneko, Y. et al. 2006, ApJS, 166, 298
* Kaneko et al. (2008) Kaneko, Y. et al. 2008, ApJ, 677, 1168
* Kocevski & Butler (2008) Kocevski, D., & Butler, N. 2008, ApJ, 680, 531
* Le & Dermer (2007) Le, T., & Dermer, C. D. 2007, ApJ, 661, 394
* Lithwick & Sari (2001) Lithwick, Y., & Sari, R. 2001, ApJ, 555, 540
* Lloyd & Petrosian (1999) Lloyd, N., & Petrosian, V. 1999, ApJ, 511, 550
* Longo et al. (2007) Longo, F. et al. 2007, AIP Conf. Proc., 906, p.147
* Malkov & Drury (2001) Malkov, M. A., & Drury, L. O’C. 2001, Rep. Prog. Phys., 64, 429
* Mészáros et al., (1993) Mészáros, P., Laguna, P., & Rees, M. J., 1993, ApJ, 414, 181
* Mészáros (2006) Mészáros, P. 2006, Rep. Prog. Phys., 69, 2259
* Milgrom & Usov (1996) Milgrom, M. & Usov, V. 1996, Astropart. Phys., 4, 365
* Murase (2007) Murase, K. 2007, Phys. Rev. D, 76, 123001
* Murase et al. (2008) Murase, K., Ioka, K., Nagataki, S., & Nakamura, T. 2008, Phys. Rev. D, 78, 023005
* Niemiec et al. (2006) Niemiec, J., Ostrowski, M., & Pohl, M. 2006, ApJ, 650, 1020
* Omodei (2006) Omodei, N. astro-ph/0603762
* Piran (2005) Piran, T. 2005, Rev. Mod. Phys., 76, 1143
* Preece et al. (1996) Preece, R. D. et al. 1996, ApJ, 473, 310
* Preece et al. (2000) Preece, R. D. et al. 2000, ApJS, 126, 19
* Razzaque & Mészáros (2006) Razzaque, S. & Mészáros, P. 2006, JCAP, 6, 006
* Rees & Mészáros (1994) Rees, M. J., & Mészáros, P. 1994, ApJ, 430, L93
* Sakamoto et al. (2005) Sakamoto, T. et al. 2005, ApJ, 629, 311
* Totani (1998) Totani, T. 1998, ApJ, 509, L81
* Vietri (1995) Vietri, M. 1995, ApJ, 453, 883
* Vietri (1997) Vietri, M. 1997, Phys. Rev. Lett., 78, 4328
* Waxman (1995) Waxman, E. 1995, Phys. Rev. Lett., 75, 386
* Waxman & Bahcall (1998) Waxman, E. & Bahcall, J. 1998, Phys. Rev. D, 59, 023002
* Wick et al. (2004) Wick, S. D., Dermer, C. D., & Atoyan, A. 2004, Astropart. Phys., 21, 125
| arxiv-papers | 2008-07-07T05:59:51 | 2024-09-04T02:48:56.560607 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Katsuaki Asano, Susumu Inoue, and Peter M\\'esz\\'aros",
"submitter": "Katsuaki Asano",
"url": "https://arxiv.org/abs/0807.0951"
} |
0807.1028 | # Density matrix of the superposition of excitation on coherent states with
thermal light and its statistical properties
Li-yun Hu Department of Physics, Shanghai Jiao Tong University, Shanghai
200030, China Hong-yi Fan Department of Physics, Shanghai Jiao Tong
University, Shanghai 200030, China
(5 July 2008)
###### Abstract
A beam’s density matrix that is described by the superposition of excitation
on coherent states with thermal noise (SECST) is presented, and its matrix
elements in Fock space are calculated. The maximum information transmitted by
the SECST beam is derived. It is more than that by coherent light beam and
increases as the excitation photon number increases. In addition, the
nonclassicality of density matrix is demonstrated by calculating its Wigner
function.
PACS numbers: 42.50.Dv, 03.67.Hk, 03.65.Ud
excitation on coherent states, thermal light, Wigner function
## I Introduction
Recently, much attention has been paid to the excitation on coherent states
(ECS) 1 ; 2 ; 3 ; 4 ; 5 ; 6 . As pointed out in Refs.2 ; 3 , the single photon
ECS causes a classical-to-quantum (nonclassical) transition. The ECSs can be
considered as a generalization of coherent states 7 ; 8 and number
eigenstates. All these states can be used as signal beams in optical
communications field, in which the nonclassicality of signals plays an
important role.
However, in reality, signal beams are usually mixed with thermal noise.
Statistical properties of the superposition of (squeezed) coherent states with
thermal light (SCST) have been investigated by calculating the photon number
matrix elements $\left\langle N\right|\rho\left|M\right\rangle$ of SCST’s
density matrix 9 ; 10 . These properties are useful in quantum optics and
quantum electronics (e.g. how lasers working well above threshold, heterodyne
detection of light, etc.) 11 . Some general properties of the density matrices
which describe coherent, squeezed and number eigenstates in thermal noise are
studied in Ref.12 . It is found that the information transmitted by the
superposition of number eigenstates with thermal light (SNET) beam is less
than that by the SCST beam 13 .
In this paper, we investigate statistical properties of the superposition of
ECS with thermal light (SECST). We present the relevant density matrix in Fock
space and derive the Mandel $Q$ parameter. The SECST field can exhibit a
significant amount of super-Poissonian photon statistics (PPS) due to the
presence of thermal noise for excitation photon number $m=0;$ while for $m\neq
0$ the SECST field can present the sub-PPS when the thermal mean photon number
is less than a threshold value. In addition, the threshold value increases as
$m$ increases. We also calculate the maximum information (channel capacity)
transmitted by the SECST beam, which increases as $m$ increases. In addition,
the nonclassicality of density matrix is also presented by calculating the
Wigner function of the SECST.
Our paper is arranged as follows. In Sec. II we present the density matrix
$\rho$ that describes the SECST and calculate its matrix elements in Fock
space by using the normal ordered form of $\rho$. The PPS distributions are
discussed in Sec III. The maximum information is calculated in Sec. IV. Sec. V
is devoted to deriving the Wigner function of the SECST and discussing its
nonclassicality in details. Conclusions are summarized in the last section.
## II Excitation on coherent states with thermal noise
Firstly, let us briefly review the excitation on coherent states (ECSs). The
ECSs, first introduced by Agarwal and Tara 1 , are the result of successive
elementary one-photon excitations of a coherent state, and is an intermediate
state in between the Fock state and the coherent state, since it exhibits the
sub-Poissonian character. Theoretically, the ECSs can be obtained by
repeatedly operating the photon creation operator $a^{{\dagger}}$ on a
coherent state, so its density operator is
$\rho_{0}=C_{\alpha,m}a^{{\dagger}m}\left|\alpha\right\rangle\left\langle\alpha\right|a^{m},$
(1)
where $C_{\alpha,m}=[m!L_{m}(-\left|\alpha\right|^{2})]^{-1}$ is the
normalization factor,
$\left|\alpha\right\rangle=\exp(-\left|\alpha\right|^{2}/2+\alpha
a^{\dagger})\left|0\right\rangle$ is the coherent state 7 ; 8 , and
$L_{m}\left(x\right)$ is the $m$th-order Laguerre polynomial.
The SECST is described by the density matrix 12
$\displaystyle\rho$
$\displaystyle=\int\frac{d^{2}z}{\pi}P\left(z\right)D\left(z\right)\rho_{0}D^{{\dagger}}\left(z\right),$
(2) $\displaystyle P\left(z\right)$
$\displaystyle=\frac{1}{\bar{n}_{t}}\exp\left[-\frac{\left|z\right|^{2}}{\bar{n}_{t}}\right],$
(3)
where $D\left(z\right)=\exp(za^{{\dagger}}-z^{\ast}a)$ is the displacement
operator, and $\bar{n}_{t}$ is the mean number of thermal photons for
$\rho_{0}\rightarrow\left|0\right\rangle\left\langle 0\right|$. We can easily
prove that $\mathtt{Tr}\rho=1,$ as it should be. In fact,
$\displaystyle\mathtt{Tr}\rho$
$\displaystyle=\int\frac{d^{2}z}{\pi}P\left(z\right)\mathtt{Tr}\left[D\left(z\right)\rho_{0}D^{{\dagger}}\left(z\right)\right]$
$\displaystyle=\int\frac{d^{2}z}{\pi}P\left(z\right)\mathtt{Tr}\left(\rho_{0}\right)$
$\displaystyle=\int\frac{d^{2}z}{\pi}P\left(z\right)=1.$ (4)
### II.1 Normal ordering form of the SECST
For the simplicity in our later calculation, we first perform the integration
in Eq.(2) by using the technique of integration within an ordered product
(IWOP) of operators 14 ; 15 . Using the normal ordering form of the vacuum
projector $\left|0\right\rangle\left\langle
0\right|=\colon\exp(-a^{{\dagger}}a)\colon,$ we can reform Eq.(2) as the
following form
$\displaystyle\rho$
$\displaystyle=C_{\alpha,m}e^{-\left|\alpha\right|^{2}}\int\frac{d^{2}z}{\pi}P\left(z\right)D\left(z\right)\colon
a^{{\dagger}m}$ $\displaystyle\times\exp\left(\alpha
a^{{\dagger}}+\alpha^{\ast}a-a^{{\dagger}}a\right)a^{m}\colon
D^{{\dagger}}\left(z\right)$
$\displaystyle=\frac{C_{\alpha,m}}{\bar{n}_{t}}\colon\exp\left(-\left|\alpha\right|^{2}-a^{{\dagger}}a+\allowbreak
a^{{\dagger}}\alpha+a\alpha^{\ast}\right)$
$\displaystyle\times\int\frac{d^{2}z}{\pi}\exp\left[-\frac{1+\bar{n}_{t}}{\bar{n}_{t}}\left|z\right|^{2}\right]$
$\displaystyle\times\exp\left[\left(a^{{\dagger}}-\alpha^{\ast}\right)\allowbreak
z+\left(a-\alpha\right)z^{\ast}\right]$
$\displaystyle\times\left(a^{{\dagger}}-z^{\ast}\right)^{m}\left(a-z\right)^{m}\colon.$
(5)
In the last step of (5), we noticed that for any operator $f(a^{{\dagger}},a)$
$D\left(z\right)f(a^{{\dagger}},a)D^{{\dagger}}\left(z\right)=f(a^{{\dagger}}-z^{\ast},a-z).$
(6)
Making two independent variable displacements,
$a^{{\dagger}}-z^{\ast}\rightarrow\beta^{\ast},a-z\rightarrow\beta,$
(note that operators $a^{{\dagger}},a$ can be considered as C-number within
the normal order $\colon\colon$), thus Eq.(5) can be rewritten as
$\displaystyle\rho$
$\displaystyle=\frac{C_{\alpha,m}}{\bar{n}_{t}}\colon\exp\left(-\left|\alpha\right|^{2}-\frac{1}{\bar{n}_{t}}a^{{\dagger}}a\right)$
$\displaystyle\times\int\frac{d^{2}\beta}{\pi}\beta^{\ast
m}\beta^{m}\exp\left[-\lambda_{t}^{-2}\left|\beta\right|^{2}\right.$
$\displaystyle\left.+\left(\allowbreak\alpha^{\ast}+\allowbreak\frac{a^{{\dagger}}}{\bar{n}_{t}}\right)\beta+\left(\frac{a}{\bar{n}_{t}}+\alpha\right)\beta^{\ast}\right]\colon$
$\displaystyle=\frac{C_{\alpha,m}}{\bar{n}_{t}}\lambda_{t}^{2m+2}\colon\exp\left(-\left|\alpha\right|^{2}-\frac{1}{\bar{n}_{t}}a^{{\dagger}}a\right)$
$\displaystyle\times\int\frac{d^{2}\beta}{\pi}\beta^{\ast
m}\beta^{m}\exp\left[-\left|\beta\right|^{2}+A^{{\dagger}}\beta+A\beta^{\ast}\right]\colon,$
(7)
where we have set $\lambda_{t}=\sqrt{\bar{n}_{t}/(1+\bar{n}_{t})}$ and
$A=\lambda_{t}(\frac{1}{\bar{n}_{t}}a+\alpha).$ Then using the integration
expression of two-variable Hermite polynomial $H_{m,n}$ 16 ,
$\displaystyle(-1)^{n}e^{-\xi\eta}H_{m,n}\left(\xi,\eta\right)$
$\displaystyle=\int\frac{d^{2}z}{\pi}z^{n}z^{\ast
m}\exp\left[-\left|z\right|^{2}+\xi z-\eta z^{\ast}\right],$ (8)
we can put Eq.(7) into
$\displaystyle\rho$
$\displaystyle=\frac{C_{\alpha,m}}{\bar{n}_{t}}\lambda_{t}^{2m+2}\colon\left(-1\right)^{m}H_{m,m}\left(A^{{\dagger}},-A\right)$
$\displaystyle\times\exp\left[-\frac{\left(a-\alpha\right)\left(a^{{\dagger}}-\alpha^{\ast}\right)}{\bar{n}_{t}+1}\right]\colon.$
(9)
In particular, when $m=0$, corresponding to the case of superposition of
coherent state with thermal noise, Eq.(9) reduces to
$\rho=\frac{1}{\bar{n}_{t}+1}D\left(\alpha\right)\colon
e^{-\frac{a^{{\dagger}}a}{\bar{n}_{t}+1}}\colon
D^{{\dagger}}\left(\alpha\right),$ (10)
which can be directly checked by using Eqs.(2) and (3) as well as noticing
$\rho_{0}=\left|\alpha\right\rangle\left\langle\alpha\right|.$
Further employing the relation between Hermite polynomial and Laguerre
polynomial 16 ,
$H_{m,n}\left(\xi,\kappa\right)=\left\\{\begin{array}[c]{cc}n!\left(-1\right)^{n}\xi^{m-n}L_{n}^{m-n}\left(\xi\kappa\right),&m>n\\\
m!\left(-1\right)^{m}\kappa^{n-m}L_{m}^{n-m}\left(\xi\kappa\right),&m<n\end{array}\right.,$
(11)
we can see that
$\displaystyle\rho$
$\displaystyle=\frac{1}{L_{m}(-\left|\alpha\right|^{2})}\frac{\bar{n}_{t}^{m}}{(1+\bar{n}_{t})^{m+1}}\colon
L_{m}\left(-A^{{\dagger}}A\right)$
$\displaystyle\times\exp\left[-\frac{\left(a-\alpha\right)\left(a^{{\dagger}}-\alpha^{\ast}\right)}{\bar{n}_{t}+1}\right]\colon.$
(12)
Eqs.(9) and (12) are the normal ordering form of the SECST. From these it is
convenient to calculate the phase space distributions, such as Q-function,
P-representation and Wigner function.
### II.2 The matrix elements $\left\langle N\right|\rho\left|M\right\rangle$
Now we calculate the matrix elements of $\rho$ in Eq.(2) between two number
states $\left\langle N\right|$ and $\left|M\right\rangle,$ i.e., $\left\langle
N\right|\rho\left|M\right\rangle.$ Employing the overcompleteness of coherent
states, one can express the matrix elements $\left\langle
N\right|\rho\left|M\right\rangle$ as
$\left\langle N\right|\rho\left|M\right\rangle=\int\frac{d^{2}\beta
d^{2}\gamma}{\pi^{2}}\left\langle
N\right.\left|\beta\right\rangle\left\langle\beta\right|\rho\left|\gamma\right\rangle\left\langle\gamma\right.\left|M\right\rangle,$
(13)
where the overlap between the coherent state and the number state is given by
$\left\langle\gamma\right.\left|M\right\rangle=\frac{1}{\sqrt{M!}}e^{-\left|\gamma\right|^{2}/2}\gamma^{\ast
M},$ (14)
and the matrix elements
$\left\langle\beta\right|\rho\left|\gamma\right\rangle$ can be obtained from
Eq.(9) due to $\rho^{\prime}$s normal ordering form,
$\displaystyle\left\langle\beta\right|\rho\left|\gamma\right\rangle$
$\displaystyle=\left(-1\right)^{m}\frac{C_{\alpha,m}}{\bar{n}_{t}}\lambda_{t}^{2m+2}e^{-\left|\alpha\right|^{2}/(\bar{n}_{t}+1)}$
$\displaystyle\times\frac{\partial^{2m}}{\partial\tau^{m}\partial\tau^{\prime
m}}\exp\left[-\tau\tau^{\prime}+\lambda_{t}\tau\alpha^{\ast}-\lambda_{t}\alpha\tau^{\prime}\right]$
$\displaystyle\times\exp\left\\{\left(\frac{\alpha+\bar{n}_{t}\allowbreak\gamma}{\bar{n}_{t}+1}+\frac{\lambda_{t}\tau}{\bar{n}_{t}}\right)\beta^{\ast}-\frac{1}{2}\left|\beta\right|^{2}\right.$
$\displaystyle-\left.\frac{1}{2}\left|\gamma\right|^{2}+\left(\frac{\alpha^{\ast}}{\bar{n}_{t}+1}-\frac{\lambda_{t}\tau^{\prime}}{\bar{n}_{t}}\right)\gamma\right\\}_{\tau=\tau^{\prime}=0},$
(15)
where we have used the generating function of two-variable Hermite polynomial
$H_{m,n},$
$H_{m,n}\left(x,y\right)=\left.\frac{\partial^{m+n}}{\partial t^{m}\partial
t^{\prime
n}}\exp\left[-tt^{\prime}+tx+t^{\prime}y\right]\right|_{t=t^{\prime}=0}.$ (16)
When $M=N,$ $\left\langle N\right|\rho\left|N\right\rangle$ is just the photon
number distribution of the SECST. Then combing with Eqs.(15), (13) and (14),
after a lengthy but straightforward calculation, one can get the matrix
elements $\left\langle N\right|\rho\left|M\right\rangle,$ (without loss of
generality, let $M\geqslant N$)
$\displaystyle\left\langle N\right|\rho\left|M\right\rangle$
$\displaystyle=\frac{\left(-1\right)^{N}}{\sqrt{M!N!}}\frac{\lambda_{t}^{2N}C_{\alpha,m}}{\bar{n}_{t}+1}e^{-\left|\alpha\right|^{2}}\frac{\partial^{2m}}{\partial\upsilon^{m}\partial\upsilon^{\prime
m}}$
$\displaystyle.\left\\{e^{\lambda_{t}^{2}\upsilon\upsilon^{\prime}}H_{M,N}\left(\frac{\upsilon^{\prime}}{\bar{n}_{t}+1},-\frac{\upsilon}{\bar{n}_{t}}\right)\right\\}_{\upsilon=\alpha,\upsilon^{\prime}=\alpha^{\ast}},$
(17)
where we have used the integral formula 17
$\int\frac{d^{2}\beta}{\pi}f\left(\beta^{\ast}\right)\exp\left\\{-\left|\beta\right|^{2}+\tau\beta\right\\}=f\left(\tau\right),$
(18)
and another expression of two-variable Hermite polynomial $H_{m,n},$
$H_{m,n}\left(\xi,\kappa\right)=\sum_{l=0}^{\min(m,n)}\frac{m!n!\left(-1\right)^{l}\xi^{m-l}\kappa^{n-l}}{l!\left(n-l\right)!\left(m-l\right)!}.$
(19)
In particular, when $m=0$, noticing $M\geqslant N$ and Eq.(11), Eq.(17)
reduces to
$\displaystyle\left\langle N\right|\rho\left|M\right\rangle$
$\displaystyle=\sqrt{\frac{N!}{M!}}\alpha^{\ast
M-N}\frac{\left(\bar{n}_{t}\right)^{N}}{\left(\bar{n}_{t}+1\right)^{M+1}}$
$\displaystyle\times
e^{-\left|\alpha\right|^{2}/(\bar{n}_{t}+1)}L_{N}^{M-N}\left[-\frac{\left|\alpha\right|^{2}}{\bar{n}_{t}\left(\bar{n}_{t}+1\right)}\right],$
(20)
which is just the Glauber-Lachs formula 9 when
$\bar{n}_{t}=(e^{\beta\omega}-1)^{-1}$. While for $\alpha=0,$ corresponding to
the case of superposition of number state with thermal light, using Eq.(19),
Eq.(17) becomes
$\left\langle N\right|\rho\left|M\right\rangle=\delta_{M,N}P_{N},$ (21)
where ($k_{0}=\max[0,m-N]$)
$P_{N}=\frac{m!N!}{\bar{n}_{t}+1}\sum_{k=k_{0}}^{m}\frac{1}{k!}\frac{\left(\frac{\bar{n}_{t}}{\bar{n}_{t}+1}\right)^{k+N}\left[\bar{n}_{t}\left(\bar{n}_{t}+1\right)\right]^{k-m}}{\left(k+N-m\right)!\left[(m-k)!\right]^{2}}.$
(22)
Eq.(21) is just the result of Ref. 13 .
## III Sub-Poissonian photon statistics
To see clearly the photon statistics properties of the SECST, in this section,
we pay our attention to the variance of the photon number operator
$\left\langle\left(\Delta\hat{n}\right)^{2}\right\rangle=\left\langle\hat{n}^{2}\right\rangle-\left\langle\hat{n}\right\rangle^{2}.$
In particular, we will examine the evolution of the Mandel $Q$ parameter
defined as
$\displaystyle Q$
$\displaystyle=\frac{\left\langle\left(a^{{\dagger}}a\right)^{2}\right\rangle}{\left\langle
a^{{\dagger}}a\right\rangle}-\left\langle a^{{\dagger}}a\right\rangle$
$\displaystyle=\frac{\left\langle
a^{2}a^{{\dagger}2}\right\rangle-\left\langle
aa^{{\dagger}}\right\rangle^{2}-\left\langle
aa^{{\dagger}}\right\rangle}{\left\langle aa^{{\dagger}}\right\rangle-1},$
(23)
which measures the derivation of the variance of the photon number
distribution of the field state under consideration from the Poissonian
distribution of the coherent state. $Q=1,Q>1$ and $Q<1$ correspond to
Poissonian photon statistics (PPS), super-PPS and sub-PPS, respectively.
In order to calculate the average value in Eq.(23), we first calculate the
value of
$\left\langle\alpha\right|a^{n}a^{{\dagger}m}\left|\alpha\right\rangle.$ In
fact, using
$\left\langle\alpha\right|a^{m+n}a^{{\dagger}m+n}\left|\alpha\right\rangle=\left(m+n\right)!L_{m+n}(-\left|\alpha\right|^{2})$
(24)
and
$\int\frac{d^{2}z}{\pi}z^{n}z^{\ast
m}P\left(z\right)=\bar{n}_{t}^{m}m!\delta_{m,n},$ (25)
we can evaluate (for writing’s convenience, let $L_{m}$ denote
$L_{m}(-\left|\alpha\right|^{2})$)
$\left\langle
a^{{\dagger}}a\right\rangle=\frac{1+m}{L_{m}}L_{m+1}+\bar{n}_{t}-1,$ (26)
and
$\left\langle
a^{2}a^{{\dagger}2}\right\rangle=2\bar{n}_{t}^{2}+\frac{m+1}{L_{m}}\left[4\bar{n}_{t}L_{m+1}+\left(m+2\right)L_{m+2}\right].$
(27)
Substituting Eqs.(26) and (27) into (23) leads to
$\displaystyle Q$
$\displaystyle=\frac{\bar{n}_{t}\left(\bar{n}_{t}-1\right)L_{m}+\left(2\bar{n}_{t}-1\right)\left(m+1\right)L_{m+1}}{\left(1+m\right)L_{m+1}+\left(\bar{n}_{t}-1\right)L_{m}}$
$\displaystyle+\frac{(m+1)(m+2)L_{m+2}-\frac{\left(m+1\right)^{2}}{L_{m}}L_{m+1}^{2}}{\left(1+m\right)L_{m+1}+\left(\bar{n}_{t}-1\right)L_{m}}.$
(28)
At the zero-temperature limit $(\bar{n}_{t}\rightarrow 0)$, Eq.(28) just
reduces to Eq.(2.20) in Ref.1 .
Figure 1: (Color online) The evolution of Mandel $Q$ parameter as a function
of ($n_{t},\left|\alpha\right|$) for different values $m.$
In Fig.1, we display the parameter $Q\left(n_{t},\left|\alpha\right|\right)$
as a function of $\left(n_{t},\left|\alpha\right|\right)$ for different values
$m.$ From Fig.1, we see that, for the excitation photon number $m=0$ (see
Fig.1 (a)), $Q\left(\bar{n}_{t}=0,\left|\alpha\right|\right)=1$ corresponding
to coherent state (a PPS); while $Q\left(\bar{n}_{t}\neq
0,\left|\alpha\right|\right)>1$, i.e., the SECST field exhibits a significant
amount of super-PPS due to the presence of $\bar{n}_{t}$. From Fig.1 (b) and
(c), we see that, when $m\neq 0,$ the SECST field presents the sub-PPS when
$\bar{n}_{t}$ is less than a threshold value for a given
$\left|\alpha\right|;$ the threshold value increases as $m$ increases. For
example, when $\left|\alpha\right|=0,$ the threshold values are about 0.414
and 0.481, respectively, for $m=1$ and $m=6$.
## IV Information transmitted by the SECST beam
According to the negentropy principle of Brillouin 18 , the maximum
information $I$ transmitted by a beam is
$I=S_{\max}-S_{act},$ (29)
in which $S_{\max}$ and $S_{act}$ represent the maximum entropy and the actual
entropy, respectively, possessed by the quantum mechanical system described by
a density matrix $\rho$. Here the maximum information $I$ is an ideal one
transmitted through an ideal optical communication system.
For the SECST system, the actual entropy is
$S_{act}=-\mathtt{Tr}\left(\rho\ln\rho\right)=-\sum_{N}\sigma_{N}\ln\sigma_{N},$
(30)
where $\rho=\sum_{N}\sigma_{N}\left|N\right\rangle\left\langle N\right|,$ and
$\sigma_{N}=\left\langle N\right|\rho\left|N\right\rangle.$ $\sigma_{N}$ can
be obtained from Eq.(17), i.e.,
$\displaystyle\sigma_{N}$
$\displaystyle=\frac{\bar{n}_{t}^{N}e^{-\left|\alpha\right|^{2}}C_{\alpha,m}}{\left(\bar{n}_{t}+1\right)^{N+1}}\frac{\partial^{2m}}{\partial\upsilon^{m}\partial\upsilon^{\prime
m}}$
$\displaystyle\times\left\\{e^{\lambda_{t}^{2}\upsilon\upsilon^{\prime}}L_{N}\left(\frac{-\upsilon\upsilon^{\prime}}{\bar{n}_{t}\left(\bar{n}_{t}+1\right)}\right)\right\\}_{\upsilon=\alpha,\upsilon^{\prime}=\alpha^{\ast}},$
(31)
which is independent of the phase of $\alpha.$ On the other hand, for a system
in thermal equilibrium, described by the density matrix $\rho_{th}$, with mean
photons number $\bar{n}_{t}$, its entropy is
$S=-\sum_{N}P_{N}\ln
P_{N}=\ln(1+\bar{n}_{t})+\bar{n}_{t}\ln\frac{\bar{n}_{t}+1}{\bar{n}_{t}},$
(32)
where $P_{N}=\bar{n}_{t}^{N}/\left(\bar{n}_{t}+1\right)^{N=1}$ obtained from
Eq.(20) under the condition $m=0,\alpha=0.$ Note that the maximum entropy of
the system is equal to the entropy of a system in thermal equilibrium, with an
equal mean number of photons. The mean photons number of the SECST is given by
Eq.(26). Therefore, using Eq.(32), we have
$\displaystyle S_{\max}$
$\displaystyle=\ln\left(\left(1+m\right)\frac{L_{m+1}}{L_{m}}+\bar{n}_{t}\right)$
$\displaystyle+\left(\left(1+m\right)\frac{L_{m+1}}{L_{m}}+\bar{n}_{t}-1\right)$
$\displaystyle\times\ln\left(\frac{\left(1+m\right)L_{m+1}+\bar{n}_{t}L_{m}}{\left(1+m\right)L_{m+1}+\left(\bar{n}_{t}-1\right)L_{m}}\right).$
(33)
Figure 2: (Color online) The maximum information
$I\left(\bar{n}_{t},\left|\alpha\right|\right)$as a function of
$\left(\bar{n}_{t},\left|\alpha\right|\right)$ for some different values (a)
$m=0,(b)$ $m=1,$(truncating the infinite sum at $N_{\max}=70$). Figure 3:
(Color online) The maximum information
$I\left(\bar{n}_{t},\left|\alpha\right|=1\right)$as a function of
$\left(\bar{n}_{t}\right)$ for some different values (a) $m=0,(b)$ $m=1,(c)$
${\small m=2}$ (truncating the infinite sum at $N_{\max}=70$).
From Eqs.(29), (30) and (33), we can calculate the maximum information
transmitted by the SECST beam. In Fig. 2, the maximum information
$I\left(\bar{n}_{t},\left|\alpha\right|\right)$ is plotted as a function of
$\left(\bar{n}_{t},\left|\alpha\right|\right)$ for some different values $m$
(truncating the infinite sum at $N_{\max}=70$). From Fig.2, we can see that,
for a given $\bar{n}_{t},$ $I\left(\bar{n}_{t},\left|\alpha\right|\right)$
increases as the value $\left|\alpha\right|$ increases; for given
$\left|\alpha\right|,$ in general,
$I\left(\bar{n}_{t},\left|\alpha\right|\right)$ grows up as the value
$\bar{n}_{t}$ increases. In order to see clearly the effect of different
parameter $m$ to $I\left(\bar{n}_{t},\left|\alpha\right|\right),$ we presented
a plot in Fig.3, from which it is obvious that
$I\left(\bar{n}_{t},\left|\alpha\right|\right)$ becomes bigger due to the
presence of $m$, and increases as $m$ increases. In other words, the maximum
information transmitted by the SECST beam is larger than that by the SCST
($m=0$). The channel of ECS can carry with more information than that of
coherent state. In Ref. 13 , Vourdas has pointed out that the coherent signals
(of known phase) can transmit more information than the number eigenvectors
signals. Thus among these three beams, the SECST beam can transmit most
information.
## V The Wigner function of the SECST
### V.1 The Wigner function
The Wigner function (WF) plays an important role in quantum optics, especially
the WF can be reconstructed from measurements 19 ; 20 . The WF is a powerful
tool to investigate the nonclassicality of optical fields. The presence of
negativity in the WF of optical field is a signature of its nonclassicality is
often used to describe the decoherence of quantum states. In this section,
using the normally ordered form of the SECST, we evaluate its WF. For a
single-mode system, the WF is given by 21
$W\left(\gamma,\gamma^{\ast}\right)=\frac{e^{2\left|\gamma\right|^{2}}}{\pi}\int\frac{d^{2}\beta}{\pi}\left\langle-\beta\right|\rho\left|\beta\right\rangle
e^{2\left(\beta^{\ast}\gamma-\beta\gamma^{\ast}\right)},$ (34)
where $\left|\beta\right\rangle$ is the coherent state and $\gamma=x+iy$. From
Eq.(34) it is easy to see that once the normal ordered form of $\rho$ is
known, we can conveniently obtain the WF of $\rho.$
On substituting Eq.(9) into Eq.(34) we obtain the WF of the SECST
$\displaystyle W\left(\gamma,\gamma^{\ast}\right)$
$\displaystyle=\frac{\left(\lambda_{t}^{2}A_{1}^{2}\right)^{m}C_{\alpha,m}}{\pi\left(2\bar{n}_{t}+1\right)}\exp\left\\{-\frac{2\left|\alpha-\gamma\right|^{2}}{2n_{t}+1}\right\\}$
$\displaystyle\times\left(-1\right)^{m}H_{m,m}\left(\frac{A_{2}^{\ast}}{A_{1}},-\frac{A_{2}}{A_{1}}\right)$
$\displaystyle=\frac{\left(\lambda_{t}^{2}A_{1}^{2}\right)^{m}\exp\left\\{-\frac{2\left|\alpha-\gamma\right|^{2}}{2\bar{n}_{t}+1}\right\\}}{\pi\left(2\bar{n}_{t}+1\right)L_{m}\left(-\left|\alpha\right|^{2}\right)}L_{m}\left(-\left|A_{2}\right|^{2}/A_{1}^{2}\right),$
(35)
where we have set
$\displaystyle A_{1}^{2}$
$\displaystyle=1-\frac{1}{\left(2\bar{n}_{t}+1\right)\bar{n}_{t}},$
$\displaystyle A_{2}$
$\displaystyle=\frac{\lambda_{t}\left(\bar{n}_{t}+1\right)}{\left(2\bar{n}_{t}+1\right)\bar{n}_{t}}\left(2\bar{n}_{t}\alpha-\alpha+2\gamma\right).$
(36)
Figure 4: (Color online) The evolution of the Wigner function of the SECST
with $\alpha=0.2+0.2i$ for several different values $m$ and $\bar{n}_{t}.$
Noticing that $L_{m}(-\left|\alpha\right|^{2})>0,$ and
$L_{m}[-\left|A_{2}\right|^{2}/A_{1}^{2}]>0$ when
$1-\frac{1}{\left(2\bar{n}_{t}+1\right)\bar{n}_{t}}>0,$ thus the WF of the
SECST is always positive under the condition of $\bar{n}_{t}>1/2$. In
particular, when $m=0$, Eq.(35) becomes
$W\left(\gamma,\gamma^{\ast}\right)=\frac{1}{\pi\left(2\bar{n}_{t}+1\right)}\exp\left\\{-\frac{2\left|\alpha-\gamma\right|^{2}}{2\bar{n}_{t}+1}\right\\},$
(37)
which corresponds to the thermal state with mean photon number $\bar{n}_{t}$.
While for $\alpha=0,$ $A_{2}\rightarrow
2\gamma\lambda_{t}\left(\bar{n}_{t}+1\right)/[\left(2\bar{n}_{t}+1\right)\bar{n}_{t}],$
$\left|A_{2}\right|^{2}/A_{1}^{2}\rightarrow
4\left|\gamma\right|^{2}\left(\bar{n}_{t}+1\right)/\\{\left(2\bar{n}_{t}+1\right)\left[\left(2\bar{n}_{t}+1\right)\bar{n}_{t}-1\right]\\}\equiv\xi$,
Eq. (35) yields
$W\left(\gamma,\gamma^{\ast}\right)=\frac{\left[\left(2\bar{n}_{t}+1\right)\bar{n}_{t}-1\right]^{m}}{\pi\left(2\bar{n}_{t}+1\right)^{m+1}\left(\bar{n}_{t}+1\right)^{m}}e^{-\frac{2\left|\gamma\right|^{2}}{2\bar{n}_{t}+1}}L_{m}\left(-\xi\right).$
(38)
At the zero-temperature limit, $T\rightarrow 0,\bar{n}_{t}\rightarrow 0,$
Eq.(37) reduces now into
$\frac{1}{\pi}\exp(-2\left|\alpha-\gamma\right|^{2}),$ i.e., the WF of
coherent state (a Guassian form), which can be seen from Eq.(2) yielding
$\rho=\left|\alpha\right\rangle\left\langle\alpha\right|$ under the condition
$m=0$; while Eq.(38) becomes
$\frac{1}{\pi}(-1)^{m}e^{-2\left|\gamma\right|^{2}}L_{m}(4\left|\gamma\right|^{2}),$
corresponding to the WF of number state.
Using Eq.(35), the WFs of the SECST are depicted in Fig.4 in phase space with
$\alpha=0.2+0.2i$ for several different values $m$ and $\bar{n}_{t}.$ It is
easy to see that the negative region of WF gradually disappears as $m$ and
$\bar{n}_{t}$ and increases.
### V.2 The marginal distributions of the SECST
We now find the probability distribution of position or momentum$|$—–the
marginal distributions, by performing the WF either over the variable $y$ or
the variable $x$, respectively. Using Eqs.(35) and (36) we can derive (denote
$\gamma=x+iy,$ $\alpha=q+ip$)
$\displaystyle\mathrm{P}\left(x,\bar{n}_{t}\right)$ $\displaystyle\equiv\int
W\left(x,y\right)dy$
$\displaystyle=\frac{\left(\lambda_{t}^{2}A_{1}^{2}\right)^{m}C_{\alpha,m}}{\sqrt{2\pi\left(2\bar{n}_{t}+1\right)}}\frac{\left[m!\right]^{2}e^{-\frac{2\left(q-x\right)^{2}}{2\bar{n}_{t}+1}}}{\left(2\bar{n}_{t}-1\right)^{m}}$
$\displaystyle\times\sum_{k=0}^{m}\frac{2^{2k-m}\bar{n}_{t}^{k}}{k!\left[(m-k)!\right]^{2}}\left|H_{m-k}\left(E_{1}\right)\right|^{2},$
(39)
where $H_{m}\left(x\right)$ is single variable Hermite polynomial and
$E_{1}=\left[\left(2\bar{n}_{t}-1\right)\alpha+2x+2ip\right]/\sqrt{2\left(2\bar{n}_{t}+1\right)}$.
Eq.(39) is the marginal distribution of WF of the SECST in “$x$-direction”.
On the other hand, performing the integration over $dx$ yields the other
marginal distribution in “$y$-direction”,
$\displaystyle\mathrm{P}\left(y,\bar{n}_{t}\right)$
$\displaystyle=\frac{\left(\lambda_{t}^{2}A_{1}^{2}\right)^{m}C_{\alpha,m}}{\sqrt{2\pi\left(2\bar{n}_{t}+1\right)}}\frac{\left[m!\right]^{2}e^{-\frac{2(p-y)^{2}}{2\bar{n}_{t}+1}}}{\left(2\bar{n}_{t}-1\right)^{m}}$
$\displaystyle\times\sum_{k=0}^{m}\frac{2^{2k-m}\bar{n}_{t}^{k}}{k!\left[(m-k)!\right]^{2}}\left|H_{m-k}\left(E_{2}\right)\right|^{2},$
(40)
where
$E_{2}=i\left(2\bar{n}_{t}\alpha-\alpha+2q+2iy\right)/[\sqrt{2\left(2\bar{n}_{t}+1\right)}]$.
As expected, the two marginal distributions are both real.
## VI Conclusions
In summary, we have investigated the photon statistics properties of the
SECST, described by the density matrix $\rho$ (2). We have calculated the
matrix elements $\left\langle N\right|\rho\left|M\right\rangle$ in Fock space
and the Mandel $Q$ parameter. It is found that the SECST field exhibits a
significant amount of super-PPS due to the presence of thermal noise
($\bar{n}_{t}$) for excitation photon number $m=0$ and that, for $m\neq 0$ and
a given $\left|\alpha\right|,$ the SECST field presents the sub-PPS when
$\bar{n}_{t}$ is less than a threshold value. In addition, the threshold value
increases as $m$ increases. We have presented the maximum information (channel
capacity) transmitted by the SECST beam. It is shown that the maximum
information transmitted increases as $m$ increases. This implies that among
the coherent signals, the eigen-number signals and the ECS in thermal light,
the last one can transmit the most information. Further, as one of the photon
statistical properties, the Wigner function and the marginal distributions of
the SECST have also been derived, from which one can clearly see the
nonclassicality. The negative region has no chance to be present when the
average photon number $\bar{n}_{t}$ of thermal noise exceeds $1/2.$ The
marginal distributions are related to the Hermite polynomial.
###### Acknowledgements.
This work is supported by the National Natural Science Foundation of China
(Grant No 10775097). L.-Y. Hu’s email address is hlyun@sjtu.edu.cn or
hlyun2008@126.com.
## References
* (1) G. S. Agarwal, K. Tara, Phys. Rev. A 43, 492 (1991).
* (2) A. Zavatta, S. Viciani, M. Bellini, Science 306, 660 (2004).
* (3) A. Zavatta, S. Viciani, M. Bellini, Phys. Rev. A 72, (2006) 023820.
* (4) Y. Li, H. Jing and M. S. Zhan, arXIV: quantum-ph/0610143v1.
* (5) D. K.alamidas, C. C. Gerry, A. Benmoussa, Phys. Lett. A 372, 1937-1940 (2008).
* (6) Truong Minh Duc, Jaewoo Noh, Opt. Commun. 281, 2842 (2008).
* (7) R. J. Glauber, Phys. Rev. 130, 2529 (1963); R. J. Glauber, Phys. Rev. 131, 2766 (1963).
* (8) J. R. Klauder and B. S. Skargerstam, Coherent States, (World Scientific, Singapore, 1985).
* (9) B. R. Mollow and R. J. Glauber, Phys. Rev. 160, 1076 (1967); G. Lachs, Phys. Rev. B 138, 1012 (1965).
* (10) A. Vourdas, Phys. Rev. A 34, 3466 (1986).
* (11) B. Saleh, Photoelectrons Statistics (Springer-Verlag, Berlin, 1978).
* (12) A. Vourdas, Phys. Rev. A 39, 206 (1989).
* (13) A. Vourdas, Phys. Rev. A 37, 3890 (1988).
* (14) H.-Y. Fan, H. R. Zai and J. R. Klauder, Phys. Rev. D 35, 1831 (1987); H.-Y. Fan, H.-L. Lu and Y. Fan, Ann. Phys. 321, 480 (2006).
* (15) A. Wünsche, J. Opt. B: Quantum Semiclass. Opt. 1, R11 (1999).
* (16) A. Wünsche, J. Phys. A: Math. and Gen. 33, 1603 (2000).
* (17) R. R. Puri, Mathematical Methods of Quantum Optics, (Springer-Verlag, Berlin, 2001), Appendix A.
* (18) L. Brillouin, J. Appl. Phys. 24, 1152 (1953); A. Wehrl, Rev. Mod. Phys. 50, 221 (1978).
* (19) K. Vogel and H. Risken, Phys. Rev. A 40, 2847 (1989).
* (20) D. T. Smithey et al., Phys. Rev. Lett. 70, 1244 (1993).
* (21) M. O. Scully and M. S. Zubairy, Quantum Optics (Cambridge: Cambridge University Press, 1997).
| arxiv-papers | 2008-07-07T14:57:36 | 2024-09-04T02:48:56.566462 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Li-yun Hu and Hong-yi Fan",
"submitter": "Liyun Hu",
"url": "https://arxiv.org/abs/0807.1028"
} |
0807.1063 | ON GEOMETRY OF THE RÖSSLER
SYSTEM OF EQUATIONS
Valery Dryuma
Institute of Mathematics and Informatics AS Moldova, Kishinev
E-mail: valery@dryuma.com; cainar@mail.md
Abstract
Theory of Riemann extensions of spaces of constant affine connection is
proposed to study the Rössler system of equations
$\frac{dx}{ds}=-(y+z),\quad\frac{dy}{ds}=x+\alpha
y,\quad\frac{dz}{ds}=\beta+xz-\nu z.$
After its presentation in a homogeneous form
${\frac{d}{ds}}\xi(s)+1/5\,{\xi}^{2}-1/5\,\xi\,\rho\,\nu+1/5\,\xi\,\rho\,\alpha+\eta\,\rho+\theta\,\rho=0$
${\frac{d}{ds}}\eta(s)+1/5\,\xi\,\eta-\xi\,\rho-1/5\,\eta\,\rho\,\nu-4/5\,\eta\,\rho\,\alpha=0$
${\frac{d}{ds}}\theta(s)-4/5\,\xi\,\theta+4/5\,\theta\,\rho\,\nu+1/5\,\theta\,\rho\,\alpha-\beta\,{\rho}^{2}=0$
${\frac{d}{ds}}\rho(s)+1/5\,\xi\,\rho-1/5\,{\rho}^{2}\nu+1/5\,{\rho}^{2}\alpha=0,$
where
$x=\frac{\xi}{\rho},\quad y=\frac{\eta}{\rho},\quad z=\frac{\theta}{\rho},$
it can be considered as geodesic equations
$\frac{d^{2}X^{i}}{ds^{2}}+\Pi^{i}_{jk}\frac{dX^{j}}{ds}\frac{dX^{k}}{ds}=0.$
of four dimensional space $M^{4}$ of constant affine connection with the
components $\Pi^{i}_{jk}=\Pi^{i}_{jk}(\alpha,\beta,\nu)$ depending on
parameters.
## 1 From the first order system of equations to the second order systems of
ODE
The systems of the first order differential equations
$\frac{dx^{i}}{ds}=c^{i}+a^{i}_{j}x^{j}+b^{i}_{jk}x^{j}x^{k}$ (1)
depending on the parameters $a,b,c$ are not suitable object of consideration
from usually point of the Riemann geometry.
The systems of the second order differential equations in form
$\frac{d^{2}x^{i}}{ds^{2}}+\Pi^{i}_{kj}(x)\frac{dx^{k}}{ds}\frac{dx^{j}}{ds}=0$
(2)
are best suited to do that.
They can be considered as geodesic of affinely connected space $M^{k}$ in
local coordinates $x^{k}$. The values $\Pi^{i}_{jk}=\Pi^{i}_{kj}$ are the
coefficients of affine connections on $M^{k}$.
With the help of such coefficients can be constructed curvature tensor and
others geometrical objects defined on variety $M^{k}$.
## 2 From affinely connected space to the Riemann space
We shall construct the Riemann space starting from a given affinely connected
space defined by the systems of the second order ODE’s.
With this aim we use the notion of the Riemann extension of nonriemannian
space which was used earlier in the articles of author.
Remind the basic properties of this construction.
With help of the coefficients of affine connection of a given n-dimensional
space can be introduced 2n-dimensional Riemann space $D^{2n}$ in local
coordinates $(x^{i},\Psi_{i})$ having the metric of form
${{}^{2n}}ds^{2}=-2\Pi^{k}_{ij}(x^{l})\Psi_{k}dx^{i}dx^{j}+2d\Psi_{k}dx^{k}$
(3)
where $\Psi_{k}$ are an additional coordinates.
The important property of such type metric is that the geodesic equations of
metric (3) decomposes into two parts
$\ddot{x}^{k}+\Pi^{k}_{ij}\dot{x}^{i}\dot{x}^{j}=0,$ (4)
and
$\frac{\delta^{2}\Psi_{k}}{ds^{2}}+R^{l}_{kji}\dot{x}^{j}\dot{x}^{i}\Psi_{l}=0,$
(5)
where
$\frac{\delta\Psi_{k}}{ds}=\frac{d\Psi_{k}}{ds}-\Pi^{l}_{jk}\Psi_{l}\frac{dx^{j}}{ds}$
and $R^{l}_{kji}$ are the curvature tensor of n-dimensional space with a given
affine connection.
The first part (4) of the full system is the system of equations for geodesic
of basic space with local coordinates $x^{i}$ and it do not contains the
supplementary coordinates $\Psi_{k}$.
The second part (5) of the system has the form of linear $N\times N$ matrix
system of second order ODE’s for supplementary coordinates $\Psi_{k}$
$\frac{d^{2}\vec{\Psi}}{ds^{2}}+A(s)\frac{d\vec{\Psi}}{ds}+B(s)\vec{\Psi}=0.$
(6)
Remark that the full system of geodesics has the first integral
$-2\Pi^{k}_{ij}(x^{l})\Psi_{k}\frac{dx^{i}}{ds}\frac{dx^{j}}{ds}+2\frac{d\Psi_{k}}{ds}\frac{dx^{k}}{ds}=\nu$
(7)
which is equivalent to the relation
$2\Psi_{k}\frac{dx^{k}}{ds}=\nu s+\mu$ (8)
where $\mu,\nu$ are parameters.
The geometry of extended space connects with geometry of basic space. For
example the property of the space to be Ricci-flat $R_{ij}=0$ or symmetrical
$R_{ijkl;m}=0$ keeps also for an extended space.
It is important to note that for extended space having the metric (3) all
scalar curvature invariants are vanished.
As consequence the properties of linear system of equation (5-6) depending
from the the invariants of $N\times N$ matrix-function
$E=B-\frac{1}{2}\frac{dA}{ds}-\frac{1}{4}A^{2}$
under change of the coordinates $\Psi_{k}$ can be of used for that.
First applications the notion of extended spaces for the studying of nonlinear
second order ODE’s connected with nonlinear dynamical systems have been
considered by author (V.Dryuma 2000-2008).
## 3 Eight-dimensional Riemann space for the Rössler system of equations
To investigation the properties of the Rössler system equations
$\frac{dx}{ds}=-(y+z),\quad\frac{dy}{ds}=x+\alpha
y,\quad\frac{dz}{ds}=\beta+xz-\nu z$ (9)
we use its presentation in homogeneous form
${\frac{d}{ds}}\xi(s)+1/5\,{\xi}^{2}-1/5\,\xi\,\rho\,\nu+1/5\,\xi\,\rho\,\alpha+\eta\,\rho+\theta\,\rho=0$
${\frac{d}{ds}}\eta(s)+1/5\,\xi\,\eta-\xi\,\rho-1/5\,\eta\,\rho\,\nu-4/5\,\eta\,\rho\,\alpha=0$
${\frac{d}{ds}}\theta(s)-4/5\,\xi\,\theta+4/5\,\theta\,\rho\,\nu+1/5\,\theta\,\rho\,\alpha-\beta\,{\rho}^{2}=0$
${\frac{d}{ds}}\rho(s)+1/5\,\xi\,\rho-1/5\,{\rho}^{2}\nu+1/5\,{\rho}^{2}\alpha=0,$
where
$x=\frac{\xi}{\rho},\quad y=\frac{\eta}{\rho},\quad z=\frac{\theta}{\rho}.$
The relation between both systems is defined by the conditions
$x(s)=\frac{\xi}{\rho},\quad y(s)=\frac{\eta}{\rho},\quad
z(s)=\frac{\theta}{\rho}.$
Remark that for a given system
$\dot{\xi}=P,\quad\dot{\eta}=Q,\quad\dot{\theta}=R,\quad\dot{\rho}=T$
the condition
$P_{\xi}+Q_{\eta}+R_{\theta}+T_{\rho}=0$
is fulfield.
Such type of the system can be rewriten in the form
$\frac{d^{2}X^{i}}{ds^{2}}+\Pi^{i}_{jk}\frac{dX^{j}}{ds}\frac{dX^{k}}{ds}=0,$
which allow us to consider it as geodesic equations of the space with constant
affine connection.
In our case nonzero components of connection are
$\Pi^{1}_{11}=\frac{1}{5},\quad\Pi^{1}_{14}=\frac{\alpha-\nu}{10},\quad\Pi^{1}_{24}=\frac{1}{2},\quad\Pi^{1}_{34}=\frac{1}{2},\quad\Pi^{2}_{12}=\frac{1}{10},$
$\Pi^{2}_{14}=-\frac{1}{2},\quad\Pi^{2}_{24}=-\frac{4\alpha+\nu}{10},\quad\Pi^{3}_{34}=\frac{4\nu+\alpha}{10},$
$\Pi^{3}_{12}=-\frac{4}{10},\quad\Pi^{3}_{44}=-\beta,\quad\Pi^{4}_{14}=\frac{1}{10},\quad\Pi^{4}_{44}=\frac{\alpha-\nu}{5}.$
The metric of corresponding Riemann space is
$^{8}ds^{2}=-2\,\Pi^{1}_{11}\,Pdx^{2}+2\,\left(-2\,\Pi^{2}_{12}\,Q-2\,\Pi^{3}_{12}\,U\right)dxdy+2\,\left(-2\,\Pi^{1}_{14}\,P-2\,\Pi^{2}_{14}\,Q-2\,\Pi^{4}_{14}\,V\right)dxdu+$
$+2\,\left(-2\,\Pi^{1}_{24}\,P-2\,\Pi^{2}_{24}\,Q\right)dydu+2\,\left(-2\,\Pi^{1}_{34}\,P-2\,\Pi^{3}_{34}\,U\right)dzdu+$
$+\left(-2\,\Pi^{3}_{44}\,U-2\,\Pi^{4}_{44}\,V\right)du^{2}+2\,dxdP+2\,dydQ+2\,dzdU+2\,dudV,$
(10)
where $(P,Q,U,V)$ are an additional coordinates.
Geodesic of the metric (3) for coordinates $(x,y,z,u)$ are
${\frac{d^{2}}{d{s}^{2}}}x(s)+1/5\,\left({\frac{d}{ds}}x(s)\right)^{2}+1/5\,\left({\frac{d}{ds}}u(s)\right)\left({\frac{d}{ds}}x(s)\right)\alpha-1/5\,\left({\frac{d}{ds}}u(s)\right)\left({\frac{d}{ds}}x(s)\right)\nu+\left({\frac{d}{ds}}u(s)\right){\frac{d}{ds}}y(s)+$
$+\left({\frac{d}{ds}}u(s)\right){\frac{d}{ds}}z(s)=0,$
${\frac{d^{2}}{d{s}^{2}}}y(s)+1/5\,\left({\frac{d}{ds}}y(s)\right){\frac{d}{ds}}x(s)-\left({\frac{d}{ds}}u(s)\right){\frac{d}{ds}}x(s)-4/5\,\left({\frac{d}{ds}}u(s)\right)\left({\frac{d}{ds}}y(s)\right)\alpha-1/5\,\left({\frac{d}{ds}}u(s)\right)\left({\frac{d}{ds}}y(s)\right)\nu=0,$
${\frac{d^{2}}{d{s}^{2}}}z(s)-4/5\,\left({\frac{d}{ds}}y(s)\right){\frac{d}{ds}}x(s)+4/5\,\left({\frac{d}{ds}}u(s)\right)\left({\frac{d}{ds}}z(s)\right)\nu+1/5\,\left({\frac{d}{ds}}u(s)\right)\left({\frac{d}{ds}}z(s)\right)\alpha-\beta\,\left({\frac{d}{ds}}u(s)\right)^{2}=0,$
${\frac{d^{2}}{d{s}^{2}}}u(s)+1/5\,\left({\frac{d}{ds}}u(s)\right){\frac{d}{ds}}x(s)+1/5\,\left({\frac{d}{ds}}u(s)\right)^{2}\alpha-1/5\,\left({\frac{d}{ds}}u(s)\right)^{2}\nu=0$
and they have a form of homogeneous Rössler system in the variables
$\xi=\frac{dx}{ds},\quad\eta=\frac{dy}{ds},\theta=\frac{dz}{ds},\quad\rho=\frac{du}{ds}.$
The system of second order differential equations for additional coordinates
can be reduced to the linear system of the first order equations with variable
coefficients
${\frac{d}{dt}}P(t)+A_{1}P(t)+B_{1}Q(t)+C_{1}U(t)+E_{1}V(t)=0,$
${\frac{d}{dt}}Q(t)+A_{2}P(t)+B_{2}Q(t)+C_{2}U(t)+E_{2}V(t)=0,$
${\frac{d}{dt}}U(t)+A_{3}P(t)+B_{3}Q(t)+C_{3}U(t)+E_{3}V(t)=0,$
${\frac{d}{dt}}V(t)+A_{4}P(t)+B_{4}Q(t)+C_{4}U(t)+E_{4}V(t)=0,$
where $A_{i},~{}B_{i},~{}C_{i},~{}E_{i}$ are the functions of the variables
$x,~{}y,~{}z,~{}u)$.
Properties of such type of the systems can be investigated with help of the
Wilczynski invariants.
## 4 Laplace operator
In theory of Riemann spaces the equation
$L\psi=g^{ij}(\frac{\partial^{2}}{\partial x^{i}\partial
x^{j}}-\Gamma^{k}_{ij}\frac{\partial}{\partial x^{k}})\psi(x)=0$ (11)
can be used to the study of the properties of spaces.
For the eight-dimensional space with the metric (3) corresponded the Rössler
system we get the equation on the function
$\psi(x,y,z,u,P,Q,U,V)=\theta(P,Q,U,V)$
$4/5\,{\frac{\partial}{\partial
P}}\theta(P,Q,U,V)+2/5\,P{\frac{\partial^{2}}{\partial{P}^{2}}}\theta(P,Q,U,V)-8/5\,\left({\frac{\partial^{2}}{\partial
P\partial Q}}\theta(P,Q,U,V)\right)U+$
$+2/5\,\left({\frac{\partial^{2}}{\partial P\partial
Q}}\theta(P,Q,U,V)\right)Q-2\,\left({\frac{\partial^{2}}{\partial P\partial
V}}\theta(P,Q,U,V)\right)Q+2/5\,\left({\frac{\partial^{2}}{\partial P\partial
V}}\theta(P,Q,U,V)\right)V+$ $+2\,\left({\frac{\partial^{2}}{\partial
Q\partial V}}\theta(P,Q,U,V)\right)P+2\,\left({\frac{\partial^{2}}{\partial
U\partial V}}\theta(P,Q,U,V)\right)P+8/5\,\left({\frac{\partial}{\partial
V}}\theta(P,Q,U,V)\right)\alpha+$ $+2/5\,\left({\frac{\partial}{\partial
V}}\theta(P,Q,U,V)\right)\nu-2/5\,\left({\frac{\partial^{2}}{\partial{V}^{2}}}\theta(P,Q,U,V)\right)V\nu+2/5\,\left({\frac{\partial^{2}}{\partial{V}^{2}}}\theta(P,Q,U,V)\right)V\alpha+$
$+2/5\,\left({\frac{\partial^{2}}{\partial U\partial
V}}\theta(P,Q,U,V)\right)U\alpha+2/5\,\left({\frac{\partial^{2}}{\partial
Q\partial
V}}\theta(P,Q,U,V)\right)Q\nu+8/5\,\left({\frac{\partial^{2}}{\partial
Q\partial V}}\theta(P,Q,U,V)\right)Q\alpha-$
$-2/5\,\left({\frac{\partial^{2}}{\partial P\partial
V}}\theta(P,Q,U,V)\right)P\nu-2\,\left({\frac{\partial^{2}}{\partial{V}^{2}}}\theta(P,Q,U,V)\right)\beta\,U+8/5\,\left({\frac{\partial^{2}}{\partial
U\partial V}}\theta(P,Q,U,V)\right)U\nu+$
$+2/5\,\left({\frac{\partial^{2}}{\partial P\partial
V}}\theta(P,Q,U,V)\right)P\alpha=0.$
This equation has varies type of particular solutions.
A simplest one is
$\psi(x,y,z,u,P,Q,U,V)={e^{\left(\nu-\alpha\right)P-5\,Q+5\,U+V}}$
at the conditions on parameters of the Rössler system
$\beta={\frac{25}{2}}\,\nu,\quad\alpha=-3/2\,\nu.$
As examples obtained by direct substitutions we get the quadratic solution
$\theta(P,Q,U,V)=$
$=1/18\,{\frac{9\,{Q}^{2}+30\,QP\alpha-18\,QV+25\,{P}^{2}{\alpha}^{2}-30\,P\alpha\,V+9\,{V}^{2}+15\,{P}^{2}+60\,U\alpha\,P+45\,\beta\,UP}{{\it\\_c}_{{2}}}}$
with conditions
$\nu=8/3\,\alpha,\quad\beta=arbitrary.$
Cubic solution
$\theta(P,Q,U,V)=-{\frac{1}{240}}\,{\frac{\left(-96\,\alpha\,{\it
l3}-2880\,{\alpha}^{2}{\it
l1}\right){V}^{3}}{{\beta}^{2}}}-{\frac{1}{240}}\,{\frac{\left(-48\,\beta\,{\it
l3}-1440\,\beta\,{\it l1}\,\alpha\right)U{V}^{2}}{{\beta}^{2}}}+$
$+\left(-{\frac{1}{240}}\,{\frac{\left(60\,\beta\,{\it l3}+1800\,\beta\,{\it
l1}\,\alpha\right){P}^{2}}{{\beta}^{2}}}+{\it l1}\,{U}^{2}\right)V+{\it
k1}\,{U}^{3}+\left({\it l3}\,P+{\it l2}\,Q\right){U}^{2}-$
$-{\frac{1}{240}}\,{\frac{\left(2250\,\beta\,{\it l1}\,\alpha+75\,\beta\,{\it
l3}\right){P}^{2}Q}{{\beta}^{2}}}+\left({\it
m2}\,{Q}^{2}-{\frac{1}{240}}\,{\frac{\left(400\,{\beta}^{2}{\it
l1}+3000\,\beta\,{\it l1}\,\alpha+100\,\beta\,{\it
l3}\right){P}^{2}}{{\beta}^{2}}}\right)U+n{Q}^{3}$
at the condition
$\nu=6\alpha$
and arbitrary coefficients ${\it l},{\it m},{\it k},{\it n}$.
A polynomial solution of degree four
$\theta(P,Q,U,V)=r{Q}^{4}+{\it k1}\,{U}^{4}-2/5\,{\it m3}\,{U}^{3}V+{\it
l2}\,U{Q}^{3}+{\it m2}\,{U}^{2}{Q}^{2}+{\it n2}\,{U}^{3}Q+{\it
m3}\,{U}^{2}{P}^{2}$
at the condition
$\nu=-{\frac{7}{13}}\,\alpha.$
Remark that the properties of of such type of solutions depend on parameters
and may be highly diversified.
More complicated solutions of the Laplace equation can be obtained by
application of the method of $(u,v)$\- transformation developed in the works
of author.
## 5 Eikonal equation
Solutions of eikonal equation
$g{{{}^{i}}{{}^{j}}}\frac{\partial F}{\partial x^{i}}\frac{\partial
F}{\partial x^{j}}=0$ (12)
also gives useful information about the properties of Riemann space.
In particular the condition
$F(x^{1},x^{2},...,x^{i})=0$
where function $F(x^{i})$ satisfies the equation $(\ref{dryuma:eq22})$,
determines $(N-1)$-dimensional hypersurface with normals forming an isotropic
vector field.
For the space with the metric (3) the eikonal equation on the function
$\psi(x,y,z,u,P,Q,U,V)=\eta(P,Q,U,V)$ takes the form
$2/5\,P\left({\frac{\partial}{\partial
P}}\eta(P,Q,U,V)\right)^{2}-8/5\,\left({\frac{\partial}{\partial
P}}\eta(P,Q,U,V)\right)\left({\frac{\partial}{\partial
Q}}\eta(P,Q,U,V)\right)U+$ $+2/5\,\left({\frac{\partial}{\partial
P}}\eta(P,Q,U,V)\right)\left({\frac{\partial}{\partial
Q}}\eta(P,Q,U,V)\right)Q-2\,\left({\frac{\partial}{\partial
P}}\eta(P,Q,U,V)\right)\left({\frac{\partial}{\partial
V}}\eta(P,Q,U,V)\right)Q+$ $+2/5\,\left({\frac{\partial}{\partial
P}}\eta(P,Q,U,V)\right)\left({\frac{\partial}{\partial
V}}\eta(P,Q,U,V)\right)P\alpha-2/5\,\left({\frac{\partial}{\partial
P}}\eta(P,Q,U,V)\right)\left({\frac{\partial}{\partial
V}}\eta(P,Q,U,V)\right)P\nu+$ $+2/5\,\left({\frac{\partial}{\partial
P}}\eta(P,Q,U,V)\right)\left({\frac{\partial}{\partial
V}}\eta(P,Q,U,V)\right)V+2\,\left({\frac{\partial}{\partial
Q}}\eta(P,Q,U,V)\right)\left({\frac{\partial}{\partial
V}}\eta(P,Q,U,V)\right)P-$ $-8/5\,\left({\frac{\partial}{\partial
Q}}\eta(P,Q,U,V)\right)\left({\frac{\partial}{\partial
V}}\eta(P,Q,U,V)\right)Q\alpha-2/5\,\left({\frac{\partial}{\partial
Q}}\eta(P,Q,U,V)\right)\left({\frac{\partial}{\partial
V}}\eta(P,Q,U,V)\right)Q\nu+$ $+2\,\left({\frac{\partial}{\partial
U}}\eta(P,Q,U,V)\right)\left({\frac{\partial}{\partial
V}}\eta(P,Q,U,V)\right)P+8/5\,\left({\frac{\partial}{\partial
U}}\eta(P,Q,U,V)\right)\left({\frac{\partial}{\partial
V}}\eta(P,Q,U,V)\right)U\nu+$ $+2/5\,\left({\frac{\partial}{\partial
U}}\eta(P,Q,U,V)\right)\left({\frac{\partial}{\partial
V}}\eta(P,Q,U,V)\right)U\alpha+2/5\,\left({\frac{\partial}{\partial
V}}\eta(P,Q,U,V)\right)^{2}V\alpha-$ $-2/5\,\left({\frac{\partial}{\partial
V}}\eta(P,Q,U,V)\right)^{2}V\nu-2\,\left({\frac{\partial}{\partial
V}}\eta(P,Q,U,V)\right)^{2}\beta\,U=0.$ (13)
A simplest solution of this equation is
$\eta(P,Q,U,V)=-{\frac{Q}{\alpha}}+{\frac{V}{\nu-\alpha}}+{\frac{U}{\alpha}}+P$
with condition on the coefficients of the Rössler system
$-5\,\beta\,\alpha-11\,\alpha\,\nu+8\,{\nu}^{2}+3\,{\alpha}^{2}=0.$ (14)
From here we find
$\nu={\frac{11}{16}}\,\alpha+1/16\,\sqrt{25\,{\alpha}^{2}+160\,\beta\,\alpha}$
To provide a more complicated solutions of the equation $(\ref{dryuma:eq22})$
we use the method of $(u,v)$-transformation.
For the sake of convenience we rewrite the equation $(\ref{dryuma:eq22})$ in
the form
$2/5\,x\left({\frac{\partial}{\partial
x}}\eta(x,y,z,p)\right)^{2}-8/5\,\left({\frac{\partial}{\partial
x}}\eta(x,y,z,p)\right)\left({\frac{\partial}{\partial
y}}\eta(x,y,z,p)\right)z+$ $+2/5\,\left({\frac{\partial}{\partial
x}}\eta(x,y,z,p)\right)\left({\frac{\partial}{\partial
y}}\eta(x,y,z,p)\right)y-2\,\left({\frac{\partial}{\partial
x}}\eta(x,y,z,p)\right)\left({\frac{\partial}{\partial
p}}\eta(x,y,z,p)\right)y+$ $+2/5\,\left({\frac{\partial}{\partial
x}}\eta(x,y,z,p)\right)\left({\frac{\partial}{\partial
p}}\eta(x,y,z,p)\right)x\alpha-2/5\,\left({\frac{\partial}{\partial
x}}\eta(x,y,z,p)\right)\left({\frac{\partial}{\partial
p}}\eta(x,y,z,p)\right)x\nu+$ $+2/5\,\left({\frac{\partial}{\partial
x}}\eta(x,y,z,p)\right)\left({\frac{\partial}{\partial
p}}\eta(x,y,z,p)\right)p+2\,\left({\frac{\partial}{\partial
y}}\eta(x,y,z,p)\right)\left({\frac{\partial}{\partial
p}}\eta(x,y,z,p)\right)x-$ $-8/5\,\left({\frac{\partial}{\partial
y}}\eta(x,y,z,p)\right)\left({\frac{\partial}{\partial
p}}\eta(x,y,z,p)\right)y\alpha-2/5\,\left({\frac{\partial}{\partial
y}}\eta(x,y,z,p)\right)\left({\frac{\partial}{\partial
p}}\eta(x,y,z,p)\right)y\nu+$ $+2\,\left({\frac{\partial}{\partial
z}}\eta(x,y,z,p)\right)\left({\frac{\partial}{\partial
p}}\eta(x,y,z,p)\right)x+8/5\,\left({\frac{\partial}{\partial
z}}\eta(x,y,z,p)\right)\left({\frac{\partial}{\partial
p}}\eta(x,y,z,p)\right)z\nu+$ $+2/5\,\left({\frac{\partial}{\partial
z}}\eta(x,y,z,p)\right)\left({\frac{\partial}{\partial
p}}\eta(x,y,z,p)\right)z\alpha+2/5\,\left({\frac{\partial}{\partial
p}}\eta(x,y,z,p)\right)^{2}p\alpha-2/5\,\left({\frac{\partial}{\partial
p}}\eta(x,y,z,p)\right)^{2}p\nu-$ $-2\,\left({\frac{\partial}{\partial
p}}\eta(x,y,z,p)\right)^{2}\beta\,z=0$ (15)
Now after change of the function and derivatives in accordance with the rules
$\eta(x,y,z,p)\rightarrow u(x,t,z,p),\quad y\rightarrow v(x,t,z,p,$
$\frac{\partial\eta(x,y,z,p)}{\partial x}\rightarrow\frac{\partial
u(x,t,z,p)}{\partial x}-\frac{\frac{\partial u(x,t,z,p)}{\partial
t}}{\frac{\partial v(x,t,z,p))}{\partial t}}\frac{\partial
v(x,t,z,p)}{\partial x},$ $\frac{\partial\eta(x,y,z,p)}{\partial
z}\rightarrow\frac{\partial u(x,t,z,p)}{\partial z}-\frac{\frac{\partial
u(x,t,z,p)}{\partial t}}{\frac{\partial v(x,t,z,p))}{\partial
t}}\frac{\partial v(x,t,z,p)}{\partial z},$
$\frac{\partial\eta(x,y,z,p)}{\partial p}\rightarrow\frac{\partial
u(x,t,z,p)}{\partial p}-\frac{\frac{\partial u(x,t,z,p)}{\partial
t}}{\frac{\partial v(x,t,z,p))}{\partial t}}\frac{\partial
v(x,t,z,p)}{\partial p},$ $\frac{\partial\eta(x,y,z,p)}{\partial
y}\rightarrow\frac{\frac{\partial u(x,t,z,p)}{\partial t}}{\frac{\partial
v(x,t,z,p))}{\partial t}},$
where
$u(x,t,z,p)=t{\frac{\partial}{\partial
t}}\omega(x,t,z,p)-\omega(x,t,z,p),\quad v(x,t,z,p)={\frac{\partial}{\partial
t}}\omega(x,t,z,p)$
we find the equation on the function $\omega(x,t,z,p)$
$\left({\frac{\partial}{\partial
x}}\omega(x,t,z,p)\right)^{2}x-5\,tx{\frac{\partial}{\partial
p}}\omega(x,t,z,p)-p\nu\,\left({\frac{\partial}{\partial
p}}\omega(x,t,z,p)\right)^{2}-5\,\beta\,z\left({\frac{\partial}{\partial
p}}\omega(x,t,z,p)\right)^{2}+$ $+p\left({\frac{\partial}{\partial
x}}\omega(x,t,z,p)\right){\frac{\partial}{\partial
p}}\omega(x,t,z,p)+4\,z\nu\,\left({\frac{\partial}{\partial
z}}\omega(x,t,z,p)\right){\frac{\partial}{\partial
p}}\omega(x,t,z,p)+4\,\left({\frac{\partial}{\partial
x}}\omega(x,t,z,p)\right)zt-$ $-\left({\frac{\partial}{\partial
x}}\omega(x,t,z,p)\right)\left({\frac{\partial}{\partial
t}}\omega(x,t,z,p)\right)t-5\,\left({\frac{\partial}{\partial
t}}\omega(x,t,z,p)\right)\left({\frac{\partial}{\partial
x}}\omega(x,t,z,p)\right){\frac{\partial}{\partial p}}\omega(x,t,z,p)+$
$+p\alpha\,\left({\frac{\partial}{\partial
p}}\omega(x,t,z,p)\right)^{2}+5\,x\left({\frac{\partial}{\partial
z}}\omega(x,t,z,p)\right){\frac{\partial}{\partial
p}}\omega(x,t,z,p)+z\alpha\,\left({\frac{\partial}{\partial
z}}\omega(x,t,z,p)\right){\frac{\partial}{\partial p}}\omega(x,t,z,p)+$
$+4\,t\left({\frac{\partial}{\partial
t}}\omega(x,t,z,p)\right)\alpha\,{\frac{\partial}{\partial
p}}\omega(x,t,z,p)+t\left({\frac{\partial}{\partial
t}}\omega(x,t,z,p)\right)\nu\,{\frac{\partial}{\partial p}}\omega(x,t,z,p)-$
$-x\nu\,\left({\frac{\partial}{\partial
x}}\omega(x,t,z,p)\right){\frac{\partial}{\partial
p}}\omega(x,t,z,p)+x\alpha\,\left({\frac{\partial}{\partial
x}}\omega(x,t,z,p)\right){\frac{\partial}{\partial p}}\omega(x,t,z,p)=0.$ (16)
In spite of the fact that this equation looks not a simple than equation (5)
its particular solutions can be find without trouble.
As example the substitution of the form
$\omega(x,t,z,p)=B(x,t,z)+kpt,$
into the equation (5) lead to expression on the function $B(x,t,z)$
$B(x,t,z)=\alpha\,tx+tz+A(t),$
with arbitrary function $A(t)$ and the conditions on coefficients of the
Rössler system
$\beta=1/5\,{\frac{\alpha\,\left(8+5\,k\right)}{{k}^{2}}},\quad\nu={\frac{\alpha\,\left(1+k\right)}{k}}$
depending from arbitrary parameter $k$.
Using the function $\omega(x,t,z,p)$ we can obtain the solution of the
equation (5) by elimination of the parameter $t$ from the relations
$\eta(x,y,z,p)-t{\frac{\partial}{\partial
t}}\omega(x,t,z,p)+\omega(x,t,z,p)=0,\quad y-{\frac{\partial}{\partial
t}}\omega(x,t,z,p)=0.$
As example at the choice $(A(t)=1+t^{2}$ we get
$\eta(x,y,z,p)=1/4\,{k}^{2}{p}^{2}+\left(-1/2\,yk+1/2\,zk+1/2\,\alpha\,xk\right)p+1/4\,{z}^{2}+\left(1/2\,\alpha\,x-1/2\,y\right)z-1+1/4\,{y}^{2}-1/2\,y\alpha\,x+1/4\,{\alpha}^{2}{x}^{2}.$
At the condition $A(t)=\ln(t)$ we get
$\eta(x,y,z,p)=1-\ln(-\left(-y+\alpha\,x+z+kp\right)^{-1}).$
Remark that parameters $\alpha,\beta,\nu$ in these cases the relation (14) is
satisfied.
To cite another example.
Substitution the expression
$\omega(x,t,z,p)=A(t)x+tz+C(t)p$
into the equation (5) lead to the system of equations on the functions
$A(t),~{}C(t)$
$\left(t\nu\,C(t)+4\,t\alpha\,C(t)-5\,A(t)C(t)-A(t)t\right){\frac{d}{dt}}A(t)+\left(A(t)\right)^{2}-\nu\,A(t)C(t)+\alpha\,A(t)C(t)=0,$
$\left(t\nu\,C(t)+4\,t\alpha\,C(t)-5\,A(t)C(t)-A(t)t\right){\frac{d}{dt}}C(t)+A(t)C(t)-\nu\,\left(C(t)\right)^{2}+\alpha\,\left(C(t)\right)^{2}=0,$
$-5\,\beta\,z\left(C(t)\right)^{2}+\left(5\,z\alpha\,t-5\,A(t)z+5\,z\nu\,t\right)C(t)+3\,A(t)zt=0.$
From this system of equations we find the condition on parameters
$\nu=3/5\,\beta-\alpha$ (17)
and expression on the function $A(t)$
$A(t)=-\beta\,C(t).$
Function $C(t)$ in this case satisfies the equation
$\left(25\,\beta\,C(t)+8\,\beta\,t+15\,t\alpha\right){\frac{d}{dt}}C(t)-8\,\beta\,C(t)+10\,\alpha\,C(t)=0.$
Its solution is defined by the relation
$-t\alpha+\left(C(t)\right)^{-1/2\,{\frac{8\,\beta+15\,\alpha}{-4\,\beta+5\,\alpha}}}{\it\\_C1}\,\alpha-\beta\,C(t)=0.$
Using the expression on the function $C(t)$ we can find the function
$\omega(x,t,z,p)$ and after elimination of the parameter $t$ from
corresponding relations it is possible to get the solution of the eikonal
equation at the condition (17) on parameters of the Rössler system of
equations.
## References
* [1] Dryuma V., The Riemann Extension in theory of differential equations and their applications,Matematicheskaya fizika, analiz, geometriya, 2003,v.10, No.3, 1–19.
* [2] Dryuma V., The Invariants, the Riemann and Einstein-Weyl geometries in theory of ODE’s, their applications and all that, New Trends in Integrability and Partial Solvability, p. 115-156, (ed. A.B.Shabat et al.), Kluwer Academic Publishers, (ArXiv: nlin: SI/0303023, 11 March, 2003, 1–37).
* [3] Dryuma V., Applications of Riemannian and Einstein-Weyl Geometry in the theory of second order ordinary differential equations,Theoretical and Mathematical Physics, 2001, V.128, N 1, 845–855.
* [4] Valery Dryuma, On theory of surfaces defined by the first order sustems of equations Buletinul Academiei de Stiinte a Republicii Moldova, (matematica), 1(56), 2008, p.161–175.
* [5] Valery Dryuma, Riemann extensions in theory of the first order systems of differential equations ArXiv:math. DG/0510526 v1 25 Oct 2005, 1–21.
* [6] Valery Dryuma, Riemann geometry in theory of the first order systems of differential equations ArXiv: 0807.01178 v1 [nlin.SI] 1 Jul 2008, 1–17.
| arxiv-papers | 2008-07-07T16:29:49 | 2024-09-04T02:48:56.571310 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Valery Dryuma",
"submitter": "Valerii Dryuma sem",
"url": "https://arxiv.org/abs/0807.1063"
} |
0807.1118 | # Enhancement of Entanglement Percolation in Quantum Networks via Lattice
Transformations
G John Lapeyre, Jr lapeyre@physics.arizona.edu Jan Wehr Department of
Mathematics, The University of Arizona, Tucson, AZ 85721-0089, USA Maciej
Lewenstein ICFO-Institut de Ciències Fotòniques, Mediterranean Technology
Park, E-08860 Castelldefels (Barcelona), Spain
ICREA-Institució Catalana de Recerca i Estudis Avançats, Lluis Companys 23,
08010 Barcelona, Spain
(March 4, 2009)
###### Abstract
We study strategies for establishing long-distance entanglement in quantum
networks. Specifically, we consider networks consisting of regular lattices of
nodes, in which the nearest neighbors share a pure, but non-maximally
entangled pair of qubits. We look for strategies that use local operations and
classical communication. We compare the classical entanglement percolation
protocol, in which every network connection is converted with a certain
probability to a singlet, with protocols in which classical entanglement
percolation is preceded by measurements designed to transform the lattice
structure in a way that enhances entanglement percolation. We analyze five
examples of such comparisons between protocols and point out certain rules and
regularities in their performance as a function of degree of entanglement and
choice of operations.
###### pacs:
03.67.-a, 03.67.Bg, 64.60.ah
## I Introduction
Entanglement is the property of states of multipartite quantum systems that is
the most important resource for quantum information processing Horodecki et
al. . For this reason, one of the most important tasks of quantum information
science is to establish entanglement at long distances in quantum networks,
and to optimize final entanglement and probability of success.
Quantum networks Cirac et al. (1997); Boozer et al. (2007) play a key role in
quantum information processing. Here we limit our attention to those networks
in which quantum states can be prepared initially and shared. That is,
entanglement can be generated between neighboring or, at least, not-too-
widely-separated nodes (or stations). There are two instances in which the
above-mentioned tasks become obviously relevant. On one hand, one can consider
macroscopic quantum communication networks, such as cryptographic networks, or
more generally quantum communication nets Ekert (1991); Bennett et al. (1993);
Poppe et al. (2008), or distributed quantum computation Cirac et al. (1999)
involving arbitrary nodes of the network. The second instance concerns
microscopic or mesoscopic networks that could constitute architectures of
quantum computers (cf. Ref. Nielsen and Chuang, 2000).
Despite enormous progress in experimental techniques (cf. Ref. Poppe et al.,
2008 and references therein), it is in principle a very hard task to establish
entanglement at large distances due to decoherence and attenuation effects.
Two remedies for this problem have been proposed:
* •
Quantum repeaters. This concept has been developed for 1D quantum
communication chains Briegel et al. (1998); Dür et al. (1999); Childress et
al. (2005); Hartmann et al. (2007). Although the simple entanglement swapping
Żukowski et al. (1993) procedure can lead to quantum communication at large
distances (see Fig. 1), for imperfect resources, the performance of such
communication chains decays exponentially with the distance (i.e. the number
of repeaters). However, one can use more sophisticated quantum repeater
protocols, which use purification and swapping methods that lead to polynomial
decay only.
* •
Entanglement percolation. Recently, our collaborators, together with two of
us, proposed using networks in which properties of the connectivity of the
network enable the establishment of, and determine the probability of,
entanglement on large distances. In Refs. Acín et al., 2007 and Perseguers et
al., 2008 we considered in particular pure-state networks on regular lattices,
where the nearest-neighbor (NN) nodes share a non-maximally entangled pair of
qubits, or more generally qudits (an entangled bond). We searched for local
operations and classical communication (LOCC) protocols that lead to
establishment of entanglement between remote nodes of the network.
Figure 1: Entanglement swapping and 1D repeater. Circles are qubits. Heavy
lines represent pure non-maximally-entangled states. Loops represent the
entanglement swapping measurement on a pair of qubits. a) Two states $\alpha$
and $\beta$ of the form specified in (1). After the operation, qubits $a$ and
$d$ may be in an entangled, mixed state. b) swapping as the first step in a
repeater. c) a 1D chain of repeaters.
At present, only a handful of studies that explicitly apply percolation theory
to problems in quantum information have appeared in the literature. These
articles, comprising Refs. Acín et al., 2007 and Perseguers et al., 2008 as
well as applications to cluster states, are reviewed in Ref. Kieling and
Eisert, 2008.
Our results leading to the present work can be summarized as follows:
* •
For 1D chains we proved that even optimal LOCC strategies only allow the
establishment of entanglement between distant nodes that decays exponentially
with both the distance, as well as the quality of entanglement of NN bonds.
* •
In 2D and higher dimensions, the possibilities for protocols are greatly
expanded. The most straightforward, naive protocol— the one we use as a
baseline to evaluate other protocols— is the one we term classical
entanglement percolation (CEP), although it does involve obviously some
quantum operations. This protocol begins with converting (using LOCC) each of
the entangled bonds into a singlet (i.e. maximally entangled state) with
probability $p$ (the so-called singlet conversion probability (SCP) Vidal
(1999); Nielsen and Vidal (2001)). After the conversion, each of the parties
(nodes) knows obviously which of the bonds are now perfectly entangled. Using
classical communications, the parties establish whether there exists an
infinite percolating cluster (or one spanning the lattice ) and who belongs to
it. Then, by performing a series of entanglement swappings, it is possible to
propagate entanglement between any two (widely) separated nodes that belong to
the percolating cluster. We call this scheme “classical” because it
essentially maps the problem onto a classical bond percolation problem
Grimmett (1999), and its success or failure is equivalent to the success or
failure of bond percolation in the same lattice. Namely, if $p>p_{c}$, where
$p_{c}$ is the lattice-dependent critical percolation threshold (or, in other
words, critical open-bond density), then entanglement between any two remote
nodes can be established with probability $P>0$, which asymptotically does not
depend on the distance.
* •
At the same time that we introduced CEP, we presented several schemes that
went beyond the simple application of singlet conversion everywhere followed
by entanglement swapping along a path. We call these quantum entanglement
percolation (QEP) protocols because they use some kind of quantum pre-
processing— for instance quantum measurements to transform one percolation
problem to another one. In these protocols CEP is also used, but is preceded
by application of certain LOCC, which remove and replace bonds in the network
lattice, resulting in a new lattice geometry with fundamentally different
long-range properties. The pre-processing in QEP may greatly improve the
possibility of entanglement percolation, either by reducing $p_{c}$, or by
increasing $P$.
In this paper we present a more systematic and thorough study of QEPs based on
lattice transformations. In particular, we explore the entire parameter space
with the addition of Monte Carlo and series-expansion methods. Although
general principles remain to be found, we do discover certain rules and
regularities governing such strategies (see Sec. III). The paper is organized
as follows. In Sec. II we formulate and describe the models and fix the
notation. Section II.1 discusses quantum networks and percolation, whereas
Sec. II.2 deals with LOCCs used for lattice transformations. Our main results
are presented in Sec. III, which contains descriptions of five examples of
transformations enhancing CEP: i) the transformation of the kagomé lattice to
the square lattice, ii) the transformation of the double-bond honeycomb
(hexagonal) lattice to the triangular lattice, iii) the transformation of the
square lattice to two decoupled copies of the square lattice, iv) the
transformation of the bowtie lattice to the decoupled triangular and square
lattices, and v) the transformation of a triangular lattice with two different
degrees of entanglement into the decoupled square and triangular lattices. The
cases ii), iii), and v) were discussed already in Ref. Perseguers et al.,
2008, but we present here more general and stronger results. We conclude in
Sec. IV.
Over the past few decades, classical percolation theory has seen the
development of a number of quite sophisticated and powerful methods, which we
have adapted to the questions at hand. In order to demonstrate the supremacy
of QEP over CEP in each of the cases i)-v) we employ these methods combined
with the methods of quantum information theory. An overview of these methods
is presented in the appendices: Appendix A presents the necessary facts from
majorization theory and singlet conversion protocols Nielsen and Vidal (2001),
Appendix B explains some details of our Monte Carlo (MC) simulations, and
finally Appendix C deals with series expansions.
## II Formulation of the models
### II.1 Quantum networks and percolation
In this section, we describe the classes of quantum networks and communication
protocols that we investigated.
#### II.1.1 Networks of bipartite states
The networks we consider consist, prior to application of communication
protocols, of a collection of qubits partitioned into pairs, each pair being
prepared in an identical pure state
$\,|\alpha\rangle\in\mathbb{C}^{2}\otimes\mathbb{C}^{2}$. With appropriate
choice of bases, any such state can be written
$\,|\alpha\rangle=\sqrt{\alpha_{0}}\,|00\rangle+\sqrt{\alpha_{1}}\,|11\rangle,$
(1)
where the Schmidt coefficients $\alpha_{0},\alpha_{1}$ satisfy
$\alpha_{0}\geq\alpha_{1}$ and $\alpha_{0}+\alpha_{1}=1$. We identify these
pairs with bonds on a two-dimensional lattice or edges on a graph, with two
spatially-separated qubits, one occupying each end of the bond. At regular
positions, a small set of these qubits are arranged near enough to one another
to allow measurements on any subset. Such a set of qubits then constitutes a
vertex (or node or site) which is incident to each edge that contributes a
qubit to the vertex. This defines a lattice to which we apply methods of
statistical physics. In particular, many results on the possibility of long-
range entanglement are described using percolation theory Grimmett (1999);
Stauffer and Aharony (1992), and depend only on the graph structure of the
system.
#### II.1.2 Classical entanglement percolation
Given a lattice prepared as described above, we search for the protocols
consisting of local operations and classical communication (LOCC) that yield
the maximum probability of achieving entanglement between two nodes separated
by an arbitrarily large distance. The answer depends on the single parameter
$\alpha_{1}$ characterizing the state $\,|\alpha\rangle$. Even with a small
palette of possible operations, finding the globally optimal solution is a
difficult task. Instead, we search for promising classes of protocols. The
simplest protocol consists of attempting to convert the state associated with
each bond to a singlet via the “Procrustean method” of entanglement
concentration, which is the optimal strategy at the level of a single bond
Vidal (1999); Nielsen and Vidal (2001). This conversion succeeds with
probability $p=2\alpha_{1}$ (See Appendix A.), while failure leaves the pair
in a state with no entanglement. In this way the system is described exactly
by a bond percolation process with open-bond density $p$. If there is a path
of open (maximally entangled) bonds connecting two nodes, a sequence of
entanglement swapping measurements, one at each intermediate node, is then
applied in order to entangle the first and last node. This protocol is the
simplest example of classical entanglement percolation.
#### II.1.3 Percolation theory for CEP.
Here we review a few fundamental ideas in percolation theory necessary to
analyze CEP. The nodes in the lattice can be partitioned into sets such that
each node within a set is connected to each other node in the set via a path
of open bonds. Such a set of nodes is called an open cluster, or sometimes
simply a cluster. The central fact of percolation theory is that percolation
processes on most commonly-studied lattices in dimension $2$ and higher
exhibit a continuous phase transition as the bond density passes through a
critical value $p_{c}$. For $p>p_{c}$ there exists with probability one a
unique (for the lattices we study here) cluster of infinite mass (number of
nodes), while for $p<p_{c}$ all clusters are finite with probability one. It
follows that improving an LOCC protocol to obtain a small change in $p$ can
have a dramatic effect on the probability of long-range entanglement. In the
supercritical phase, long-range entanglement is possible, while in the
subcritical phase it is not possible. Serving as the order parameter is the
density of the infinite cluster $\theta(p)$ which we define via
$\theta(p)=P[A\in{\cal C}],$ (2)
the probability that a fixed node $A$ (say, the node at the origin) is in the
the infinite cluster ${\cal C}$. When referring to $\theta$ on a specific
lattice, we use a symbol representing the lattice as a subscript. The
probability that two selected nodes are members of the same cluster decays
roughly exponentially in their separation distance to an asymptotic value of
$\theta^{2}(p)$. (The length scale of decay is the correlation length
$\xi(p)$.) This means, for the problem at hand, that in CEP the probability
that information can be propagated between two nodes is asymptotically
$\theta^{2}(p)$.
### II.2 Transformations of lattice structure
It was proven in Refs. Acín et al., 2007 and Perseguers et al., 2008 that CEP
is not the optimal strategy for establishing long-distance entanglement. The
demonstration is based on applying certain LOCC prior to CEP. All of these
pre-processing LOCC act on pairs of qubits, and they either transform the
state on a given bond, or they replace two adjacent pure-state bonds by one,
in general, mixed state bond.
There are essentially three types of generalized measurement used:
* •
Singlet conversion The optimal LOCC singlet protocol Nielsen and Vidal (2001)
is used in three situations. In CEP with single-bond lattices we apply it
directly to the state (1), which results in the singlet conversion probability
$p=2\alpha_{1}$. If the protocol is successful, the bond that is converted to
the singlet can be used for entanglement propagation (swapping), otherwise it
is useless. In CEP with a double-bond lattice we apply it directly to the two
copies of the state (1), which lives in ${\cal C}^{4}\otimes{\cal C}^{4}$, and
has Schmidt coefficients
$\alpha_{0},\sqrt{\alpha_{0}\alpha_{1}},\sqrt{\alpha_{0}\alpha_{1}},\alpha_{1}$,
ergo the singlet conversion probability is
$p=\min\left\\{1,2(1-\alpha_{0}^{2})\right\\}.$ (3)
Finally, in all of the QEP strategies that we study, we also apply singlet
conversion to any remaining untouched bonds after the lattice transformation.
* •
Entanglement swapping This protocol Żukowski et al. (1993), illustrated in
Fig. 1a, consists of performing the so–called Bell measurement on a pair of
qubits (b and c) in a node, i.e. a von Neumann measurement in a basis of 4
maximally entangled orthonormal states (in the computational basis). It allows
conversion of a pair of adjacent singlets into a singlet connecting end points
(a and d) with probability 1, i.e. allows for perfect entanglement propagation
in a connected cluster of singlets. At the same time, when applied to a pair
of imperfect states (1), it produces a mixed state, which, amazingly, has the
average singlet conversion probability equal to $p=2\alpha_{1}$.
Unfortunately, this effect cannot be iterated. When applied to the mixed
states, entanglement swapping reduces the singlet conversion probability. This
last point places a significant constraint on our choice of lattice
transformations.
We use entanglement swapping in both CEP and QEP, but the two uses have very
different effects. In the the case of CEP, entanglement swapping is used to
locally move entanglement between neighboring nodes, an operation that is
repeated in hopes of transporting entanglement over long distances (This is,
roughly speaking, a brute-force method). But with QEP, before attempting to
transfer entanglement, we search for a way to selectively apply entanglement
swappings to alter the geometry of the lattice and hence its long-range
connectivity properties as given by percolation theory. The goal of this paper
and future work is to enumerate the rich possibilities and point a way towards
a general description of QEP.
* •
Worst case entanglement Finally, in Refs. Acín et al., 2007 and Perseguers et
al., 2008 the worst case protocol was used, which maximizes minimal
entanglement over all measurement outcomes. This protocol consists also of
Bell measurement, but the basis is computational for one qubit, and
corresponding to eigenstates of $\sigma_{x}$ for the second. When applied to
qubits, it produces a mixed state with the property that for all measurement
outcomes, the resulting pure states have the same singlet conversion
probability. We will not use this protocol here.
In order to describe the lattice transformations involved in QEP, we need a
more general formulation of the lattice than the one given in Sec. II.1.1 for
CEP. It is useful to define the percolation processes a bit more precisely. We
begin with a graph, that is, a set of edges (bonds) $E$ and a set of vertices
(nodes) $V$. We consider embedding the graph in $\mathbb{R}^{2}$ in order to
treat geometric properties. In fact, the important properties don’t change if
we force the vertices to occupy points in $\mathbb{Z}^{2}$. The configuration
of open and closed bonds can be described by a probability sample space
$\Omega=\prod_{e\in E}\\{0,1\\}$, with points $\omega=(\omega(e):e\in E)$,
where $\omega(e)$ takes the values $0$ and $1$. We allow each bond to be open
with a different probability, i.e. $\boldsymbol{p}=(p(e):e\in E)$. The
appropriate measure is a product measure on $\Omega$ with marginal
probabilities defined by $\mu_{e}(\omega(e)=1)=p(e)$,
$\mu_{e}(\omega(e)=0)=1-p(e)$. To allow all transformations possible via LOCC,
we take the graph to be complete— that is, all possible edges $e=\langle
v,w\rangle$ with $v,w\in V$ are present (In fact, sometimes we need double
bonds, as well.) For percolation properties, any graph with fewer edges can be
identified with this complete graph by setting the appropriate $p(e)$ to zero.
Figure 2: (Color online) a) Mapping a quantum network to a percolation
problem. Qubits are represented by small circles. Each vertex (node) contains
six qubits. b) using entanglement swapping to transform the lattice structure.
The two red (grey) bonds in a) are replaced by the red bond in b). This
process can be continued to produce a double-bond hexagonal lattice.
Now we describe the correspondence of entanglement on the physical system of
qubits to this percolation formulation. We denote the set of the indices of
all qubits by $A$. Each qubit $a\in A$ is assigned to a vertex $v=V(a)$, with,
in general, multiple qubits assigned to each vertex. For example, a portion of
a network that can be described by a triangular lattice is shown in figure
Fig. 2a, where vertex $v_{1}$ contains the six circled qubits. For every pair
of qubits $a,b\in A$ with $V(a)\neq V(b)$, we denote the reduced state by
$\rho_{ab}$, and by $S(a,b)$ the singlet conversion probability (SCP), that
is, the probability of conversion of $\rho_{ab}$ to a singlet, maximized over
all possible measurements. Then we assign $p(e)=S(a,b)$ for every $e=\langle
V(a),V(b)\rangle$. Initially, all reduced bipartite states have either
$p(e)=0$ for separable states, or $p(e)=p=2\alpha_{1}$ for the prepared, pure,
partially entangled states. In Fig. 2a the edges with $p(e)=p$ are shown in
black, while edges with $p(e)=0$ are simply absent in the diagram. Two bonds
are colored red (grey) only to show that they will be replaced by another bond
(in the sense of altering the SCP) via entanglement swapping. For QEP, a
successful preprocessing, entanglement swapping operation such as that
depicted in Fig. 1, alters the percolation process by setting $p(\langle
V(a),V(b)\rangle)=0$ and $p(\langle V(c),V(d)\rangle)=0$ and $p(\langle
V(a),V(d)\rangle)=p$. Likewise, in the example in Fig. 2, an entanglement
swapping measurement on qubits $q_{1}$ and $q_{2}$ sets $p(\langle
V(q_{1}),V(q_{3})\rangle)=0$, $p(\langle V(q_{2}),V(q_{4})\rangle)=0$, and
$p(\langle V(q_{3}),V(q_{4})\rangle)=p$. Each of the examples discussed below
conforms to this description.
## III Examples
We consider five examples of lattice transformations in this paper, each
exhibiting a different combination of features, with implications for the
analysis of protocols. In particular, two transformations convert one lattice
to another, while the other three convert a lattice into two decoupled
lattices. Four transformations involve lattices with single bonds, while the
fifth involves double bonds. Quantities arising in the analysis, such as
$\theta(p)$, are analytic about $p=1$. In fact, the lowest-order term in the
expansion about $p=1$ is typically much larger than the remaining terms even
relatively far from $p=1$, with the result that most of the interesting
crossover behavior in comparing protocols occurs for smaller values of $p$.
But this behavior does not appear when using the distillation procedure for
double bonds that leads to (3). The reason is that distillation produces a
saturation point for $\theta(p)$ smaller than $p=1$. This results in a much
stronger difference between classical and quantum protocols in the high-
density regime than does singlet conversion on single bonds (two-qubit pure
states). Four of the transformations produce a smaller critical density on at
least one of the resulting lattices, which gives the most pronounced advantage
in the regime near the critical density. The fifth example shows that on some
lattices where a particular QEP strategy is not advantageous, allowing bonds
of different strengths $p$ and $p^{\prime}$ can produce regions in the phase
space $(p,p^{\prime})$ where QEP is indeed advantageous. Finally, consider
comparing CEP in which singlet conversion is applied to each bond separately,
with QEP that results in a single transformed lattice. In every such case we
find that QEP is better than CEP over the entire (non-trivial) range of $p$.
It is an open question whether this is generic behavior.
We analyze the results of the transformation of each lattice in three regimes:
near the critical density (or densities); near $p=1$; and between these two
regimes. Arguments near the critical densities typically rely simply on the
fact that long-range entanglement is impossible on a lattice with density
below the critical density. These results are the most insensitive to details
of the definitions of entanglement and connectivity. Near $p=1$ we compute
high-density expansions for $\theta(p)$ and related quantities (see Sec. C.)
Often the difference between the performance of CEP and QEP in this regime is
marginal. It is important nonetheless to carry out the analysis if we hope to
make general statements about transformations that hold for all $p$. Between
the critical regime and high-density regime there are some techniques widely
used in percolation theory that may be useful, such as Russo’s formula, which
is used to prove inequalities in the rate of change of $\theta(p)$. However,
we leave these techniques for future work and use Monte Carlo computations in
the present paper.
### III.1 Kagomé lattice to square lattice
In our first example we compare CEP on the kagomé lattice to quantum
entanglement percolation consisting of transformation of the kagomé lattice to
the square lattice via entanglement swapping at nodes specified in Fig. 3.
Although we do not treat them here, there are at least two more ways to
transform the kagomé lattice to the square lattice using the same kind of
entanglement swapping. Rigorous bounds have been obtained for
$p_{c}(\text{kagom\'{e}})$ Grimmett (1999) (as well as a high-precision Monte
Carlo estimate Ziff and Suding (1997)) while $p_{c}(\square)$ is known
exactly, proving that the transformation gives an advantage for $p$ lying
between $p_{c}(\square)$ and the lower bound for $p_{c}(\text{kagom\'{e}})$.
As shown in Table 2, the series for $\theta_{\square}$ and
$\theta_{\text{kag}}$ are the same to the first non-trivial order in $q=1-p$.
But the next term shows that $\theta_{\square}>\theta_{\text{kag}}$ for $p$
close enough to $1$. The MC data provides strong evidence that
$\theta_{\square}>\theta_{\text{kag}}$ everywhere except near $p=1$, where the
statistical error is too large to distinguish the curves. But as is evident
from the lower plot in Fig. 3, the terms in the expansions that we computed
dominate $\theta_{\square}-\theta_{\text{kag}}$ already at values of $p$ for
which the MC data is still accurate. Thus we find that QEP is advantageous
over the entire range of $p$.
Figure 3: One of three transformations of the kagomé lattice to the square
lattice. Pairs of qubits that are subjected to entanglement swapping are
marked with loops.
Figure 4: Upper plot: Monte Carlo of $1-\theta(p)$ v.s. $p=2\alpha_{1}$. The
solid line is on the square lattice. The dashed line is on the kagomé lattice.
Lower plot: Normalized difference between $\theta_{\square}$ and
$\theta_{\text{kag}}$. The solid line is computed from MC data. The dashed
line is computed from expansions from Table 2.
### III.2 Double-bond hexagonal lattice to triangular lattice
The hexagonal lattice with double bonds, each in the state specified by (1)
(see Fig. 5a), was discussed in Refs. Acín et al., 2007 and Perseguers et al.,
2008, in which it was shown that a transformation of the lattice to a
triangular lattice offers an advantage over CEP protocols for values of $p$
between $p_{c}(\triangle)$ and another critical value defined below.
Figure 5: (a) Double-bond hexagonal lattice. (b) transformed into the
triangular lattice (b). Pairs of qubits that are subjected to entanglement
swapping are marked with loops.
Here we extend the analysis to determine over the entire range of $p$ which of
three protocols gives the highest probability of long-range entanglement. The
first of the three protocols, which we call CEP I, consists of performing an
optimum singlet conversion to each bond separately so that the probability of
getting at least one singlet connecting two nodes is $p^{\prime}=1-(1-p)^{2}$.
Communication on the resulting lattice is then determined by bond percolation
on the hexagonal lattice with bond density $p^{\prime}$. In particular, the
critical density (see Table 1) is $p=p_{c}(\text{CEP I})\approx 0.4107$. In
the second protocol, we perform a more efficient conversion, namely
distillation, on all four qubits in a double bond. This protocol succeeds in
producing a maximally entangled pair with probability
$p^{\prime\prime}=\min\\{1,2(1-\alpha_{0}^{2})\\}$ (See (3) and Appendix A.)
This results in bond percolation on the hexagonal lattice with density
$p^{\prime\prime}=\min\left\\{1,2\left[1-\left(1-\frac{p}{2}\right)^{2}\right]\right\\}.$
(4)
We refer to this method as CEP II, with critical density $p=p_{c}(\text{CEP
II})\approx 0.358$. In the third method, the entanglement swapping that
maximizes SCP is applied at every other node (see Fig. 5), converting the
double-bond hexagonal lattice to the triangular lattice with bond density $p$,
with $p_{c}(\triangle)\approx 0.347$.
Figure 6: Monte Carlo and series expansions about $p=1$ of $1-\theta(p)$ v.s.
$p=2\alpha_{1}$ for various strategies on the double-bond hexagonal lattice.
(a) Solid lines are Monte Carlo data. Dashed lines are series expansions. (b)
Solid lines are MC. Series expansions are indistinguishable from MC over the
entire plot. (c) Monte Carlo data.
The three protocols are compared in Fig. 6. We first observe that each method
fails below its corresponding critical density, and these are known exactly.
To compare the methods away from the critical points, we computed series
expansions of $\theta(p)$ about $q=0$ with results listed in Table 2. Defining
$\tilde{\theta}(q)=\theta(1-q)$ we see that
$\tilde{\theta}_{\text{CEP
I}}(q)=\tilde{\theta}_{\hexagon}(q^{2})=1-q^{6}-3q^{8}+\ldots$ (5)
Comparing (5) to $\theta_{\triangle}(p)$ from Table 2 we see that for $p\to 1$
the conversion to the triangular lattice is better than CEP I. As is evident
from Figs. 6a and 6b, the two curves have the same leading behavior as $p\to
1$, with the result that the the Monte Carlo cannot distinguish them in this
region. However, the series expansion lies well within the statistical error
of the Monte Carlo points, even relatively far from $p=1$, so we can be
confident that $\theta_{\triangle}>\theta_{\text{CEP I}}$ even in the region
where the MC cannot distinguish the curves. Finally, the MC clearly shows that
$\theta_{\triangle}>\theta_{\text{CEP I}}$ for smaller $p$, where the
expansion fails. Turning now to CEP II, we know that
$\theta_{\triangle}>\theta_{\text{CEP II}}$ for
$p_{c}(\triangle)<p<p_{c}(\text{CEP II})$ and from (4) that $\theta_{\text{CEP
II}}=1$ for $p\geq 2-\sqrt{2}$. It follows that there must be a crossover
point. The MC data suggests that this occurs for $p\approx 0.375$. In summary,
we see that QEP gives an advantage over CEP I for all $p$, but that CEP II,
which is more efficient in its use of double bonds, is better than QEP at high
densities.
### III.3 Square lattice to two decoupled copies of the square lattice
This transformation replaces the square lattice with two decoupled copies of
the square lattice. To effect the transformation, at selected nodes the two
horizontally opposing bonds are joined into one bond and likewise with the
vertically opposing bonds. This procedure is applied at every other node,
while staggering by one node when shifting rows as shown in Fig. 7.
Figure 7: Doubling the square lattice. Large circles represent nodes at which
entanglement swapping is performed. Figure 8: Monte Carlo and series
expansions for doubling the square lattice. Lower points: $1-P_{\text{doub}}$
v.s. $p$. Upper points: $1-{\pi_{\square}}^{2}$ v.s. $p$. Solid lines are
series expansions. Figure 9: Monte Carlo for doubling the square lattice.
Upper points: $1-\theta_{\square}^{2}$ v.s. $p$. Lower points:
$(2-\omega)^{2}-1$ v.s. $p$.
Because the transformation splits the original lattice into two disjoint
lattices, each taking half the surviving nodes, we cannot compare connection
between a single pair of nodes before and after the transformation. Rather, we
consider on the original lattice connection between, on the one hand, a pair
of nodes $A,A^{\prime}$ separated by a small distance, and on the other hand,
the same pair translated a distance much larger than the correlation length to
a pair $B,B^{\prime}$. We choose $A$ and $A^{\prime}$ so that each one goes to
a different lattice in the transformation. After the transformation, there is
the possibility of connection between $A$ and $B$ on one lattice and between
$A^{\prime}$ and $B^{\prime}$ on the other. We choose the nodes as shown in
Fig. 7. In addition to the complication of splitting, this example is subtle
because the critical density is the same on all three lattices. In Ref.
Perseguers et al., 2008 it was shown that just above $p_{c}(\square)=1/2$, the
transformation gives an advantage over CEP as measured by the choice of nodes
$A,A^{\prime}$.
Here we examine the connectivity for all $p>p_{c}(\square)$. The probability
that at least one of the pairs on the decoupled lattices is connected is
$P_{\text{double}}=1-(1-\theta_{\square}^{2})^{2}=\theta_{\square}^{2}(2-\theta_{\square}^{2}).$
We compare $P_{\text{double}}$ to the probability on the original lattice that
at least one of $A$ and $A^{\prime}$ is connected to at least one of $B$ and
$B^{\prime}$. This probability is $\pi_{\square}^{2}$ where
${\pi_{\square}}=P[A\in{\cal C}\text{ or }A^{\prime}\in{\cal C}].$ (6)
(In this paper ‘or’ in ‘$X$ or $Y$’ does not mean exclusive or.) We prove
that the doubling is advantageous in the limit $p\to 1$ using the power series
about $q=0$ for $\theta_{\square}$ and $\pi_{\square}$ given in Table 2. This
follows from comparing
$\pi_{\square}^{2}=1-8q^{8}-36q^{10}+\ldots$
and
$\theta_{\square}^{2}(2-\theta_{\square}^{2})=1-4q^{8}-32q^{10}+\ldots.$
Moreover, Monte Carlo evidence suggests that the transformation improves on
CEP for all $p>p_{c}(\square).$ In Fig. 8 one sees from the accuracy of the
series expansion in the region of high-quality MC data that that the advantage
is maintained even when the Monte Carlo becomes noisy near $p=1$. The same MC
data is not useful near $p_{c}(\square)$ for two reasons. Firstly, both
$P_{\text{doub}}$ and $\pi_{\square}^{2}$ vanish at the same critical point
with infinite slope. Secondly, the systematic error due to finite lattice size
and large fluctuations in cluster statistics further complicate distinguishing
the curves. We instead make use of an alternate expression for $\pi$, that is
$\pi_{\square}=\theta_{\square}(2-\omega)$ where $\omega=P[A\in{\cal
C}|A^{\prime}\in{\cal C}]$. The condition for advantage over CEP then becomes
$(2-\omega)^{2}<2-\theta_{\square}^{2}.$ (7)
The MC data for $\omega$ was generated by considering the largest cluster in
the finite lattice to represent the infinite cluster ${\cal C}$ even if it is
not a spanning cluster. The MC data supports (7), and furthermore, shows no
evidence of non-analyticity at $p_{c}(\square)$, as is evident in Fig. 9.
The foregoing analysis of the doubled square lattice was based on the choice
of $A,A^{\prime}$ shown in Fig. 7. But this is not the only reasonable choice.
In fact the question of whether the doubling transformation is better than CEP
depends somewhat on the details of how the nodes are chosen. Although this is
an extreme example, similar questions arise in analyzing other
transformations. Thus, the ambiguity in the measures comparing the various
protocols must be addressed in future work.
### III.4 Bowtie lattice to decoupled triangular and square lattices
Figure 10: Transformation of the bowtie lattice to decoupled square and
triangular lattices. Loops marking pairs of qubits represent swapping
measurements.
This is our second example that transforms a lattice into two decoupled
lattices. Figure 10 shows the measurements that decouple the bowtie lattice
into the square lattice and the triangular lattice. For $p$ satisfying
$p_{c}(\triangle)<p<p_{c}(\bowtie)$ (the exact critical values are listed in
Table 1), the transformation is obviously advantageous.
In carrying out further analysis, the bowtie lattice presents a complication
not present in other lattices studied in this paper: While the other lattices
are regular in the sense that each vertex has the same environment up to
rotations and reflections, the bowtie lattice has two kinds of vertices in
this sense. Notice first that all the surviving nodes on the square and
triangular lattices shown in Fig. 10 were generated from nodes of coordination
number $z=6$ on the bowtie lattice, with all nodes of $z=4$ disappearing. Each
node of $z=6$ on the bowtie lattice has four nearest neighbors of $z=6$,
discounting nodes of $z=4$ (for example node $A$ in Fig. 10.) Furthermore, two
of these nearest neighbors are connected by a diagonal bond (node
$A^{\prime\prime}$) and two are not (node $A^{\prime}$.)
In treating the example of doubling the square lattice in Sec. III.3, the fact
that the lattice decouples forced us to consider connections between two
widely separated pairs of nodes. We treat the present example in the same way,
except that the more complicated local structure of the bowtie lattice forces
us to consider a cluster of three nodes $A,A^{\prime},A^{\prime\prime}$ rather
than a pair. In Fig. 10b we see that $A^{\prime},A^{\prime\prime}$ are sent to
the square lattice and $A$ is sent to the triangular lattice. We consider a
distant cluster of nodes $B,B^{\prime},B^{\prime\prime}$ related to
$A,A^{\prime},A^{\prime\prime}$ by translation, and examine the probability
that a connection exists between at least one of
$A,A^{\prime},A^{\prime\prime}$ and at least one of
$B,B^{\prime},B^{\prime\prime}$. On the bowtie lattice, this probability is
$\pi_{\bowtie}^{2}$ where
$\pi_{\bowtie}=P_{\bowtie}[A\in{\cal C}\text{ or }A^{\prime}\in{\cal C}\text{
or }A^{\prime\prime}\in{\cal C}].$ (8)
If instead we decouple the lattices via the lattice transformation, this
probability is given by
$P_{\text{doub}\triangle-\square}=\theta_{\triangle}^{2}+{\pi_{\square\text{b}}}^{2}-\theta_{\triangle}^{2}{\pi_{\square\text{b}}}^{2},$
where
${\pi_{\square\text{b}}}=P_{\square}[A^{\prime}\in{\cal C}\text{ or
}A^{\prime\prime}\in{\cal C}].$ (9)
A comparison of Monte Carlo estimates and high-density expansions of these
quantities is shown in Fig. 11.
Figure 11: Monte Carlo estimates of $\pi_{\bowtie}$ and
$P_{\text{doub}\triangle-\square}$. The inset is the same data on a linear
scale. Solid lines are high-density expansions.
The MC data shows that for small densities, the lattice transformation gives
an advantage, while for higher densities CEP is the better protocol. The
cross-over occurs at $p\approx 0.425$. As with the other examples, the curves
representing the series expansions suggests that no cross-over occurs in the
high-density region.
Lattice | $p_{c}$ for bond percolation
---|---
triangular | $2\sin(\pi/18)\approx 0.347$
bowtie | $\approx 0.4045$
square | $0.5$
kagomé | $\approx 0.5244053$ MC estimate
hexagonal | $1-2\sin(\pi/18)\approx 0.653$
Table 1: $p_{c}$ for bond percolation on some lattices. $p_{c}(\bowtie)$ is the unique root of $1-p-6p^{2}+6p^{3}-p^{5}$. All critical densities are exactGrimmett (1999) except for $p_{c}(\text{kagom\'{e}})$Ziff and Suding (1997). $\theta_{\square}(p)$ | $1-q^{4}-4q^{6}$
---|---
${\pi_{\square}}(p)$ | $1-4q^{8}-18q^{10}$
${\pi_{\square\text{b}}}(p)$ | $1-q^{6}+q^{7}-8q^{8}$
$\theta_{\triangle}(p)$ | $1-q^{6}-6q^{10}+6q^{11}-6q^{12}-21q^{14}+42q^{15}$
$\theta_{\hexagon}(p)$ | $1-q^{3}-3q^{4}-6q^{5}-25q^{6}$
$\theta_{\text{kag}}(p)$ | $1-q^{4}-6q^{6}$
$\pi_{\bowtie}$ | $1-4q^{14}$
Table 2: A few terms in series expansions about $q=0$ of $\pi$ and $\theta$
for various lattices.
$\theta_{\square},\theta_{\triangle},\theta_{\hexagon},\theta_{\text{kag}}$
are defined via (2). ${\pi_{\square\text{b}}}$ is defined via (9),
${\pi_{\square}}$ via (6) , and $\pi_{\bowtie}$ via (8).
### III.5 Asymmetric triangular lattice
Figure 12: a) The asymmetric triangular lattice. Solid bonds represent the
state $\,|\alpha\rangle$ (density $p$). Dashed bonds represent the state
$\,|\alpha^{\prime}\rangle$ ( density $p^{\prime}$). b) Entanglement swapping
is performed on pairs of solid bonds resulting in a triangular lattice with
bond density $p$. The remaining dashed bonds form a kagomé lattice which is
then transformed into a square lattice with bond density $p^{\prime}$ as in
Sec. III.1.
With this example we demonstrate that QEP can succeed when the initial bonds
are not all in the same state, but fail if they are in the same state.
Consider the lattice in Fig. 12a composed of two different kinds of bonds,
each of the form given by (1). The solid and dashed bonds represent the states
$\,|\alpha\rangle$ (with $p=2\alpha_{1}$) and $\,|\alpha^{\prime}\rangle$
(with $p^{\prime}=2\alpha_{1}^{\prime}$) respectively. In general, $p\neq
p^{\prime}$. We consider two entanglement distribution protocols: a) classical
entanglement percolation on the original lattice and b) QEP consisting of
creating two decoupled lattices (the square and triangular lattices) via
entanglement swapping, followed by CEP on each of the resulting lattices. The
transformation is described in Fig. 12. Note that, for $p^{\prime}=0$, the
initial lattice is a triangular lattice with serial double bonds and therefore
the critical point $p=\sqrt{p_{c}(\triangle)}\approx 0.589$. On the other hand
for $p=0$ and $p^{\prime}\neq 0$ the initial lattice is the kagomé lattice.
We first examine the most robust measure of a protocol’s effectiveness— the
binary measure that tells whether long-range entanglement is possible or not.
The phase diagrams for this example before and after the transformation are
shown in Fig. 13. We see that there are regions in the phase space for which
QEP is better than CEP, and vice versa.
In the region in which both protocols allow long-range entanglement, a more
detailed measure similar to those in previous sections is necessary. For
example, for $p=p^{\prime}$, we performed an analysis similar to the one used
for doubling the square lattice. The two critical boundaries intersect at
$p=p^{\prime}=p_{c}(\triangle)$, but comparing connectivity just above the
critical point shows that the transformation does not improve the probability
of long-range entanglement. Likewise for $p=p^{\prime}$ and $p$ near $1$,
series expansions showed that the transformation is not an improvement on CEP.
We have not yet determined whether QEP in this scenario succeeds for some
other $p=p^{\prime}$, but this seems unlikely.
Figure 13: Phase diagram for the asymmetric triangular lattice. The heavy
solid curve (obtained by Monte Carlo) separates the supercritical and
subcritical regions on the original asymmetric triangular lattice (CEP). The
dashed curve composed of two line segments separates the supercritical and
subcritical regions after the transformation (QEP). After transformation, the
supercritical region is defined to be the region for which at least one of the
disjoint lattices is supercritical. In the light gray regions, long-range
entanglement is possible with QEP, but not with CEP. In the dark gray region
it is possible with CEP, but not with QEP.
## IV Conclusions
In this paper we have considered the problem of establishing long-distance
entanglement in quantum pure-state networks on regular 2D lattices. We have
discussed in detail several examples of quantum entanglement percolation
strategies that are better than the corresponding classical strategies, i.e.
those consisting of a direct attempt to convert bonds into singlets. Our
results illustrate nicely the interplay between quantum information theory and
classical percolation theory. Despite the fact that we do find certain rules
and regularities governing QEP strategies, many questions remain open. For
instance, we cannot say anything about optimality of our QEP protocols— and
most probably they are not optimal. All of our protocols involve LOCC acting
on pairs of qubits only; can multipartite LOCC, which leads inevitably to
creation of multipartite entanglement, help? We know that for sufficiently
large initial entanglement, perfect entanglement between remote nodes may be
established with distance-independent probability $P>0$. Is it also possible
for arbitrarily small initial entanglement?
## Acknowledgements
We thank A. Acín, J.I. Cirac, and D. Cavalcanti for discussions. We thank S.
Perseguers for kindly supplying figures. We acknowledge support of the EU IP
Programme “SCALA”, and Spanish MEC grants (FIS 2005-04627, Consolider Ingenio
2010 “QOIT”). J.W. thanks Lluis Torner and ICFO for hospitality. He was
partially supported by NSF grant DMS 0623941. M. L. thanks also the Alexander
von Humboldt Foundation for support.
## Appendix A Majorization Theory
Majorization theory has been applied to questions of transforming one
bipartite pure state to another by LOCC. In particular, a theorem due to
VidalNielsen and Vidal (2001) gives the probability that such a transformation
can be achieved via an optimal protocol (without specifying that protocol.) We
state the result and apply it to the distillation procedure used in this
paper. We must first introduce a certain partial order on vectors. Consider
two real, $d$-dimensional vectors ${\boldsymbol{r}}$ and ${\boldsymbol{s}}$.
We define the vector ${\boldsymbol{r^{\uparrow}}}$ by reordering the elements
of ${\boldsymbol{r}}$ into non-decreasing order, and likewise with
${\boldsymbol{s}}$. We say that ${\boldsymbol{r}}$ is submajorized by
${\boldsymbol{s}}$, denoted by ${\boldsymbol{r}}\prec^{w}{\boldsymbol{s}}$, if
$\sum_{i=0}^{k}r_{i}^{\uparrow}\geq\sum_{i=0}^{k}s_{i}^{\uparrow},$ (10)
for all $k=0,\ldots,d-1$. Denoting the vector of Schmidt coefficients of
$\psi$ by ${\boldsymbol{\lambda}}(\psi)$, the theorem states that
$\,|\psi\rangle$ can be transformed into $\,|\phi\rangle$ with probability
$p$, where $p$ is the largest number on $[0,1]$ such that
${\boldsymbol{\lambda}}(\psi)\prec^{w}p{\boldsymbol{\lambda}}(\phi)$.
A simple, relevant application is the computation of the probability that
optimal conversion of a state
$\,|\psi\rangle\in\mathbb{C}^{d}\otimes\mathbb{C}^{d}$ to a singlet will
succeed. Application of (10) gives $p=\min\\{1,2(1-\alpha_{0})\\}$.
Now we consider distillation, defined here as the optimal protocol for
converting $n$ pure states
$\,|\alpha_{i}\rangle\in\mathbb{C}^{2}\otimes\mathbb{C}^{2}$ to $n-1$ product
states and one maximally entangled state in the Schmidt bases. This operation
is useful, for instance, in attempting to get a single, maximally-entangled
bond from $n$ bonds connecting two nodes. Explicitly,
$\,|\alpha_{i}\rangle=\sqrt{\alpha_{i,0}}\,|00\rangle_{i}+\sqrt{\alpha_{i,1}}\,|11\rangle_{i},$
with $\alpha_{i,0}\geq\alpha_{i,1}$ and elements of the ordered set
$(\alpha_{0,0},\alpha_{1,0},\ldots,\alpha_{n-1,0})$ non-increasing. As usual,
$\,|jk\rangle_{i}$ is shorthand for
$\,|j\rangle_{i,0}\otimes\,|k\rangle_{i,1}$. For
${\boldsymbol{j}}\in\\{0,1\\}^{n}$ define the bijective numeration
$k({\boldsymbol{j}})=\sum_{i=0}^{n-1}10^{i}j_{i}$ and
$\gamma_{k}=\gamma_{k}({\boldsymbol{\alpha}})=\alpha_{0,j_{0}}\alpha_{1,j_{1}}\cdots\alpha_{n-1,j_{n-1}}$.
Then we can write
$\bigotimes_{i=0}^{n-1}\,|\alpha_{i}\rangle=\sum_{{\boldsymbol{j}}:k({\boldsymbol{j}})=0}^{2^{n}-1}\sqrt{\gamma_{k}}\bigotimes_{m=0}^{n-1}\,|j_{m}j_{m}\rangle_{m}.$
(11)
In order to apply the theorem, we need to show that this state can be written
as a bipartite state with Schmidt coefficients that can be chosen to satisfy
our distillation condition. To this end collect all the first qubits of the
bipartite states and all the second qubits, defining
$\,|k({\boldsymbol{j}})\rangle_{a}=\bigotimes_{i=0}^{n-1}\,|j_{i}\rangle_{i,0}\
\text{ and
}\,|k({\boldsymbol{j}})\rangle_{b}=\bigotimes_{i=0}^{n-1}\,|j_{i}\rangle_{i,1},$
so that (11) becomes
$\bigotimes_{i=0}^{n-1}\,|\alpha_{i}\rangle=\sum_{{\boldsymbol{j}}:k({\boldsymbol{j}})=0}^{2^{n}-1}\sqrt{\gamma_{k}}\,|k({\boldsymbol{j}})\rangle_{a}\,|k({\boldsymbol{j}})\rangle_{b}.$
(12)
Note that
$\sum_{k=0}^{2^{n}-1}\gamma_{k}=\prod_{i=0}^{n-1}(\alpha_{i,0}+\alpha_{i,1})=1^{n}=1$
so that (12) defines a state in $\mathbb{C}^{2n}\otimes\mathbb{C}^{2n}$
already in Schmidt form. The submajorization condition is
$(\gamma_{2^{n}-1},\gamma_{2^{n}-2},\ldots,\gamma_{0})\prec^{w}\left(0,0,\ldots,\frac{p}{2},\frac{p}{2}\right),$
for which the only nontrivial inequality is the penultimate one
$1-\gamma_{0}\geq p/2$. Thus the maximum distillation probability is
$p=\min\left\\{1,2(1-\alpha_{0,0}\alpha_{1,0}\cdots\alpha_{n-1,0})\right\\}.$
Note that this result agrees with the special case in (3).
## Appendix B Monte Carlo estimates of $\theta(p)$ and $\pi(p)$
We computed Monte Carlo estimates of $\theta(p)$ and $\pi(p)$ using the
Hoshen-Kopelmann Hoshen and Kopelman (1976) algorithm with modifications for
efficiency Nakanishi and Stanley (1980) and the Mersenne twister Nishimura
(2000) random number generator. Together with series expansions and exactly
known critical densities, the quality of the data we obtained is more than
sufficient to determine which transformations are advantageous. We estimated
$\theta(p)$ by computing the mean density of the largest cluster on an
$L\times L$ lattice. Near the critical density we typically used lattices of
size $L=2$–$5\times 10^{5}$. We took the value of $L$ at the inflection point
of plots of $\theta(p)$ v.s. $L$ at fixed $p$ as estimates of the correlation
length $\xi(p)$. We typically found that $\xi(p)>L$ for $p-p_{c}<0.0005$. We
only expect significant systematic error in this region, but this poses no
problem because the curves are never close to one another in these regions.
(We reformulated the problem when this is the case.) We computed statistical
error, but the error bars are at most barely visible on the plots, so we
omitted them. Exceptions are very near $p=p_{c}$ where fluctuations in the
size of the largest cluster become large, and near $p=1$, where finite
clusters are rare, so collecting sufficient samples to distinguish curves is
too expensive. We discuss the effects of these errors in the main body of the
paper.
We wrote a single computer code to study all the lattices. The code supports
lattices with vertices that occupy points on $\mathbb{Z}^{2}$ with bonds
connecting each pair of nearest neighbors as well as a diagonal bond from
$(i,j)$ to $(i+1,j+1)$. Because the connectivity properties that we calculated
depend only on the graph structure of the lattice, we embedded the graph of
each lattice in the square lattice plus diagonals described above. Depending
on the lattice being modeled, some of the bonds in the underlying lattice are
closed with probability one, and some are open with probability one, with the
vertices identified in the embedded lattice. The graph structures of all
lattices appearing in this paper were modeled in this way.
## Appendix C Series Expansions
We follow the ideas of the perimeter method Domb (1959); Domb and Sykes
(1961); Blease et al. (1978) to compute high-density series expansions of
$\theta(p)$ and $\pi(p)$ listed in Table 2. Here we discuss the method for
computing the series for $\theta(p)$, but our method for computing $\pi(p)$ is
similar. Here, a cluster is any connected subgraph that contains at least one
vertex. We denote by ${\cal S}_{0}$ the collection of all finite clusters that
include the vertex at the origin, and by ${\cal S}$ the partition of ${\cal
S}_{0}$ induced by equivalence under translation, in other words the
collection of free clusters. Choosing an enumeration $\alpha_{j}$ of clusters
in ${\cal S}$, the probability that a randomly selected site is in the
infinite open cluster is easily seen to be
$\theta(p)=1-\sum_{j=1}^{\infty}s_{j}(1-q)^{b_{j}}q^{t_{j}},$
where $s_{j}$ is the number of sites and $b_{j}$ the number of bonds in
cluster $\alpha_{j}$, and $t_{j}$ is the number of perimeter bonds (bonds
adjacent to cluster $\alpha_{j}$). Clearly, we can find the series expansion
in $q$ by enumerating the clusters in an order that is non-decreasing in
$t_{j}$. For instance, on the hexagonal lattice, only the cluster consisting
of an isolated site has a perimeter $t$ of size less than or equal to $3$ (see
Fig. 14), so that to lowest non-trivial order $\theta(p)=1-q^{3}$. Percolation
theory is a difficult subject precisely because the full enumeration is
difficult. The series for $\theta_{\triangle}$ (for bond percolation) was
calculated by machine to high order in Ref. Blease et al., 1978. Although
tables of cluster numbers have been published, all the others that we are
aware of are either for site percolation or for bond percolation with cluster
size measured by the number of bonds rather than sites (neither of which can
be mapped to our problem.)
For our results we counted a few small clusters by hand, which is not so
difficult. In practice however, we find that it is also essential to
categorize the clusters by symmetry. As an example, the clusters contributing
to $\theta$ on the hexagonal lattice to sixth order in $q$ are shown in Fig.
14.
Figure 14: Clusters on the hexagonal lattice contributing to $\theta$ to
sixth order in $q$. Symmetries and multiplicities $m$ are: none–$12$, one
axis–$6$, rotation–$4$, two axes–$3$, rotation and axis–$2$, all–$1$. The
number of contributing clusters per figure is given by $n=ms/2$, the factor of
$1/2$ accounting for the fact that each site only supports half the rotations.
The contribution to $\theta$ is then $-ms(1-q)^{b}q^{t}/2$.
## References
* (1) R. Horodecki, P. Horodecki, M. Horodecki, and K. Horodecki, Rev. Mod. Phys. , in press, eprint arXiv:quant-ph/0702225v2.
* Cirac et al. (1997) J. I. Cirac, P. Zoller, H. J. Kimble, and H. Mabuchi, Phys. Rev. Lett. 78, 3221 (1997), eprint arXiv:quant-ph/9611017v1.
* Boozer et al. (2007) A. D. Boozer, A. Boca, R. Miller, T. E. Northup, and H. J. Kimble, Phys. Rev. Lett. 98, 193601 (pages 4) (2007), eprint arXiv:quant-ph/0702248v1, URL http://link.aps.org/abstract/PRL/v98/e193601.
* Ekert (1991) A. K. Ekert, Phys. Rev. Lett. 67, 661 (1991).
* Bennett et al. (1993) C. H. Bennett, G. Brassard, C. Crépeau, R. Jozsa, A. Peres, and W. K. Wootters, Phys. Rev. Lett. 70, 1895 (1993).
* Poppe et al. (2008) A. Poppe, M. Peev, O. Maurhart, and (on behalf of the Integrated European Project SECOQC), Int. J. Quant. Inf. 6, 209 (2008), eprint arXiv:quant-ph/0701168.
* Cirac et al. (1999) J. I. Cirac, A. K. Ekert, S. F. Huelga, and C. Macchiavello, Phys. Rev. A 59, 4249 (1999), eprint arXiv:quant-ph/9803017v2.
* Nielsen and Chuang (2000) M. Nielsen and I. Chuang, _Quantum Computation and Quantum Information_ (Cambridge University Press, Cambridge, 2000).
* Briegel et al. (1998) H.-J. Briegel, W. Dür, J. I. Cirac, and P. Zoller, Phys. Rev. Lett. 81, 5932 (1998), eprint arXiv:quant-ph/9803056v1.
* Dür et al. (1999) W. Dür, H.-J. Briegel, J. I. Cirac, and P. Zoller, Phys. Rev. A 59, 169 (1999), eprint arXiv:quant-ph/9808065v1.
* Childress et al. (2005) L. Childress, J. M. Taylor, A. S. Sorensen, and M. D. Lukin, Phys. Rev. A 72, 052330 (pages 16) (2005), eprint arXiv:quant-ph/0502112, URL http://link.aps.org/abstract/PRA/v72/e052330.
* Hartmann et al. (2007) L. Hartmann, B. Kraus, H.-J. Briegel, and W. Dur, Phys. Rev. A 75, 032310 (pages 17) (2007), eprint arXiv:quant-ph/0610113, URL http://link.aps.org/abstract/PRA/v75/e032310.
* Żukowski et al. (1993) M. Żukowski, A. Zeilinger, M. A. Horne, and A. K. Ekert, Phys. Rev. Lett. 71, 4287 (1993).
* Acín et al. (2007) A. Acín, J. I. Cirac, and M. Lewenstein, Nature Physics 3, 256 (2007), eprint arXiv:quant-ph/0612167, URL http://dx.doi.org/10.1038/nphys549.
* Perseguers et al. (2008) S. Perseguers, J. I. Cirac, A. Acín, M. Lewenstein, and J. Wehr, Phys. Rev. A 77, 022308 (pages 14) (2008), eprint arXiv:0708.1025, URL http://link.aps.org/abstract/PRA/v77/e022308.
* Kieling and Eisert (2008) K. Kieling and J. Eisert, in _Quantum Percolation and Breakdown_ (Springer, Heidelberg, 2008), in press, eprint arXiv:0712.1836.
* Vidal (1999) G. Vidal, Phys. Rev. Lett. 83, 1046 (1999), eprint arXiv:quant-ph/9902033.
* Nielsen and Vidal (2001) M. A. Nielsen and G. Vidal, Quantum Inf. Comput. 1, 76 (2001).
* Grimmett (1999) G. Grimmett, _Percolation_ (Springer-Verlag, Berlin, 1999), 2nd ed.
* Stauffer and Aharony (1992) D. Stauffer and A. Aharony, _Introduction to Percolation Theory_ (Taylor & Francis, London, 1992), 2nd ed.
* Ziff and Suding (1997) R. M. Ziff and P. N. Suding, J. Phys. A 30, 5351 (1997), eprint arXiv:cond-mat/9707110.
* Hoshen and Kopelman (1976) J. Hoshen and R. Kopelman, Phys. Rev. B 14, 3438 (1976).
* Nakanishi and Stanley (1980) H. Nakanishi and H. E. Stanley, Phys. Rev. B 22, 2466 (1980).
* Nishimura (2000) T. Nishimura, ACM Trans. on Modeling and Computer Simulation 10, 348 (2000).
* Domb (1959) C. Domb, Nature 184, 589 (1959).
* Domb and Sykes (1961) C. Domb and M. F. Sykes, Phys. Rev. 122, 77 (1961).
* Blease et al. (1978) J. Blease, J. W. Essam, and C. M. Place, J. Phys. C. 11, 4009 (1978).
| arxiv-papers | 2008-07-08T19:51:46 | 2024-09-04T02:48:56.577411 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "G. John Lapeyre Jr, Jan Wehr, Maciej Lewenstein",
"submitter": "Gerald Lapeyre Jr.",
"url": "https://arxiv.org/abs/0807.1118"
} |
0807.1121 | $Id:espcrc2.tex,v1.22004/02/2411:22:11speppingExp$ Anomalous quartic gauge
couplings in $\gamma\gamma$ interactions at the LHC T. Pierzchała and K.
Piotrzkowski
# Sensitivity to anomalous quartic gauge couplings in photon-photon
interactions at the LHC
Contribution to the CERN workshop on ”High energy photon collisions at the
LHC”, 22-25th April 2008
T. Pierzchała and K. Piotrzkowski[UCL] Email: Tomasz.Pierzchala@uclouvain.be
Université catholique de Louvain, Center for Particle Physics and
Phenomenology (CP3)
Chemin du Cyclotron 2, 1348 Louvain-la-Neuve, Belgium
###### Abstract
The exclusive two-photon production at the LHC of pairs of W and Z bosons
provides a novel and unique test-ground for the electroweak gauge boson
sector. In particular it offers, thanks to high $\gamma\gamma$ center-of-mass
energies, large and direct sensitivity to the anomalous quartic gauge
couplings otherwise very difficult to investigate at the LHC. An initial
analysis has been performed assuming leptonic decays and generic acceptance
cuts. Simulation of a simple counting experiment has shown for the integrated
luminosity of 10 fb-1 at least four thousand times larger sensitivity to the
genuine quartic couplings, $a_{0}^{W}$, $a_{0}^{Z}$, $a_{C}^{W}$ and
$a_{C}^{Z}$, than those obtained at LEP. The impact of the unitarity
constraints on the estimated limits has been studied using the dipole form-
factors. Finally, differential distributions of the decay leptons have been
provided to illustrate the potential for further improvements of the
sensitivities.
## 1 LHC as a photon collider
The $\gamma\gamma$ exclusive production of pairs of charged particles offers
interesting potential for signals of new physics at the LHC. In a recent paper
[1], the initial comprehensive studies of high energy photon interactions at
the LHC were reported. In the present contribution, the selected results on
the gauge boson pairs discussed in Ref. [1] are introduced and supplemented by
new results.
The exclusive two-photon production, $pp\rightarrow pXp$, provides clean
experimental conditions, thanks to absence of the proton remnants. Well
defined final states can be then selected, and precisely reconstructed.
Moreover, detection of the two final state protons, scattered at almost zero-
degree angle, in the dedicated very forward detectors (VFDs), provides another
striking signature, effective also at high luminosity and with large event
pile-up [2, 3]. In addition, the photon energies can be then measured and used
for the event kinematics reconstruction. Finally, virtualities of the
exchanged photons are on average very small, and are limited from above due to
the proton electromagnetic form-factors, allowing for treating the LHC protons
as sources of quasi-real photons.
The cross sections of two-photon pair production are in general determined by
the mass, spin and charge of the produced particles, so the rate of produced
particles at the LHC can be well predicted using the Equivalent Photon
Approximation for the equivalent photon fluxes [4]. For proton-proton
collisions at $\sqrt{s}=14$ TeV the EPA predicts the photon-photon luminosity
spectrum as shown in Fig. 1, where $dL_{\gamma\gamma}/dW_{\gamma\gamma}$ is
defined by the relation between the proton-proton and $\gamma\gamma$ cross-
sections assuming a minimal center-of-mass energy $W_{0}$:
$\sigma_{pp}=\int_{W_{0}}^{\sqrt{s}}\sigma_{\gamma\gamma}~{}\frac{dL_{\gamma\gamma}}{dW_{\gamma\gamma}}~{}dW_{\gamma\gamma}~{}.$
Figure 1: Elastic luminosity spectrum of photon-photon collisions at the LHC
assuming the maximal photon virtuality $Q_{max}^{2}=2~{}\textrm{GeV}^{2}$
(solid line). The luminosity spectrum assuming the photon tagging range
$20~{}\textrm{GeV}~{}<~{}E_{\gamma}~{}<~{}900~{}\textrm{GeV}$ is also shown
(dashed line).
A set of VFDs at 220 m or 420 m from the LHC interaction points will be
capable of tagging photon interactions within the wide photon energy range of
$20~{}\textrm{GeV}<E_{\gamma}<900~{}\textrm{GeV}$ [5, 6]. In Fig. 1 also the
$\gamma\gamma$ luminosity spectrum is shown assuming double tagging (i.e.
requesting both forward protons to be detected). One should note that apart
from such $elastic$ two-photon processes where both protons have survived the
interaction, the $inelastic$ production can also be considered, when at least
one of the two protons dissociates into a low mass state. The corresponding
two-photon luminosity increases then by about a factor of three [2]. By
integrating the luminosity spectrum above some minimal center-of-mass energy
$W_{0}$, one can introduce the relative photon-photon luminosity
$L_{\gamma\gamma}$, shown in Fig. 2. Effectively, $L_{\gamma\gamma}$ gives a
fraction of the proton-proton luminosity which is available for $\gamma\gamma$
collisions, and is especially useful if a given photon-photon cross-section is
approximately constant as a function of $W_{\gamma\gamma}$. For example, the
relative photon-photon luminosity at the LHC is equal to 1% for $W_{0}=23$ GeV
(i.e. for $W_{\gamma\gamma}>23$ GeV), and 0.1% for $W_{0}=225$ GeV. Given the
very large LHC luminosity, this leads to significant event rates of high-
energy processes with relatively small photon-photon cross sections. In the
following only the proton cross-sections are quoted.
Figure 2: Relative elastic $\gamma\gamma$ luminosity for photon collisions at
the center-of-mass energy above $W_{0}$, obtained using the EPA for proton
collisions at $\sqrt{s}=14~{}TeV$, and assuming the maximal photon virtuality
$Q_{max}^{2}=2~{}\textrm{GeV}^{2}$.
One needs however to consider corrections beyond the EPA due to a possibility
of strong interactions between protons, or the so-called rescattering effects.
The resulting suppression of the cross sections weakly depends on the
invariant mass of the exclusively produced state $X$, and for the processes
discussed in the following, as $pp\rightarrow pWWp$, it is estimated to be
about 15% [7]. This correction is ignored in the present analysis. In
addition, one should stress that potentially dangerous background due to the
exclusive diffractive production is heavily suppressed for non strongly-
interacting particles and can be safely neglected. For example, the gluon
mediated exclusive production of W boson pairs, is about 100 times smaller
than the two-photon production at the LHC [8].
Anomalous quartic couplings of the electroweak gauge bosons could directly
reveal the exchange of new heavy bosons [9, 10], and the two-photon exclusive
production of WW and ZZ pairs is particularly well suited for studies of the
quartic couplings $\gamma\gamma WW$ and $\gamma\gamma ZZ$. In the Standard
Model (SM), the cross section of the exclusive two-photon production
$pp\rightarrow pWWp$ is large about 108 fb-1 [1]. This means that unique, high
statistics tests of the gauge boson sector in the SM can be performed at the
LHC. In contrast, the cross section of the exclusive two-photon production
$pp\rightarrow pZZp$ is very small in the SM, since this proccess is not
allowed at the tree level. It means that an observation of even a few events
of this type could signal the new physics.
## 2 Anomalous quartic gauge couplings
The anomalous quartic gauge couplings (AQGCs) can be introduced in various
ways, usually by building an effective lagrangian which models a low energy
behavior of a wide class of possible extensions of the SM. For example, one
can introduce new terms in such a lagrangian, which are allowed by the local
$\mathrm{U}(1)_{Y}\times\mathrm{SU}(2)_{L}$ gauge invariance, and then
consider two scenarios, with [11] and without [12] the Higgs boson. In the
present analysis, the phenomenological lagrangians are used, which allow for
genuine anomalous quartic vector boson couplings, without need for associated
trilinear gauge couplings. First, the simplest lagrangian term of power six in
energy for two photon interaction with weak bosons has to conserve local
$\mathrm{U}(1)_{em}$ and custodial SU(2)c. Then, by imposing conservation of
discrete C and P symmetries one finally obtains two new terms [9, 13]:
$L^{0}_{6}=-\frac{e^{2}}{16}\frac{a_{0}}{\Lambda^{2}}F_{\mu\nu}F^{\mu\nu}\vec{W}^{\alpha}\cdot\vec{W}_{\alpha}$
$L^{C}_{6}=-\frac{e^{2}}{16}\frac{a_{C}}{\Lambda^{2}}F_{\mu\alpha}F^{\mu\beta}\vec{W}^{\alpha}\cdot\vec{W}_{\beta},$
(1)
where $e$ is the electron charge and $\Lambda$ is the energy scale of the new
physics. When the global SU(2)c symmetry is not imposed then one can
distinguish the neutral and charged couplings of the W and Z bosons. Since
this more general model has been assumed for studies at LEP [14], to allow for
a direct comparison of sensitivities, this is also assumed in the following:
$\displaystyle\mathcal{L}_{6}^{0}$ $\displaystyle~{}~{}=~{}~{}$
$\displaystyle-\frac{e^{2}}{8}{\bf\frac{a_{0}^{\mathrm{W}}}{\Lambda^{2}}}F_{\mu\nu}F^{\mu\nu}{W}^{+\alpha}{W}^{-}_{\alpha}$
$\displaystyle-\frac{e^{2}}{16\cos^{2}\theta_{W}}{\bf\frac{a_{0}^{\mathrm{Z}}}{\Lambda^{2}}}F_{\mu\nu}F^{\mu\nu}{Z}^{\alpha}{Z}_{\alpha},$
$\displaystyle\mathcal{L}_{6}^{\mathrm{c}}$ $\displaystyle~{}~{}=~{}~{}$
$\displaystyle-\frac{e^{2}}{16}{\bf\frac{a_{\mathrm{c}}^{\mathrm{W}}}{\Lambda^{2}}}F_{\mu\alpha}F^{\mu\beta}({W}^{+\alpha}{W}^{-}_{\beta}+W^{-\alpha}W^{+}_{\beta})$
(2)
$\displaystyle-\frac{e^{2}}{16\cos^{2}\theta_{W}}{\bf\frac{a_{\mathrm{c}}^{\mathrm{Z}}}{\Lambda^{2}}}F_{\mu\alpha}F^{\mu\beta}{Z}^{\alpha}{Z}_{\beta}.$
Using this formalism, one obtains a general relation for the $WW$ and $ZZ$
cross sections (which is also valid after applying the acceptance cuts) as a
function of the anomalous parameters:
$\sigma=\sigma_{SM}+\sigma_{0}a_{0}+\sigma_{00}a^{2}_{0}+\sigma_{c}a_{c}+\sigma_{cc}a^{2}_{c}+\sigma_{0c}a_{0}a_{c}~{}~{}~{}\\\
$ (3)
which corresponds, for a fixed cross section $\sigma$, to an ellipse on the
$a_{0}$, $a_{c}$ plane.
## 3 Estimation of sensitivity
Simulation of the exclusive two-photon pair production at the LHC was
performed using the modified MadGraph/MadEvent [16] and CalcHep [17] packages.
The generated events were then passed to the modified Pythia generator [18] to
allow for decays and hadronisation. The sensitivity to the anomalous quartic
vector boson couplings at the LHC has been investigated using the signature of
two opposite charge leptons ($e$ or $\mu$) within the generic lepton
acceptance window – $|\eta|<2.5$ and $p_{T}>10$ GeV. In the $WW$ case it
corresponds to the subprocesses $\gamma\gamma\rightarrow W^{+}W^{-}\rightarrow
l^{+}l^{-}\nu\bar{\nu}$ while in the $ZZ$ case to the subprocesses
$\gamma\gamma\rightarrow ZZ\rightarrow l^{+}l^{-}jj$. It is assumed that both
processes are background free. Under this condition, the upper limits of
number of events $\lambda^{up}$ at the 95% confidence level (CL) were
calculated assuming the number of observed events equal to the $SM$ prediction
$\mathrm{N_{obs}=\sigma^{\textsc{sm}}_{acc}~{}L}$ for a given integrated
luminosity L.
$\mathrm{\sum^{N_{obs}}_{k=0}P_{Poisson}(\lambda^{up}=\sigma^{up}~{}L;k)=1-CL}$
(4)
This is a simplified approach and for more precise analysis one should use
formula from [15], however the difference in the obtained limits is small,
less than 20%.
The expected 95$\%$ CL limits $\mathrm{\lambda^{up}}$ have been then used to
calculate the upper limits on the observed cross section
$\mathrm{\sigma^{up}}$ at the 95% CL for the integrated luminosity L = 1 fb-1
and L = 10 fb-1, shown in the Table 1.
Table 1: Expected 95$\%$ CL upper limits for the cross sections after acceptance cuts for chosen subprocesses $\gamma\gamma\rightarrow W^{+}W^{-}\rightarrow l^{+}l^{-}\nu\bar{\nu}$ and $\gamma\gamma\rightarrow ZZ\rightarrow l^{+}l^{-}jj$. The assumed numbers of the observed $WW$ events correspond to the $SM$ cross section calculated using MG/ME and the acceptance cuts. | $\gamma\gamma\rightarrow W^{+}W^{-}$ | $\gamma\gamma\rightarrow ZZ$
---|---|---
$\mathrm{\sigma^{up}~{}[fb]}$ | | $\mathrm{N_{obs}=0}$
| $\mathrm{\sigma^{\textsc{sm}}_{acc}=4.081~{}fb}$ | $\mathrm{\lambda^{up}=2.996}$
$\mathrm{L=1~{}fb^{-1}}$ | $\mathrm{9.2}$ | $\mathrm{3.0}$
$\mathrm{L=10~{}fb^{-1}}$ | $\mathrm{5.3}$ | $\mathrm{0.30}$
The calculated cross section upper limits can be directly converted to the
limits of the anomalous quartic couplings as presented at the Fig. 3 where the
95% CL contours are shown. In Fig. 4 one parameter limits (with the other
anomalous coupling set to zero) are shown, and the obtained limits are quoted
in the Table 2.
Figure 3: Profiles of the 95$\%$ CL upper limits of the cross section after the acceptance cuts for $pp(\gamma\gamma\rightarrow W^{+}W^{-}\rightarrow$ $l^{+}l^{-}\nu\bar{\nu})pp$ and $pp(\gamma\gamma\rightarrow ZZ\rightarrow$ $l^{+}l^{-}jj)pp$ as a function of relevant anomalous couplings $\mathrm{a_{0}^{W}/\Lambda^{2}}$, $\mathrm{a_{\mathrm{c}}^{W}/\Lambda^{2}}$, $\mathrm{a_{0}^{Z}/\Lambda^{2}}$ and $\mathrm{a_{\mathrm{c}}^{Z}/\Lambda^{2}}$. The contours are shown assuming two values of the integrated $pp$ luminosity. Figure 4: Cross sections of $pp(\gamma\gamma\rightarrow W^{+}W^{-}\rightarrow$ $l^{+}l^{-}\nu\bar{\nu})pp$ and $pp(\gamma\gamma\rightarrow ZZ\rightarrow$ $l^{+}l^{-}jj)pp$ after the acceptance cuts are shown as a function of the genuine anomalous quartic vector boson couplings $\mathrm{a_{0}^{W}/\Lambda^{2}}$ and $\mathrm{a_{0}^{Z}/\Lambda^{2}}$ (for $\mathrm{a_{C}^{W}/\Lambda^{2}}=\mathrm{a_{C}^{Z}/\Lambda^{2}=0}$), together with the upper cross section limits at CL=95$\%$ (horizontal lines). Table 2: Expected one-parameter limits for anomalous quartic vector boson couplings at 95$\%$ CL for two values of the integrated luminosity. Coupling | Limits $\mathrm{[10^{-6}~{}\textrm{GeV}^{-2}~{}]}$
---|---
| $\mathrm{L=1~{}fb^{-1}}$ | $\mathrm{L=10~{}fb^{-1}}$
$\mathrm{|a^{Z}_{0}/\Lambda^{2}|}$ | $0.49$ | $0.16$
$\mathrm{|a^{Z}_{C}/\Lambda^{2}|}$ | $1.84$ | $0.58$
$\mathrm{|a^{W}_{0}/\Lambda^{2}|}$ | $0.54$ | $0.27$
$\mathrm{|a^{W}_{C}/\Lambda^{2}|}$ | $2.02$ | $0.99$
The obtained limits are about 40 000 times better than the best limits
established at LEP2 [14] clearly showing large and unique potential of such
studies at the LHC.
## 4 Impact of unitarity bound condition
The lagrangian terms in Eq.(2), do not preserve the SU(2)L local symmetry. In
consequence, for the center-of-mass energies $W_{\gamma\gamma}\gg 2M_{W}$ the
scattering amplitude will grow and eventually will violate the unitarity
condition. It is therefore necessary to investigate the impact of the
unitarity constraint on the derived limits. For simplicity, the unitarity
violation is checked only for the process $\gamma\gamma\rightarrow
W^{+}W^{-}$. The $\gamma\gamma\rightarrow\gamma\gamma$ and
$\gamma\gamma\rightarrow ZZ$ (i.e. the anomalous coupling $\gamma\gamma ZZ$ is
set to zero) processes are neglected, as well as the two-photon exclusive
production of fermion pairs as it drops fast with $W_{\gamma\gamma}$ [1]. For
spin one particles one obtains the partial wave amplitudes $a_{J}$ using the
Legendre polynomial $P_{J}$:
$a_{J}(\sqrt{s})=\frac{1}{32\pi}\int_{-1}^{1}du~{}{\cal
M}(\sqrt{s},u,a^{W}_{0},a^{W}_{C})P_{J}(u),~{}$ (5)
where u is the cosine of the W boson polar angle in the $\gamma\gamma$ center-
of-mass system, and ${\cal M}$ is the amplitude of the process
$\gamma\gamma\rightarrow W^{+}W^{-}$. Then, the unitarity bound condition has
a form:
$\beta\sum_{pol}|a_{J}(\sqrt{s},a^{W}_{0},a^{W}_{C})|^{2}\leq(1/2)^{2}$ (6)
where $\beta=\sqrt{1-4m_{W}^{2}/s}$ is the Lorentz velocity of a W boson in
the cms frame and pol stands for polarization states in
$\gamma\gamma\rightarrow W^{+}W^{-}$.
Figure 5: Unitarity limit for the W neutral anomalous coupling
$a_{0}^{W}/\Lambda^{2}$ calculated according to Eq. 6 for J=0 (full line), and
assuming that the left hand side of Eq. 6 is smaller than $(0.4)^{2}$ (dashed
line). The arrow indicates the obtained limit at 95% CL from Tab. 2 for $10$
fb-1.
Unitarity bounds of Eq. 6 for J=0 as a function of $W_{\gamma\gamma}$ are
shown for both neutral, in Fig. 5, and charged, in Fig. 6, AQGCs together with
the obtained limits of anomalous couplings for the integrated luminosity
$L=10$ fb-1. It shows that for these limits the unitarity will be violated
when $W_{\gamma\gamma}$ is above 2 TeV, also if one allows for a contribution
of some other channels, by requesting the left hand side of Eq. 6 to be
smaller than $(0.4)^{2}$ instead of $(0.5)^{2}$. Other $WW$ states with
different total angular momentum do not violate the unitarity earlier, and
$a_{1}$ is close to 0 when $a_{2}$ is much smaller than $a_{0}$.
Figure 6: Analogous plot to Fig. 5 but for the W charged anomalous coupling
$a_{C}^{W}/\Lambda^{2}$. Also in this case, unitarity will be violated for
$W_{\gamma\gamma}$ above 2 TeV.
The unitarity violation occurs due to the increasing contribution of
longitudinally polarized W bosons, which in the SM is fully compensated thanks
to the local SU(2)L symmetry and the trilinear gauge couplings. As a
consequence, one has to verify if the limits quoted in Tab. 2 are not driven
by $\gamma\gamma$ interactions at $W_{\gamma\gamma}$ above 2 TeV. As one can
see in Fig. 7, this is actually the case, since the anomalous cross-section is
large for the energies well above 2 TeV. This can be avoided in two ways. The
double tagging using very forward detectors would limit the maximal
$\gamma\gamma$ center-of-mass energy to 1.8 TeV [6]. One should note that the
tagging efficiency in this energy range is high, about 80% [19]. In other
case, one can introduce dipole form-factors for each anomalous coupling, as in
Eq. 7. The form-factors suppress the anomalous amplitudes when the energies
are close to the new physics energy scale $\Lambda$:
${\bf a}\rightarrow\frac{\bf
a}{\left(1+W_{\gamma\gamma}^{2}/{\Lambda^{2}}\right)^{X}},$ (7)
where the power $X$ is chosen so to preserve the unitarity – in the present
analysis $X=2$.
Figure 7: Differential cross section $d\sigma/dW_{\gamma\gamma}$ for
$\gamma\gamma\rightarrow W^{+}W^{-}\rightarrow l^{+}l^{-}$ after acceptance
cuts, for the SM (solid line) and for the AQGC case (dashed line), assuming
$a_{0}^{W}/\Lambda^{2}=0.54\cdot 10^{-6}$ GeV-2 and $a_{C}^{W}/\Lambda^{2}=0$.
## 5 Expected limits for anomalous quartic gauge couplings
The sensitivity analysis has been repeated for the process
$\gamma\gamma\rightarrow W^{+}W^{-}$ using the above form-factors. In Fig. 8
and in Fig. 9 the new sensitivity limits for the integrated luminosity $L=10$
fb-1 are shown.
Figure 8: The 95% CL limits of the W neutral anomalous quartic coupling
(dashed line) calculated for the integrated luminosity $L=10$ fb-1 with the
dipole form-factor. The unitarity curve (solid line) is calculated for
$W_{\gamma\gamma}=\Lambda$, and due to inclusion of the from-factor is four
times higher than in Fig. 5. The solid arrows indicate the strongest limit
without violating the unitarity, and the corresponding energy scale $\Lambda$
of the new physics. Dashed arrow recalls the limit obtained without the form-
factors.
The limits were calculated for 5 different $\Lambda$ values and linearly
interpolated. Moreover, the unitarity limit (solid line in Figs. 8,9) is
calculated at $W_{\gamma\gamma}=\Lambda$, after correcting for the form-factor
suppression. This choice can be explained in the following way. If one obtains
a limit of an anomalous coupling for $W_{\gamma\gamma}=\Lambda$ than for the
energies below $\Lambda$ the unitarity is automatically preserved. For higher
energies, bigger than $\Lambda$ the form-factor suppression should be strong
enough to keep the unitarity unbroken.
Figure 9: The sensitivity plot for the W charged anomalous quartic coupling.
The plot is analogous to Fig. 8.
By repeating the sensitivity analysis from Sec. 3, the results shown in Fig. 8
and in Fig. 9 are obtained, and the following 95% CL limits for the anomalous
quartic couplings can be derived for $\gamma\gamma\rightarrow W^{+}W^{-}$
fully leptonic events, for the integrated luminosity $L=10$ fb-1:
$\displaystyle a_{0}^{W}/\Lambda^{2}$ $\displaystyle<$ $\displaystyle 2.5\cdot
10^{-6}~{}\mathrm{GeV}^{-2}$ (8) $\displaystyle a_{C}^{W}/\Lambda^{2}$
$\displaystyle<$ $\displaystyle 9\cdot 10^{-6}~{}\mathrm{GeV}^{-2},$ (9)
which are about 4 000 times stronger than achieved at LEP [14], and are 10
times weaker than results from Tab. 2, obtained ignoring the unitarity
constraints.
## 6 Summary and perspectives
The initial analysis of sensitivity to genuine AQGCs in photon-photon
collisions at the LHC has been presented. A simple method of event counting
has provided very promising results, showing improvements of the LEP AQGC
limits by a factor of about 4 000, for the integrated luminosity 10 fb-1. The
effects of the form-factors, introduced to preserve unitarity constraints, are
taken into account. In addition, the analysis demonstrates the sensitivity to
the new physics energy scale up to 2 TeV.
There are still several possible ways to improve the sensitivity. First, one
can include also semi-leptonic channels in $\gamma\gamma\rightarrow WW$, which
will increase statistics about 6 times.
Figure 10: Differential cross sections $d\sigma/d\eta$ for
$\gamma\gamma\rightarrow W^{+}W^{-}\rightarrow l^{+}l^{-}$ after acceptance
cuts for the SM (solid line) and for the anomalous quartic coupling (dashed
line), $a_{0}^{W}/\Lambda^{2}=2.5\cdot 10^{-6}$ GeV-2 and $\Lambda=2$ TeV.
In addition, one can increase sensitivity by studing the differential
distributions, like the lepton pseudo-rapidity (see Fig. 10), or the lepton
acoplanarity $\delta\phi=\pi-Min(2\pi-\Delta\phi,\Delta\phi)$ (see Fig. 11)
distributions.
Figure 11: Differential cross section $d\sigma/d(\delta\phi)$ for
$\gamma\gamma\rightarrow W^{+}W^{-}\rightarrow l^{+}l^{-}$ after acceptance
cuts, for the SM (solid line) and for the anomalous quartic coupling (dashed
line), $a_{0}^{W}/\Lambda^{2}=2.5\cdot 10^{-6}$ GeV-2 with $\Lambda=2$ TeV.
Finally, assuming the custodial SU(2)c symmetry, one can convert the limits of
$a_{0}^{W}$ and $a_{0}^{Z}$ into the single $a_{0}$ limit, and similarly the
$a_{C}^{W}$ and $a_{C}^{Z}$ limits into the single $a_{C}$ limit (see. Eq. 1).
This will be the subject of a future publication.
## References
* [1] J. de Favereau _et al._ , High energy photon interactions at the LHC, CP3–08–04, July 2008, to be submitted to EPJC.
* [2] K. Piotrzkowski, Phys. Rev. D 63 (2001) 071502.
* [3] M. Albrow _et al._ , The FP420 R$\&$D Project: Higgs and New Physics with forward protons at the LHC FP420, arXiv:0806.0302 [hep-ex].
* [4] V. M. Budnev _et al._ , Phys. Rept. 15 (1974) 181.
* [5] J. de Favereau de Jeneret, X. Rouby and K. Piotrzkowski, JINST 2: P09005 (2007), arXiv:0707.1198v1 [physics.acc-ph] (CP3–07–13).
* [6] X. Rouby, Tagging photon interactions at the LHC, these proceedings, arXiv:0806.4941 [hep-ex].
* [7] V. A. Khoze _et al._ , Eur. Phys. J. C 23 (2002) 311; private communication from V. Khoze.
* [8] B. E. Cox _et al._ , Eur. Phys. J. C 45 (2006) 401.
* [9] G. Belanger and F. Boudjema, Phys. Lett. B 288 (1992) 201.
* [10] C. Arzt, M. B. Einhorn and J. Wudka, Nucl. Phys. B 433 (1995) 41.
* [11] M. Maniatis, A. von Manteuffel and O. Nachtmann, Anomalous couplings in $\gamma\gamma\rightarrow W^{+}W^{-}$ at the LHC and ILC, these proceedings.
* [12] O. Eboli, LHC potential to study the quartic electroweak gauge boson couplings, these proceedings.
* [13] O. J. P. Eboli, M. C. Gonzalez-Garcia, S. M. Lietti and S. F. Novaes, Phys. Rev. D 63 (2001) 075008 [arXiv:hep-ph/0009262].
* [14] G. Abbiendi et al. [OPAL Collaboration], Phys. Rev. D 70 (2004) 032005 [arXiv:hep-ex/0402021].
* [15] G. Zech, Nucl. Instrum. Meth. A 277 (1989) 608.
* [16] J. Alwall _et al._ , JHEP 0709:028 (2007).
* [17] A. Pukhov, Nucl. Inst. Meth A 502 (2003) 596.
* [18] T. Sjöstrand _et al._ , Comput. Phys. Commun. 135 (2001) 238.
* [19] N. Schul and K. Piotrzkowski, Detection of two-photon exclusive production of supersymmetric pairs at the LHC, these proceedings, arXiv:0806.1097 [hep-ph].
| arxiv-papers | 2008-07-07T20:36:10 | 2024-09-04T02:48:56.583727 | {
"license": "Public Domain",
"authors": "Tomasz Pierzchala, Krzysztof Piotrzkowski",
"submitter": "Tomasz Pierzcha{\\l}a",
"url": "https://arxiv.org/abs/0807.1121"
} |
0807.1134 | # Variations on Debris Disks: Icy Planet Formation at 30–150 AU for 1–3 M⊙
Main Sequence Stars
Scott J. Kenyon Smithsonian Astrophysical Observatory, 60 Garden Street,
Cambridge, MA 02138 e-mail: skenyon@cfa.harvard.edu Benjamin C. Bromley
Department of Physics, University of Utah, 201 JFB, Salt Lake City, UT 84112
e-mail: bromley@physics.utah.edu
###### Abstract
We describe calculations for the formation of icy planets and debris disks at
30–150 AU around 1–3 M⊙ stars. Debris disk formation coincides with the
formation of planetary systems. As protoplanets grow, they stir leftover
planetesimals to large velocities. A cascade of collisions then grinds the
leftovers to dust, forming an observable debris disk. Stellar lifetimes and
the collisional cascade limit the growth of protoplanets. The maximum radius
of icy planets, $r_{max}\approx$ 1750 km, is remarkably independent of initial
disk mass, stellar mass, and stellar age. These objects contain $\lesssim$
3%–4% of the initial mass in solid material. Collisional cascades produce
debris disks with maximum luminosity $\sim 2\times 10^{-3}$ times the stellar
luminosity. The peak 24 $\mu$m excess varies from $\sim$ 1% times the stellar
photospheric flux for 1 M⊙ stars to $\sim$ 50 times the stellar photospheric
flux for 3 M⊙ stars. The peak 70–850 $\mu$m excesses are $\sim$ 30–100 times
the stellar photospheric flux. For all stars, the 24–160 $\mu$m excesses rise
at stellar ages of 5–20 Myr, peak at 10–50 Myr, and then decline. The decline
is roughly a power law, $f\propto t^{-n}$ with $n\approx$ 0.6–1.0. This
predicted evolution agrees with published observations of A-type and solar-
type stars. The observed far-IR color evolution of A-type stars also matches
model predictions.
planetary systems – solar system: formation – stars: formation – circumstellar
matter – infrared: stars
## 1 INTRODUCTION
During the past 25 years, observations from IRAS, ISO, and Spitzer have
revealed substantial mid-infrared (mid-IR) excesses associated with hundreds
of normal main sequence stars (e.g., Backman & Paresce, 1993; Habing et al.,
2001; Rieke et al., 2005; Bryden et al., 2006; Moór et al., 2006; Rhee et al.,
2007a). Current samples include stars with spectral types A–M and ages $\sim$
5 Myr to $\sim$ 10 Gyr (e.g., Chen et al., 2005; Kim et al., 2005; Rieke et
al., 2005; Beichman et al., 2006; Su et al., 2006; Hillenbrand et al., 2008).
Although binary stars and single stars in dense clusters and in the field are
roughly equally likely to have IR excesses (Stauffer et al., 2005; Su et al.,
2006; Bryden et al., 2006; Gorlova et al., 2006; Currie et al., 2007a; Siegler
et al., 2007; Trilling et al., 2007), the frequency of excess emission
declines from $\sim$ 30%–40% for A-type stars (Su et al., 2006) to $\sim$
10%–20% for solar-type stars (Greaves & Wyatt, 2003; Trilling et al., 2008;
Meyer et al., 2008). Thus, this phenomenon is common among main sequence stars
and may depend on stellar mass.
High quality images demonstrate that dust orbiting the central star produces
the excesses (Smith & Terrile, 1984; Brandeker et al., 2004; Stapelfeldt et
al., 2004; Kalas, 2005; Meyer et al., 2007). In $\beta$ Pic and AU Mic, the
dust is in a geometrically thin, edge-on disk with an outer radius of $a\sim$
200–1000 AU (Smith & Terrile, 1984; Telesco et al., 1988; Golimowski et al.,
1993; Kalas et al., 2004; Liu, 2004; Augereau & Beust, 2006). In these disks,
the small scale height of the dust, $H/a\sim$ 0.1, is consistent with material
in roughly circular orbits (e.g., Backman & Paresce, 1993; Kalas et al., 2004,
and references therein). Although broad tori of dust are visible in many other
systems (e.g. Greaves et al., 1998; Augereau et al., 1999; Holland et al.,
2003; Kalas et al., 2006; Su et al., 2006; Fitzgerald et al., 2007), narrow
rings of dust produce the emission in $\alpha$ PsA and HR 4796A (Jayawardhana
et al., 1998; Schneider et al., 1999; Greaves et al., 2000; Telesco et al.,
2000; Kalas et al., 2005). For systems with face-on rings and tori, the total
emission constrains the scale height, $H/a\sim$ 0.1. Thus, the dust in these
systems is as highly flattened as the structures in $\beta$ Pic and AU Mic.
Broadband spectral energy distributions constrain the luminosity, size,
temperature, and total mass of the dust (Backman & Paresce, 1993; Lagrange et
al., 2000; Dent et al., 2000; Wolf & Hillenbrand, 2003). Some stars have
excesses from grains plausibly associated with the terrestrial zone (e.g.,
Beichman et al., 2005; Absil et al., 2006; Currie et al., 2007b; Rhee et al.,
2007b; Meyer et al., 2008; Lisse et al., 2007c). Optically thin emission from
cooler material with temperature $T\sim$ 20–150 K is more typical (e.g. Su et
al., 2006; Trilling et al., 2008; Hillenbrand et al., 2008). For systems with
submm observations, the measured fluxes suggest grains with sizes $\sim$ 1
$\mu$m–1 cm and total mass $\sim$ 0.01 M⊕ (Liu et al., 2004; Najita &
Williams, 2005; Chen et al., 2006; Williams & Andrews, 2006). The grains have
compositions similar to dust in the asteroid belt, comets, or the trans-
Neptunian region of the Solar System (Grün et al., 1995; Brownlee et al.,
1997; Lisse et al., 2007a, b). Because the dust mass in these systems lies
between the initial mass of solids in protostellar disks ($\sim$ 100–1000 M⊕;
Natta et al., 2000; Andrews & Williams, 2005) and the dust mass in the Solar
System ($\lesssim 10^{-4}$ M⊕; Hahn et al., 2002; Landgraf et al., 2002;
Nesvorný et al., 2006), the dusty structures in these systems are often called
‘debris disks’ (Backman & Paresce, 1993; Lagrange et al., 2000).
In addition to the dust properties, several other observations suggest
plausible links between debris disks and the formation of planetary systems.
Observations of A-type stars suggest a ‘rise and fall’ of debris disk emission
(Currie et al., 2008a), with a clear increase in the typical 24 $\mu$m excess
at 5–10 Myr, a peak at 10–20 Myr, and a decline for $t\gtrsim$ 20–30 Myr. The
rise in debris disk emission roughly coincides with the disappearance of
optically thick emission from protostellar disks (e.g. Haisch, Lada, & Lada,
2001; Sicilia-Aguilar et al., 2006; Hernández et al., 2006; Currie et al.,
2007a, c). The broad plateau occurs at a time when radiometric dating (Yin et
al., 2002) and theory (Chambers, 2001; Kenyon & Bromley, 2006) suggest the
Earth contained $\sim$ 90% of its final mass. The decline of dusty debris
around A-type stars is at roughly the same time as a gradual decrease in the
cratering rate of objects in the Solar System (Swindle, 1993; Melosh, Vickery,
& Tonks, 1993; Wadhwa & Russell, 2000; Koeberl, 2003). These results suggest
that the evolution of dust in debris disks parallels the evolution of larger
solid objects in the Solar System.
Simple physical arguments also link debris disks with the formation of
planetary systems. Because radiation removes 1–100 $\mu$m grains on timescales
shorter than the stellar age, some process replenishes the dust. To maintain
the observed dust masses for long timescales, normal stars must have a large
reservoir, $\sim$ 10–100 M⊕, of unseen objects that continuously collide at
large velocities and fragment into smaller objects. Remnant material from
planet formation satisfies both needs. The growth of 1000 km or larger planets
in a disk of small grains naturally leaves behind an ensemble of ‘leftover’
1–10 km ‘planetesimals’ on eccentric orbits (Kenyon & Bromley, 2004a, 2004b).
For a mass of $\sim$ 10–100 M⊕ in leftovers, high velocity collisions produce
enough dust for most debris disks (e.g., Backman & Paresce, 1993; Habing et
al., 2001; Kenyon & Bromley, 2004b). If this interpretation is correct, debris
disks provide conclusive evidence for the formation of Pluto-mass or larger
planets around many, if not most, main sequence stars.
In addition to these considerations, numerical calculations suggest that an
evolving swarm of 1–10 km planetesimals explains several observed trends in
the properties of debris disks. Starting with an ensemble of $\lesssim$ 1 km-
sized planetesimals, Kenyon & Bromley (2002b, 2004a, b, 2005) show that
collisions and mergers form 500–1000 km-sized objects in 1–50 Myr. These
protoplanets stir up leftover planetesimals along their orbits. Destructive
collisions among the leftovers then produce a collisional cascade – where
collisions gradually grind large objects into smaller ones – along with
copious amounts of dust (see also Williams & Wetherill, 1994; Durda & Dermott,
1997; Quillen et al., 2007). Dominik & Decin (2003), Wyatt et al. (2007a), and
Löhne et al. (2008) show that collisional evolution in a belt of high velocity
planetesimals naturally produces a dust luminosity that declines roughly
inversely with time (see also Kenyon & Bromley, 2002a, 2004b, 2005),
explaining the observed time evolution – $L_{d}\propto t^{-n}$, with
$n\approx$ 0.5–1 – suggested by recent observations of A-type stars (e.g.,
Kalas, 1998; Habing et al., 2001; Decin et al., 2003; Greaves & Wyatt, 2003;
Rieke et al., 2005; Rhee et al., 2007a). To account for the large observed
range of IR excesses among stars of similar ages, Wyatt et al. (2007a) propose
belts with a range of initial masses and semimajor axes, as suggested from
submm observations of protostellar disks (Andrews & Williams, 2005, 2007b).
Here, we continue to explore the evolution of dusty debris arising from planet
formation in a disk of icy planetesimals. Our suite of calculations for disks
at 30–150 AU around 1–3 M⊙ stars yields robust predictions for the maximum
sizes of icy planets as a function of semimajor axis and stellar age. Results
for the long-term evolution of IR excesses account for many fundamental
aspects of the data. These calculations are the first to explain the ‘rise and
fall of debris disks’ around A-type stars (Currie et al., 2008a) and the
apparent peak in the 70–160 $\mu$m excesses of G-type stars with ages of
$\sim$ 100 Myr (Hillenbrand et al., 2008). Comparisons between our models and
current observations suggest that the minimum stable grain size and the slope
of the IR emissivity law are critical parameters.
The models suggest a set of further critical observations. Spatially resolved
images of debris disks around A-type and solar-type stars can improve our
understanding of the minimum stable grain size. Larger samples of debris disks
with high quality submm data from ALMA, Herschel, and SOFIA can place better
constraints on the slope of the emissivity law. Together, these data can test
our predictions for the time evolution of debris disk emission around 1–3 M⊙
stars and provide input for more complete calculations that include the
formation and dynamical evolution of giant planets.
We outline our model in §2. We describe results for the formation of icy
planets in §3 and the evolution of debris disks in §4. After discussing
several applications of our calculations in §5, we conclude with a brief
summary in §6.
## 2 THE MODEL
Kenyon & Bromley (2001, 2002a, 2004a, 2004c) and Bromley & Kenyon (2006)
describe our hybrid multiannulus numerical model for planetesimal growth.
Kenyon & Luu (1998, 1999), Kenyon & Bromley (2001, 2002a), and Bromley &
Kenyon (2006) compare results with analytical and numerical calculations. We
adopt the Safronov (1969) statistical approach to calculate the collisional
evolution of an ensemble of planetesimals in orbit around a star of mass
$M_{\star}$ (see also Spaute et al., 1991; Weidenschilling et al., 1997;
Krivov et al., 2006; Thébault & Augereau, 2007; Löhne et al., 2008). The model
grid contains $N$ concentric annuli with widths $\delta a_{i}$ centered at
semimajor axes $a_{i}$. Calculations begin with a differential mass
distribution $n(m_{ik}$) of objects with horizontal and vertical velocities
$h_{ik}(t)$ and $v_{ik}(t)$ relative to a circular orbit. The horizontal
velocity is related to the orbital eccentricity, $e_{ik}^{2}(t)$ = 1.6
$(h_{ik}(t)/V_{K,i})^{2}$, where $V_{K,i}$ is the circular orbital velocity in
annulus $i$. The orbital inclination depends on the vertical velocity,
$i_{ik}^{2}(t)$ = sin${}^{-1}(2(v_{ik}(t)/V_{K,i})^{2})$.
The mass and velocity distributions evolve in time due to inelastic
collisions, drag forces, and gravitational forces. For inelastic collisions,
we solve the coagulation equations for a particle in mass batch $k$ of annulus
$i$ colliding with another particle in mass batch $l$ of annulus $j$,
$\delta n_{i^{\prime}k^{\prime}}=\delta
t\left[\epsilon_{ijkl}A_{ijkl}n_{ik}n_{jl}~{}-~{}n_{i^{\prime}k^{\prime}}A_{i^{\prime}jk^{\prime}l}n_{jl}\right]~{}+~{}\delta
n_{i^{\prime}k^{\prime},f}~{}-~{}\delta n_{i^{\prime}k^{\prime},gd}$ (1)
$\delta M_{i^{\prime}k^{\prime}}=\delta
t~{}m_{i^{\prime}k^{\prime}}\left[\epsilon_{ijkl}A_{ijkl}n_{ik}n_{jl}~{}-~{}n_{i^{\prime}k^{\prime}}A_{i^{\prime}jk^{\prime}l}n_{jl}\right]~{}+~{}\delta
M_{i^{\prime}k^{\prime},f}-~{}\delta M_{i^{\prime}k^{\prime},gd}$ (2)
where $t$ is time, $M_{i^{\prime}k^{\prime}}$ is the total mass in mass bin
$k^{\prime}$ in annulus $i^{\prime}$, $A_{ijkl}$ is the cross-section,
$\epsilon_{ijkl}=1/2$ for $i=j$ and $k=l$, and $\epsilon_{ijkl}=1$ for $k\neq
l$ and any $i,j$. The terms in these equations represent (i) mergers of
$m_{ik}$ and $m_{jl}$ into a body of mass
$m_{i^{\prime}k^{\prime}}=m_{ik}+m_{jl}-m_{e,ijkl}$, (ii) loss of
$m_{i^{\prime}k^{\prime}}$ through mergers with other objects, (iii) addition
of mass from debris produced by the collisions of other objects (Kenyon & Luu,
1999), and (iv) loss of mass by gas drag (Kenyon & Luu, 1998). In each
equation, the second term includes the possibility that a collision can
produce debris but no merger (rebounds; see Davis et al., 1985; Kenyon & Luu,
1999, and references therein).
The collision cross-section is
$A_{ijkl}=\alpha_{coll}~{}\left(\frac{1}{4~{}H_{ijkl}~{}\langle
a_{ij}\rangle~{}\langle\Delta
a_{ij}\rangle}\right)~{}V_{ijkl}~{}F_{g,ijkl}~{}(r_{ik}+r_{jl})^{2}~{},$ (3)
where $\alpha_{coll}$ is a constant (Wetherill & Stewart, 1993; Kenyon & Luu,
1998), $H_{ijkl}$ =
$\sqrt{2~{}(v_{ik}^{2}+v_{jl}^{2})}/\langle\Omega_{ij}\rangle$ is the mutual
scale height, $\langle a_{ij}\rangle$ and $\langle\Delta a_{ij}\rangle$ are
the average heliocentric distance and width for the two interacting annuli,
$\langle Omega_{ij}\rangle$ is the average angular velocity, $V_{ijkl}$ is the
relative particle velocity, $F_{g,ijkl}$ is the gravitational focusing factor,
and $r_{ik}$ and $r_{jl}$ are the particle radii. We adopt the piecewise
analytic approximation of Spaute et al. (1991) for the gravitational focusing
factor in the dispersion regime and the collisional cross-sections of
Greenberg et al. (1991) in the shear-dominated regime (see also Greenzweig &
Lissauer, 1990, 1992). For more details of this algorithm, see Kenyon & Luu
(1998), Kenyon & Bromley (2002a), Kenyon & Bromley (2004a), and Bromley &
Kenyon (2006).
To choose among possible collision outcomes, we use an energy-scaling
algorithm. If $Q_{d}^{*}$ is the collision energy needed to eject half the
mass of a pair of colliding planetesimals and $Q_{c}$ is the center of mass
collision energy, the mass of the ejecta is
$m_{e,ijkl}=0.5~{}(m_{ik}+m_{jl})\left(\frac{Q_{c}}{Q_{d}^{*}}\right)^{9/8}~{},$
(4)
where $m_{ik}$ and $m_{jl}$ are the masses of the colliding planetesimals.
This approach allows us to derive ejected masses for catastrophic collisions
with $Q_{c}\sim Q_{d}^{*}$ and cratering collisions with $Q_{c}\ll Q_{d}^{*}$
(see also Wetherill & Stewart, 1993; Stern & Colwell, 1997; Kenyon & Luu,
1999). Consistent with N-body simulations of collision outcomes (e.g., Benz &
Asphaug, 1999), we set
$Q_{d}^{*}=Q_{b}r_{ijkl}^{\beta_{b}}+Q_{g}\rho_{g}r_{ijkl}^{\beta_{g}}$ (5)
where $r_{ijkl}$ is the radius of a merged object with mass $m_{ik}$ \+
$m_{jl}$, $\rho_{g}$ is the mass density of a planetesimal,
$Q_{b}r^{\beta_{b}}$ is the bulk component of the binding energy, and
$Q_{g}\rho_{g}r^{\beta_{g}}$ is the gravity component of the binding energy.
Kenyon & Bromley (2005) and Kenyon et al. (2008) describe how collisional
evolution depends on various choices for $Q_{d}^{*}$. For icy objects,
detailed numerical collision simulations yield $Q_{b}\lesssim 10^{7}$ erg
cm${}^{-\beta_{b}}$ g-1, $-0.5\lesssim\beta_{b}\lesssim$ 0, $\rho_{g}\approx$
1–2 g cm-3, $Q_{g}\lesssim$ 1–2 erg cm${}^{3-\beta_{g}}$ g-2, and $\beta_{g}$
$\approx$ 1–2 (e.g., Benz & Asphaug, 1999; Leinhardt et al., 2008).
Calculations for the breakup of comet Shoemaker-Levy 9 suggest a smaller
component of the bulk strength, $Q_{b}r^{\beta_{b}}\sim 10^{3}$ erg g-1 (e.g.,
Asphaug & Benz, 1996), which yields smaller disruption energies for smaller
objects. Because nearly all models for collisional disruption yield similar
results for objects with $r\gtrsim$ 1 km (e.g., Kenyon & Bromley, 2004c;
Kenyon et al., 2008), collisional evolution is relatively independent of these
uncertainties as planetesimals grow into larger objects. Thus, we choose
standard values – $Q_{g}$ = 1.5 erg cm1.75 g-2, $\rho_{g}$ = 1.5 g cm-3, and
$\beta_{g}$ = 1.25 – for the gravity component of $Q_{d}^{*}$. To check how
the evolution of the small planetesimals depends on $Q_{d}^{*}$, we consider a
broad range in the bulk component of the strength, $Q_{b}$ = 1–$10^{5}$ erg
g-1 with $\beta_{b}$ = 0 (Pan & Sari, 2005; Kenyon & Bromley, 2004c, 2005;
Kenyon et al., 2008).
To compute velocity evolution, we include collisional damping from inelastic
collisions, gas drag, and gravitational interactions. Our equations for the
evolution of the velocity dispersion are
$\frac{dh_{ik}^{2}}{dt}=\frac{dh_{in,ik}^{2}}{dt}+\frac{dh_{gd,ik}^{2}}{dt}+\frac{dh_{lr,ik}^{2}}{dt}+\frac{dh_{sr,ik}^{2}}{dt}$
(6)
for the horizontal component and
$\frac{dv_{ik}^{2}}{dt}=\frac{dv_{in,ik}^{2}}{dt}+\frac{dv_{gd,ik}^{2}}{dt}+\frac{dv_{sr,ik}^{2}}{dt}$
(7)
for the vertical component, where the subscripts refer to the contributions
from collisional damping (‘in’), gas drag (‘gd’), and long-range (‘lr’) and
short-range (‘sr’) gravitational interactions.
For collisional damping, we adopt
$\frac{dh_{in,ik}^{2}}{dt}=\sum_{j=0}^{j=N}\sum_{l=0}^{l=l_{max}}\frac{C_{in}}{2}~{}(m_{jl}h_{jl}^{2}-m_{ik}h_{ik}^{2}-(m_{ik}+m_{jl})h_{ik}^{2})~{}I_{e}(\beta_{ijkl})$
(8)
and
$\frac{dv_{in,ik}^{2}}{dt}=\sum_{j=0}^{j=N}\sum_{l=0}^{l=l_{max}}\frac{C_{in}}{\beta_{ijkl}^{2}}~{}(m_{jl}v_{jl}^{2}-m_{ik}v_{ik}^{2}-(m_{ik}+m_{jl})v_{ik}^{2})~{}I_{i}(\beta_{ijkl})$
(9)
where
$C_{in}=\alpha_{coll}~{}f_{ijkl}~{}\epsilon_{ijkl}~{}\rho_{jl}~{}V_{ijkl}~{}F_{g,ijkl}~{}(r_{ik}+r_{jl})^{2}$,
$\beta_{ijkl}^{2}=(i_{ik}^{2}+i_{jl}^{2})/(e_{ik}^{2}+e_{jl}^{2})$, and
$\rho_{jl}$ is the volume density of planetesimals with mass $m_{jl}$ in
annulus $j$ (Ohtsuki, 1992; Wetherill & Stewart, 1993). In the second
summation, $l_{max}=k$ when $i=j$; $l_{max}$ = $M$ when $i\neq j$ (see also
Kenyon & Luu, 1998, 1999). We add a term, $f_{ijkl}$, to treat the overlap
between adjacent zones; $f_{ijkl}$ = 1 when $i=j$ and $f_{ijkl}\leq 1$ when
$i\neq$ j (Kenyon & Bromley, 2001). The integrals $I_{e}$ and $I_{i}$ are
elliptic integrals described in previous publications (Wetherill & Stewart,
1993; Stewart & Ida, 2000; Ohtsuki, Stewart, & Ida, 2002).
For velocity damping from gas drag, we follow Wetherill & Stewart (1993) and
write
$\frac{dh_{gd,ik}}{dt}=-\beta_{ik}~{}\frac{\pi
C_{D}}{2m_{ik}}\rho_{gas}V_{gas}^{2}r_{ik}^{2},$ (10)
and
$\frac{dv_{gd,ik}}{dt}=-(1-\beta_{ik})~{}\frac{\pi
C_{D}}{2m_{ik}}\rho_{gas}V_{gas}^{2}r_{ik}^{2},$ (11)
where $C_{D}$ = 0.5 is the drag coefficient,
$\beta_{ik}=h_{ik}/(h_{ik}^{2}+v_{ik}^{2})^{1/2}$, $\rho_{gas}$ is the gas
density, $\eta$ is the relative gas velocity, and
$V_{gas}=(V_{ik}(V_{ik}+\eta))^{1/2}$ is the mean relative velocity of the gas
(see Adachi et al., 1976; Weidenschilling, 1977b; Wetherill & Stewart, 1993).
For gravitational interactions, we compute long-range stirring from distant
oligarchs (Weidenschilling, 1989) and short-range stirring from the swarm
(Ohtsuki, Stewart, & Ida, 2002). The long-range stirring only has a horizontal
component,
$\frac{dh_{lr,ik}^{2}}{dt}=\sum_{j=1}^{j=N}\sum_{l=1}^{l=M}C_{lr,e}~{}x_{ijkl}~{}\frac{G^{2}\rho_{jl}M_{jl}}{\langle\Omega_{ij}\rangle}\left(\frac{{\rm
tan^{-1}}(H_{ijkl}/D_{min})}{D_{min}}-\frac{{\rm
tan^{-1}}(H_{ijkl}/D_{max})}{D_{max}}\right)$ (12)
for continuum objects and
$\frac{dh_{lr,ik}^{2}}{dt}=\sum_{j=1}^{j=N}\sum_{l=1}^{l=M}\frac{G^{2}}{\pi\Omega
a}\left(\frac{C_{lr,e}^{\prime}m_{jl}^{2}}{(\delta
a^{2}+0.5H_{jl}^{2})^{2}}\right)$ (13)
for individual oligarchs, where $x_{ijkl}$ is the fraction of objects with
mass $m_{ik}$ in annulus $i$ that approach no closer than 2.4 $R_{H}$ of the
objects with mass $m_{l}$ in annulus $j$, $D_{min}={\rm
max}(2.4R_{H},1.6(h_{ik}^{2}+h_{jl}^{2})^{1/2})$, $D_{max}=0.5~{}{\rm
max}(w_{ik},w_{jl})$, $\delta a$ = $|a_{i}-a_{j}|$, $C_{lr,e}$ = 23.5, and
$C_{lr,e}^{\prime}$ = 5.9 (see also Kenyon & Bromley, 2001).
For short-range gravitational interactions, the stirring depends on the ratio
of the relative collision velocity to the mutual Hill velocity,
$v_{H}\approx\langle\Omega_{ij}\rangle~{}\langle
a_{ij}\rangle~{}[(m_{ik}+m_{jl})/3M_{\star}]^{1/3}~{}.$ (14)
In the high velocity regime, the collision velocity exceeds the Hill velocity.
Statistical solutions to the Fokker-Planck equation then yield accurate
estimates for the stirring rates (e.g., Hornung et al., 1985; Wetherill &
Stewart, 1993; Stewart & Ida, 2000; Kenyon & Bromley, 2001). At low
velocities, $n$-body calculations provide good estimates. We follow Ohtsuki,
Stewart, & Ida (2002) and write the stirring as the sum of rates in the two
regimes:
$\frac{dh_{sr,ik}^{2}}{dt}=\frac{dh_{high,ik}^{2}}{dt}~{}+~{}\frac{dh_{low,ik}^{2}}{dt}$
(15)
and
$\frac{dv_{sr,ik}^{2}}{dt}=\frac{dv_{high,ik}^{2}}{dt}~{}+~{}\frac{dv_{low,ik}^{2}}{dt}~{},$
(16)
where the subscripts ‘high’ and ‘low’ indicate the velocity regime (e.g., Eq.
(25) of Ohtsuki, Stewart, & Ida, 2002).
In the high velocity regime, the stirring is (e.g. Stewart & Ida, 2000; Kenyon
& Bromley, 2001)
$\frac{dh_{high,ik}^{2}}{dt}=\sum_{j=1}^{j=N}\sum_{l=1}^{l=M}f_{ijkl}~{}C_{high}~{}((h_{ik}^{2}+h_{jl}^{2})~{}m_{jl}~{}J_{e}(\beta_{ijkl})+1.4~{}(m_{jl}h_{jl}^{2}-m_{ik}h_{ik}^{2})~{}H_{e}(\beta_{ijkl}))$
(17)
and
$\frac{dv_{high,ik}^{2}}{dt}=\sum_{j=1}^{j=N}\sum_{l=1}^{l=M}f_{ijkl}~{}\frac{C_{high}}{\beta_{ijkl}^{2}}~{}((v_{ik}^{2}+v_{jl}^{2})~{}m_{jl}~{}J_{z}(\beta_{ijkl})+1.4~{}(m_{jl}v_{jl}^{2}-m_{ik}v_{ik}^{2})~{}H_{z}(\beta_{ijkl}))~{},$
(18)
where $f_{ijkl}$ the fraction of objects with mass $m_{ik}$ in annulus $i$
that approach within 2.4 $R_{H}$ of the objects with mass $m_{jl}$ in annulus
$j$ and $C_{high}$ = 0.28
$A_{\Lambda}~{}G^{2}~{}\rho_{jl}/((h_{ik}^{2}+h_{jl}^{2})^{3/2})$. In the
expression for $C_{high}$, $A_{\Lambda}$ = ln $(\Lambda^{2}+1)$ and
$\Lambda=\left(\frac{0.19~{}(h_{ik}^{2}+h_{jl}^{2}~{}+~{}1.25~{}(v_{ik}^{2}+v_{jl}^{2}))~{}(v_{ik}^{2}+v_{jl}^{2})^{1/2}}{v_{H}^{3}}\right)^{2}~{}.$
(19)
The functions $H_{e}$, $H_{z}$, $J_{e}$, and $J_{z}$ are definite integrals
defined in Stewart & Ida (2000).
In the low velocity regime, the evolution equations are (Ohtsuki, Stewart, &
Ida, 2002):
$\frac{dh_{low,ik}^{2}}{dt}=\sum_{j=1}^{j=N}\sum_{l=1}^{l=M}f_{ijkl}~{}C_{low}~{}(m_{jl}~{}\chi_{1}~{}+~{}(m_{jl}h_{jl}^{2}~{}-~{}m_{ik}h_{ik}^{2})~{}\chi_{3})$
(20)
and
$\frac{dv_{low,ik}^{2}}{dt}=\sum_{j=1}^{j=N}\sum_{l=1}^{l=M}f_{ijkl}~{}C_{low}~{}(m_{jl}~{}\chi_{2}~{}+~{}(m_{jl}h_{jl}^{2}-m_{ik}h_{ik}^{2})~{}\chi_{4})$
(21)
where the $\chi$’s are simple functions of the Hill radius
$r_{H,ijkl}=a[(2~{}(m_{ik}+m_{jl}))/3M_{\star}]^{1/3}~{}.$ (22)
and the normalized eccentricity and inclination (Ohtsuki, Stewart, & Ida,
2002, see also Ida 1990; Ida & Makino 1992). For the low velocity limit of the
horizontal velocity
$\chi_{1}=73~{}C_{1}~{}r_{H,ijkl}^{4}$ (23)
and
$\chi_{2}=C_{2}~{}(4~{}\tilde{i}_{ij}^{2}+0.2~{}\tilde{e}_{ij}^{2}~{}(\tilde{e}_{ij}^{2}\tilde{i}_{ij}^{2})^{1/2})~{}r_{H,ijkl}^{4}~{}.$
(24)
For the low velocity limit of the vertical velocity
$\chi_{3}=10~{}C_{3}~{}\tilde{e}_{ij}^{2}~{}r_{H,ijkl}^{4}/(h_{ik}^{2}+h_{jl}^{2})$
(25)
and
$\chi_{4}=10~{}C_{3}~{}\tilde{i}_{ij}^{2}~{}r_{H,ijkl}^{4}/(h_{ik}^{2}+h_{jl}^{2})~{}.$
(26)
Here, $C_{low}=0.625~{}\langle
a_{ij}\rangle~{}\rho_{jl}~{}H_{ijkl}~{}V_{K,i}^{3}/(m_{ik}+m_{jl})^{2}$,
$\tilde{e}_{ij}^{2}=(e_{ik}^{2}+e_{jl}^{2})/r_{H,ijkl}^{2}$, and
$\tilde{i}_{ij}^{2}=(i_{ik}^{2}+i_{jl}^{2})/r_{H,ijkl}^{2}$. The constants
$C_{1}$, $C_{2}$, and $C_{3}$ are identical to those in Eq. (26) of Ohtsuki,
Stewart, & Ida (2002).
Several tests indicate that these expressions provide an accurate treatment of
velocity evolution for planetesimals in the high and low velocity regimes.
Figs. 5–7 of Ohtsuki, Stewart, & Ida (2002) show comparisons with results from
$n$-body simulations (see also Ida, 1990; Ida & Makino, 1992). Our simulations
confirm this analysis. Weidenschilling et al. (1997) and Kenyon & Bromley
(2001) compare variants of this formalism with other $n$-body calculations.
Goldreich, Lithwick, & Sari (2004) demonstrate that our numerical results
agree with analytic estimates.
To follow the evolution of the most massive objects more accurately, our code
includes an $n$-body algorithm. When objects have masses exceeding a
‘promotion mass’ $m_{pro}$, we promote them into an $n$-body code that
directly solves their orbits. The $n$-body code incorporates algorithms to
allow collisions among $n$-bodies and interactions between $n$-bodies and
coagulation particles. Bromley & Kenyon (2006) describe this code in detail.
Because dynamical interactions among large oligarchs are rare and occur at
late stages in the evolution, we set $m_{pro}=10^{26}$ g. We describe several
test calculations with smaller $m_{pro}$ in §3.3.
To treat the time evolution of the gas volume density $\rho_{gas}$, we use a
simple nebular model with gas surface density
$\Sigma_{gas}(a,t)=\Sigma_{gas,0}a^{-3/2}e^{-t/t_{gas}}$, gas-to-solids ratio
$\Sigma_{gas,0}(a)/\Sigma(a)$ = 100 – where $\Sigma$ is the surface density of
solids, and scale height $H_{gas}(a)=H_{gas,0}(a/a_{0})^{1.125}$ (Kenyon &
Hartmann, 1987). To approximate gas removal on a timescale $t_{gas}$, the gas
density declines exponentially with time. We set $t_{gas}$ = 10 Myr. During
the early stages of calculations at 30–150 AU, velocity damping is important
for particles with $r\lesssim$ 100 m. However, particle losses from gas drag
are small, $\sim$ 1% or less of the initial mass. By the time viscous stirring
dominates the velocity evolution of small objects, the gas disk has dispersed.
Inward drift and velocity damping are then negligible (see also Wetherill &
Stewart, 1993).
The initial conditions for these calculations are appropriate for a disk with
an age of $\lesssim$ 1–2 Myr (e.g. Dullemond & Dominik, 2005; Nomura &
Nakagawa, 2006; Ciesla, 2007; Garaud, 2007). We consider systems of $N$ annuli
in disks with $a_{i}$ = 30–150 AU and $\delta a_{i}/a_{i}$ = 0.025. The disk
is composed of small planetesimals with radii of $\sim$ 1–1000 m and an
initial mass distribution $n_{i}(m_{ik})$ in each annulus. The mass ratio
between adjacent bins is $\delta=m_{ik+1}/m_{ik}$ = 1.4–2. At the start of the
calculations, each bin has the same total mass, eccentricity
$e_{0}=1-3~{}\times~{}10^{-4}$, and inclination $i_{0}=e_{0}/2$. We assume a
power law variation of the initial surface density of solid material with
semimajor axis,
$\Sigma_{i}=\Sigma_{0}(M_{\star})~{}x_{m}~{}(a_{i}/a_{0})^{-3/2}~{},$ (27)
where $x_{m}$ is a scaling factor. For a 1 $M_{\odot}$ central star, models
with $\Sigma_{0}\approx$ 0.1–0.2 g cm-2 at $a_{0}$ = 30 AU have a mass in icy
solids comparable to the minimum mass solar nebula (MMSN hereafter;
Weidenschilling, 1977a; Hayashi, 1981). Consistent with observations of disks
surrounding pre-main sequence stars (e.g., Natta et al., 2000; Scholz et al.,
2006), we scale the reference surface density with the stellar mass,
$\Sigma_{0}(M_{\star})$ = 0.18 $(M_{\star}/M_{\odot}$) g cm-2.
Table 1 lists the ranges in $M_{\star}$ and $x_{m}$ we consider. The table
also lists the main sequence lifetime, $t_{ms}$, defined as the time to reach
core hydrogen exhaustion in the $X$ = 0.71, $Y$ = 0.27, and $Z$ = 0.02 stellar
evolution models of Demarque et al. (2004), where $X$ is the initial mass
fraction of hydrogen, $Y$ is the mass fraction of helium, and $Z$ is the
metallicity. For most of our calculations, the number of annuli in the disk is
$N$ = 64. To check these results, we also calculated models for disks with $N$
= 32 around 1 $M_{\odot}$ stars. Because the growth of planets has large
stochastic variations, we repeated the calculations 5–12 times for each set of
starting conditions, $M_{\star}$, $N$, $x_{m}$, and $Q_{b}$. Table 1 lists the
number of calculations for each ($M_{\star}$, $x_{m}$) pair.
Our calculations follow the time evolution of the mass and velocity
distributions of objects with a range of radii, $r_{ik}=r_{min}$ to
$r_{ik}=r_{max}$. The upper limit $r_{max}$ is always larger than the largest
object in each annulus. To save computer time in our main calculation, we do
not consider small objects which do not affect significantly the dynamics and
growth of larger objects, $r_{min}$ = 100 cm. Erosive collisions produce
objects with $r_{ik}$ $<r_{min}$ which are ‘lost’ to the model grid. Lost
objects are more likely to be ground down into smaller objects than to collide
with larger objects in the grid (see Kenyon & Bromley, 2002a, 2004a).
To estimate the amount of dusty debris produced by planet formation, we
perform a second calculation. Each main calculation yields $\dot{M}_{i}(t)$,
the amount of mass lost to the grid per annulus per timestep, and $H_{i0}(t)$,
the scale height of the smallest particle ($r=r_{min}$) in each annulus of the
coagulation grid. Objects with sizes smaller than $r_{min}$ contain a small
fraction of the mass in each annulus; thus, the scale height for objects with
$r<r_{min}$ is $H_{i0}(t)$ (Goldreich, Lithwick, & Sari, 2004). The total
amount of mass lost from the planetesimal grid in each timestep is
$\dot{M}(t)=\sum_{i=1}^{N}\dot{M}_{i}(t)$. The debris has a known size
distribution, $n^{\prime}_{ij}=n^{\prime}_{i0}r_{i}^{-\beta}$, where $\beta$
is a constant (see Stern & Colwell, 1997; Kenyon & Luu, 1999, and references
therein). The normalization constant $n_{i0}^{\prime}$ depends only on $\beta$
and $\dot{M}(t)$, which we derive at each timestep in the main calculation. To
evolve the dust distribution in time, we use a simple collision algorithm that
includes Poynting-Robertson drag and radiation pressure111Because the
collisional cascade begins after the gas disk dissipates, we ignore gas drag..
The optical depth $\tau$ of the dust follows from integrals over the size
distribution in each annulus. The optical depth and a radiative transfer
solution then yield the luminosity and radial surface brightness of the dust
as a function of time. Kenyon & Bromley (2004a) describe this calculation in
more detail.
Throughout the text, we use simple scaling relations to show how our results
depend on initial conditions and the properties of the grid. For each set of
calculations (Table 1), we derive median results for the size distribution,
the size of the largest object as a function of $a$ and $t$, and other
physical variables. Substituting the inter-quartile range for the dispersion,
we then perform least-squares fits to relate these median results to input
parameters (e.g., $x_{m}$) and the properties of the grid (e.g., $a$). For
parameters where analytic theory predicts a relation (e.g., the growth time as
a function of $a$), we derive the best-fitting coefficient, test whether
different fitting functions provide better fits to our results, and keep the
result that minimizes $\chi^{2}$ per degree of freedom. When analytic theory
provides no guidance, we derive fitting functions that yield the sensitivity
of our results to all important physical variables. Thus, our fits test some
aspects of analytic theory and guide other aspects.
## 3 PLANET FORMATION CALCULATIONS
### 3.1 Icy Planet Formation in Disks around 1 $M_{\odot}$ Stars
We begin with a discussion of planet formation in disks at 30–150 AU around a
1 $M_{\odot}$ star. For most disks around low mass stars, the timescale for
planet formation is shorter than the main sequence lifetime. Thus, the growth
of planetesimals into planets and the outcome of the collisional cascade
depend more on the physics of solid objects than on stellar physics. Here, we
review the stages in planet growth and describe the outcome of the collisional
cascade. For a suite of calculations of planet formation in disks of different
masses, we derive basic relations for the growth time and the maximum planet
mass as a function of initial disk mass. We also show how the dust production
rate and the mass in small objects depend on initial disk mass and time.
The next section compares these results with calculations for 1.5–3
$M_{\odot}$ stars. For disks around more massive stars, the planet formation
timescale is comparable to the main sequence lifetime. Thus, the central star
evolves off the main sequence before planet formation and the collisional
cascade reach a natural end-state. During post-main sequence evolution, the
star brightens considerably (e.g., Demarque et al., 2004) and develops a
powerful stellar wind (e.g., Knapp & Morris, 1985, and references therein),
melting icy objects in the inner disk and ejecting small grains throughout the
disk. Thus, we halt our calculations when the star evolves off the main
sequence. Here, we show how the physics of main sequence stars changes the
results derived for planet formation around 1 M⊙ stars.
#### 3.1.1 Growth of Large Objects
The formation of icy planets in the outer regions of a quiescent planetesimal
disk has three distinct stages (Kenyon & Luu, 1999; Kenyon & Bromley, 2004a).
Planetesimals with $r\lesssim$ 1 km grow slowly. As they grow, dynamical
friction damps $e$ for the largest objects; dynamical friction and viscous
stirring raise $e$ for the smallest objects (e.g., Greenberg et al., 1984;
Wetherill & Stewart, 1993; Goldreich, Lithwick, & Sari, 2004). After $\sim$
0.1–1 Myr, gravitational focusing enhances the collision cross-sections by
factors of 10–100. Slow, orderly growth ends. Runaway growth begins. At the
inner edge of the disk, the largest objects take $\sim$ 3 Myr to grow to
$\sim$ 100 km and $\sim$ 30 Myr to grow to $\sim$ 1000 km. Throughout runaway
growth, the gas disk dissipates. Thus, velocity damping by gas drag ceases;
dynamical friction and viscous stirring increase $e$ for the smallest objects.
Stirring reduces gravitational focusing factors, slowing the growth of the
largest objects relative to one another and relative to leftover planetesimals
(Ida & Makino, 1993; Wetherill & Stewart, 1993). Runaway growth ends;
oligarchic growth begins (Kokubo & Ida, 1998; Rafikov, 2003; Chambers, 2006;
Nagasawa et al., 2007). After 30–100 Myr, the largest objects – oligarchs –
grow slowly and contain an ever increasing fraction of the remaining mass in
the disk.
During the transition from runaway to oligarchic growth, collisions start to
produce copious amount of dust. Once oligarchs reach sizes $\sim$ 500 km,
collisions between 1–10 km objects produce debris instead of mergers (Kenyon
et al., 2008, and references therein). Once fragmentation begins, continued
stirring leads to a collisional cascade, where leftover planetesimals are
ground to dust. For dust grains with sizes $\gtrsim$ 10 $\mu$m, the collision
time is much shorter than the time to remove particles by gas drag (Adachi et
al., 1976) or by Poynting-Robertson drag (Burns, Lamy, & Soter, 1979). Thus,
the cascade proceeds to particle sizes $\lesssim$ 1–10 $\mu$m, where radiation
pressure removes material on the dynamical time scale (Burns, Lamy, & Soter,
1979). Because runaway growth leaves most of the mass in 1–10 km objects, the
collisional cascade effectively removes a significant fraction of the solid
material in the disk.
Fig. 1 shows the time evolution of the eccentricity and the mass distributions
for objects in the innermost 8 annuli of a disk with initial mass distribution
similar to the MMSN. To minimize stochastic variations, these plots show
median results for 15 calculations. During slow growth and the early stages of
runaway growth, dynamical friction damps $e$ for the largest objects and
raises $e$ for the smallest objects (Fig. 1; right panel, 10 Myr). The mass
distribution develops a pronounced shoulder from 10 km to $\sim$ 300–500 km.
As the evolution proceeds, growth concentrates more mass in the largest
objects; stirring excites the orbits of the smallest objects. After 100 Myr,
the collisional cascade removes mass efficiently from the smallest objects but
leaves the mass distribution of the largest objects unchanged. By $\sim$ 5–10
Gyr, almost all of the small objects have been removed.
In these calculations, the rate of planet formation is very sensitive to
semimajor axis (Fig. 2). For collisional processes, the growth time in the
disk is $t\propto P/\Sigma$, where $P$ is the orbital period (see the
Appendix; also Lissauer, 1987; Kenyon & Luu, 1998; Goldreich, Lithwick, &
Sari, 2004). For $P\propto a^{3/2}$ and $\Sigma\propto x_{m}a^{-3/2}$ (Eq.
(27)), the growth time is $t\propto a^{3}x_{m}^{-1}$. Thus, although it takes
only $\sim$ 10 Myr for the largest planets to reach radii of 300–600 km at 30
AU, the largest objects at $a\gtrsim$ 100 AU still have $r\sim$ 3–5 km. By 100
Myr, 100 km objects form at 75–80 AU. After 1 Gyr, 100 km objects form beyond
125 AU. By the end of the calculation at 10 Gyr, 1000 km objects form
throughout the disk.
The formation rate also depends on the initial disk mass (Fig. 3). For an
expected growth time $t\propto a^{3}x_{m}^{-1}$, planets grow faster in more
massive disks. At 100 AU, planets with $r\sim$ 2000 km form in a massive disk
($x_{m}$ = 3) within 1 Gyr. In a low mass disk with $x_{m}\sim$ 1/3, the
largest planet at 100 AU grows to $r\sim$ 300 km in 1 Gyr and $r\sim$ 2000 km
in 10 Gyr. For all simulations of disks around 1 $M_{\odot}$ stars, the median
timescale for the formation of the first 1000 km object is
$t_{1000}\approx 475~{}x_{m}^{-1.15}~{}\left(\frac{a}{\rm
80~{}AU}\right)^{3}~{}{\rm Myr}~{}.$ (28)
This relation fits our results for the median growth time to $\approx$ 5% for
$a$ = 30–150 AU and for $x_{m}$ = 1/3 to 3. For each initial disk mass, the
inter-quartile range for the formation time is $\sim$ 20%. Thus, there is a
modest range of outcomes for identical starting conditions.
In equation 28, there is a small difference between the result expected from
simple theory ($t\propto x_{m}^{-1}$) and the result derived in our
calculations ($t\propto x_{m}^{-1.15}$). We show in the Appendix how gas drag
during runaway and oligarchic growth can modify the simple theory and explain
the result of our calculations.
Although the timescale to produce the first 1000 km object is a strong
function of initial disk mass and semimajor axis, the evolution at late times
is less sensitive to the starting conditions. To derive a simple relation for
the median size $r_{max}$ of the largest object as a function of initial disk
mass and semimajor axis, we adopt a simple function
$r_{max}(a)=r_{0}~{}e^{-(a_{i}/a_{0})^{\alpha_{r}}}$ (29)
and use an amoeba algorithm (Press et al., 1992) to derive the fitting
parameters $a_{0}$, $r_{0}$, and $\alpha_{r}$ as a function of time. For
stellar ages 1 Gyr $\lesssim t_{\star}\lesssim$ 10 Gyr, the ensemble of
calculations yields
$r_{0}\approx 1650~{}x_{m}^{0.2}\left(\frac{t}{\rm
3~{}Gyr}\right)^{0.06}~{}{\rm~{}km}$ (30)
for the radius of the largest object,
$a_{0}\approx 190~{}x_{m}^{0.1}\left(\frac{t}{\rm
3~{}Gyr}\right)^{0.1}~{}{\rm~{}AU}$ (31)
for the scale length, and $\alpha_{r}\approx$ 5–6 for the exponent. These
relations match our results to $\pm$5%. The uncertainties are $\pm$3% in the
coefficients and $\pm$5% in the exponents.
These calculations produce relatively low mass icy planets with radii $\sim$
50% larger than the radius of Pluto (Young & Binzel, 1994; Elliot et al.,
2003, 2007). Although these objects form relatively rapidly in the inner disk,
they grow very slowly at late times. Between 1–10 Gyr, most large object grow
by $\sim$ 10–20% in radius ($\sim$ 50% in mass). Because the size of the
largest object depends weakly on the initial disk mass, nearly all disks form
Pluto-mass planets. These objects stir leftover planetesimals effectively;
thus, nearly all disks develop a collisional cascade.
Despite the general formation of Pluto-mass planets in any disk, the lowest
mass disks ($x_{m}\lesssim$ 1/3) form Plutos inefficiently. The scale length,
$a_{0}\gtrsim$ 150 AU, derived from our calculations exceeds the outer radius
of the disk. Thus, planet formation does not proceed to completion at large
$a$ for the lowest mass disks. In these systems, the largest icy planets at
$a\approx$ 125–150 AU are factors of 3–10 smaller than $r_{0}$. The large
exponent, $\alpha_{r}\sim$ 5–6, derived in our fits implies a rapid transition
– $\sim$ 10–20 AU – between disk regions where the largest objects are planets
with $r_{max}$ $\approx r_{0}$ and where the largest objects have
$r_{max}\lesssim$ 300–500 km.
In our calculations, the collisional cascade limits the size of the largest
objects. Once a few objects have radii $\gtrsim$ 1000 km, they stir up
leftover planetesimals to the disruption velocity. When the collisional
cascade begins, the timescale for 1 km planetesimals to collide and fragment
into smaller objects is shorter than the timescale for oligarchs to accrete
leftover planetesimals. Thus, the growth of the largest objects stalls at
$\sim$ 1000–2000 km ($\sim$ 0.01–0.02 $M_{\oplus}$). Occasionally, runaway and
oligarchic growth produce a very large object with $r\sim$ 5000 km ($\sim$ 0.1
$M_{\oplus}$), but these objects form in only $\sim$ 5–10% of the simulations.
These objects form at random semimajor axes and tend to form in more massive
disks.
The large radial variation in the formation time produces dramatic differences
in the mass distribution as a function of semimajor axis (Fig. 4). In the
inner disk, rapid growth leads to many objects with $r\gtrsim$ 1000 km (Table
2). With many large objects stirring the leftover planetesimals in the inner
disk, the collisional cascade removes most of the mass in small objects (Fig.
4, left panel). In the outer disk, slow growth results in a handful of Pluto-
mass objects. A few large objects cannot stir leftover planetesimals
efficiently. Thus, the collisional cascade is weak and leaves a substantial
amount of mass in 1–10 km planetesimals (Fig. 4, right panel).
The growth of objects as a function of semimajor axis and time is not
sensitive to the size of the model grid (Fig. 5). For two sets of calculations
with 32 annuli (cyan and magenta points), the median radius of the largest
object in each annulus is nearly identical to results for calculations with 64
annuli (black points). The results for Eqs. (28–34) are also independent of
the grid. In principle, long-range stirring from planets at small $a$ can
influence runaway growth of objects at large $a$. Our results suggest that icy
planet formation at large semimajor axes is not influenced by the formation of
small icy planets at small semimajor axes.
To conclude this discussion of the formation of large objects in a
planetesimal disk, we quote several simple relations for the amount of solid
material in small and large objects as a function of initial disk mass and
semimajor axis for the ensemble of calculations around a 1 M⊙ star. At 3–10
Gyr, the median fraction of solids remaining in the disk is
$f_{s}\approx 0.3\left(\frac{a}{\rm
100~{}AU}\right)\left(\frac{1}{x_{m}}\right)^{1/4}~{}.$ (32)
For a MMSN with $x_{m}$ = 1, the amount of mass remaining in the disk at 3–10
Gyr ranges from 9% of the initial mass at 30 AU to roughly 50% of the initial
mass at 150 AU. Thus, the inner disk is substantially depleted, while the
outer disk contains a significant fraction of its initial mass.
For each $x_{m}$, the median fraction of the initial disk mass in 1000 km and
larger objects is
$f_{1000}=0.035\left(\frac{\rm 30~{}AU}{a}\right)~{}.$ (33)
The median fraction of the mass in 100 km and larger objects is roughly 50%
larger,
$f_{100}=0.06\left(\frac{\rm 30~{}AU}{a}\right)~{}.$ (34)
For the ensemble of calculations, the typical inter-quartile range is $\sim$
0.1 for $f_{s}$ and $\sim$ 20% for $f_{100}$ and $f_{1000}$. Thus, the mass
distributions in our calculations are top-heavy, with more mass in 1000+ km
objects than in 100–1000 km objects.
These relations demonstrate that planet formation at 30–150 AU is very
inefficient. For all disk radii in this range, only $\sim$ 6% or less of the
initial population of 1 km objects is incorporated into large objects with
radii exceeding 100 km. In the inner disk (30–50 AU), the collisional cascade
is very efficient at removing leftover 1–10 km objects. Thus, at 3–10 Gyr, the
large objects contain most of the mass in the inner disk. In the outer disk
(100–150 AU), the collisional cascade does not have enough time to remove
leftover planetesimals. Thus, small objects with radii of 1–10 km contain most
of the remaining mass at 100–150 AU.
#### 3.1.2 Evolution of Dust
At all semimajor axes, the collisional cascade converts a large fraction of
the initial mass in solids into small dust grains. Because oligarchs and
leftover planetesimals are unobservable with current techniques, dust emission
provides the sole observational diagnostic of the growth of icy planets at
30–150 AU around other stars. Here, we describe the evolution of these small
particles and demonstrate that the collisional cascade is observable.
Two physical processes set the visibility of dust grains in a debris disk.
Once significant fragmentation begins, collisions gradually grind the
fragments to dust. When dust grains are small enough, radiative processes
remove them from the disk. For disks at 30–150 AU, radiation pressure
dominates mass loss for $t\lesssim$ 1–3 Gyr and removes 65% to 70% of the
total mass loss. Poynting-Robertson drag removes material at late times and is
responsible for 30% to 35% of the total mass loss. Because the gas density is
negligible once the collisional cascade begins, gas drag is unimportant.
To describe our results, we divide dusty debris into ‘large grains’ with 1 mm
$\lesssim r\lesssim$ 1 m, ‘small grains’ with 1 $\mu$m $\lesssim r\lesssim$ 1
mm, and ‘very small grains’ with $r\lesssim$ 1 $\mu$m. Collisions dominate the
evolution of large grains at all times. For $t\lesssim$ 1–3 Gyr, collisions
dominate the evolution of small grains; Poynting-Robertson drag then removes
grains with radii of 1–100 $\mu$m on Gyr timescales. Radiation pressure
removes very small grains on the local dynamical timescale. Thus, radiation
pressure produces a ‘wind’ of very small grains in the disk midplane.
Fig. 6 shows the time evolution of the dust production rate for very small
grains as a function of initial disk mass. At the start of each calculation,
dynamical friction and collisions damp orbital eccentricities. Thus,
collisions produce less and less debris; the dust production rate declines
with time. As oligarchs reach radii of $\sim$ 500 km, they stir leftover
planetesimals along their orbits. Dust production increases. Because oligarchs
continue to grow, they stir leftover planetesimals to larger and larger
velocities. Collision energies rapidly exceed the critical disruption energy;
the dust production rate increases dramatically (Eq. 4–5). When oligarchs
start to reach their maximum radii in the inner disk (Eq. (29)), the dust
production rate peaks. As oligarchs grow at larger and larger disk radii, the
dust production rate slowly declines.
Although the outer disk contains more mass, the global dust production rate
declines with time for two reasons. Large oligarchs form at late times in the
outer disk (Fig. 2), but the smaller disk surface density and the longer
orbital periods lead to smaller collision rates. Smaller collision rates yield
smaller dust production rates. Initially, collisions dominate Poynting-
Robertson drag; thus, radiation pressure ejects the smallest grains on the
local orbital timescale (e.g., Krivov et al., 2000; Wyatt, 2005). The dust
production rate then declines roughly as $t^{-1}$. At late times, the
collision rates decline. Poynting-Robertson drag then removes larger grains
from the disk, which reduces the population of very small grains from erosive
collisions. The dust production rate then declines with evolution time as
$t^{-2}$ (see also Dominik & Decin, 2003; Kenyon & Bromley, 2004a, 2005;
Wyatt, 2005; Wyatt et al., 2007a, b).
The time evolution of the collision rate in the disk yields a simple relation
between the maximum dust production rate $\dot{M}_{max}$ and the initial disk
mass. For the complete set of calculations,
$\dot{M}_{max}\approx 6.5\times 10^{20}~{}x_{m}^{2}~{}{\rm g~{}yr^{-1}}~{}.$
(35)
For a MMSN with $x_{m}$ = 1, the maximum rate is $\sim$ 0.1 $M_{\oplus}$ every
million years. The collision rate scales with the square of the number density
of objects; thus, the dust production rate grows as the square of the initial
disk mass (e.g., $\dot{M}_{max}\propto x_{m}^{2}$).
The timescale for the peak in dust production is shorter than the timescale
for the production of 1000 km objects in the disk,
$t_{\dot{M}_{max}}\approx 14~{}x_{m}^{-1}~{}{\rm Myr}~{}.$ (36)
This time corresponds roughly to the time of peak stirring of leftover
planetesimals by oligarchs growing in the inner disk, starting the collisional
cascade. Oligarchs form faster in more massive disks; thus, the dust
production rate peaks earlier in more massive disks.
Fig. 7 shows the time evolution of the mass in small grains for disks with a
range of initial masses. Initially, the dust production rates are small (Fig.
6) and the dust mass increases slowly with time. Once large oligarchs form in
the inner disk, the dust production rate – and thus the mass in small grains –
grows rapidly. For all disks, it takes $\sim$ 5–10 Myr for the mass in small
grains to grow from $10^{-6}$ $M_{\oplus}$ (which is unobservable with current
techology) to $\sim$ $1-10\times 10^{-4}$ $M_{\oplus}$ (which is observable).
After this rapid rise, oligarchs form at larger and larger disk radii, leading
to enhanced dust production farther and farther out in the disk. The dust mass
then rises slowly with time. Once oligarchs form at the outer edge of the
disk, the collisional cascade removes more and more solid material throughout
the disk. The dust mass then begins to decline.
The maximum mass in small grains scales with the initial disk mass,
$M_{max,small}\approx 0.013~{}x_{m}~{}M_{\oplus}~{}.$ (37)
Because the derived size distributions are dominated by collisional processes,
the maximum mass in large grains is roughly 40 times larger (e.g., Dohnanyi,
1969; Williams & Wetherill, 1994; Kenyon & Bromley, 2004c; Krivov et al.,
2006),
$M_{max,large}\approx 0.5~{}x_{m}~{}M_{\oplus}~{}.$ (38)
In both cases, the larger collision rate in more massive disks leads to more
dust. Although these dust masses are significant, they are small compared to
the mass in objects with $r\gtrsim$ 100 km. The typical mass in large grains
is $\lesssim$ 10% of the mass in 100 km and larger objects. The mass in small
grains is $\lesssim$ 0.25% of the mass in the largest objects.
The timescale to reach the maximum dust mass is a function of the particle
size. For the small grains,
$t_{max,small}\approx 270~{}x_{m}^{-1/2}~{}{\rm Myr}~{}.$ (39)
For the large grains,
$t_{max,large}\approx 600~{}x_{m}^{-1/2}~{}{\rm Myr}~{}.$ (40)
Several features of the collisional cascade set these timescales. Early in the
evolution, the collision timescale for all particle sizes is smaller than the
timescale for Poynting-Robertson drag. Thus, the cascade erodes small
particles until radiation pressure ejects them. As planet formation propagates
out through the disk, collisions produce more and more small grains. Because
the mass in grains is set by a balance between the collision time, which
scales as $x_{m}^{-1}$ and the local dynamical time, which scales as
$x_{m}^{-1/2}$, the timescale to reach the maximum grain mass is proportional
to $x_{m}^{-1/2}$. As the collision rate declines, Poynting-Robertson drag
starts to remove mass from the disk. This drag removes smaller particles from
the disk more effectively than it removes large particles. Thus, the mass in
small grains peaks before the mass in larger grains.
These results suggest that the mass in collisional debris is large, roughly a
lunar mass in 0.001–1 mm grains, throughout the lifetime of a 1 $M_{\odot}$
star. Although the Solar System has much less dust (e.g., Landgraf et al.,
2002; Nesvorný et al., 2006, and references therein), these large disk masses
are comparable to the mass in dust grains detected in many debris disks around
other stars (e.g., Beichman et al., 2006; Trilling et al., 2008; Moro-Martin
et al., 2007). Because our dust production rates are observable, the large
range in dust masses as a function of initial disk mass and time implies a
correspondingly large range in the observable properties of debris disks, such
as the disk luminosity and the IR excess, at fixed stellar age. Because the
dust production rate declines with time, we expect the disk luminosity and IR
excesses to decline with time. We derive detailed predictions for this
evolution in §4 and compare these results with observations in §5.
### 3.2 Icy Planet Formation in 1.5–3 $M_{\odot}$ Stars
Stellar evolution is an important feature of icy planet formation at 30–150
AU. Because the main sequence lifetime ($t_{ms}\propto M_{\star}^{-n}$, with
$n\approx$ 3–3.5; e.g., Iben, 1967; Demarque et al., 2004) is more sensitive
to stellar mass than the timescale to produce large icy planets ($t\propto
M_{\star}^{-3/2}$; see below), massive stars evolve off the main sequence
before oligarchic growth and the collisional cascade remove solid material in
the outer disk. After a 1–3 M⊙ star evolves off the main sequence, it becomes
more luminous (as a red giant) and hotter (as a post-AGB star). During this
evolution, icy planetesimals and planets melt, decreasing collision rates and
changing the outcome of the collisional cascade222We assume that melting is
accompanied by a loss of volatiles and an increase in the mass density of
leftover planetesimals.. Short main sequence lifetimes of massive stars thus
lead to clear differences in the amount of solid material in large and small
objects in the outer disk.
The stellar mass also affects the outcome of icy planet formation. The
timescale for planet formation scales with orbital period and the surface
density, $t\propto P/\Sigma$ (see the Appendix). For a disk with
$\Sigma=\Sigma_{0}~{}x_{m}~{}a^{-3/2}$ (Eq. (27)) and fixed
$\Sigma_{0}~{}x_{m}$, the formation time is $t\propto
a^{3}~{}M_{\star}^{-1/2}$. Thus, at fixed $a$ in disks with identical surface
density distributions, planets form faster around more massive stars. However,
disk masses in the youngest stars scale with stellar mass (e.g., $M_{d}\propto
M_{\star}$; Natta et al., 2000; Scholz et al., 2006). Thus, $\Sigma_{0}$
scales with stellar mass, $\Sigma_{0}\propto M_{d}\propto M_{\star}$.
Combining these relations leads to a formation time $t\propto
a^{3}~{}M_{\star}^{-3/2}$. Thus, at fixed $a$ in typical disks, icy planets
form $\sim$ 5 times faster around 3 M⊙ stars than around 1 M⊙ stars.
To illustrate how stellar mass and evolution affect planet formation, we begin
with the growth of large objects at 40 AU and at 100 AU (Fig. 8). For disks
with identical initial surface density distributions, planets at the same $a$
in disks around 3 M⊙ stars grow $\sim$ 1.7 times faster than planets around 1
M⊙ stars. Fig. 8 also shows the clear scaling of growth time with semimajor
axis, $t\propto a^{3}$ for a disk with $\Sigma\propto a^{-3/2}$. The simple
scaling of the growth time with disk mass and orbital period in our
calculations leads to a general relation for the median timescale for the
formation of the first 1000 km object in disks at 30–150 AU,
$t_{1000}\approx 145~{}x_{m}^{-1.15}~{}\left(\frac{a}{\rm
80~{}AU}\right)^{3}~{}\left(\frac{2~{}M_{\odot}}{M_{\star}}\right)^{3/2}{\rm
Myr}~{}.$ (41)
For 1–3 M⊙ stars, this relation fits our results for the median growth time to
$\approx$ 7% for $a$ = 30–150 AU and for $x_{m}$ = 1/3 to 3. For all initial
disk masses, the inter-quartile range for the formation time is $\sim$ 20%.
Thus, there is a modest range of outcomes for identical starting conditions
around 1–3 M⊙ stars.
Aside from the extra factor of $x_{m}^{-0.15}$, this relation follows the
prediction of $t\propto x_{m}^{-1}M_{\star}^{-3/2}a^{3}$ from standard
coagulation theory. As outlined in the Appendix, velocity damping from gas
drag can speed up planet formation in more massive disks.
Fig. 9 further shows how the growth time varies with $a$ and $M_{\star}$ for
constant $x_{m}$. At 100 Myr, icy planets are close to their maximum sizes at
30–50 AU in the inner disk. At large disk radii ($a\sim$ 100–150 AU), planet
formation is clearly faster around more massive stars. For all stars with
$M_{\star}$ = 1–3 M⊙, the $r_{max}(a)$ relations have a similar morphology,
consisting of a plateau at $r_{max}\approx$ 1000–2000 km and a steep decline
of $r_{max}$ with increasing $a$. As in §3.1.1, we fit our results to a simple
relation between $r_{max}$, $a$, and time (Eq. (29)). For all of our
calculations, we derive an exponent $\alpha_{r}\approx$ 5–6 and a
characteristic maximum radius
$r_{0}\approx
1750~{}x_{m}^{b}\left(\frac{M_{\star}}{2~{}M_{\odot}}\right)^{0.09}\left(\frac{3t}{t_{ms}}\right)^{0.06}~{}{\rm~{}km}$
(42)
with the exponent $b=0.22+0.033~{}M_{\star}/M_{\odot}$. The disk scale length
is
$a_{0}\approx
190~{}x_{m}^{0.1}\left(\frac{3t}{t_{ms}}\right)^{0.1}~{}{\rm~{}AU}$ (43)
These results are valid for late times, $t\approx$ 0.1–1 $t_{ms}$.
Eq. (42) shows that the maximum sizes of icy planets at 30–150 AU are
relatively insensitive to initial disk mass, stellar mass, or time. For disks
with identical $x_{m}$ around 1–3 M⊙ stars, the largest icy planets around 3
M⊙ stars are only $\sim$ 10%–20% larger than the largest icy planets around
solar-type stars. This range is comparable to the range in maximum sizes for
planets formed in identical disks around stars of identical mass (§3.1.1).
Disks with a factor of 10 range in $x_{m}$ yield planets with a 20%–30% range
in radii (a factor of $\sim$ 2 in mass). Thus, our calculations predict that
the largest icy planets at 30–150 AU around 1–3 M⊙ stars have masses
comparable to Pluto and other large Kuiper belt objects in the Solar System
beyond the orbit of Neptune.
Although the maximum sizes of planets are fairly independent of initial
conditions, the number of Pluto-mass objects $n_{P}$ is sensitive to disk mass
and stellar mass (Table 3). In the inner disk (30–60 AU), $n_{P}$ scales
roughly with initial disk mass and stellar mass. In the outer disk (100–150
AU), the formation timescale for icy planets is comparable to the main
sequence lifetime. Thus, $n_{P}$ scales with initial disk mass and stellar
mass only for the most massive disks. In lower mass disks, stars evolve off
the main sequence before disks can produce large numbers of Pluto-mass
objects.
Cumulative mass distributions provide another useful comparison of icy planet
formation as a function of disk mass and stellar mass. For disks with
identical initial surface density distributions at $a$ = 30–37 AU, the shape
of the mass distribution is fairly independent of stellar mass at 100 Myr
(Fig. 10; left panel). Because large icy planets form first in disks around
more massive stars, disks of fixed age around 3 M⊙ stars have more mass in
larger planets and are more collisionally depleted than disks around 1 M⊙
stars.
For calculations in scaled MMSN (Fig. 10; right panel), growth is a stronger
function of stellar mass. As predicted by the simple scaling relations, disks
around 3 M⊙ stars have $\sim$ 5 times more mass in large objects than 1 M⊙
stars. More mass in large objects produces more stirring, enhancing mass loss
in the collisional cascade. Despite large difference in initial disk mass, the
mass distributions of disks around 1–3 M⊙ stars are very similar for $r<$ 1 km
at similar times.
Although planet formation proceeds faster with increasing stellar mass,
stellar evolution halts the collisional cascade and the growth of planets in
more massive stars (Fig. 11). Planets reach their maximum radii in the inner
disks for all 1–3 M⊙ stars; however, the timescale for the collisional cascade
to run to completion is long compared to the main sequence lifetimes of 2–3 M⊙
stars. Thus, the collisional cascade removes a larger fraction of material
from the inner disks around 1 M⊙ stars than from the inner disks of 2–3 M⊙
stars (Fig. 11; left panel). In the outer disk, the growth time for 1000 km
planets is large compared to the main sequence lifetime for 2–3 M⊙ stars.
Thus, planets do not reach their maximum radii of $\sim$ 1000–2000 km in the
outer disks of 2–3 M⊙ stars (Fig. 11; right panel)
To conclude our comparison of icy planet formation around stars of different
masses, we consider the long-term evolution of all solids in the disk. For
$t=0.3-1.0~{}t_{ms}$, the median fraction of solids remaining in the disk is
$f_{s}\approx 0.38\left(\frac{a}{\rm
100~{}AU}\right)^{1.25}\left(\frac{1}{x_{m}}\right)^{1/4}~{}.$ (44)
for the ensemble of calculations for 1.5 M⊙ stars and
$f_{s}\approx 0.6\left(\frac{a}{\rm
100~{}AU}\right)^{1.6}\left(\frac{1}{x_{m}}\right)^{1/3}~{}.$ (45)
for calculations for 2–3 M⊙ stars. At 30–50 AU, all stars lose roughly the
same fraction of mass from the disk. At larger disk radii, more massive stars
evolve off the main sequence before the collisional cascade removes most of
the leftover 1–10 km planetesimals. Thus, icy planet formation around lower
mass stars converts a larger fraction of the initial solid mass into dusty
debris.
Although planet formation around massive stars converts a smaller fraction of
the initial mass into dusty debris, icy planet formation is equally efficient
at producing massive objects around all 1–3 M⊙ stars. For all disks in our
calculations, the median fraction of the initial disk mass in 1000 km and
larger objects is
$f_{1000}=0.035\left(\frac{\rm 30~{}AU}{a}\right)~{}.$ (46)
The median fraction of the mass in 100 km and larger objects is $\sim$ 50%
larger,
$f_{100}=0.06\left(\frac{\rm 30~{}AU}{a}\right)~{}.$ (47)
Mass distributions for icy planets are top heavy for all 1–3 M⊙ stars. As for
calculations around 1 M⊙ stars, the typical inter-quartile ranges are $\sim$
0.1 for $f_{s}$ and $\sim$ 20% for $f_{100}$ and $f_{1000}$. Thus, identical
starting conditions lead to a modest range of outcomes.
#### 3.2.1 Evolution of Dust
The evolution of dusty debris in disks around 1.5–3 M⊙ stars generally follows
the evolution for 1 M⊙ stars. As oligarchs form, stirring leads to a
collisional cascade that converts 10 km and smaller objects into small dust
grains. Because planets form more rapidly around more massive stars, disks
around massive stars produce dust sooner than disks around lower mass stars.
In our calculations, the disk mass scales with the stellar mass. Thus, disks
around massive stars also produce more dust than disks around lower mass
stars. However, massive stars do not live as long as lower mass stars,
preventing the collisional cascade from removing all of the small objects from
the disk (Fig. 10). Over the lifetime of the star, disks around lower mass
stars form more dust than disks around more massive stars.
To illustrate these points, Fig. 12 shows the time evolution of the dust
production rate for a scaled MMSN ($x_{m}$ = 1) around 1 M⊙, 2 M⊙, and 3 M⊙
stars. During runaway growth, destructive collisions are rare; thus, the dust
production rate declines with time. As runaway growth ends, there are three
distinct phases in dust production: (i) an exponential rise when runaway
objects start to stir leftover planetesimals in the inner disk, (ii) a long
plateau as oligarchs form farther and farther out in the disk, and (iii) a
long decline in dust production as the collisional cascade depletes the disk
of 1–10 km objects.
Because planets grow more rapidly around more massive stars, the exponential
growth in dust production occurs first around more massive stars. The
timescale for the onset of dust production also scales inversely with disk
mass; thus, more massive disks produce dust faster than low mass disks.
When runaway growth ends and oligarchic growth begins, the dust production
rate reaches a clear plateau (Fig. 12). We define the onset of the plateau
phase as the time of maximum dust production333Because there are small
fluctuations in the dust production rate during the plateau phase, we define
the maximum dust production as the time when the derivative $d\dot{M}(t)/dt$
first changes sign.. For our calculations, there is a simple relation between
the timescale of maximum dust production and the masses of the disk and
central star,
$t_{\dot{M}_{max}}\approx
5~{}x_{m}^{-1}~{}\left(\frac{2~{}M_{\odot}}{M_{\star}}\right)^{1.5}{\rm
Myr}~{}.$ (48)
At this time, our simulations yield a simple relation between the maximum dust
production rate and the masses of the disk and the central star,
$\dot{M}_{max}\approx 3.5\times
10^{21}~{}x_{m}^{2}~{}\left(\frac{M_{\star}}{2~{}M_{\odot}}\right)^{2.5}~{}{\rm
g~{}yr^{-1}}~{}.$ (49)
Each of these scaling laws has a simple physical origin. The maximum dust
production rate, $\dot{M}_{max}$, depends on the collision rate, the square of
the disk mass divided by the orbital period. Thus, $\dot{M}_{max}$
$\propto\Sigma^{2}/P$ $\propto\Sigma^{2}M_{\star}^{1/2}$. For disks where the
surface density scales with stellar mass (Eq. (27)), $\Sigma\propto
x_{m}M_{\star}$. Thus, $\dot{M}_{max}$ $\propto x_{m}^{2}M_{\star}^{5/2}$. The
timescale to reach this rate is the orbital period divided by the disk mass.
Thus, $t_{\dot{M}_{max}}$ $\propto\Sigma^{-1}M_{\star}^{-1/2}$ $\propto
x_{m}^{-1}M_{\star}^{-3/2}$.
Once oligarchs form in the outer disk, the dust production rate declines
rapidly. Stellar evolution sets the duration of this decline. In massive
stars, the short main sequence lifetime halts the evolution before the
collisional cascade depletes the disk of 1–10 km objects. Thus, the dust
production rate declines by roughly an order of magnitude before the central
star evolves off the main sequence. For lower mass stars, the long main
sequence lifetime allows the collisional cascade to remove some material in
the outer disk. Thus, the dust production rate declines by $\sim$ two orders
of magnitude before the central star evolves into a red giant. We show in §4
how the long-term evolution of the dust production rate as a function of
stellar mass produces observable differences in the IR excesses of massive and
low mass stars.
Despite the large differences in dust production rates, there are smaller
variations in the amount of dust as a function of disk mass and stellar mass.
Fig. 13 shows the time evolution of the median mass in small grains for scaled
MMSN around 1–3 M⊙ stars. Once the exponential rise in dust production begins,
the dust masses rapidly evolve to similar configurations with $\sim 10^{26}$ g
in small grains. For our set of calculations, the maximum mass in small grains
is
$M_{max,small}\approx
0.026~{}x_{m}~{}\left(\frac{M_{\star}}{2~{}M_{\odot}}\right)M_{\oplus}~{}.$
(50)
The coefficient in this equation, $0.026M_{\oplus}$, is roughly twice the mass
of the Moon. The maximum mass in large particles is
$M_{max,large}\approx
1.0~{}x_{m}~{}\left(\frac{M_{\star}}{2~{}M_{\odot}}\right)M_{\oplus}~{}.$ (51)
The timescale to reach the maximum dust mass is a function of the particle
size. For the small grains,
$t_{max,small}\approx
135~{}x_{m}^{-1/2}~{}\left(\frac{M_{\star}}{2~{}M_{\odot}}\right)^{-1}{\rm
Myr}~{}.$ (52)
For the large grains,
$t_{max,large}\approx
300~{}x_{m}^{-1/2}~{}\left(\frac{M_{\star}}{2~{}M_{\odot}}\right)^{-1}{\rm
Myr}~{}.$ (53)
As described in §3.1.2, the collision rate, the dynamical timescale, and
Poynting-Robertson drag combine to produce the shorter timescale for smaller
dust grains.
### 3.3 Limitations of the Calculations
In previous papers, we have described limitations to multiannulus (Bromley &
Kenyon, 2006; Kenyon & Bromley, 2001, 2002a, 2004b, 2005, 2006) and single
annulus (Kenyon & Luu, 1998, 1999) coagulation calculations. Here, we review
how several of these limitations affect results for the simulations described
above.
As long as the statistical assumptions underlying the formalism are met,
coagulation calculations provide a reasonable representation of real collision
evolution (Wetherill, 1980; Greenberg et al., 1984; Davis et al., 1985; Barge
& Pellat, 1991; Spaute et al., 1991; Lissauer & Stewart, 1993; Wetherill &
Stewart, 1993; Stern & Colwell, 1997; Weidenschilling et al., 1997; Kenyon &
Luu, 1998; Inaba et al., 2001). For calculations at 30–150 AU around 1–3 M⊙
stars, the spacing of mass bins in an annulus and the spacing of annuli in the
disk limit the accuracy of the results. Our standard mass spacing, $\delta=2$,
lengthens the evolution time by 10% to 20% relative to more accurate
calculations with $\delta\lesssim$ 1.4 (see Kenyon & Luu, 1998, and references
therein). The radial resolution, $\Delta a_{i}/a_{i}$ = 0.025, also lengthens
the evolution time. Compared to calculations described in Kenyon & Bromley
(2004b), improvements in our treatment of interactions among particles in
neighboring annuli reduce lags by a factor of two, from $\sim$ 20% to $\sim$
10%. Combining the lags for mass spacing and radial resolution, our evolution
timescales are $\sim$ 20% to 30% longer than the actual evolution times. This
lag is comparable to the dispersion in timescales derived from multiple
calculations with identical starting conditions. Thus, improvements in
resolution are unlikely to alter our results significantly.
#### 3.3.1 Dynamical interactions
The coagulation algorithm begins to break down when (i) a few large objects
contain most of mass in the grid and (ii) the gravity of these objects
dominates the stirring. For $r\lesssim$ 500–1000 km, the largest objects
contain a small fraction of the mass in an annulus; individual dynamical
interactions are much smaller than the Fokker-Planck stirring rates. Thus,
kinetic theory yields good estimates for collisions and stirring among small
objects. As objects grow beyond $\sim$ 1000 km, however, both assumptions of
our statistical approach begin to fail: (i) the collisional cascade removes
leftover planetesimals, increasing the fraction of mass in the largest objects
and (ii) individual interactions among the largest objects in an annulus can
deflect large objects into neighboring annuli, raising collision and stirring
rates significantly. With $\sim$ 100–1000 Pluto-mass objects at 30–150 AU (see
Tables 2–3), interactions among the largest objects could play a significant
role in the late-time evolution of our models.
Dynamical interactions among an ensemble of Pluto-mass planets occur when the
radial spacing is $\Delta a\sim 2\sqrt{3}r_{H}$, where $r_{H}$ is the Hill
radius in Eq. (22). For planets with mass $M_{p}\sim 6\times 10^{24}$ g and
$M_{\star}$ = 1 M⊙, $r_{H}$ = 0.001 $a$. Thus, dynamical interactions among
the largest objects are inevitable when $n_{P}\approx$ 50–100 in a region with
a radial extent $\Delta a/a\approx$ 0.2 (Goldreich, Lithwick, & Sari, 2004;
Kenyon & Bromley, 2006). Many of our calculations yield such large numbers of
Pluto-mass objects.
To save computer time, we did not calculate the typical long-term evolution of
hundreds of Plutos using our hybrid evolution code (Bromley & Kenyon, 2006).
However, we can infer the outcome from scaling the results of calculations for
the formation of the Earth at 1 AU (e.g., Chambers, 2001; Bromley & Kenyon,
2006; Kenyon & Bromley, 2006). At 0.4–2 AU, dynamical evolution of 100–200
lunar mass objects produces several Earth-mass planets in 10–30 Myr. The
spacing of lunar mass objects in these calculations is $\sim$ 70% of the
critical spacing $\Delta a\sim 2\sqrt{3}r_{H}$, similar to the spacing of
Pluto-mass objects at late times in our calculations at 30–150 AU. Scaling the
evolution times by the ratio of orbital periods suggests that 100–200 Pluto
mass objects collide to form planets with masses $\sim$ 0.1 M⊕ on 1–3 Gyr
timescales (see also Levison & Stewart, 2001; Goldreich, Lithwick, & Sari,
2004).
This analysis suggests that dynamical interactions between large numbers of
Plutos at 30–150 AU are interesting only for low mass stars. For 2–3 M⊙ stars,
the main sequence lifetimes are shorter than the time needed for Plutos to
interact, to collide, and to grow into Mars-mass planets. For lower mass
stars, several test calculations with our hybrid code confirm that ensembles
of 100–200 Plutos can grow into several Mars-mass planets on timescales of 2–5
Gyr444For $a\gtrsim$ 75 AU, the escape velocity of Mars-mass planets exceeds
the orbital velocity. Although dynamical interactions among Mars-mass objects
can produce ejections in these circumstances (Goldreich, Lithwick, & Sari,
2004), damping by leftover planetesimals limits ejections in our simulations..
Although this evolution leads to some extra stirring of leftover low mass
planetesimals, there are only small changes in the dust production rate and
the total mass in small grains. Thus, dynamical interactions among Plutos have
little impact on our general results.
#### 3.3.2 Fragmentation parameters
Fragmentation is another uncertainty in our calculations. We treat destructive
collisions with an energy-scaling algorithm that uses (i) the ratio of the
center of mass collision energy $Q_{c}$ to the critical disruption energy
$Q_{d}^{*}$ and (ii) a simple power-law size distribution to apportion ejected
material into lower mass bins. Although the energy-scaling algorithm yields a
reasonable treatment of collisions in coagulation calculations, the disruption
energy $Q_{d}^{*}$ sets the strength of the collisional cascade. Large
$Q_{d}^{*}$ leads to a weak cascade with little debris; small $Q_{d}^{*}$
allows a strong cascade with significant debris. Because $Q_{d}^{*}$ and the
size distribution of the ejecta set the amount of material in small grains, we
now discuss how our choices for these input parameters affect our results.
Detailed comparisons of various approaches suggest that the size distribution
of the ejected mass has little impact on our results. For the large collision
rates in our calculations, all methods for dividing ejected mass among lower
mass bins – including dividing the ejected mass equally among 2–3 lower mass
bins – leads to a power-law mass distribution with a characteristic slope of
$dn/dm\approx-0.8$ (Dohnanyi, 1969; Williams & Wetherill, 1994). Thus, the
adopted mass distribution for the ejecta is relatively unimportant.
Kenyon & Bromley (2005) and Kenyon et al. (2008) describe how the form of
$Q_{d}^{*}$ in Eq. (5) impacts collisional evolution of icy objects at 30–150
AU. Here, we divide $Q_{d}^{*}$ into a bulk component (the first term of Eq.
(5)) and a gravity component (the second term of Eq. (5)). Gravity provides
nearly all of the binding energy for large objects with $r\gtrsim$ 10 km; the
bulk component of $Q_{d}^{*}$ provides most of the binding energy for small
objects with $r\lesssim$ 1–10 km. For icy objects with maximum sizes
$r_{max}\sim$ 2000 km, stirring never leads to orbital motions large enough to
disrupt objects with $r\gtrsim$ 10–20 km. Thus, our choices for the gravity
component of $Q_{d}^{*}$ have little impact on our results. Although both
components of the bulk strength – $Q_{b}$ and $\beta_{b}$ – contribute to
$Q_{d}^{*}$, quoted uncertainties in $\beta_{b}$ derived from theoretical
simulations lead to unimportant variations in $Q_{d}^{*}$ as a function of
$r$. Thus, we concentrate on $Q_{b}$.
To quantify the impact of $Q_{b}$ on our results, we first consider the
evolution of the dust production rate and the amount of material in large and
small grains. During runaway growth, the dust production rates for models with
$Q_{b}\leq 10^{4}$ erg g-1 are 10%–20% larger than dust production rates for
models with $Q_{b}\geq 10^{5}$ erg g-1. At the same time, the total mass in
large and small grains is $\sim$ 10 times larger for models with small $Q_{b}$
than for models with large $Q_{b}$. In both cases, models with the smallest
initial disk mass have the largest differences as a function of $Q_{b}$.
During oligarchic growth, these differences disappear. For models with
$Q_{b}=1-10^{6}$ erg g-1, the dispersion in dust production rates near the
time of maximum dust production is $\sim$ 5% or less for all disks around 1–3
M⊙ stars. Although the dispersion in the total mass in large and small grains
is a factor of $\sim$ 3 during the early stages of oligarchic growth, the
dispersion in dust masses declines to 10% or less at late times when the dust
masses reach their maximum values (Kenyon & Bromley, 2004b).
The time variation in dust production rate and total dust mass as a function
of $Q_{b}$ has a simple physical origin (see also Kenyon et al., 2008). During
runaway and oligarchic growth, the collision energies of small objects scale
with the mass of the largest objects in the grid (see also Goldreich,
Lithwick, & Sari, 2004). Thus, small objects have larger and larger collision
energies $Q_{c}$ at later and later evolution times. Because this feature of
the evolution depends only on gravitational stirring, it is independent of
$Q_{b}$. Throughout the evolution, the mass ejected in a collision scales with
$Q_{c}/Q_{b}$ (Eq. (4)). Thus, calculations with small $Q_{b}$ eject more
material at early times compared to models with large $Q_{b}$, leading to a
large dispersion in the dust production rate and total dust masses early in
the evolution. At late times, all calculations produce objects with
$r_{max}\approx$ 1500–2000 km. These large objects stir all leftover small
planetesimals to large random velocities, where the collision energies
$Q_{c}\gg Q_{d}^{*}$ for all $Q_{b}$. All collisions then lead to copious mass
loss, which eliminates the dispersion in dust production rates and total dust
masses at late times (see also Kenyon & Bromley, 2004b).
In addition to the small late-time dispersion in dust production rates and
total dust masses, our results yield negligible differences in the late-time
fractions of mass in large objects ($r\gtrsim$ 100 km) as a function of
$Q_{b}$. The median radius of the largest object and the median number of
Pluto mass objects are also independent of $Q_{b}$. Thus, our analysis
suggests that the fragmentation parameters have a small impact on observable
quantities. For low mass disks at $t\lesssim$ 10 Myr, destructive collisions
between planetesimals with small $Q_{b}$ produce more dust than objects with
large $Q_{b}$. Although these differences are probably large enough to be
observable, they disappear at late times when planets reach their maximum
sizes.
#### 3.3.3 Treatment of small particles
Our algorithm for deriving the evolution of small particles with
$r_{min}\lesssim$ 1 m is a final uncertainty in our calculations. To follow
the evolution of sizes and orbits for large objects in a reasonable amount of
computer time, we do not calculate the evolution of small particles directly.
Instead, we use the known production rate of small particles from the detailed
calculation $\dot{M}(t)$, an adopted power-law size distribution, and a simple
collision algorithm to evolve the small particle size distribution with time.
Because we include radiation pressure and Poynting-Robertson drag in this
simple treatment of collisional evolution, our predicted size distributions
consist of three distinct pieces. For particle sizes where the collisional
timescale is shorter than the timescale for Poynting-Robertson drag, $n\propto
r^{-0.8}$. For very small sizes where radiation pressure ejects grains, we
adopt $n\propto r^{-0.8}$ for grains in a constant velocity outflowing wind
(see the Appendix). For intermediate sizes, Poynting-Robertson drag can remove
grains faster than collisions replenish them. Thus, the particle number
$n\rightarrow$ 0\. To conserve mass, we solve a continuity equation to derive
the number density of grains dominated by Poynting-Robertson drag.
Although our solution for the evolution of small particles is efficient, it
does not consider how fluctuations in the collision and fragmentation rates
might modify the size distribution. Campo Bagatin et al. (1994) note that
size-dependent fluctuations can produce wavy size distributions for 0.1–10 mm
particles. In their simulations of the $\beta$ Pic disk, Thébault, Augereau, &
Beust (2003) derive steady-state size distributions with substantial deficits
of 0.1–10 mm particles compared to a standard $n\propto r^{-0.8}$ power-law
(see also Krivov et al., 2006; Thébault & Augereau, 2007; Löhne et al., 2008).
If these deficits are typical, then our algorithm seriously overestimates the
mass in small dust grains and thus the infrared fluxes of debris disks.
To check for this possibility, we computed several models with a simple
version of our multiannulus coagulation code. In these tests, we extracted a
complete disk model near the peak of the collisional cascade, extended the
lower end of the size distribution from $r_{min}$ = 1 m to 1 $\mu$m using a
power law $n\propto r^{-\alpha}$, and continued the calculation for $\sim$ 100
Myr with collisions and Poynting-Robertson drag but without our Fokker-Planck
velocity evolution. To estimate the range of errors in our simple algorithm,
we varied the power law exponent for the size distribution, $\alpha\approx$
0.6–1, the power law exponent for the fragmentation law, $\beta_{b}\approx$
-0.5–0, and the magnitude of the bulk strength $Q_{b}$ = 1–$10^{5}$ erg g-1.
For a range of disk masses around a 2 M⊙ star, this approach provides a
straightforward estimate for the accuracy of our results for small particles.
These tests confirm that the simple collision algorithm yields results
reasonably close to more detailed coagulation calculations. For models with
$\beta_{b}\approx$ 0, $Q_{b}\gtrsim$ $10^{3}$ erg g-1, and $\alpha\approx$
0.6–1, the derived size distributions are within $\sim$ 20% of those predicted
by the simple model for all particles with $r\approx$ 0.01–100 mm. Although
calculations with $Q_{b}\lesssim 10^{2}$ erg g-1 yield larger deviations from
the simple model, these are small compared to those quoted by Thébault &
Augereau (2007). Because particles with small $\beta_{b}$ are harder to
fragment, calculations with $\beta_{b}\lesssim-0.25$ tend to produce smaller
departures for a power law size distribution than those with
$\beta_{b}\gtrsim-0.25$.
Several features of our calculations combine to minimize wavy size
distributions for small particles in disks at 30–150 AU. Because icy planet
formation is inefficient, the collisional cascade begins when most of the
initial disk mass is in 1–10 km planetesimals. Fragmentation of the leftovers
leads to a very large production rate of 1 m and smaller objects. Continued
fragmentation of these objects tends to wash out wavy size distributions
produced by a low mass cutoff (Campo Bagatin et al., 1994; Thébault, Augereau,
& Beust, 2003). In our Fokker-Planck treatment of velocity evolution, leftover
planetesimals are also in dynamical equilibrium with larger protoplanets that
are ‘safe’ from fragmentation. Thus, the dust production rate from the
collisional cascade is well-matched to the dynamical state of the system and
tends to sustain a power-law size distribution for the smallest objects.
### 3.4 Highlights of Icy Planet Formation Around 1–3 M⊙ Stars
Starting with a disk of 1 km planetesimals, icy planet formation at 30–150 AU
follows the same path for all 1–3 M⊙ stars. This evolution has six main
features.
* •
It takes 5–30 Myr for runaway growth to produce an ensemble of oligarchs with
radii of 500–1000 km. Throughout runaway growth, oligarchs stir up the orbits
of leftover planetesimals. Collisions between leftover planetesimals produce
more and more debris.
* •
From $\sim$ 10 Myr to the main sequence turnoff, planets slowly grow to a
characteristic radius. For a broad range of input parameters, the maximum size
of an icy planet is $\sim$ 1750 km at 30–150 AU. Because the timescale for
planet formation at 100–150 AU is similar to the main sequence lifetime of a
1–3 M⊙ star, the inner disk contains more 1500–2000 km planets than the outer
disk.
* •
As planets grow slowly, a collisional cascade grinds leftover planetesimals to
dust. Early on, radiation pressure ejects the smallest grains in an outflowing
wind. Later, Poynting-Robertson drag also removes larger grains from the disk.
In our calculations, radiation pressure removes roughly twice as much mass
from the disk as Poynting-Robertson drag. The timescale for the collisional
cascade to remove leftover planetesimals is close to the main sequence
lifetime of the central star. Thus, the cascade removes more material from the
inner disk than from the outer disk.
* •
Icy planet formation is inefficient. In our calculations, icy planets with
radii exceeding 1000 km contain $\lesssim$ 3–4% of the initial mass in solid
material. Objects with radii $\sim$ 100–1000 km contain $\sim$ 2–3% of the
initial mass. Because short stellar lifetimes limit the growth of planets in
the outer disk, the mass in large objects declines linearly with increasing
distance from the central star. Thus, the inner region of the disk contains
many more Pluto-mass objects than the outer region.
* •
The dust produced by the collisional cascade is observable. For disks around
1–3 M⊙ stars, the maximum mass in small dust grains with radii of 1 $\mu$m to
1 mm is $\sim$ 1–2 lunar masses. This mass is comparable to the masses derived
for the most luminous debris disks around A-type and G-type stars. The time
evolution of the dust production rate and the mass in small dust grains
suggest the dust luminosity declines with time.
* •
Dusty debris is a signature of the formation of a planetary system. This
debris is present throughout the lifetime of the central star.
## 4 DEBRIS DISK EVOLUTION
To convert our derived size distributions into observable quantities, we
perform a radiative transfer calculation. For each evolution time $t$, we
derive the luminosity $L_{\star}$ and effective temperature $T_{\star}$ of the
central star from the $Y^{2}$ stellar evolution models (Demarque et al.,
2004). We then compute the optical depth $\tau(a)$ of each annulus in our
model grid. The optical depth allows us to derive the fraction of the stellar
luminosity $L_{\star}$ absorbed by each annulus. For each grain size in each
annulus, we derive an equilibrium grain temperature $T(r,a)$ and an emitted
spectrum. Summing the emitted spectra over $r$ and $a$ yields the predicted
spectral energy distribution (SED) and the total dust luminosity $L_{d}$ as a
function of time. The Appendix describes this calculation in more detail (see
also Kenyon & Bromley, 2004b).
In our calculation of observable quantities, the most important input
parameters are the smallest stable grain size $r_{2}$ (also known as the
‘blowout’ radius; see Backman & Paresce 1993) and the slope $q$ of the
emissivity law for small grains. Although several estimates for the minimum
grain size suggest $r_{2}\approx 0.5-2M_{\star}^{3}$ $\mu$m for 1–3 M⊙ stars
(e.g. Burns, Lamy, & Soter, 1979; Artymowicz, 1988; Backman & Paresce, 1993;
Kimura et al., 2002), the coefficient and the scaling relation are sensitive
to the composition, internal structure, and radiative properties of the
grains. Because observations allow few tests of this relation, we adopt
$r_{2}$ = 1 $\mu$m for all stars. If more luminous stars have larger $r_{2}$,
our calculations overestimate the optical depth in small grains. Thus, we
overestimate the mid-IR and submm excesses. For the emissivity, submm data
suggest $q\approx$ 0.6–1 from a handful of debris disks (Najita & Williams,
2005; Williams & Andrews, 2006). To provide some balance for our likely
underestimate of $r_{2}$, we adopt $q$ = 1. Grains with smaller $q$ emit more
efficiently at longer wavelengths; our models then underestimate mid-IR and
submm excesses.
To describe the evolution of observable quantities with time, we focus on the
dust luminosity $L_{d}$ and the excesses at IR and submm wavelengths. The
fractional dust luminosity $L_{d}/L_{\star}$ provides a clear measure of the
relative luminosity of the debris disk. For excesses at specific wavelengths,
we quote the total emission of the disk and the central star relative to the
emission from the stellar photosphere, $F_{\lambda}/F_{\lambda,0}$. With this
definition, disks that produce no excess have $F_{\lambda}/F_{\lambda,0}$ = 1;
disks where the excess emission is comparable to the emission from the central
star have $F_{\lambda}/F_{\lambda,0}$ = 2.
We begin this section with a discussion of excess emission for 1 M⊙ stars.
After discussing results for 1.5–3 M⊙ stars, we conclude this section with a
brief summary. To facilitate comparisons of our results with observations,
Tables 4–8 list results for the fractional dust luminosity and excesses at
24–850 $\mu$m. The paper version lists the first five lines of results for
$x_{m}$ = 1/3, 1, and 3. The electronic version includes all results for these
$x_{m}$.
### 4.1 Evolution for 1 M⊙ stars
Fig. 14 shows the evolution of the fractional dust luminosity
$L_{d}/L_{\star}$ for disks surrounding a 1 M⊙ star. Early in the evolution,
collisions produce mergers instead of debris. For an ensemble of growing
planetesimals, the surface area per unit mass (and hence the opacity)
decreases with time. Thus, $L_{d}/L_{\star}$ declines with time. Less massive
disks have smaller dust masses and smaller dust luminosities. As oligarchic
growth begins, the dust luminosity rises rapidly and reaches a peak
$L_{d}/L_{\star}\approx 2\times 10^{-3}$ in 30–100 Myr. More massive disks
reach larger peak luminosities earlier than less massive disks. At late times,
all disks converge to the same dust luminosity, $L_{d}/L_{\star}\approx
10^{-4}$ at $\sim$ 10 Gyr.
Despite their small fractional dust luminosities, these disks produce large
excesses at 70 $\mu$m (Fig. 15; left panel). For massive disks with $x_{m}$ =
2–3, the 70 $\mu$m excess rises from $F_{70}/F_{70,0}$ $\sim$ 2–3 at 3 Myr to
$F_{70}/F_{70,0}$ $\sim$ 30–50 at 30 Myr. Lower mass disks with $x_{m}$ = 1/3
to 1/2 produce smaller peak excesses at later times, $F_{70}/F_{70,0}$ $\sim$
10 at $\sim$ 100 Myr. For all disk masses, the 70 $\mu$m excess is close to
its maximum value for a short period when planet formation peaks in the inner
disk. The excess then declines with time. The rapid decline leads to modest
excesses at late times, $F_{70}/F_{70,0}$ $\sim$ 3–5 at $\sim$ 1 Gyr and
$F_{70}/F_{70,0}$ $\sim$ 2 at $\sim$ 3–10 Gyr.
The large excesses at 70 $\mu$m are a simple consequence of blackbody
radiation from small grains at 30–50 AU around a solar-type star. With typical
temperatures $\sim$ 40–60 K, these grains emit most of their radiation at
$\sim$ 50–70 $\mu$m. The peak flux from a blackbody grain at temperature T is
$F_{\lambda,max}\propto T^{5}$ (Allen, 1976). Setting the total disk
luminosity $L_{d}\propto T^{4}$ yields $F_{\lambda,max}\propto L_{d}T$. At
this wavelength, the stellar flux follows a Rayleigh-Jeans law,
$F_{\lambda}\propto T_{\star}\lambda^{-4}$. Combining these relations and
including a correction factor for inefficient radiation from small grains
yields a simple relation for the 70 $\mu$m flux from the disk and central
star,
$F_{70}/F_{70,0}\approx 1+10\left(\frac{L_{d}/L_{\star}}{10^{-3}}\right)~{}.$
(54)
For the luminosities in Fig. 14, this relation accounts for the 70 $\mu$m
excesses in Fig. 15 at all times.
At longer wavelengths, the disks in our calculations achieve larger peak
excesses and stay close to the peak excess for longer periods of time (Fig.
15; right panel). Disks reach their peak excesses at 850 $\mu$m on timescales
similar to those at 70 $\mu$m, $\sim$ 30 Myr for disks with $x_{m}$ = 2–3 and
$\sim$ 100 Myr for disks with $x_{m}$ = 1/3. The fractional excesses at 850
$\mu$m are a factor of $\sim$ 2 larger than the excesses at 70 $\mu$m. Because
the emitting region evolves more slowly, these disks are luminous for $\sim$ 1
Gyr and then decline with time. Despite the rapid decline, the excesses are
significant at late times, with $F_{850}/F_{850,0}$ $\sim$ 3–10 at $\sim$ 3–10
Gyr.
The time variation of IR excess also depends on the outer radius of the disk.
For solar mass stars, grains at 30–50 AU in the inner disk produce most of the
flux at 50–100 $\mu$m. Thus, disks with outer radii of 70 AU and 150 AU
produce similar 70 $\mu$m excesses for $t\lesssim$ 30 Myr (Fig. 16; left
panel). Once the smaller disk reaches peak emission, the 70 $\mu$m excess
begins a dramatic decline. The larger disk maintains the peak excess for
$\sim$ 30 Myr and then declines more slowly with time. For $t\gtrsim$ 100 Myr,
the smaller disk is a factor of 2–3 fainter at 70 $\mu$m than the larger disk.
The evolution of disks with different sizes is more dramatic at 850 $\mu$m
(Fig. 16; right panel). For typical grain temperatures $\sim$ 20–60 K, long
wavelength emission from the disk follows the Rayleigh-Jeans tail of a set of
blackbodies. The radiation from each disk annulus is then $\propto a^{2}T$.
Because the outer disk produces more long wavelength emission than the inner
disk, the 850 $\mu$m excess scales with the outer disk radius. For $t\gtrsim$
1 Gyr, we derive $F_{850}/F_{850,0}\propto$ $a_{out}^{n}$, where $a_{out}$ is
the outer radius of the disk and $n\approx$ 3–4. Thus, doubling the outer disk
radius increases the predicted 850 $\mu$m excess by a factor of $\sim$ 10 at
late times.
### 4.2 Evolution for 1.5-3 M⊙ stars
Several factors change the evolution of the dust luminosity and the IR/submm
excesses in stars more massive than 1 M⊙. More massive stars are hotter; for
the Y2 stellar evolution isochrones $T_{\star}\propto M_{\star}$ (Demarque et
al., 2004). Thus, grains in the inner disks around massive stars are warmer,
emit more short wavelength radiation, and produce bluer colors than disks
around less massive stars. More massive stars also evolve faster,
$t_{ms}\propto M_{\star}^{-3}$. Because the evolutionary timescales for solids
in the disk are much less sensitive to stellar mass, $t\propto
M_{\star}^{-1}$, massive stars have more dust at the end of their main
sequence lifetime than low mass stars (e.g., Fig. 13). Thus, these systems
have relatively large IR excesses when their central stars evolve off the main
sequence.
To compare the evolution of dust emission in debris disks around 1–3 M⊙ stars,
we begin with the evolution of the dust luminosity (Fig. 17). Planets grow
faster around more massive stars; thus, the dust luminosity rises earlier for
more massive stars. Once the collisional cascade begins, the timescale to
reach the peak luminosity depends only on the initial disk mass and the
stellar mass,
$t_{d,max}\approx
25~{}x_{m}^{-2/3}~{}\left(\frac{2~{}M_{\odot}}{M_{\star}}\right){\rm Myr}.$
(55)
This timescale is similar to the timescale required to produce the first
Pluto-mass object in the inner disk (Eq. (41)). The peak luminosity depends
only on the initial disk mass
$L_{d,max}/L_{\star}\approx 2\times 10^{-3}x_{m}~{}.$ (56)
The luminosity remains close to the peak for $\sim$ 10–30 Myr and then
declines with time.
Stellar evolution has a clear impact on the evolution of the dust luminosity.
For 1 M⊙ stars, the dust luminosity declines by a factor of $\sim$ 20 before
the star evolves off the main sequence (Fig. 14). For 3 M⊙ stars, the typical
decline in $L_{d}/L_{\star}$ is only a factor of $\sim$ 4\. Because debris
disks have roughly the same peak luminosities, an ensemble of debris disks
around middle-aged low mass stars should be systematically less luminous than
disks around middle-aged high mass stars.
Stellar physics also produces dramatic differences in the behavior of the 24
$\mu$m excess with stellar mass (Fig. 18; lower left panel). At 30–50 AU, the
grain temperatures range from $\sim$ 40–60 K for 1 M⊙ stars to $\sim$ 80–120 K
for 3 M⊙ stars. For these temperatures, radiation at 24 $\mu$m is on the Wien
side of the blackbody peak and thus varies exponentially with temperature. Our
calculations for 1 M⊙ stars produce very little 24 $\mu$m radiation from
material at 30–150 AU. However, the peak 24 $\mu$m excesses reach
$F_{24}/F_{24,0}\sim$ 20 for disks around 3 M⊙ stars. For all 1–3 M⊙ stars,
our results yield
${\rm log}~{}F_{24,max}/F_{24,0}\approx
0.74~{}(M_{\star}-1M_{\odot})+0.27~{}(M_{\star}/M_{\odot})~{}{\rm
log}~{}x_{m}~{}.$ (57)
This maximum flux occurs at roughly the same time as the peak dust luminosity.
At longer wavelengths, the excesses are less sensitive to stellar mass.
Radiation at 70 $\mu$m is at the blackbody peak for grains in the inner disk.
Thus, the inner disk produces most of the 70 $\mu$m excess. The peak excess is
then independent of the stellar luminosity and depends only on the total disk
mass (Fig. 18; upper left panel),
$F_{70,max}/F_{70,0}\approx
55~{}x_{m}^{0.90}\left(\frac{M_{\star}}{2~{}M_{\odot}}\right)~{}.$ (58)
For grains at 30–150 AU, radiation at longer wavelengths is on the Rayleigh-
Jeans tail of the blackbody. Thus, observations at 160–850 $\mu$m probe
material throughout the disk. At 160 $\mu$m, extra emission from hotter grains
in disks around more massive stars is balanced by more flux from the hotter
central star. Thus, the excess is independent of stellar mass and depends only
on $x_{m}$ (Fig. 18; upper right panel),
$F_{160,max}/F_{160,0}\approx 65~{}x_{m}^{0.90}~{}.$ (59)
At 850 $\mu$m, grains in disks around 1 M⊙ stars are closer to their blackbody
peaks than grains in disks around more massive stars. Thus, the 850 $\mu$m
excesses are larger for 1 M⊙ stars (Fig. 18; lower right panel),
$F_{850,max}/F_{850,0}\approx\left\\{\begin{array}[]{l
l}40~{}x_{m}^{0.9}&~{}~{}~{}~{}~{}M_{\star}=1~{}M_{\odot}\\\ \\\
25~{}x_{m}^{0.9}&~{}~{}~{}~{}~{}M_{\star}=1.5-3.0~{}M_{\odot}\\\
\end{array}\right\\}$ (60)
At 70–850 $\mu$m, the time of peak excess is similar to the maximum in the
dust luminosity. Thus, all excesses at 24–850 $\mu$m peak at $\sim$ 20–30 Myr
for 1–3 M⊙ stars.
Following the peak in the excess at 20–30 Myr, the relative disk luminosity
and the excesses at 24–850 $\mu$m decrease monotonically with time. For this
evolution, simple debris disk models predict a power law decline,
$L_{d}/L_{\star}\propto t^{-n}$ with $n\approx$ 1–2 (e.g. Dominik & Decin,
2003; Wyatt et al., 2007a, b). To compare our results for $t\gtrsim t_{d,max}$
with these predictions, we adopt
$f_{d}\equiv L_{d}/L_{\star}\propto t^{-n_{d}}$ (61)
and
$f_{\lambda}\equiv F_{\lambda}/F_{\lambda,0}\propto t^{-n_{\lambda}}$ (62)
and derive the power law exponents $n_{d}=d~{}{\rm log}~{}f_{d}/d~{}{\rm
log}~{}t$ and $n_{\lambda}=d~{}{\rm log}~{}f_{\lambda}/d~{}{\rm log}~{}t$ from
all of our calculations as a function of disk mass, stellar mass, and time.
Throughout the evolution of all our debris disk models, $n_{d}$ changes
continuously with time. For $t\gtrsim t_{d,max}$, collisions and radiation
pressure dominate the removal of small grains. As collision rates slowly
decline with time, the exponent increases slowly from $n_{d}\approx$ 0 to
$n_{d}\approx$ 1\. When the central star approaches the end of its main
sequence lifetime, Poynting-Robertson drag starts to dominate collisions. The
disk luminosity then decreases rapidly; $n_{d}$ increases from $\sim$ 1 to
$\sim$ 2\. Because most systems are collisionally-dominated, our calculations
yield a typical $n_{d}\approx$ 0.6–0.8.
For $\lambda\approx$ 24–850 $\mu$m, the exponents $n_{\lambda}$ follow the
evolution of $n_{d}$. Because collision rates are larger in the warmer, inner
disk than in the colder outer disk, $n_{\lambda}$ increases slowly with
$\lambda$. Thus, the typical $n_{24}\approx$ 0.6–0.8 is smaller than the
typical $n_{850}\approx$ 0.8–1.0.
The exponents $n_{d}$ and $n_{\lambda}$ are somewhat sensitive to the disk
mass and the stellar mass. At fixed stellar mass, more massive disks evolve
faster. Thus, $n_{d}$ changes faster for more massive disks and is larger at
the main sequence turnoff. For fixed disk mass, lower mass stars live longer
and have more time to reach the Poynting-Robertson drag-dominated regime. Our
results suggest a 0.1–0.2 range in $n_{d}$ and $n_{\lambda}$ for a factor of
10 range in $x_{m}$ and a factor of 3 range in stellar mass.
In addition to excesses at specific wavelengths, the evolution of color
excesses yield interesting trends with stellar mass and time. Because the 24
$\mu$m excess is sensitive to stellar mass, the [24]–[70] color cleanly
distinguishes debris disks around stars of different masses (Fig. 19). For 2–3
M⊙ stars, [24]–[70] rises rapidly to [24]–[70] $\approx$ 1–2 at $\sim$ 1 Myr
and then rises slowly throughout the main sequence lifetime of the central
star. For 1–1.5 M⊙ stars, the color rises later, reaches [24]–[70] $\approx$
3–4 at 30–100 Myr, and then declines slowly.
For disks at 30–150 AU, the variation of [24]–[70] with $M_{\star}$ depends
solely on the properties of the central star. Because more massive stars are
hotter, their disks are warmer. Warmer disks produce bluer colors. Thus, the
peak [24]–[70] scales with $M_{\star}$.
The mass-dependent color evolution of debris disks at 30–150 AU suggests that
color-color diagrams can discriminate masses of the central star. In Fig. 20,
color-color tracks for scaled MMSN around 1.5 M⊙ stars are clearly distinct
from tracks for scaled MMSN around 2 M⊙ and 3 M⊙ stars. In Fig. 21, tracks for
a range of disks around 2 M⊙ stars define a triangle-shaped locus distinct
from the tracks for 1.5 M⊙ and 3 M⊙ stars.
To establish a triangular debris disk locus for each stellar mass, we define
two vectors. Adopting a vertex, x0,y0, the upper boundary of the locus is a
vector connecting the vertex with an upper point, xu,yu. The lower boundary is
a second vector connecting the vertex with a lower point, xl,yl. Table 9 lists
our results for the vertex and the upper/lower points as a function of stellar
mass. For each stellar mass, colors for debris disks at 30–150 AU lie within
the area defined by the two vectors. More massive disks produce redder colors.
Within each locus, the initial disk mass scales with distance from the vertex.
When dust inside $\sim$ 30 AU produces a small IR excess, this color-color
diagram provides a useful discriminant of stellar mass. For disks at 30–150 AU
around 1–3 M⊙ stars, the typical [5.8]–[8] color is small, with [5.8]–[8]
$\lesssim$ 0.1 at all times. Predicted colors for terrestrial debris disks are
much larger. For 3 M⊙ (1.5 M⊙) stars, we predict maximum colors [5.8]–[8]
$\sim$ 0.5–1 (0.2–0.5) (e.g., Kenyon & Bromley, 2004a, 2005). Thus, mid-IR
color-color diagrams are useful diagnostics of the outer disk for [5.8]–[8]
$\lesssim$ 0.1.
### 4.3 Summary
Planet formation and stellar evolution combine to produce several robust
trends in the time evolution of the dust luminosity and IR/submm excesses from
debris disks around 1–3 M⊙ stars.
For scaled MMSN, the maximum dust luminosity is $L_{d,max}\sim 2\times
10^{-3}$. For an ensemble of debris disks, the range in the peak dust
luminosity scales with the initial mass of solid material in the disk. The
dust luminosity reaches this peak at roughly the time when the first Pluto-
mass objects form at 30–50 AU. Following this peak, the luminosity declines as
$t^{-n_{d}}$ with $n_{d}\approx$ 0.6–0.8. Because lower mass stars have longer
main sequence lifetimes, debris disks around lower mass stars reach smaller
fractional dust luminosities at late times.
The IR/submm excesses from debris disks at 30–150 AU are sensitive to the mass
of the central star. At 24 $\mu$m, disks around more massive stars produce
larger excesses; disks around stars with $M_{\star}\lesssim$ 1 M⊙ produce
negligible excesses at 24 $\mu$m. At 70 $\mu$m, the excess is a simple
function of the total dust luminosity, $F_{70}/F_{70,0}\approx$ 1 +
$10^{4}L_{d}/L_{\star}$. At 850 $\mu$m, debris disks around 1 M⊙ stars produce
larger peak excesses than disks around more massive stars. At late times,
however, the typical 850 $\mu$m excess is fairly independent of stellar mass,
with $F_{850}/F_{850,0}\approx$ 3–5 for stars with ages $t\sim t_{ms}$.
Among stars with different masses, mid-IR colors provide a sensitive
discriminant of debris disk evolution when the [5.8]–[8] color is small (Fig.
19–21). For 2–3 M⊙ stars, [24]–[70] slowly becomes redder with the age of the
central star; for 1–1.5 M⊙ stars, [24]–[70] rises more rapidly, remains at
peak color for 300 Myr to 1 Gyr, and then declines rapidly with time. For all
stars, [8]–[24] and [24]–[70] correlate with stellar mass. Debris disks around
2–3 M⊙ (1–2 M⊙) stars have redder (bluer) [8]–[24] colors and bluer (redder)
[24]–[70] colors. Thus, an [8]–[24] vs [24]–[70] color-color diagram provides
a way to analyze debris disks around stars with different masses (Table 9).
## 5 APPLICATIONS
To test whether our predictions provide a reasonable match to observations, we
now consider several applications of our models to real systems. For these
calculations, the broad trends in the evolution of IR excesses and colors are
sensitive to the physics of planet formation and the collisional cascade.
Thus, our main goal is to compare our results with observed trends of excesses
and colors for large samples of main sequence stars observed with the IRAS,
ISO, and Spitzer satellites. In addition to long-term trends, the absolute
level of the excesses depends on $r_{2}$ and $q$. Thus, our second goal is to
learn whether our assumptions yield mid-IR and submm excesses similar to those
observed.
We begin with an analysis of Spitzer data for the prototypical debris disk,
Vega. After demonstrating that our models can explain the mid-IR fluxes and
morphology of Spitzer images for this system, we show that our predictions
provide a good match to observations of mid-IR excesses for a sample of A-type
stars (Rieke et al., 2005; Su et al., 2006) and a sample of solar-type stars
(Beichman et al., 2006; Hillenbrand et al., 2008).
### 5.1 The Vega Disk
Observations of Vega with IRAS first revealed a large excess of emission above
the A-type photosphere for wavelengths exceeding 12 $\mu$m (Aumann et al.,
1984). The best-fitting single temperature blackbody to the IRAS data yields a
temperature of $\sim$ 85 K, a fractional luminosity of $\sim 2.5\times
10^{-5}$ relative to the central star, and a radius of $\sim$ 150–200 AU for
the emitting material. Because the lifetime for small grains at 150–200 AU is
much shorter than the age of Vega, Aumann et al. (1984) concluded that the
grains have sizes larger than 1 mm. Thus, Vega provided the first direct
evidence for grain growth outside the Solar System.
Since the Aumann et al. (1984) discovery, Vega has become the prototypical
debris disk (e.g., Backman & Paresce, 1993; Artymowicz, 1997; Lagrange et al.,
2000). The debris consists of a bright torus with small-scale clumps at
80–1000 AU from the central star (Holland et al., 1998; Wilner et al., 2002;
Liu et al., 2004; Su et al., 2005) and a smaller disk of debris at $\sim$ 1 AU
from the central star (Absil et al., 2006). Dust in the small disk is hot
($\sim$ 1500 K), luminous ($L_{d}/L_{\star}$ $\sim 5\times 10^{-4}$), and
mostly confined to a narrow ring with a diameter of $\sim$ 0.5–1 AU (Absil et
al., 2006). This dust might be a result of collisions between larger objects
at 1 AU or grains lost from icy comets at 80–100 AU in the outer disk.
Recently, Su et al. (2005) analyzed high quality Spitzer images at 24, 70, and
160 $\mu$m. Their results demonstrate that the large-scale debris consists of
a bright ring at 80–200 AU and a smooth ‘halo’ that extends to $\sim$ 1000 AU
at 160 $\mu$m. The halo has an $a^{-2}$ radial density profile, consistent
with a wind of small grains ejected by radiation pressure. Fits to the radial
surface brightness profiles and the spectral energy distribution suggest the
grains in the wind have sizes of 1–50 $\mu$m and a total mass of
$M_{d,1-50}\sim 3\times 10^{-3}$ M⊕. The grains in the bright ring are larger,
with typical sizes of $\sim$ 240 $\mu$m, and have a total mass of
$M_{d,240}\sim 2\times 10^{-3}$ M⊕ (see also Marsh et al., 2006). For an
adopted residence time of $\sim 10^{3}$ yr in the wind, the mass in small
grains implies that larger grains in the ring produce smaller dust particles
at a rate of $\sim 10^{15}$ g s-1.
To check whether our model predictions can match the MIPS data for Vega, we
make a simple comparison with median results from several calculations. In
addition to the observed fluxes at 24, 70, 160, and 850 $\mu$m (Su et al.,
2005), we adopt published values for the age (200 Myr; Su et al., 2006),
luminosity (37 L⊙; Aufdenberg et al., 2006), and mass (2.3 M⊙; Aufdenberg et
al., 2006) of the central star. Su et al. (2005) separate the observed fluxes
into contributions from the debris disk and the central star. The first row of
Table 10 lists these results, along with their derived values for the mass in
1–50 $\mu$m dust grains and the dust production rate. The rest of Table 10
lists predictions for four of our debris disk models around 2–2.5 M⊙ stars.
The comparison in Table 10 suggests a reasonable match to the data.
Predictions for the total mass in 1–50 $\mu$m dust grains and the fluxes at
160 $\mu$m and at 850 $\mu$m bracket the observed values. Our models also
predict a bright ring in the 70–160 $\mu$m dust emission at 80–130 AU, close
to the observed position of the bright ring inferred from Spitzer images
(85–200 AU; Su et al., 2005). The mass of 0.1–1 mm particles in this ring,
$\sim 3-5\times 10^{-3}$ M⊕, also agrees with the mass in 240 $\mu$m grains
derived from the Spitzer data. However, our models overpredict the fluxes at
24 $\mu$m and at 70 $\mu$m by a factor of 2–10 and underpredict the dust
production rate by a similar factor.
To understand possible origins for the mismatches between the data and the
models, we consider the evolution of small grains in our calculations. When
the collisional cascade starts removing material from the disk, the most
destructive collisions involve grains with comparable masses. These collisions
gradually erode the parent objects and produce modest amounts of debris in
smaller particles. Because (i) the collision timescale is much shorter than
the timescale for Poynting-Robertson drag and (ii) the ratio of the radiation
force to the gravitational force is $\beta_{rad}\equiv F_{rad}/F_{grav}\propto
r^{-1}$, erosion continues until particles reach a size $r_{2}$ where
$\beta_{rad}\gtrsim$ 0.5–1 (Burns, Lamy, & Soter, 1979). Particles with
$r\lesssim r_{2}$ are ejected. For simplicity, we assume $r_{2}\approx$ 1
$\mu$m for all of our calculations.
In this picture for the collisional cascade, the Su et al. (2005) results
provide a simple solution for the overprediction of the 24–70 $\mu$m fluxes in
our calculations. If grains with $r_{2}\gg$ 1 $\mu$m are in the wind, our
‘Vega models’ underestimate the mass in the wind and overestimate the mass
left behind in the disk. More mass at larger distances from the central star
lowers the optical depth in the inner disk, reducing the predicted fluxes at
short wavelengths. Thus, increasing our adopted $r_{2}$ for Vega models should
provide a better match between observed and predicted fluxes at 24 $\mu$m and
at 70 $\mu$m. For an adopted $L_{\star}$ = 60 L⊙, Backman & Paresce (1993)
estimated $r_{2}$ = 14 $\mu$m. Scaling this result for our adopted $L_{\star}$
= 37 L⊙, $r_{2}\approx$ 8.5 $\mu$m. Several test calculations with $r_{2}=$ 10
$\mu$m yield predicted 24 $\mu$m and 70 $\mu$m fluxes close to the observed
values.
Reconciling the estimated dust production rate with our predictions requires a
more rigorous analysis of dust production and ejection in the Vega debris
disk. To derive the dust production rate, Su et al. (2005) assume that (i) the
240 $\mu$m grains in the bright ring are bound, with $\beta_{rad}\sim$ 0 and
residence times $\gg 10^{3}$ yr, and (ii) the 1–50 $\mu$m grains in the wind
are unbound, with $\beta_{rad}\geq$ 1 and residence times $\sim 10^{3}$ yr. If
the collisional cascade proceeds as a gradual erosion of larger objects into
smaller objects, however, we expect a more gradual transition from grains with
$\beta_{rad}\sim$ 0 to grains with $\beta_{rad}\sim$ 1 (see also Burns, Lamy,
& Soter, 1979; Artymowicz, 1988). Allowing the residence time to change
gradually from the bound 240 $\mu$m grains to the unbound 1 $\mu$m grains
provides a way to lower the apparent dust production rate and to resolve the
mismatch between our models and the observations.
To provide an alternate estimate for the residence time of grains in the Vega
disk, we consider the collision times in the ring and the wind. We adopt the
dust masses derived from the Spitzer images ($M_{d,1-50}$ and $M_{d,240}$) and
a typical particle size $\langle r\rangle$. For a ring at $a\sim$ 150 AU with
a width $\Delta a\sim$ 50 AU, the collision time for a single grain is
$t_{c}\sim 10^{2}~{}P~{}\left(\frac{\langle r\rangle}{10~{}\mu{\rm
m}}\right)~{},$ (63)
where $P$ is the local orbital period in yr (Lissauer, 1987; Wetherill &
Stewart, 1993; Kenyon & Luu, 1998). Collision times in the wind are similar.
For $P\sim 10^{3}$ yr at 150 AU, the collision times range from $\sim$
$10^{4}$ yr for 1 $\mu$m grains to $2-3\times 10^{6}$ yr for 240 $\mu$m
grains.
If we assume that the residence times are comparable to the collision times,
we can construct a self-consistent picture for the collisional cascade in the
Vega disk. Collisions in the bright ring gradually erode 200–300 $\mu$m grains
until they reach sizes $\lesssim$ 100 $\mu$m, when they become incorporated
into the wind. Collisions in the wind gradually erode the smaller grains until
they reach sizes $\sim$ 1 $\mu$m, when they are ejected rapidly from the
system.
As long as the 200–300 $\mu$m grains are replenished from a reservoir of
larger grains, this cascade can remain in a quasi-steady state over the main
sequence lifetime of Vega. The required mass for the reservoir of larger
objects is $\sim$ 100–1000 times the current mass in 240 $\mu$m grains, $\sim$
1–5 M⊕. This mass is small compared to the initial mass of solid material in a
torus at 80–200 AU in a scaled MMSN, $\sim$ 50–100 M⊕ (Eq. (27)). Because the
optical depth of this reservoir is small, it produces a small IR excess
compared to the emission from smaller grains.
This picture relies on two features of the collisional cascade. We need an
approximate equivalence in mass between the large grains in the ring and the
small grains in the wind. The Su et al. (2005) mass estimates support this
feature. We also need a gradual change in grain lifetime from the $\sim
10^{6}$ yr collision timescale of the large grains to the $10^{3}-10^{4}$ yr
dynamical lifetime of the smallest grains. Otherwise, collisions in the broad
torus cannot occur fast enough to maintain the current smooth structure of the
wind for timescales longer than $\sim 10^{3}-10^{4}$ yr. Current theoretical
analyses support this idea (Burns, Lamy, & Soter, 1979; Artymowicz, 1988;
Takeuchi & Artymowicz, 2001; Grigorieva et al., 2007). Numerical simulations
of a collisional cascade with a careful treatment of the interactions between
the radiation field and the small grains could test this proposal in detail
(e.g., Grigorieva et al., 2007).
We conclude that our calculations provide a reasonable match to observations
of Vega. If we adopt $r_{2}\sim$ 10 $\mu$m, the data are consistent with a
standard collisional cascade within a broad torus at 80–200 AU. The cascade
feeds an outflowing wind of small grains with sizes 1–50 $\mu$m. If the grain
lifetime changes smoothly from $\sim 10^{6}$ yr for large grains to $\sim$
$10^{3}-10^{4}$ yr for small grains, the cascade can maintain the wind
indefinitely.
### 5.2 Debris Disks Around A-type Stars
Since the discovery of the Vega debris disk, IRAS, ISO, and Spitzer
observations have revealed debris around dozens of nearby A-type stars
(Backman & Paresce, 1993; Lagrange et al., 2000; Rieke et al., 2005; Su et
al., 2006). Like Vega, several of these disks are resolved and thus provide
important information on the radial structure of the dusty disk (e.g., Smith &
Terrile, 1984; Stapelfeldt et al., 2004; Kalas, 2005; Meyer et al., 2007; Su
et al., 2008). Although most A-type stars with debris disks are unresolved,
the sample is large enough to probe the time evolution of debris around 1.5–3
M⊙ stars. We now consider whether our calculations can explain this evolution.
To compare our model predictions with observations, we examine data for nearby
A-type stars from Rieke et al. (2005) and Su et al. (2006). Rieke et al.
(2005) combined 24–25 $\mu$m data from IRAS and ISO with new 24 $\mu$m
photometry from Spitzer to investigate the decay of planetary debris disks
around 266 A-type stars. Su et al. (2006) analyze a sample of $\sim$ 160
A-type stars with high quality 24 $\mu$m and/or 70 $\mu$m data acquired with
MIPS on Spitzer. The combined sample has 319 (160) stars with 24 $\mu$m (70
$\mu$m) observations, spectral types B7–A6, and ages 5–850 Myr. From the
Kenyon & Hartmann (1995) table of stellar effective temperatures and spectral
types and the Demarque et al. (2004) stellar evolution tracks, $\sim$ 75%
($\sim$ 85%) of the stars in the Rieke et al. (2005) (Su et al., 2006) sample
have masses of 1.7–2.5 M⊙. Thus, we compare these data with our results for
debris disk evolution around 2 M⊙ stars.
The observed 24–70 $\mu$m excesses of A-type stars show a clear trend with the
age of the star (Fig. 22–23). Although the statistics are poor, the data
suggest a rise in the 24 $\mu$m excess at 5–10 Myr. The larger sample of young
stars with 70 $\mu$m excesses provides better evidence for this rise. At both
wavelengths, the excess has a broad peak for stars with ages of 10–30 Myr. At
later times, the excess declines with time as $t^{-n}$ with $n\approx$ 0.5–1
(see also Decin et al., 2003; Greaves & Wyatt, 2003; Rieke et al., 2005; Rhee
et al., 2007a).
To improve the statistics for 24 $\mu$m excesses around younger stars, Currie
et al. (2008a) added Spitzer data for many young clusters to the Rieke et al.
(2005) sample. This expanded set of data provides unambiguous evidence for a
rise in the typical 24 $\mu$m excess at stellar ages of 5–10 Myr and a robust
peak in the excess at stellar ages of 10–15 Myr. As in Fig. 22, the 24 $\mu$m
excesses for this larger sample of A-type stars decline with age from $\sim$
20 Myr to 1 Gyr.
In addition to the long-term time evolution of mid-IR excess, the data also
indicate a large range in the 24–70 $\mu$m excess at fixed stellar age (Rieke
et al., 2005; Carpenter et al., 2006; Su et al., 2006; Currie et al., 2008a).
Although younger stars are more likely to have mid-IR excesses than older
stars, there are many stars without excesses at every age. For ages $\lesssim$
200 Myr, stars are equally likely to have any excess between zero and the
maximum excess at that age. As stars age, they are less likely to have an
excess close to the maximum excess at that age. Thus, the dispersion in the
excess declines with time.
Our calculations provide a good match to the time evolution of the amplitude
of the 24–70 $\mu$m excesses. At both wavelengths, the models explain the rise
in the amplitude at 5–10 Myr, the maximum at 10–20 Myr, and the slope of the
power-law decline at late times. For models with $x_{m}$ = 1–3, the predicted
excesses also agree with the maximum observed excesses. Although there are a
few stars with excesses larger than the model predictions, more than 99% of
the A stars in this sample have excesses within the range predicted in our
calculations.
Our calculations also provide a natural explanation for a large range in the
observed 24–70 $\mu$m excesses at fixed stellar age. At 70 $\mu$m, the maximum
excess is roughly proportional to the initial disk mass (Eq. (58)). Thus, a
factor of ten range in initial disk masses yields nearly a factor of ten range
in the maximum excess at 70 $\mu$m. For stars with ages 10–300 Myr, the
Spitzer observations suggest a factor of $\sim$ 100 range in the 70 $\mu$m
excess. If this range is set by the initial disk mass, our models suggest
initial disk masses with $x_{m}$ = 0.03–3.
Variations in the initial disk radius can also produce a range in 24–70 $\mu$m
excesses at fixed stellar age (e.g., Fig. 16). For 2 M⊙ stars with ages $\sim$
400–800 Myr, our results suggest that a factor of three variation in the outer
disk radius (e.g., 50–150 AU) yields a factor of two (five) variation in the
amplitude of the 24 $\mu$m (70 $\mu$m) excess. Although the observed range in
the amplitude of the 24 $\mu$m excess for older A stars agrees with this
prediction, the range at 70 $\mu$m is much larger. Thus, variations in the
initial disk radius can explain some of the observed range of excesses at
24–70 $\mu$m.
Observations of the youngest stars support a large range in initial disk
masses and disk radii. Submillimeter observations of dusty disks in the nearby
Ophiuchus and Taurus-Auriga star-forming regions indicate a 2–3 ($\sim$ 1)
order of magnitude range in the masses (radii) of disks surrounding young
stars with typical ages of $\sim$ 1 Myr (e.g., Osterloh & Beckwith, 1995;
Motte & André, 2001; Andrews & Williams, 2005, 2007a,b). Our models with
$x_{m}$ = 3 have disk masses a little smaller than the maximum dust masses
derived from the submillimeter surveys. Thus, the submillimeter data imply
disks with initial masses $0.01\lesssim x_{m}\lesssim$ 5 and initial disk
radii 50 AU $\lesssim a_{out}\lesssim$ 1000 AU. Disks with this range of
initial masses and outer radii can produce the range of 24–70 $\mu$m excesses
observed around nearby A-type stars.
Variations in the initial surface density distribution can also lead to a
range in the 24–70 $\mu$m excess. In our calculations, we adopted a ‘standard’
surface density relation with $\Sigma\propto a^{-3/2}$. Compared to this
model, disks with shallower (steeper) surface density distributions have
relatively more (less) mass at large semimajor axes. The outer disk has cooler
grains than the inner disk; thus, disks with shallower (steeper) surface
density distributions should produce more (less) flux at longer wavelengths
than our standard models. Because the relative fluxes at 24 $\mu$m and at 70
$\mu$m provide a measure of the relative disk masses at different semimajor
axes, color indices provide a natural measure of the gradient of the surface
density distribution.
Fig. 24 compares the predicted color evolution for disks around 2 M⊙ stars
with data from Su et al. (2006). Although current observations do not probe
the evolution well at 1–10 Myr, disks have a large color range, [24]–[70]
$\sim$ 1–3, for stars with ages $\sim$ 10 Myr. For older stars, the data
suggest a slow rise in the maximum color from [24]–[70] $\sim$ 3 at 10 Myr to
[24]–[70] $\sim$ 3.5 at $\sim$ 100 Myr. After $\sim$ 300 Myr, the maximum
color declines. For all stars older than $\sim$ 100 Myr, the range in color is
$\sim$ 3–4 mag.
Our models match the observed color evolution. At all ages, the predicted
colors for calculations with $x_{m}$ = 3 provide a clear upper envelope to the
observed colors. The predicted colors also explain the slow rise in the
maximum observed color for stars with ages of 10–100 Myr. To explain the full
range in observed colors for 100 Myr to 1 Gyr old stars, we require disks with
initial masses $x_{m}\approx$ 0.01–3. This range is similar to the range
required for the time evolution of the 24 $\mu$m and 70 $\mu$m excesses.
The good match to the color observations suggests that the typical initial
surface density distribution is reasonably close to our adopted $\Sigma\propto
a^{-3/2}$. For disks with shallower gradients, we expect redder colors at
later times. A few stars lie above our model predictions; however, most stars
have bluer colors than models with $x_{m}$ = 3. Thus, few disks in these
samples require shallower surface density distributions. Disks with steeper
surface density distributions can produce stars with blue colors, [24]–[70]
$\sim$ 1, at late times. A large sample of A-type stars with [70]–[160] colors
and spatially resolved observations of the radial dust distributions of these
stars would provide a constraint on the initial surface density gradient.
Despite our success in matching these observations, other physical processes
may be needed to explain the full diversity of debris disk properties for A
stars with similar ages and luminosities. In their analysis of the large
debris disk surrounding $\gamma$ Oph, Su et al. (2008) examine a dozen main
sequence stars with A0–A3 spectral types, ages of 150–400 Myr, and fractional
disk luminosities $L_{d}/L_{\star}\approx$ $10^{-5}$ to $10^{-4}$ (see also Su
et al., 2006). Although all of these stars have dust with $T\approx$ 50–100 K,
Fomalhaut has a bright torus of dust with weak or negligible emission from a
wind of small grains, Vega has a bright torus with a luminous wind of small
grains, and $\gamma$ Oph has an extended disk ($a_{out}\approx$ 500 AU) of
dust apparently bound to the star. Some A0–A3 stars have warm inner disks with
dust temperatures $\sim$ 100 K to $\gtrsim$ 200 K; other A stars have no
obvious warm dust emission. Su et al. (2008) conclude that collisional
cascades in disks with a range of masses and other processes, such as the
formation of giant planets or recent catastrophic collisions, combine to
produce the wide range of observed properties in this sample.
In principle, our models can explain some of this diversity. The Su et al.
(2008) sample contains A stars with a factor of five range in $L_{\star}$.
Thus, these stars probably have a factor of five range in the blowout radius
$r_{2}$ (see also §5.1; Artymowicz, 1988; Backman & Paresce, 1993). If the
protostellar disks around these stars had properties similar to those observed
in Taurus-Auriga (Andrews & Williams, 2005, 2007a), they probably had a factor
of ten range in initial disk mass, a factor of three range in initial disk
radius, and a 50% range in the slope of the initial surface density
distribution. Coupled with a similar dispersion in initial conditions for the
terrestrial zones of these stars (Kenyon & Bromley, 2005), our results suggest
that this range in initial conditions can produce a broad diversity of debris
disks. We have started a suite of calculations to address this issue. Larger
samples of A stars with resolved disks will provide crucial tests of these
calculations.
Other aspects of planet formation are also important. If the cores of gas
giant planets form before their parent stars reach the main sequence, we
expect gas giants at 20–30 AU around 2–3 M⊙ stars (e.g., Kennedy & Kenyon,
2008). Gas giants rapidly remove debris in the inner disk and impose structure
in the debris beyond 30 AU (e.g., Wilner et al., 2002; Moro-Martín & Malhotra,
2005). Because gas giants are common around evolved A stars (Johnson et al.,
2007), gas giants probably play a significant role in the evolution of debris
disks around A stars.
Catastrophic collisions may also produce diversity among A star debris disks
(e.g., Wyatt & Dent, 2002; Su et al., 2005). Although debris from complete
disruption of colliding planetesimals is unobservable in our simulations (see
also Kenyon & Bromley, 2005), dynamical events similar to those that produced
the Late Heavy Bombardment in the Solar System probably are visible (e.g.,
Gomes et al., 2005). Testing this idea requires numerical calculations that
link the dynamics of massive planets with the collisional evolution of smaller
objects (e.g., Charnoz & Morbidelli, 2003; Kenyon & Bromley, 2006).
We conclude that our debris disk models can explain the overall time evolution
of the IR excesses and IR colors of A-type main sequence stars at 24 $\mu$m
and at 70 $\mu$m. Our calculations for disks with $x_{m}$ = 1/3 to 3 around 2
M⊙ stars fit the overall level of the excesses and the trends with stellar
age. Explaining the full range of observed IR excesses and IR colors requires
a set of disks with $x_{m}$ = 0.01–3, as suggested from observations of disks
around the youngest stars. Matching other properties of these stars –
including the relative amount of emission from a warm inner disk, an
outflowing wind of small grains, and a large outer disk – requires
calculations that include a broader range of initial disk radii and gas giant
and terrestrial planet formation at $a_{i}<$ 30 AU (e.g., Kenyon & Bromley,
2005; Nagasawa et al., 2007; Kennedy & Kenyon, 2008; Ida & Lin, 2008; Kretke
et al., 2008).
### 5.3 Debris Disks Around Solar-type Stars
Although most of the debris disks discovered with IRAS and ISO have A-type
central stars, a few have F-type or G-type central stars with masses of 1–1.5
M⊙ (Backman & Paresce, 1993; Lagrange et al., 2000; Decin et al., 2003; Song
et al., 2005; Rhee et al., 2007a). More recent Spitzer observations reveal
debris disks around many solar-type stars (Bryden et al., 2006; Beichman et
al., 2006; Meyer et al., 2006; Trilling et al., 2008; Hillenbrand et al.,
2008). Although several Spitzer programs concentrate on older solar-type stars
as preparation for detailed planet searches, the range of ages is large enough
to provide an initial test of our predictions.
To compare our model predictions with observations, we consider data for
nearby solar-type stars from Beichman et al. (2006) and Hillenbrand et al.
(2008). Beichman et al. (2006) observed $\sim$ 80 solar-type stars at 24
$\mu$m and at 70 $\mu$m using MIPS on Spitzer. Hillenbrand et al. (2008)
analyze $\sim$ 30 stars with 70 $\mu$m excesses out of a sample of 328 stars
from the Spitzer Legacy Science Program, “Formation and Evolution of Planetary
Systems” (Meyer et al., 2006). After eliminating K-type and M-type stars from
the Beichman et al. (2006) study, the two programs contain $\sim$ 80 stars
with ages of $\sim$ 10 Myr to $\sim$ 10 Gyr. Most of these stars have masses
of 0.8–1.5 M⊙. Thus, we compare these data with our results for debris disk
evolution around 1 M⊙ stars.
The observed 70–160 $\mu$m excesses of solar-type stars show trends similar to
those observed in the evolution of A-type stars at 24–70 $\mu$m (Fig. 25–26).
Although the statistics for solar-type stars are poor for the youngest stars,
the data suggest a rise in the 70–160 $\mu$m excess at 10–100 Myr. The maximum
in the 70 $\mu$m excess is comparable in magnitude but a factor of $\sim$ 10
later in time than the maximum 70 $\mu$m excess for A-type stars. At 70 $\mu$m
and at 160 $\mu$m, the excess follows a roughly power-law decline with time
for older stars. Solar-type stars also have a large range in excess at all
ages, with $F_{70}/F_{70,0}\approx$ 1–300 at 100 Myr and
$F_{70}/F_{70,0}\approx$ 1–30 at 1–3 Gyr. For stars with similar ages, the
range in the 70 $\mu$m excess is larger for solar-type stars than for A-type
stars.
Our models match the observed trends for the IR excesses of solar-type stars.
At 70 $\mu$m, 65%–75% of the observations lie within model predictions; at 160
$\mu$m, more than half of the observations are within model predictions. For a
range of initial disk masses ($x_{m}\approx$ 0.01–3) and outer radii
($a_{out}\approx$ 70–150 AU), we predict a large range of excesses at all
ages, as observed. These model also explain the larger 70 $\mu$m excesses
observed for solar-type stars relative to A-type stars, the apparent maximum
in the 70–160 $\mu$m excess at 30–100 Myr, and the general power-law decline
in the excess flux for the oldest stars.
Despite this general success, however, the models underpredict the largest
observed excesses. At 70 $\mu$m, the brightest disks are a factor of 5–10
brighter than disks with $x_{m}$ = 3. At 160 $\mu$m, the brightest systems are
3–5 times brighter than our most luminous disks. Although the sample of 160
$\mu$m sources is small, our models underpredict the largest observed fluxes
at all ages.
Changing two assumptions in our models yields a better match to the observed
fluxes at 70–160 $\mu$m. For realistic grain properties, Burns, Lamy, & Soter
(1979) show that radiation pressure from the Sun cannot eject small grains
from the Solar System. Reducing the minimum stable grain size from $r_{2}$ = 1
$\mu$m to $r_{2}$ = 0.1 $\mu$m increases our predicted 70 $\mu$m (160 $\mu$m)
fluxes by a factor of 2–3 (1.5–2). Submm observations of several debris disks
imply $q\approx$ 0.6–1 for the slope of the radiative emissivity. If we adopt
$q=0.7$ instead of $q=1$, our predicted 70–160 $\mu$m fluxes increase by
factors of 2–4. Combining these two modifications increase our predicted
fluxes by a factor of $\sim$ 5–10 at both wavelengths.
Several observations could check whether these modifications of our standard
model are reasonable. By analogy with the Spitzer Vega data, detection of an
outflowing wind of small grains in a debris disk around a solar-type star
provides a clean measurement of $r_{2}$ and a better constraint on our
predicted IR excesses. Measurements of $q$ for larger samples of debris disks
allows a better assessment of our assumptions for the grain emissivity.
## 6 CONCLUSIONS
Our calculations provide a robust picture for the formation of planets and
debris disks from a disk of icy planetesimals and set the context for the
evolution of dusty debris in a dynamic system of planets. The results of this
study provide a framework for interpreting existing observations of debris
disks around 1–3 M⊙ stars and suggest new observational tests of this picture.
We describe a suite of numerical calculations of planets growing from
ensembles of icy planetesimals at 30–150 AU in disks around 1–3 M⊙ stars.
Using our hybrid multiannulus coagulation code, we solve for the evolution of
sizes and orbits of objects with radii of $\sim$ 1 m to $\gtrsim$ 1000 km over
the main sequence lifetime of the central star. These results allow us to
constrain the growth of planets as a function of disk mass, stellar mass, and
semimajor axis.
Debris disk formation is coincident with the formation of a planetary system.
All calculations of icy planet formation at 30–150 AU lead to a collisional
cascade which produces copious amounts of dust on timescales of 5–30 Myr. This
dust is observable throughout the lifetime of the central star. Because we
consider a broad range of input parameters, we derive the time evolution of
(i) dust produced in the collisional cascade and (ii) the IR and submm
emission from this dust as a function of disk mass, stellar mass, and time.
We divide the rest of this section into (i) theoretical considerations, (ii)
observable consequences, and (iii) observational tests. The theoretical
considerations build on the highlights of icy planet formation in §3.4.
Observable consequences of the calculations follow from the discussion in §4.
The observational tests of the models are described in §5.
### 6.1 Theoretical Considerations
* •
Icy planet formation at 30–150 AU is self-limiting. Starting with a swarm of
$\lesssim$ 1 km planetesimals, runaway growth produces a set of 100–500 km
protoplanets. As the protoplanets grow, they stir up leftover planetesimals
along their orbits. When the leftovers reach high $e$, collisions produce
debris instead of mergers. Because protoplanets cannot accrete leftovers
rapidly, a cascade of destructive collisions grinds the leftovers to dust.
Poynting-Robertson drag and radiation pressure then remove the dust from the
disk.
* •
The maximum sizes of icy planets at 30–150 AU are remarkably independent of
initial disk mass, stellar mass, and stellar age. For disks with $x_{m}$ = 1/3
to 3 around 1–3 M⊙ stars with ages $t=0.1-1t_{ms}$, the typical planet has
$r_{max}\sim$ 1750 km and $m_{max}\sim 0.005$ M⊕. These objects contain
$\lesssim$ 3%–4% of the initial disk mass. Although this result is also
independent of the fragmentation parameters, the finite main sequence
lifetimes of 1–3 M⊙ stars limits the formation of many large planets in the
outer disk. Thus, the inner disk produces many more Pluto-mass planets than
the outer disk (Tables 2 and 3).
* •
For stars close to the main sequence turnoff, stellar lifetimes and the
collisional cascade limit the mass in solid objects at 30–150 AU. In the inner
disk, the collisional cascade removes most of the leftover planetesimals
before the central star evolves off the main sequence. Thus, the typical mass
in small objects is $\sim$ 10% of the initial mass at 30–40 AU. In the outer
disk, smaller collision rates produce a slower cascade. Thus, the central star
evolves off the main sequence with $\sim$ 50% of the initial mass remaining in
1–10 km planetesimals at 125–150 AU.
* •
The collisional cascade produces copious amounts of dust. Dust begins to form
during the transition from runaway to oligarchic growth ($t$ = 5–10 Myr),
peaks when the first objects reach their maximum sizes ($t$ = 10–30 Myr), and
then slowly declines $t\gtrsim$ 30–50 Myr). The peak mass in 0.001–1 mm
(0.001–1 m) particles is $\sim$ 1–2 lunar masses ($\sim$ 1 M⊕). Disks with
initial masses $x_{m}$ = 1/3 to 3 reach these peak masses when the age of the
star is $\sim$ 10% to 20% of its main sequence lifetime. Because the timescale
to form dust is short ($\sim$ 10–20 Myr), stars are surrounded by large disks
of debris at 30–150 AU throughout their main sequence lifetimes.
* •
Radiative processes remove large amounts of mass from debris disks. Radiation
pressure produces a radial wind of small particles containing $\sim$ 60% to
70% of the mass removed from the disk. Poynting-Robertson drag pulls the rest
of the lost mass into the inner disk. Because radiation pressure is more
important than Poynting-Robertson drag when collision rates are large, we
expect more wind (inner disk) emission earlier (later) in the evolution.
### 6.2 Observable Consequences
We derive clear observational consequences of the collisional cascade.
* •
The dusty debris from the collisional cascade is directly observable. For
disks around 1–3 M⊙ stars, the maximum fractional dust luminosity of
$L_{d}/L_{\star}\sim 2\times 10^{-3}$ is comparable to the maximum dust
luminosities of known debris disks (Backman & Paresce, 1993; Rieke et al.,
2005; Su et al., 2006; Rhee et al., 2007a). The dust temperature at the inner
edge of a 30–150 AU disk scales with the temperature of the central star;
thus, the predicted 24 $\mu$m excess is very sensitive to the stellar mass. At
70 $\mu$m, the predicted excesses scale roughly linearly with disk mass and
stellar mass. The predicted 160–850 $\mu$m excesses depend on the disk mass
but are nearly independent of the stellar mass.
* •
For systems with little or no emission from terrestrial dust ([5.8]–[8]
$\lesssim$ 0.1), mid-IR color-color diagrams clearly distinguish debris disks
around stars of different masses. In a [8]–[24] vs [24]–[70] diagram, 2–3 M⊙
(1–2 M⊙) stars have red (blue) [8]–[24] and blue (red) [24]–[70] (Fig 20). In
both cases, the color scales with the initial disk mass (Fig 21). Optical
colors and spectra generally provide good estimates for stellar mass; thus,
these diagrams provide good tests of our model predictions.
### 6.3 Observational Tests
We compare our predictions with observations of A-type stars and solar-type
stars.
* •
For A-type stars, our calculations are the first to explain the observed rise
and fall of debris disk fluxes at 24 $\mu$m (Fig. 22; Currie et al., 2008a,
b). In our picture, the rise in debris disk emission corresponds to the
transition from runaway growth – when mergers of small planetesimals produce
larger protoplanets – to oligarchic growth – when the collisional cascade
begins to grind leftover planetesimals into dust. When oligarchs in the inner
disk are close to their maximum sizes of $\sim$ 1750 km, the collisional
cascade produces a maximum in debris disk emission. For a wide range of
initial conditions, this maximum occurs at 10–20 Myr. As the collisional
cascade moves out through the disk, smaller collision rates produce less dust
which emits at lower temperatures. Thus, the 24 $\mu$m excess falls with time.
The predicted rate of decline, $t^{-n}$ with $n\approx$ 0.6–0.8, is close to
the observed rate ($n\approx$ 0.5–1; Greaves & Wyatt, 2003; Rieke et al.,
2005; Rhee et al., 2007a).
* •
At longer wavelengths, the maximum excess is larger and lasts longer than at
24 $\mu$m. Predicted mid-IR colors also increase slowly with time. Although
larger samples of A-type stars with 8 $\mu$m photometry would provide a better
test of our models, current data for the 70 $\mu$m excess and the evolution of
the [24]–[70] color agree with our predictions (Fig. 23–24). For 2–3 M⊙ stars
near the main-sequence turnoff, our calculations also yield a clear maximum in
the 850 $\mu$m flux. Large surveys, such as the proposed JCMT Legacy Survey
(Matthews et al., 2007) and submm observations with ALMA, Herschel and SOFIA,
can test this prediction.
* •
For solar-type stars, our models match observations of most sources. The
predicted evolution of the 70–160 $\mu$m excesses follows the observed rise at
10–100 Myr, the peak at $\sim$ 30–100 Myr, and the decline at $\gtrsim$ 300
Myr. Although $\sim$ 70% (55%) of observed debris disks have fluxes that lie
within model predictions, our models underpredict fluxes for the brightest
sources by a factor of 5–10. Fluxes for models with $r_{2}\approx$ 0.1 $\mu$m
and $q\lesssim$ 0.7 provide better matches to these observations. To guide our
choices for $r_{2}$ and $q$, we require spatially resolved images and submm
fluxes for these objects.
* •
For 1–3 M⊙ stars with ages $\sim$ 0.1–1 Gyr, current data suggest that solar-
type stars have a larger range of far-IR excesses than A-type stars. In our
models, faster debris disk evolution around A-type stars produces a smaller
dispersion in far-IR excesses and colors for stars with ages of 100 Myr to 1
Gyr. Larger samples of debris disks can test this prediction in more detail.
* •
We also consider observations of Vega, the prototypical debris disk. If we
adopt models with a blowout radius $r_{2}$ = 10 $\mu$m, we can match
observations with a standard collisional cascade within a broad torus at
80–200 AU. If the torus contains $\sim$ 1–5 M⊕ in large objects with
$r\gtrsim$ 1 cm, the cascade can generate (i) the observed ensemble of grains
with $r\sim$ 200–300 $\mu$m within the torus and (ii) an outflowing wind of
small grains with $r\sim$ 1–50 $\mu$m.
This conclusion differs from Su et al. (2006), who postulate a recent
catastrophic collision between two large objects as the source of the dusty
Vega wind. Although the complete destruction of two large icy objects can
produce a massive outflowing wind, our results suggest that the dusty wind is
short-lived and cannot be rapidly replenished by the observed population of
larger objects. We show that a steady-state collisional cascade can explain
the Spitzer data (see also Kenyon & Bromley, 2005). If our interpretation is
correct, sensitive observations at 1–10 mm should detect our proposed
reservoir of larger objects.
Matching other observations of debris disks requires more realism in our
planet formation calculations. Adding binary companions and giant planets
provides ways to modify the evolution of the collisional cascade and to impose
structure on rings and tori (e.g. Wilner et al., 2002; Moro-Martín & Malhotra,
2005; Quillen, 2006). Extending the coagulation calculations to smaller sizes
allows studies of the formation of winds and other large structures. Although
these calculations have been prohibitively expensive in computing time, rapid
advances in computing technology will make these additions possible in the
next few years.
Based on the results described here and in Kenyon & Bromley (2004b), we
conclude that debris disks are the inevitable outcome of icy planet formation
in a disk of solid objects. The basic structures produced by this model –
broad tori and narrow rings of dust that propagate out through the disk
(Kenyon & Bromley, 2004b) – are consistent with observations (e.g.
Jayawardhana et al., 1998; Kalas, 2005; Su et al., 2006; Fitzgerald et al.,
2007). The model also explains the time evolution of mid-IR colors and fluxes
for debris disks around A-type and solar-type stars.
We acknowledge a generous allotment, $\sim$ 1000 cpu days, of computer time on
the 1024 cpu Dell Xeon cluster ‘cosmos’ at the Jet Propulsion Laboratory
through funding from the NASA Offices of Mission to Planet Earth, Aeronautics,
and Space Science. We thank M. Werner for his strong support of this project.
We also acknowledge use of $\sim$ 250 cpu days on the CfA cluster ‘hydra.’
Advice and comments from T. Currie, M. Geller, G. Kennedy, M. Meyer, G. Rieke,
K. Su, and an anonymous referee greatly improved our presentation. Portions of
this project were supported by the NASA Astrophysics Theory Program, through
grant NAG5-13278, the NASA TPF Foundation Science Program, through grant
NNG06GH25G, and the Spitzer Guest Observer Program, through grant 20132.
## Appendix A APPENDIX
### A.1 Growth rates
In standard coagulation theory, protoplanets accrete material from a swarm of
planetesimals at a rate (e.g., Safronov, 1969; Lissauer, 1987; Wetherill &
Stewart, 1993)
$\dot{M}\propto\Sigma~{}\Omega~{}r^{2}~{}\left[1+(v_{esc}/v)^{2}\right]~{},$
(A1)
where $r$ is the radius of a planetesimal, $\Omega$ is the angular frequency
of material in the disk, $v$ is the random velocity of planetesimals, and
$v_{esc}$ is the escape velocity of the protoplanet. The $1+(v_{esc}/v)^{2}$
term is the gravitational focusing factor.
To derive the accretion time, we set $t=M/\dot{M}$ and substitute the orbital
period for the angular frequency,
$t\propto(\rho~{}r~{}P/\Sigma)~{}\left[1+(v_{esc}/v)^{2}\right]^{-1}~{},$ (A2)
where $\rho$ is the mass density of a planetesimal. Throughout runaway growth
and the early stages of oligarchic growth $v_{esc}/v\gg 1$. Because we are
interested in the time to produce planets with the same $r$ and $\rho$ in
disks with different $P$ and $\Sigma$, we eliminate $\rho$ and $r$. Thus, the
growth time is roughly
$t\propto(P/\Sigma)(v/v_{esc})^{2}~{}.$ (A3)
This equation sets the typical timescale for planet growth in a disk of
planetesimals. If $\Sigma\sim\Sigma_{0}x_{m}a^{-3/2}$ (Eq. (27)) and
$v/v_{esc}$ $\sim$ constant (Fig. 1; Wetherill & Stewart, 1993; Goldreich,
Lithwick, & Sari, 2004),
$t\propto a^{3}x_{m}^{-1}\Sigma_{0}^{-1}~{}.$ (A4)
This result is close to the $t\propto a^{3}x_{m}^{-1.15}\Sigma_{0}^{-1}$
derived for the formation of the first 1000 km object in our calculations
(e.g., Eq. (41)).
To evaluate possible sources for the extra factor of $x_{m}^{-0.15}$ in our
derived accretion times, we consider the random velocity $v$ of accreted
planetesimals. Shorter growth times require smaller random velocities. Thus,
we consider processes that damp planetesimal velocities. In our calculations,
collisions and gas drag can reduce $v$; dynamical friction and viscous
stirring increase $v$. At 30–150 AU, gas drag damps random velocities $\sim$
10–20 times more rapidly than collisions (Goldreich, Lithwick, & Sari, 2004).
Thus, we ignore collisional damping and concentrate on gas drag.
Rafikov (2004) investigated the dynamics of small planetesimals and growing
protoplanets in a gaseous nebula. For the early stages of oligarchic growth,
the random velocity of planetesimals is
$v/v_{esc}\propto\Sigma_{gas}^{-\gamma_{1}}~{}.$ (A5)
Substituting this expression into Eq. (A4) and adopting a constant gas-to-dust
ratio, $\Sigma_{gas}\sim\Sigma$, we derive
$t\propto a^{3}x_{m}^{-\gamma 2}\Sigma_{0}^{-1}~{},$ (A6)
with $\gamma_{2}=2\gamma_{1}+1$. For typical conditions in planetesimal disks,
Rafikov (2004) derived $\gamma_{1}\approx$ 1/6 to 1/5. Thus,
$\gamma_{2}\approx$ $1.3$ to $1.4$, close to the exponent of $1.15$ derived in
our calculations.
Our treatment of gas drag probably reduces the exponent of $x_{m}$ in Eq. (A6)
from the predicted $1.3$–$1.4$ to $1.15$. In our simulations, we assume the
gas density declines exponentially on a timescale $t_{gas}$ = 10 Myr. With
typical growth times of 20–40 Myr, the gas density is $\sim$ 1% to 10% of its
initial value when the first 1000 km objects form in the inner disk. Thus, gas
drag cannot reduce planetesimal random velocities as efficiently as predicted
in Eq. A6. Reducing drag lowers the exponent. With gas depletion timescales
$\sim$ 25% to 50% of the growth time, we expect an exponent of
$\gamma_{2}\approx$ $1.1$ to $1.2$, similar to the $\gamma_{2}=1.15$ in our
calculations.
### A.2 Radiation from dust
In the Appendix of Kenyon & Bromley (2004a), we briefly described our simple
algorithm for the evolution of particles with sizes smaller than the smallest
object – $r\sim$ 1 m – followed in the multiannulus coagulation code. This
algorithm yields the optical depth in very small grains ejected from the
system and the optical depth in larger grains evolving under the influence of
collisions and Poynting-Robertson drag. The optical depth in both grain
populations allows us to derive the time evolution of the disk luminosity and
surface brightness in bolometric units. Here, we describe the derivation of
grain temperature for these populations that yields the predicted time
evolution of the broadband spectral energy distributions of debris disks.
As in Kenyon & Bromley (2004a), we divide objects with sizes smaller than
$\sim$ 1 m into very small grains, small grains, and large grains. In each
annulus $k$ of our calculation, radiation pressure ejects very small grains
with radii between $r_{1}$ and $r_{2}$. If $\rho_{g}$ is the mass density of
these grains and $\dot{M}_{k}$ is the production rate of very small grains in
each annulus, the very small grains have an integrated optical depth
$\tau_{s}=\frac{3(\sqrt{r_{2}/r_{1}}-1)}{8\pi\rho_{g}r_{2}(1-\sqrt{r_{1}/r_{2}})}\sum_{i=1}^{N}~{}\left[\sum_{k=1}^{i}\left(\frac{\dot{M}_{k}}{v_{Kk}h_{k}}\right)\left(\frac{1}{a_{b,k}}-\frac{1}{a_{b,k+1}}\right)\right]~{},$
(A7)
where $a_{b,k}$ is the inner boundary of an annulus centered at $a_{k}$,
$h_{k}$ is the vertical scale height in units of the semimajor axis, and
$v_{K,k}$ is the orbital velocity in annulus $k$.
For small ($r$ = $r_{2}$ to 1 mm) and large grains ($r$ = 1 mm to 1 m), we
derive the optical depth $\tau_{k}$ in each annulus. To derive the radial
surface brightness and total disk luminosity, we follow Kenyon & Hartmann
(1987) and derive the amount of stellar radiation absorbed by each annulus. We
assume a spherical, limb-darkened star with radius $R_{\star}$, luminosity
$L_{\star}$, and limb-darkening parameter $\epsilon_{0}$ = 0.6. For a point
$P$ at the outer boundary of annulus $k$ with height $h_{P}$ above the disk
midplane, rays from the star enter the annulus at a scale height $h_{in}$
above (below) the midplane. We compute the length $l$ of the path through the
disk and derive the optical depth along this path as $\tau_{p}$ = ($l/\Delta
a_{k}$)$\tau_{k}$, where $\Delta a_{k}$ is the width of the annulus. The
radiation absorbed along this path is $e^{-\tau_{p}}I_{0}$, where $I_{0}$ is
the flux incident on the boundary of the annulus. Numerical integrations over
the stellar surface and the vertical extent of an annulus yield the amount of
flux absorbed by each annulus, which we convert to relative surface
brightness. A final numerical integration over the radial extent of the disk
yields the ratio of the disk luminosity to the stellar luminosity,
$L_{d}/L_{\star}$.
To derive the spectral energy distribution of the disk, we make several
assumptions. Consistent with observations of scattered light from resolved
debris disks (Backman & Paresce, 1993; Lagrange et al., 2000), we adopt a
single albedo $\omega$ = 0.25 for all grains. For all $\lambda$, the
luminosity in scattered light is then $\omega L_{d}/L_{\star}$; the thermal
luminosity emitted by all grains is $(1-\omega)L_{d}/L_{\star}$. In each
annulus $k$, we assume grains emit at a temperature $T_{i,k}$, where the index
$i$ refers to discrete bins in grain size. To derive equilibrium temperatures
for these grains, we assume the grains have an absorption efficiency
$\epsilon_{a}\propto(\lambda/\lambda_{0})^{p}$ and radiative efficiency
$\epsilon_{r}\propto(\lambda/\lambda_{0})^{q}$. For most grains in our
calculations, the grain size is larger than the peak wavelength of radiation
emitted by the central star. Thus, the grains efficiently absorb stellar
photons and $p$ = 0. Large grains with $r\gg\lambda$ emit as blackbodies and
have $q$ = 0. Smaller grains radiate less efficiently and have $q\approx$ 1.
## References
* Absil et al. (2006) Absil, O., et al. 2006, A&A, 452, 237
* Adachi et al. (1976) Adachi, I., Hayashi, C., & Nakazawa, K. 1976, Progress of Theoretical Physics 56, 1756
* Allen (1976) Allen, C. W. 1976, Astrophysical Quantities, Athlone, London, p. 197
* Andrews & Williams (2005) Andrews, S. M., & Williams, J. P. 2005, ApJ, 631, 1134
* Andrews & Williams (2007a) Andrews, S. M., & Williams, J. P. 2007a, ApJ, 659, 705
* Andrews & Williams (2007b) Andrews, S. M., & Williams, J. P. 2007b, ApJ, 671, 1800
* Artymowicz (1988) Artymowicz, P. 1988, ApJ, 335, L79
* Artymowicz (1997) Artymowicz, P. 1997, ARE&PS, 25, 175
* Asphaug & Benz (1996) Asphaug, E., & Benz, W. 1996, Icarus, 121, 225
* Aufdenberg et al. (2006) Aufdenberg, J. P., et al. 2006, ApJ, 645, 664
* Augereau & Beust (2006) Augereau, J.-C., & Beust, H. 2006, A&A, 455, 987
* Augereau et al. (1999) Augereau, J. C., Lagrange, A.-M., Mouillet, D., Papaloizou, J. C. B., & Grorod, P. A. 1999, A&A, 348, 557
* Aumann et al. (1984) Aumann, H. H., et al. 1984, ApJ, 278, L23
* Backman & Paresce (1993) Backman, D. E., & Paresce, F. 1993, in Protostars and Planets III, eds. E. H. Levy & J. I. Lunine, Tucson, Univ of Arizona, p. 1253
* Barge & Pellat (1991) Barge, P., & Pellat, R. 1991, Icarus, 93, 270
* Beichman et al. (2005) Beichman, C. A., et al. 2005, ApJ, 626, 1061
* Beichman et al. (2006) Beichman, C. A., et al. 2006, ApJ, 652, 1674
* Benz & Asphaug (1999) Benz, W., & Asphaug, E. 1999, Icarus, 142, 5
* Brandeker et al. (2004) Brandeker, A., Liseau, R., Olofsson, G., & Fridlund, M. 2004, A&A, 413, 681
* Bromley & Kenyon (2006) Bromley, B., & Kenyon, S. J. 2006, AJ, 131, 2737
* Brownlee et al. (1997) Brownlee, D. E., et al. 1997, Meteoritics & Planetary Science, vol. 32, page A22, 32, 22
* Bryden et al. (2006) Bryden, G., et al. 2006, ApJ, 636, 1098
* Burns, Lamy, & Soter (1979) Burns, J. A., Lamy, P. L., & Soter, S. 1979, Icarus, 40, 1
* Campo Bagatin et al. (1994) Campo Bagatin, A., Cellino, A., Davis, D. R., Farinella, P., & Paolicchi, P. 1994
* Carpenter et al. (2006) Carpenter, J. M., Mamajek, E. E., Hillenbrand, L. A., & Meyer, M. R. 2006, ApJ, 651, L49
* Chambers (2001) Chambers, J. E. 2001, Icarus, 152, 205
* Chambers (2006) Chambers, J. 2006, Icarus, 180, 496
* Charnoz & Morbidelli (2003) Charnoz, S., & Morbidelli, A. 2003, Icarus, 166, 141
* Chen et al. (2005) Chen, C. H., et al. 2005, ApJ, 634, 1372
* Chen et al. (2006) Chen, C. H., et al. 2006, ApJS, 166, 351
* Ciesla (2007) Ciesla, F. J. 2007, ApJ, 654, L159
* Currie et al. (2007a) Currie, T., et al. 2007, ApJ, 659, 599
* Currie et al. (2007b) Currie, T., Kenyon, S. J., Rieke, G., Balog, Z., & Bromley, B. C. 2007, ApJ, 663, L105
* Currie et al. (2007c) Currie, T., Kenyon, S. J., Balog, Z., Bragg, A., & Tokarz, S. 2007, ApJ, 669, L33
* Currie et al. (2008a) Currie, T., Kenyon, S. J., Balog, Z., Rieke, G., Bragg, A., & Bromley, B. 2008, ApJ, 672, 558
* Currie et al. (2008b) Currie, T., Plavchan, P., & Kenyon, S. J. 2008b, ApJ, submitted
* Davis et al. (1985) Davis, D. R., Chapman, C. R., Weidenschilling, S. J., & Greenberg, R. 1985, Icarus, 62, 30
* Decin et al. (2003) Decin, G., Dominik, C., Waters, L. B. F. M., & Waelkens, C. 2003, ApJ, 598, 636
* Demarque et al. (2004) Demarque, P., Woo, J.-H., Kim, Y.-C., & Yi, S. K. 2004, ApJS, 155, 667
* Dent et al. (2000) Dent, W. R. F., Walker, H. J., Holland, W. S., & Greaves, J. S. 2000, MNRAS, 314, 702
* Dohnanyi (1969) Dohnanyi, J. W. 1969, J. Geophys. Res., 74, 2531
* Dominik & Decin (2003) Dominik, C., & Decin, G. 2003, ApJ, 598, 626
* Dullemond & Dominik (2005) Dullemond, C. P., & Dominik, C. 2005, A&A, 434, 971
* Durda & Dermott (1997) Durda, D. D., & Dermott, S. F. 1997, Icarus, 130, 140
* Elliot et al. (2003) Elliot, J. L., Person, M. J., & Qu, S. 2003, AJ, 126, 1041
* Elliot et al. (2007) Elliot, J. L., et al. 2007, AJ, 134, 1
* Fitzgerald et al. (2007) Fitzgerald, M. P., Kalas, P. G., & Graham, J. R. 2007, ApJ, 670, 557
* Garaud (2007) Garaud, P. 2007, ApJ, 671, 2091
* Goldreich, Lithwick, & Sari (2004) Goldreich, P., Lithwick, Y., & Sari, R. 2004, ARA&A, 42, 549
* Golimowski et al. (1993) Golimowski, D. A., Durrance, S. T., & Clampin, M. 1993, ApJ, 411, L41
* Gomes et al. (2005) Gomes, R., Levison, H. F., Tsiganis, K., & Morbidelli, A. 2005, Nature, 435, 466
* Gorlova et al. (2007) Gorlova, N., Balog, Z., Rieke, G. H., Muzerolle, J., Su, K. Y. L., Ivanov, V. D., & Young, E. T. 2007, ApJ, 670, 516
* Gorlova et al. (2006) Gorlova, N., Rieke, G. H., Muzerolle, J., Stauffer, J. R., Siegler, N., Young, E. T., & Stansberry, J. H. 2006, ApJ, 649, 1028
* Greaves et al. (1998) Greaves, J. S. et al. 1998, ApJ, 506, L133
* Greaves et al. (2000) Greaves J. S., Mannings V. & Holland, W. S. 2000b, Icarus, 143, 155
* Greaves & Wyatt (2003) Greaves, J. S., & Wyatt, M. C. 2003, MNRAS, 345, 1212
* Greenberg et al. (1984) Greenberg, R., Weidenschilling, S. J., Chapman, C. R., & Davis, D. R. 1984, Icarus, 59, 87
* Greenberg et al. (1991) Greenberg, R., Bottke, W., Carusi, A., Valsecchi, G. B. 1991, Icarus, 94, 98
* Greenzweig & Lissauer (1990) Greenzweig, Y., & Lissauer, J. J. 1990, Icarus, 87, 40
* Greenzweig & Lissauer (1992) Greenzweig, Y., & Lissauer, J. J. 1992, Icarus, 100, 440
* Grigorieva et al. (2007) Grigorieva, A., Artymowicz, P., & Thébault, P. 2007, A&A, 461, 537
* Grün et al. (1995) Grün, E., et al. 1995, Planet. Space Sci., 43, 971
* Habing et al. (2001) Habing, H. J., et al. 2001, A&A, 365, 545
* Hahn et al. (2002) Hahn, J. M., Zook, H. A., Cooper, B., & Sunkara, B. 2002, Icarus, 158, 360
* Haisch, Lada, & Lada (2001) Haisch, K., Lada, E. A., & Lada, C. J. 2001, ApJ, 553, 153
* Hayashi (1981) Hayashi, C. 1981, Prog Theor Phys Suppl, 70, 35
* Hernández et al. (2006) Hernández, J., Briceño, C., Calvet, N., Hartmann, L., Muzerolle, J., & Quintero, A. 2006, ApJ, 652, 472
* Hillenbrand et al. (2008) Hillenbrand, L. A., et al. 2008, ArXiv e-prints, 801, arXiv:0801.0163
* Holland et al. (1998) Holland, W. S., et al. 1998, Nature, 392, 788
* Holland et al. (2003) Holland, W. S., et al. 2003, ApJ, 582, 1141
* Hornung et al. (1985) Hornung, P., Pellat, R., & Barge, P. 1985, Icarus, 64, 295
* Iben (1967) Iben, I. Jr, 1967, ARA&A, 5, 571
* Ida (1990) Ida, S. 1990, Icarus, 88, 129
* Ida & Lin (2008) Ida, S., & Lin, D. N. C. 2008, ArXiv e-prints, 802, arXiv:0802.1114
* Ida & Makino (1992) Ida, S., & Makino, J. 1992, Icarus, 96, 107
* Ida & Makino (1993) Ida, S., & Makino, J. 1993, Icarus, 106, 210
* Inaba et al. (2001) Inaba, S. H., Tanaka, H., Nakazawa, K., Wetherill, G. W., & Kokubo, E. 2001, Icarus, 149, 235
* Jayawardhana et al. (1998) Jayawardhana, R. et al. 1998, ApJ, 503, L79
* Johnson et al. (2007) Johnson, J. A., et al. 2007, ApJ, 665, 785
* Kalas (1998) Kalas, P. 1998, Earth, Moon, & Planets, 81, 27
* Kalas (2005) Kalas, P. 2005, ApJ, 635, L169
* Kalas et al. (2005) Kalas, P., Graham, J. R., & Clampin, M. 2005, Nature, 435, 1067
* Kalas et al. (2006) Kalas, P., Graham, J. R., Clampin, M. C., & Fitzgerald, M. P. 2006, ApJ, 637, L57
* Kalas et al. (2004) Kalas, P., Liu, M. C., & Matthews, B. C. 2004, Science, 303, 1990
* Kennedy & Kenyon (2008) Kennedy, G. M., & Kenyon, S. J. 2008, ApJ, 673, 502
* Kenyon & Bromley (2001) Kenyon, S. J., & Bromley, B. C. 2001, AJ, 121, 538
* Kenyon & Bromley (2002a) Kenyon, S. J., & Bromley, B. C. 2002a, AJ, 123, 1757
* Kenyon & Bromley (2002b) Kenyon, S. J., & Bromley, B. C. 2002b, ApJ, 577, L35
* Kenyon & Bromley (2004a) Kenyon, S. J., & Bromley, B. C., 2004a, AJ, 127, 513
* Kenyon & Bromley (2004b) Kenyon, S. J., & Bromley, B. C., 2004b, ApJ, 602, L133
* Kenyon & Bromley (2004c) Kenyon, S. J., & Bromley, B. C., 2004c, AJ, 128, 1916
* Kenyon & Bromley (2005) Kenyon, S. J., & Bromley, B. C. 2005, AJ, 130, 269
* Kenyon & Bromley (2006) Kenyon, S. J., & Bromley, B. C. 2006, AJ, 131, 1837
* Kenyon et al. (2008) Kenyon, S. J., Bromley, B. C., O’Brien, D. C., & Davis, D. R. 2008, to appear in The Solar System Beyond Neptune, edited by A. Barucci, H. Boehnhardt, D. Cruikshank, & A. Morbidelli, Tucson, Univ. of Arizona Press, in press
* Kenyon & Hartmann (1987) Kenyon, S. J., & Hartmann, L. 1987, ApJ, 323, 714
* Kenyon & Hartmann (1995) Kenyon, S. J., & Hartmann, L. W., 1995, ApJS, 101, 117
* Kenyon & Luu (1998) Kenyon, S. J., & Luu, J. X. 1998, AJ, 115, 2136
* Kenyon & Luu (1999) Kenyon, S. J., & Luu, J. X. 1999, AJ, 118, 1101
* Kenyon et al. (1999) Kenyon, S. J., Wood, K., Whitney, B. A., & Wolff, M. 1999, ApJ, 524, L119
* Kim et al. (2005) Kim, J. S., et al. 2005, ApJ, 632, 659
* Kimura et al. (2002) Kimura, H., Okamoto, H., & Mukai, T. 2002, Icarus, 157, 349
* Knapp & Morris (1985) Knapp, G. R., & Morris, M. 1985, ApJ, 292, 640
* Kobayashi & Ida (2001) Kobayashi, H., & Ida, S. 2001, Icarus, 153, 416
* Koeberl (2003) Koeberl, C. 2003, EM&P, 92, 79
* Koerner et al. (1998) Koerner, D. W., Ressler, M. E., Werner, M. W., & Backman, D. E. 1998, ApJ, 503, L83
* Koerner, Sargent, & Ostroff (2001) Koerner, D. W., Sargent, A. I., & Ostroff, N. A. 2001, ApJ, 560, L181
* Kokubo & Ida (1998) Kokubo, E., & Ida, S. 1998, Icarus, 131, 171
* Kretke et al. (2008) Kretke, K. A., Lin, D. N. C., Garaud, P., & Turner, N. J. 2008, ArXiv e-prints, 806, arXiv:0806.1521
* Krivov et al. (2000) Krivov, A. V., Mann, I., & Krivova, N. A. 2000, A&A, 362, 1127
* Krivov et al. (2006) Krivov, A. V., Löhne, T., & Sremčević, M. 2006, A&A, 455, 509
* Kuchner & Holman (2003) Kuchner, M. J., & Holman, M. J. 2003, ApJ, 588, 1110
* Lada (1999) Lada, C. J. 1999, in The Physics of Star Formation and Early Stellar Evolution, edited by C. J. Lada and N. Kylafis, Dordrecht, Kluwer, p. 143
* Lagrange et al. (2000) Lagrange, A.-M., Backman, D., & Artymowicz, P. 2000, in Protostars & Planets IV, eds. V. Mannings, A. P. Boss, & S. S. Russell, Tucson, Univ. of Arizona, in press
* Landgraf et al. (2002) Landgraf, M., Liou, J.-C., Zook, H. A., & Grün, E. 2002, AJ, 123, 2857
* Larwood (1997) Larwood, J. D. 1997, MNRAS, 290, 490
* Larwood & Kalas (2001) Larwood, J. D., & Kalas, P. G. 2001, MNRAS, 323, 402
* Leinhardt et al. (2008) Leinhardt, Z. M., Stewart, S. T., & Schultz, P. H. 2008, to appear in The Solar System Beyond Neptune, edited by A. Barucci, H. Boehnhardt, D. Cruikshank, & A. Morbidelli, Tucson, Univ. of Arizona Press, in press (ArXiv e-prints, 705, arXiv:0705.3943)
* Levison & Stewart (2001) Levison, H. F., & Stewart, G. R. 2001, Icarus, 153, 224
* Lin & Papaloizou (1979) Lin, D. N. C., & Papaloizou, J. C. B. 1979, MNRAS, 186, 799
* Lissauer (1987) Lissauer, J. J. 1987, Icarus, 69, 249
* Lissauer & Stewart (1993) Lissauer, J. J., & Stewart, G. R. 1993, In Protostars and Planets III, edited by E. H. Levy and J. I. Lunine, U. of Arizona Press, Tucson, 1061
* Lisse et al. (2007a) Lisse, C. M., Beichman, C. A., Bryden, G., & Wyatt, M. C. 2007, ApJ, 658, 584
* Lisse et al. (2007c) Lisse, C. M., Chen, C. H., Wyatt, M. C., & Morlok, A. 2007, ArXiv e-prints, 710, arXiv:0710.0839
* Lisse et al. (2007b) Lisse, C. M., Kraemer, K. E., Nuth, J. A., Li, A., & Joswiak, D. 2007, Icarus, 191, 223
* Liu (2004) Liu, M. C. 2004, Science, 305, 1442
* Liu et al. (2004) Liu, M. C., Matthews, B. C., Williams, J. P., & Kalas, P. G. 2004, ApJ, 608, 526
* Liu et al. (2004) Liu, W. M., et al. 2004, ApJ, 610, L125
* Löhne et al. (2008) Löhne, T., Krivov, A. V., & Rodmann, J. 2008, ApJ, 673, 1123
* Marsh et al. (2006) Marsh, K. A., Dowell, C. D., Velusamy, T., Grogan, K., & Beichman, C. A. 2006, ApJ, 646, L77
* Matthews et al. (2007) Matthews, B. C., et al. 2007, PASP, 119, 842
* Melosh, Vickery, & Tonks (1993) Melosh, H. J., Vockery, A. M., & Tonks, W. B. 1993, in Protostars and Planets III, eds. E. H. Levy & J. I. Lunine, Tucson, Univ of Arizona, p. 1339
* Meyer et al. (2007) Meyer, M. R., Backman, D. E., Weinberger, A. J., & Wyatt, M. C. 2007, Protostars and Planets V, 573
* Meyer et al. (2006) Meyer, M. R., et al. 2006, PASP, 118, 1690
* Meyer et al. (2008) Meyer, M. R., et al. 2008, ApJ, 673, L181
* Moór et al. (2006) Moór, A., Ábrahám, P., Derekas, A., Kiss, C., Kiss, L. L., Apai, D., Grady, C., & Henning, T. 2006, ApJ, 644, 525
* Moro-Martín & Malhotra (2005) Moro-Martín, A., & Malhotra, R. 2005, ApJ, 633, 1150
* Moro-Martin et al. (2007) Moro-Martin, A., Wyatt, M. C., Malhotra, R., & Trilling, D. E. 2007, ArXiv Astrophysics e-prints, arXiv:astro-ph/0703383
* Motte & André (2001) Motte, F., & André, P. 2001, A&A, 365, 440
* Mouillet et al. (1997) Mouillet, D., Larwood, J. D., Papaloizou, J. C. B., & Lagrange, A.-M. 1997, MNRAS, 292, 896
* Nagasawa et al. (2007) Nagasawa, M., Thommes, E. W., Kenyon, S. J., Bromley, B. C., & Lin, D. N. C. 2007, Protostars and Planets V, 639
* Najita & Williams (2005) Najita, J., & Williams, J. P. 2005, ApJ, 635, 625
* Natta et al. (2000) Natta, A., Grinin, V., & Mannings, V. 2000, Protostars and Planets IV, 559
* Nesvorný et al. (2006) Nesvorný, D., Vokrouhlický, D., Bottke, W. F., & Sykes, M. 2006, Icarus, 181, 107
* Nomura & Nakagawa (2006) Nomura, H., & Nakagawa, Y. 2006, ApJ, 640, 1099
* Ohtsuki (1992) Ohtsuki, K. 1992, Icarus, 98, 20
* Ohtsuki, Stewart, & Ida (2002) Ohtsuki, K., Stewart, G. R., & Ida, S. 2002, Icarus, 155, 436
* Osterloh & Beckwith (1995) Osterloh, M., & Beckwith, S. V. W. 1995, ApJ, 439, 288
* Pan & Sari (2005) Pan, M., & Sari, R. 2005, Icarus, 173, 342
* Press et al. (1992) Press, W. H., Flannery, B. P., Teukolsky, S. A., & Vetterling, W. T. 1992, Numerical Recipes, The Art of Scientific Computing, Cambridge, Cambridge
* Quillen (2006) Quillen, A. C. 2006, MNRAS, 372, L14
* Quillen et al. (2007) Quillen, A. C., Morbidelli, A., & Moore, A. 2007, MNRAS, 380, 1642
* Rafikov (2003) Rafikov, R. R. 2003, AJ, 125, 942
* Rafikov (2004) Rafikov, R. R. 2004, AJ, 128, 1348
* Rhee et al. (2007a) Rhee, J. H., Song, I., Zuckerman, B., & McElwain, M. 2007, ApJ, 660, 1556
* Rhee et al. (2007b) Rhee, J. H., Song, I., & Zuckerman, B. 2007, ArXiv e-prints, 711, arXiv:0711.2111
* Rieke et al. (2005) Rieke, G. H., Su, K. Y. L., Stansberry, J. A., Trilling, D., Bryden, G., Muzerolle, J., White, B., Gorlova, N., Young, E. T., Beichman, C. A., Stapelfeldt, K. R., & hines, D. C. 2005, ApJ, 620, 1010
* Safronov (1969) Safronov, V. S. 1969, Evolution of the Protoplanetary Cloud and Formation of the Earth and Planets, Nauka, Moscow [Translation 1972, NASA TT F-677]
* Schneider et al. (1999) Schneider, G., et al. 1999, ApJ, 513, L127
* Scholz et al. (2006) Scholz, A., Jayawardhana, R., & Wood, K. 2006, ApJ, 645, 1498
* Sicilia-Aguilar et al. (2006) Sicilia-Aguilar, A., et al. 2006, ApJ, 638, 897
* Siegler et al. (2007) Siegler, N., Muzerolle, J., Young, E. T., Rieke, G. H., Mamajek, E. E., Trilling, D. E., Gorlova, N., & Su, K. Y. L. 2007, ApJ, 654, 580
* Smith & Terrile (1984) Smith, B. A., & Terrile, R. J. 1984, Science, 226, 1421
* Song et al. (2005) Song, I., Zuckerman, B., Weinberger, A. J., & Becklin, E. E. 2005, Nature, 436, 363
* Spaute et al. (1991) Spaute, D., Weidenschilling, S. J., Davis, D. R., & Marzari, F. 1991, Icarus, 92, 147
* Stapelfeldt et al. (2004) Stapelfeldt, K. R., et al. 2004, ApJS, 154, 458
* Stauffer et al. (2005) Stauffer, J. R., et al. 2005, AJ, 130, 1834
* Stern & Colwell (1997) Stern, S. A., & Colwell, J. E. 1997a, AJ, 114, 841
* Stewart & Ida (2000) Stewart, G. R., & Ida, S. 2000, Icarus, 143, 28
* Su et al. (2005) Su, K. Y. L., et al. 2005, ApJ, 628, 487
* Su et al. (2006) Su, K. Y. L., et al. 2006, ApJ, 653, 675
* Su et al. (2008) Su, K. Y. L., Rieke, G. H., Stapelfeldt, K. R., Smith, P. S., Bryden, G., Chen, C. H., & Trilling, D. E. 2008, ApJ, 679, L125
* Swindle (1993) Swindle, T. D. 1993, in Protostars and Planets III, eds. E. H. Levy & J. I. Lunine, Tucson, Univ of Arizona, p. 867
* Takeuchi & Artymowicz (2001) Takeuchi, T., & Artymowicz, P. 2001, ApJ, 557, 990
* Telesco et al. (1988) Telesco, C. M., Decher, R., Becklin, E. E., & Wolstencroft, R. D. 1988, Nature, 335, 51
* Telesco et al. (2000) Telesco, C. M., et al. 2000, ApJ, 530, 329
* Thébault & Augereau (2007) Thébault, P., & Augereau, J.-C. 2007, A&A, 472, 169
* Thébault, Augereau, & Beust (2003) Thébault, P., Augereau, J. C., & Beust, H. 2003, A&A, 408, 775
* Trilling et al. (2007) Trilling, D. E., et al. 2007, ApJ, 658, 1289
* Trilling et al. (2008) Trilling, D. E., et al. 2008, ApJ, 674, 1086
* Wadhwa & Russell (2000) Wadhwa, M., & Russell, S. S. 2000, in Protostars abd Planets IV, eds. V. Mannings, A. P. Boss, & S. S. Russell, Tucson, Univ. of Arizona, p. 995
* Weidenschilling (1977a) Weidenschilling, S. J. 1977a, Astrophys Sp Sci, 51, 153
* Weidenschilling (1977b) Weidenschilling, S. J. 1977b, MNRAS, 180, 57
* Weidenschilling (1989) Weidenschilling, S. J. 1989, Icarus, 80, 179
* Weidenschilling et al. (1997) Weidenschilling, S. J., Spaute, D., Davis, D. R., Marzari, F., & Ohtsuki, K. 1997, Icarus, 128, 429
* Wetherill (1980) Wetherill, G. W. 1980, ARA&A, 18, 77
* Wetherill & Stewart (1993) Wetherill, G. W., & Stewart, G. R. 1993, Icarus, 106, 190
* Williams & Wetherill (1994) Williams, D. R., & Wetherill, G. W. 1994, Icarus, 107, 117
* Williams & Andrews (2006) Williams, J. P., & Andrews, S. M. 2006, ApJ, 653, 1480
* Wilner et al. (2002) Wilner, D. J., Holman, M. J., Kuchner, M. J., & Ho, P. T. P. 2002, ApJ, 569, 115
* Wolf & Hillenbrand (2003) Wolf, S., & Hillenbrand, L. A. 2003, ApJ, 596, 603
* Wood et al. (2002) Wood, K., Lada, C. J., Bjorkman, J. E., Kenyon, S. J., Whitney, B., Wolff, M. J. 2002, ApJ, 567, 1183
* Wyatt (2003) Wyatt, M. C. 2003, ApJ, 598, 1321
* Wyatt (2005) Wyatt, M. C. 2005, A&A, 433, 1007
* Wyatt & Dent (2002) Wyatt, M. C., & Dent, W. R. F. 2002, MNRAS, 334, 589
* Wyatt, Dent, & Greaves (2003) Wyatt, M. C., Dent, W. R. F., & Greaves, J. S. 2003, MNRAS, 342, 867
* Wyatt et al. (1999) Wyatt, M. C., Dermott, S. F., Telesco, C. M., Fisher, R. S., Grogan, K., Holmes, E. K., & Piña, R. K. 1999, ApJ, 527, 918
* Wyatt et al. (2007a) Wyatt, M. C., Smith, R., Greaves, J. S., Beichman, C. A., Bryden, G., & Lisse, C. M. 2007, ApJ, 658, 569
* Wyatt et al. (2007b) Wyatt, M. C., Smith, R., Su, K. Y. L., Rieke, G. H., Greaves, J. S., Beichman, C. A., & Bryden, G. 2007, ApJ, 663, 365
* Yin et al. (2002) Yin, Q., Jacobsen, S. B., Yamashita, K., Blichert-Toft, J., Télouk, P.; Albarède, F. 2002, Nature, 418, 949
* Young & Binzel (1994) Young, E. F., & Binzel, R. P. 1994, Icarus, 108, 219
* Zuckerman (2001) Zuckerman, B. 2001, ARA&A, 39, 549
Table 1: Grid of Debris Disk CalculationsaaNumber of independent calculations for each combination of $M_{\star},x_{m}$ | Stellar Mass in $M_{\odot}$
---|---
$x_{m}$ | 1.0bb32 annulus models at 30–70 AU and at 70–150 AU | 1.0cc64 annulus models at 30–150 AU | 1.5 | 2.0 | 2.5 | 3.0
0.33 | 41 | 19 | 15 | 20 | 18 | 18
0.50 | 49 | 18 | 18 | 17 | 17 | 17
1.00 | 49 | 15 | 22 | 17 | 15 | 15
2.00 | 41 | 15 | 30 | 17 | 16 | 15
3.00 | 45 | 18 | 12 | 22 | 15 | 21
$t_{ms}$ddMain sequence lifetime in Gyr Demarque et al. (2004) | 10.00 | 10.00 | 2.90 | 1.22 | 0.65 | 0.39
Table 2: Median number of Plutos at t = $t_{ms}$/3 for disks around 1 M⊙ stars $x_{m}$ | 30–37 AU | 37–45 AU | 45–55 AU | 55–67 AU | 67–82 AU | 82–100 AU | 100–123 AU | 123–146 AU
---|---|---|---|---|---|---|---|---
0.33 | 40 | 44 | 33 | 32 | 19 | 12 | 3 | 1
0.50 | 62 | 65 | 39 | 49 | 33 | 25 | 10 | 1
1.00 | 111 | 110 | 73 | 73 | 47 | 55 | 26 | 5
2.00 | 172 | 194 | 134 | 155 | 116 | 84 | 58 | 33
3.00 | 165 | 260 | 172 | 251 | 137 | 109 | 85 | 44
Table 3: Median number of Plutos at t = $t_{ms}$/3 for disks around 1.5–3 M⊙ stars $x_{m}$ | 30–37 AU | 37–45 AU | 45–55 AU | 55–67 AU | 67–82 AU | 82–100 AU | 100–123 AU | 123–146 AU
---|---|---|---|---|---|---|---|---
1.5 M⊙
0.33 | 46 | 46 | 45 | 28 | 16 | 3 | 1 | 1
0.50 | 69 | 85 | 68 | 48 | 53 | 25 | 2 | 1
1.00 | 102 | 136 | 128 | 101 | 98 | 67 | 28 | 2
2.00 | 158 | 243 | 240 | 261 | 211 | 123 | 169 | 47
3.00 | 201 | 239 | 301 | 295 | 381 | 165 | 198 | 97
2.0 M⊙
0.33 | 55 | 51 | 48 | 30 | 2 | 1 | 1 | 3
0.50 | 115 | 87 | 85 | 51 | 20 | 10 | 1 | 2
1.00 | 172 | 187 | 231 | 123 | 82 | 50 | 1 | 3
2.00 | 261 | 230 | 366 | 236 | 324 | 204 | 77 | 5
3.00 | 259 | 262 | 398 | 264 | 446 | 295 | 171 | 48
2.5 M⊙
0.33 | 105 | 92 | 63 | 55 | 19 | 1 | 5 | 2
0.50 | 152 | 139 | 123 | 100 | 53 | 10 | 5 | 5
1.00 | 164 | 174 | 234 | 198 | 125 | 90 | 7 | 6
2.00 | 223 | 230 | 278 | 243 | 158 | 108 | 10 | 8
3.00 | 353 | 434 | 490 | 495 | 477 | 615 | 313 | 157
3.0 M⊙
0.33 | 133 | 121 | 62 | 36 | 6 | 5 | 1 | 1
0.50 | 127 | 153 | 129 | 103 | 50 | 2 | 2 | 1
1.00 | 199 | 230 | 239 | 258 | 171 | 66 | 6 | 5
2.00 | 224 | 353 | 376 | 392 | 342 | 265 | 173 | 5
3.00 | 428 | 598 | 479 | 657 | 570 | 756 | 578 | 272
Table 4: Predicted Excesses for Disks Around 1 M⊙ StarsaaThe electronic version of this paper contains the complete version of this Table. log $t$ (yr) | log $L_{d}/L_{\star}$ | log $F_{24}/F_{24,0}$ | log $F_{70}/F_{70,0}$ | log $F_{160}/F_{160,0}$ | log $F_{850}/F_{850,0}$
---|---|---|---|---|---
$x_{m}$ = 0.33
5.05 | -4.57 | 0.000 | 0.053 | 0.145 | 0.090
5.15 | -4.57 | 0.000 | 0.053 | 0.145 | 0.090
5.25 | -4.63 | 0.000 | 0.047 | 0.131 | 0.082
5.35 | -4.65 | 0.000 | 0.045 | 0.127 | 0.080
5.45 | -4.67 | 0.000 | 0.043 | 0.123 | 0.077
$x_{m}$ = 1.0
5.05 | -4.18 | 0.000 | 0.096 | 0.255 | 0.172
5.15 | -4.18 | 0.000 | 0.096 | 0.255 | 0.172
5.25 | -4.29 | 0.000 | 0.090 | 0.243 | 0.164
5.35 | -4.31 | 0.000 | 0.088 | 0.238 | 0.161
5.45 | -4.33 | 0.000 | 0.086 | 0.233 | 0.158
$x_{m}$ = 3.0
5.05 | -4.02 | 0.000 | 0.162 | 0.464 | 0.357
5.15 | -4.06 | 0.000 | 0.150 | 0.410 | 0.307
5.25 | -4.10 | 0.000 | 0.138 | 0.355 | 0.257
5.35 | -4.12 | 0.000 | 0.133 | 0.354 | 0.248
5.45 | -4.14 | 0.000 | 0.127 | 0.332 | 0.238
Table 5: Predicted Excesses for Disks Around 1.5 M⊙ StarsaaThe electronic version of this paper contains the complete version of this Table. log $t$ (yr) | log $L_{d}/L_{\star}$ | log $F_{24}/F_{24,0}$ | log $F_{70}/F_{70,0}$ | log $F_{160}/F_{160,0}$ | log $F_{850}/F_{850,0}$
---|---|---|---|---|---
$x_{m}$ = 0.33
5.05 | -4.40 | 0.001 | 0.100 | 0.202 | 0.110
5.15 | -4.45 | 0.001 | 0.090 | 0.184 | 0.099
5.25 | -4.51 | 0.001 | 0.078 | 0.163 | 0.088
5.35 | -4.52 | 0.001 | 0.074 | 0.157 | 0.085
5.45 | -4.53 | 0.001 | 0.070 | 0.151 | 0.082
$x_{m}$ = 1.0
5.05 | -4.01 | 0.001 | 0.162 | 0.321 | 0.188
5.15 | -4.19 | 0.001 | 0.148 | 0.299 | 0.178
5.25 | -4.22 | 0.001 | 0.140 | 0.287 | 0.170
5.35 | -4.24 | 0.001 | 0.133 | 0.275 | 0.163
5.45 | -4.24 | 0.001 | 0.128 | 0.266 | 0.157
$x_{m}$ = 3.0
5.05 | -3.73 | 0.002 | 0.206 | 0.406 | 0.261
5.15 | -4.01 | 0.002 | 0.199 | 0.394 | 0.251
5.25 | -4.04 | 0.002 | 0.192 | 0.382 | 0.242
5.35 | -4.05 | 0.002 | 0.181 | 0.361 | 0.227
5.45 | -4.08 | 0.002 | 0.167 | 0.339 | 0.211
Table 6: Predicted Excesses for Disks Around 2.0 M⊙ StarsaaThe electronic version of this paper contains the complete version of this Table. log $t$ (yr) | log $L_{d}/L_{\star}$ | log $F_{24}/F_{24,0}$ | log $F_{70}/F_{70,0}$ | log $F_{160}/F_{160,0}$ | log $F_{850}/F_{850,0}$
---|---|---|---|---|---
$x_{m}$ = 0.33
5.05 | -4.28 | 0.002 | 0.128 | 0.215 | 0.109
5.15 | -4.34 | 0.002 | 0.118 | 0.201 | 0.102
5.25 | -4.41 | 0.002 | 0.108 | 0.191 | 0.096
5.35 | -4.43 | 0.002 | 0.103 | 0.183 | 0.092
5.45 | -4.46 | 0.002 | 0.098 | 0.175 | 0.087
$x_{m}$ = 1.0
5.05 | -3.89 | 0.003 | 0.195 | 0.334 | 0.185
5.15 | -4.00 | 0.003 | 0.186 | 0.320 | 0.177
5.25 | 4.11 | 0.003 | 0.177 | 0.309 | 0.170
5.35 | -4.16 | 0.003 | 0.175 | 0.307 | 0.168
5.45 | -4.18 | 0.003 | 0.161 | 0.286 | 0.156
$x_{m}$ = 3.0
5.05 | -3.71 | 0.005 | 0.248 | 0.421 | 0.250
5.15 | -3.98 | 0.005 | 0.239 | 0.409 | 0.240
5.25 | -4.02 | 0.004 | 0.227 | 0.389 | 0.226
5.35 | -4.02 | 0.004 | 0.207 | 0.358 | 0.205
5.45 | -4.07 | 0.004 | 0.192 | 0.333 | 0.188
Table 7: Predicted Excesses for Disks Around 2.5 M⊙ StarsaaThe electronic version of this paper contains the complete version of this Table. log $t$ (yr) | log $L_{d}/L_{\star}$ | log $F_{24}/F_{24,0}$ | log $F_{70}/F_{70,0}$ | log $F_{160}/F_{160,0}$ | log $F_{850}/F_{850,0}$
---|---|---|---|---|---
$x_{m}$ = 0.33
5.05 | -4.18 | 0.004 | 0.170 | 0.259 | 0.120
5.15 | -4.22 | 0.004 | 0.153 | 0.233 | 0.108
5.25 | -4.34 | 0.004 | 0.137 | 0.207 | 0.096
5.35 | -4.36 | 0.004 | 0.130 | 0.197 | 0.093
5.45 | -4.38 | 0.004 | 0.123 | 0.188 | 0.086
$x_{m}$ = 1.0
5.05 | -3.97 | 0.008 | 0.237 | 0.352 | 0.181
5.15 | -4.08 | 0.007 | 0.219 | 0.329 | 0.168
5.25 | -4.09 | 0.000 | 0.000 | 0.000 | 0.000
5.35 | -4.14 | 0.006 | 0.201 | 0.306 | 0.154
5.45 | -4.14 | 0.005 | 0.187 | 0.287 | 0.144
$x_{m}$ = 3.0
5.05 | -3.83 | 0.010 | 0.285 | 0.424 | 0.230
5.15 | -3.95 | 0.009 | 0.270 | 0.403 | 0.217
5.25 | -3.98 | 0.009 | 0.258 | 0.387 | 0.206
5.35 | -4.03 | 0.009 | 0.248 | 0.372 | 0.197
5.45 | -3.48 | 0.087 | 0.749 | 0.887 | 0.558
Table 8: Predicted Excesses for Disks Around 3.0 M⊙ StarsaaThe electronic version of this paper contains the complete version of this Table. log $t$ (yr) | log $L_{d}/L_{\star}$ | log $F_{24}/F_{24,0}$ | log $F_{70}/F_{70,0}$ | log $F_{160}/F_{160,0}$ | log $F_{850}/F_{850,0}$
---|---|---|---|---|---
$x_{m}$ = 0.33
5.05 | -4.08 | 0.011 | 0.173 | 0.229 | 0.099
5.15 | -4.11 | 0.010 | 0.164 | 0.213 | 0.091
5.25 | -4.14 | 0.009 | 0.152 | 0.198 | 0.083
5.35 | -4.16 | 0.008 | 0.140 | 0.183 | 0.076
5.45 | -4.19 | 0.007 | 0.134 | 0.174 | 0.072
$x_{m}$ = 1.0
5.05 | -3.96 | 0.015 | 0.254 | 0.326 | 0.150
5.15 | -4.01 | 0.014 | 0.239 | 0.310 | 0.142
5.25 | -4.04 | 0.013 | 0.227 | 0.297 | 0.135
5.35 | -4.07 | 0.012 | 0.216 | 0.282 | 0.127
5.45 | -4.10 | 0.011 | 0.204 | 0.269 | 0.120
$x_{m}$ = 3.0
5.05 | -3.83 | 0.020 | 0.318 | 0.410 | 0.199
5.15 | -3.88 | 0.019 | 0.301 | 0.390 | 0.189
5.25 | -3.92 | 0.030 | 0.321 | 0.389 | 0.187
5.35 | -3.38 | 0.113 | 0.661 | 0.740 | 0.416
5.45 | -3.01 | 0.226 | 0.934 | 0.987 | 0.592
Table 9: Debris disk loci in color-color space $M_{\star}$ (M⊙) | x0,y0 | xu,yu | xl,yl
---|---|---|---
1.0 | 0.00,0.0 | 4.0,0.1 | 5.00,0.00
1.5 | 1.50,0.0 | 4.0,1.5 | 4.50,0.25
2.0 | 1.25,0.0 | 2.5,2.5 | 3.50,0.50
2.5 | 1.00,0.0 | 2.0,4.0 | 3.00,1.00
3.0 | 1.00,0.0 | 1.5,5.0 | 2.25,1.75
Table 10: Vega Debris Disk ModelaaFluxes ($F$) are in units of Jy; dust mass in 1–50 $\mu$m particles ($M_{d,1-50}$) is in units of $10^{-3}$ M⊕; dust production rate ($\dot{M}$) is in units of $10^{21}$ g yr-1. Source | $F_{24,\star}$ | $F_{70,\star}$ | $F_{160,\star}$ | $F_{850,\star}$ | $F_{24,disk}$ | $F_{70,disk}$ | $F_{160,disk}$ | $F_{850,disk}$ | $M_{d,1-50}$ | $\dot{M}$
---|---|---|---|---|---|---|---|---|---|---
Vega | 7.2 | 0.8 | 0.16 | 0.006 | 1.5 | 7.0 | 4.0 | 0.091 | 3.0 | 30bbDust production rate from Su et al. (2005). Our analysis suggests a smaller dust production rate, $\dot{M}\gtrsim 0.3\times 10^{21}$ g yr-1.
Model 1ccDebris disk model with $M_{\star}$ = 2 M⊙, $x_{m}$ = 1/3, $t$ = 200 Myr | 7.2 | 0.8 | 0.16 | 0.006 | 4.2 | 10.0 | 3.0 | 0.05 | 0.9 | 0.2
Model 2ddAs in note (c) for $M_{\star}$ = 2 M⊙, $x_{m}$ = 1, $t$ = 200 Myr | 7.2 | 0.8 | 0.16 | 0.006 | 7.2 | 23.0 | 8.0 | 0.11 | 3.8 | 0.6
Model 3eeAs in note (c) for $M_{\star}$ = 2.5 M⊙, $x_{m}$ = 1/2, $t$ = 200 Myr | 7.2 | 0.8 | 0.16 | 0.006 | 10.9 | 15.0 | 4.0 | 0.05 | 2.4 | 0.3
Model 4ffAs in note (c) for $M_{\star}$ = 2.5 M⊙, $x_{m}$ = 1, $t$ = 200 Myr | 7.2 | 0.8 | 0.16 | 0.006 | 15.5 | 23.0 | 5.0 | 0.07 | 4.8 | 0.8
Figure 1: Evolution of a multiannulus coagulation model with
$\Sigma=0.18(a_{i}/{\rm 30~{}AU})^{-3/2}$ g cm-2 at 30–37 AU around a 1 M⊙
star. Left: median cumulative mass distribution at times indicated in the
legend. Right: median eccentricity distributions at $t$ = 0 (light solid
line), $t$ = 10 Myr (filled circles), $t$ = 100 Myr (open boxes), $t$ = 1 Gyr
(filled triangles), and $t$ = 10 Gyr (open diamonds). As large objects grow in
the disk, they stir up the leftover planetesimals to $e\sim$ 0.1. Disruptive
collisions then deplete the population of 0.1–10 km planetesimals, which
limits the growth of the largest objects.
Figure 2: Evolution of the radius of the largest object in each annulus for a
MMSN disk around a 1 $M_{\odot}$ star. The number to the right of each set of
points indicates log $t$ in yr from the start of the calculation. Large
objects with $r\sim$ 1000 km form at the inner edge of the disk in 10–100 Myr,
in the middle of the disk in 0.3–1 Gyr, and at the edge of the disk in 10 Gyr.
Figure 3: Median radii of the largest objects at 1 Gyr for disks around a 1
$M_{\odot}$ star. The numbers to the right of each set of points indicate
$x_{m}$, the disk mass in units of the MMSN. Planets form earlier in the inner
portions of the most massive disks. Icy planet formation saturates at maximum
radii $r\sim$ 1500 km.
Figure 4: Median cumulative mass distributions at 1 Gyr for annuli at 30–37 AU
(left panel) and at 123–146 AU (right panel) around a 1 M⊙ star. The legend
indicates $x_{m}$, the initial disk mass in units of the scaled MMSN. In the
inner disk, many large planets form and the collisional cascade removes nearly
all of the material in objects with $r\lesssim$ 1–10 km. In the outer disk,
few large planets form; collisions are inefficient at removing material in
small objects.
Figure 5: Radius of the largest object in each annulus at 10 Myr and at 100
Myr for a MMSN disk around a 1 $M_{\odot}$ star. The black points indicate
results for calculations with 64 annuli; the magenta and cyan points show
results for calculations with 32 annuli. The good agreement between the 32
annulus and 64 annulus calculations shows that planet formation is not
sensitive to the size of the grid. Figure 6: Median production rate of 0.01–1
$\mu$m objects as a function of time for 30–150 AU disks around a 1 M⊙ central
star. The legend indicates $x_{m}$, the initial disk mass in units of the
scaled MMSN. As large objects grow during the early stages of the evolution,
the dust production rate declines. Once large objects start to stir up the
leftover planetesimals, debris production rises rapidly. After dust production
peaks, the collision rate and dust production decline slowly with time. For
all stars, more massive disks eject more material into a wind of small
particles. Figure 7: Median mass in 0.001–1 mm objects as a function of time
for 30–150 AU disks around a 1 M⊙ central star. The legend indicates $x_{m}$,
the initial disk mass in units of the scaled MMSN. During runaway growth, the
median dust mass is small and roughly constant in time. As planet formation
propagates through the disk, the dust mass grows with time. Once planets form
in the outer disk, collisions and Poynting-Robertson drag removes small grains
from the disk. Figure 8: Time evolution of the radius of the largest object
at 40 AU and at 100 AU for identical disks around 1 M⊙ (dot-dashed curves), 2
M⊙ (solid curves), and 3 M⊙ (triple dot-dashed curves) stars. Planets grow
faster around more massive stars and in the inner disks of all stars.
Figure 9: Median radius of the largest object at 100 Myr in each annulus for a
scaled MMSN disk ($x_{m}$ = 1) around 1–3 $M_{\odot}$ stars. The number to the
right of each set of points indicates the stellar mass in M⊙. At all disk
radii, large objects form faster around more massive stars. At 30–60 AU,
planet formation saturates at radii $r\sim$ 1000–2000 km on relatively short
timescales, $t\sim$ 100 Myr (see also Eq. (41)). At larger disk radii, planets
form more slowly and do not reach the maximum radius unless the formation time
is shorter than the main sequence lifetime.
Figure 10: Median cumulative mass distributions at 100 Myr for planet
formation calculations at 30–37 AU around 1–3 M⊙ stars. Left: Results for
models with $\Sigma=0.18~{}(a_{i}/{\rm 30~{}AU})^{-3/2}$ g cm-2. The light
solid line indicates the initial mass distribution. The dashed (1 M⊙, $x_{m}$
= 1), dot-dashed (2 M⊙, $x_{m}$ = 1/2), and heavy solid (3 M⊙, $x_{m}$ = 1/3)
lines show median results for the same initial conditions. Right: Results for
models with a scaled surface density ($x_{m}$ = 1), $\Sigma=0.18~{}(a_{i}/{\rm
30~{}AU})^{-3/2}$ $(M_{\star}/M_{\odot})$ g cm-2, and different stellar masses
(1 M⊙: dashed line, 2 M⊙: dot-dashed line, and 3 M⊙: heavy solid line).
Although more massive planets form around more massive stars, the collisional
cascade leads to a small dispersion in total disk mass at late times.
Figure 11: Median cumulative mass distributions at $t=t_{ms}$ for annuli at
30–37 AU (left panel) and at 123–146 AU (right panel) for identical disks
($\Sigma=0.18~{}(a_{i}/{\rm 30~{}AU})^{-3/2}$ g cm-2) around 1–3 M⊙ stars. The
legend indicates the stellar mass in M⊙. In the inner disk, many large planets
form and the collisional cascade removes a large fraction of the material in
objects with $r\lesssim$ 1–10 km. In the outer disk, few large planets form;
collisions are inefficient at removing material in small objects.
Figure 12: Median production rate of 0.01–1 $\mu$m objects at 30–150 AU as a
function of time for scaled MMSN ($x_{m}$ = 1) around 1–3 M⊙ central stars.
The legend indicates the stellar mass in M⊙. For scaled MMSN, disks around
more massive stars eject much more material in very small grains at early
times ($t\lesssim$ 10–100 Myr). In an ensemble of stars with a variety of disk
masses, there is wide range of dust production rates. Figure 13: Median mass
in 0.001–1 mm objects as a function of time for scaled MMSN ($x_{m}$ = 1) at
30–150 AU around 1–3 M⊙ central stars. The legend indicates the stellar mass
in M⊙. For $t\lesssim$ 1–3 Myr, icy planet formation produces little dust. At
10–100 Myr, the mass in small grains is $\sim$ 1 lunar mass for most disks. At
late times, the mass in small grains slowly declines to currently undetectable
levels.
Figure 14: Time evolution of the median $L_{d}/L_{\star}$ (dust luminosity
relative to the luminosity of the central star) for disks surrounding a 1
$M_{\odot}$ star. The legend indicates the disk mass in units of the MMSN.
More massive disks reach larger peak dust luminosities earlier than less
massive disks. The typical peak dust luminosity is comparable to the dust
luminosity of the most luminous debris disks associated with solar-type stars.
Figure 15: As in Fig. 14 for the median 70 $\mu$m excess (left panel) and the
median 850 $\mu$m excess (right panel). At both wavelengths, dust emission
begins to increase at 5–10 Myr. Peak dust emission occurs at 30–100 Myr (70
$\mu$m) and 100–300 Myr (850 $\mu$m). When the central star evolves off the
main sequence, the typical excess at 70 $\mu$m (850 $\mu$m) is $\sim$ 2–3
(3–10) times the flux from the central star.
Figure 16: Variation of dust excess with disk size. Left panel: time evolution
of the median 70 $\mu$m excess for MMSN disks with outer radii of 70 AU
(dashed line) and 150 AU (solid line). At late times, smaller disks produce
smaller IR excesses. it Right panel: as in the left panel for the median 850
$\mu$m excess.
Figure 17: Time evolution of the median $L_{d}/L_{\star}$ for MMSN disks
surrounding 2 $M_{\odot}$ stars (left panel) and 3 $M_{\odot}$ stars (right
panel). The legend indicates the initial disk mass in units of the scaled
MMSN. The typical maximum dust luminosity, $L_{d}/L_{\star}\sim 10^{-3}$ is
comparable to the dust luminosity of the brightest debris disks around A-type
stars. Figure 18: Time evolution of median IR excesses for MMSN disks around
1–3 $M_{\odot}$ stars. The legend in the lower left panel indicates the
stellar mass in solar masses for each curve in all panels. Lower left panel:
24 $\mu$m excess. Upper left panel: 70 $\mu$m excess. Upper right panel: 160
$\mu$m excess. Lower right panel: 850 $\mu$m excess. At 24 $\mu$m, the peak
excess increases dramatically with the temperature of the central star. Thus,
hotter stars produce much larger 24 $\mu$m excesses. At longer wavelengths,
the magnitude of the excess is correlated with the mass of the central star.
Roughly independent of stellar mass, the magnitude of the excess at 24–850
$\mu$m peaks at 10–30 Myr as observed in debris disks around A-type stars
(Currie et al., 2008a). At late times, the 160–850 $\mu$m excesses for all
stars are $\sim$ 3–5 times the flux from the stellar photosphere. Figure 19:
Time evolution of the median [24]–[70] color as a function of time for 1.5 M⊙
(dashed line), 2.0 M⊙ (solid line), and 3.0 (triple dot-dashed line) M⊙ stars.
Debris disks around lower mass stars have redder [24]–[70] colors than disks
around more massive stars. For massive stars ($\gtrsim$ 2 M⊙), the [24]–[70]
color increases slowly throughout the main sequence lifetime and then declines
just before central star evolves off the main sequence. For lower mass stars,
the [24]–[70] color reaches a broad maximum at 300 Myr to 1 Gyr and then
declines. Figure 20: Evolution of debris disks with $x_{m}$ = 1 in color-
color space. For icy planet formation at 30–150 AU, debris disks around
massive stars are hotter than debris disks around less massive stars. Thus,
debris disks around stars of different masses occupy specific regions of the
[8]–[24] vs [24]–[70] color-color diagram. Figure 21: As in Fig. 20 for disks
with different initial masses around 2 M⊙ stars. Boxes: $x_{m}$ = 1/3, filled
circles: $x_{m}$ = 1, diamonds: $x_{m}$ = 3. Although more massive disks have
redder [8]–[24] and [24]–[70], the shape of the color-color track is
independent of mass. Thus, the color-color diagram isolates stars of different
masses. Figure 22: Observations of the 24 $\mu$m excess for nearby A-type
stars with known ages (Rieke et al., 2005; Su et al., 2006). The lines show
the predicted evolution of the excess for debris disk models around 2 M⊙ stars
(dot-dashed line: $x_{m}$ = 1/3; solid line: $x_{m}$ = 1, triple dot-dashed
line: $x_{m}$ = 3). Observations for all but four stars (including one 8 Myr
old star with log $F_{24}/F_{24,0}\approx$ 2) fall within loci defined by our
debris disk calculations. Model predictions are also consistent with
observational evidence for a peak in the 24 $\mu$m excess at 10–20 Myr (see
also Currie et al., 2008a). Figure 23: As in Fig. 22 for the 70 $\mu$m
excess. Observations for all but 2–3 stars fall within the loci defined by the
model tracks. Consistent with model predictions, the data suggest a peak in
the 70 $\mu$m excess at 10–20 Myr. Figure 24: Observations of the [24] - [70]
color for nearby A-type stars with known ages (Su et al., 2006). The lines
show the predicted evolution of the excess for debris disk models around 1.5
M⊙ stars (triple dot-dashed line: $x_{m}$ = 2) and for 2 M⊙ stars (dot-dashed
line: $x_{m}$ = 1/3; solid line: $x_{m}$ = 1). Observations for all but 2–3
stars fall within the model predictions. Figure 25: Observations of the 70
$\mu$m excess for nearby solar-type stars with known ages (Beichman et al.,
2006; Hillenbrand et al., 2008). The lines show the predicted evolution of the
excess for debris disk models around 1 M⊙ stars (dot-dashed line: $x_{m}$ =
1/3; solid line: $x_{m}$ = 1, triple dot-dashed line: $x_{m}$ = 3). Most stars
fall within the loci defined by the calculations, but many stars are 3–10
times brighter than model predictions. Figure 26: As in Fig. 25 for the 160
$\mu$m excess.
| arxiv-papers | 2008-07-08T18:44:53 | 2024-09-04T02:48:56.592226 | {
"license": "Public Domain",
"authors": "Scott J. Kenyon and Benjamin C. Bromley",
"submitter": "Scott J. Kenyon",
"url": "https://arxiv.org/abs/0807.1134"
} |
0807.1135 | # On fake $\mathbb{Z}_{p}$-extensions of number fields
Luca Caputo and Filippo Alberto Edoardo Nuccio
(August 27, 2024)
Abstract -For an odd prime number $p$, let $L_{\infty}$ be the
$\mathbb{Z}_{p}$-anticyclotomic extension of an imaginary quadratic field $L$.
We focus on the non-normal subextension $K_{\infty}$ of $L_{\infty}$ fixed by
a subgroup of order $2$ in $\mathrm{Gal}(L_{\infty}/\mathbb{Q})$. After
providing a general result for dihedral extensions, we study the growth of the
$p$-part of the class group of the subfields of $K_{\infty}/\mathbb{Q}$,
providing a formula of Iwasawa type. Furthermore, we describe the structure of
the projective limit of these class groups.
2000 Mathematical Subject Classification: Primary $11R23$ Secondary $11R20$
## 1 Introduction
The aim of the present paper is to study the growth of class numbers along a
tower of extensions which is not Galois over the ground field. More precisely,
let $p$ be an odd prime, let $L$ be a CM field and let $L_{\infty}/L$ be a
$\mathbb{Z}_{p}$-extension such that $L_{\infty}/L^{+}$ is pro-$p$-dihedral
(meaning that $\mathrm{Gal}(L_{\infty}/F^{+})$ is a projective limit of
dihedral groups of order $2p^{n},n\geqslant 1$). We set $K=L^{+}$. Hence the
situation is as follows:
---
$\textstyle{L_{\infty}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{2}$$\textstyle{K_{\infty}=\bigcup_{n\in\mathbb{N}}K_{n}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{L_{n}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{2}$$\scriptstyle{p^{n}}$$\textstyle{K_{n}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{p^{n}}$$\textstyle{L=L_{0}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{2}$$\textstyle{L^{+}=K}$
Such an extension always exists and, under Leopoldt conjecture for $L$ with
respect to the prime $p$, there are precisely $n/2$ of them if
$n=[L:\mathbb{Q}]$. Note that $\mathrm{Gal}(L_{\infty}/K)$ is the semidirect
product of $\mathrm{Gal}(L_{\infty}/L)\rtimes\Delta$ where
$\Delta=\mathrm{Gal}(L/K)$. For every $m\geqslant 1$, denote by $K_{m}$ the
subfield of $L_{\infty}$ which is fixed by
$\mathrm{Gal}(L_{\infty}/L_{m})\rtimes\Delta$. Note that
$\mathrm{Gal}(L_{m}/K)$ is a dihedral group (isomorphic to $D_{p^{m}}$).
Setting $K_{\infty}=\cup K_{m}$, the extension $K_{\infty}/K$ shares some
similarities with $\mathbb{Z}_{p}$-extensions, still behaving in a different
way. In particular it can be seen as a particular case of what may be called a
_fake_ $\mathbb{Z}_{p}$-extension. Here is the definition that we propose
###### Definition.
Let $p$ be a prime number, let $K$ be a number field and let $K_{\infty}/K$ be
a _non_ Galois extension. Suppose that there exists a Galois extension $L/K$
disjoint from $K_{\infty}/K$ such that $LK_{\infty}$ is a Galois closure of
$K_{\infty}/K$. If $LK_{\infty}/L$ is a $\mathbb{Z}_{p}$-extension, then
$K_{\infty}/K$ is called a fake $\mathbb{Z}_{p}$-extension.
Our strategy to study the growth is to use a class number formula at finite
levels and then to pass to the limit. This formula is not new (see for example
[HK], [Ja1], [Le]…): anyway, we shall give a proof of it which seems to be
different from others that can be found in the literature. For a number field
$M$, let $h_{M}$ denote its class number, $R_{M}$ its regulator and $E_{M}$
the group of units of $M$ modulo torsion. In Section 2 we prove by analytic
means (essentially Brauer formula for Artin $L$-functions) the following
result:
###### Theorem.
Let $q$ be an odd natural number and let $F/K$ be a Galois extension whose
Galois group is isomorphic to the dihedral group with $2q$ elements $D_{q}$.
Let $L$ (resp. $k$) be the field fixed by the cyclic subgroup of order $q$
(resp. by one of the subgroups of order $2$) of $\mathrm{Gal}(F/K)$. Then
$h_{F}=h_{L}\frac{h_{k}^{2}}{h_{K}^{2}}\cdot\frac{R_{k}^{2}R_{L}}{R_{K}^{2}R_{F}}\;.$
In order to pass to the limit we need to give an algebraic interpretation to
the ratio of regulators which appears in the theorem. This is done in Section
3, essentially by an elementary but rather technical linear algebra
computation. The result is as follows:
###### Proposition.
With notations as above, let $k^{\prime}=\rho(k)$ where $\rho$ is a generator
of the cyclic subgroup of order $q$ in $\mathrm{Gal}(F/K)$. Then the following
equality holds:
$[E_{F}:E_{k}E_{k^{\prime}}E_{L}]=\frac{qR_{k}^{2}2^{n-1}}{QR_{F}R_{K}}\>.$
where $n=[K:\mathbb{Q}]$.
Putting togheter the preceding theorem and the last proposition, we get a
formula in Theorem 3.4 relating the class numbers of $L$, $F$, $K$ and $k$
involving only algebraic objects.
In Section 4, we take $K=\mathbb{Q}$. $L$ is therefore an imaginary quadratic
field and there is only one $\mathbb{Z}_{p}$-extension of $L$ which is
pro-$p$-dihedral over $K$, the so-called _anticyclotomic_
$\mathbb{Z}_{p}$-extension of $L$, which we denote by $L_{\infty}$. The main
result of the section is then (notation as in the diagram at the beginning)
###### Theorem.
Let $p^{\varepsilon_{m}}$ be the order of the $p$-Sylow class group of
$K_{m}$. Then there exist integers $\mu_{K},\lambda_{K},\nu_{K}$ such that
$2\varepsilon_{m}=\mu_{K}p^{m}+\lambda_{K}m+\nu_{K}\qquad\mathrm{for}\quad
m\gg 0\;.$
The main ingredients of the proof are the $p$-part of the formula proved in
Section 2 and Section 3, Iwasawa’s formula for $L_{\infty}/L$ and the
interpretation of a quotient of units as a cohomology group (see Proposition
4.4). The more “Iwasawa Theory” approach of passing to the limit on this
quotient and then descending fails here as the characteristic power series
involved is $T$, as discusses after Proposition 4.4. We also give an
intepretation of the invariants $\mu_{K}$ and $\lambda_{K}$ in terms of the
invariants $\mu_{L}$ and $\lambda_{L}$ relative to $L_{\infty}/L$ (in fact we
also get a proof of the parity of $\lambda_{K}$). In particular we find
$\mu_{L}=\mu_{K}\quad\textrm{and}\quad\lambda_{K}=\lambda_{L}+\lambda_{\mathfrak{P}}$
where $\lambda_{\mathfrak{P}}$ is the Iwasawa $\lambda$-invariant relative to
the $\Lambda$-module ($\Lambda=\mathbb{Z}_{p}[\\![T]\\!]$) which is the
projective limit of the cyclic subgroups of $Cl_{L_{m}}$ generated by the
classes of the products of all prime ideals of $L_{m}$ which lie over $p$. It
is worth mentioning that R. Gillard proved in [Gi] that
$\lambda_{L}\equiv\mu_{L}\pmod{2}\quad\textrm{and}\quad\mu_{L}\leqslant 1\;,$
the latter inequality becoming an equality if and only if $p$ splits in $L$.
Section 5 is devoted to the study of the exact sequence
$0\to\mathrm{Ker}(\iota_{m})\longrightarrow A_{K_{m}}\oplus
A_{K^{\prime}_{m}}\stackrel{{\scriptstyle\iota_{m}}}{{\longrightarrow}}A_{L_{m}}\longrightarrow
A_{L_{m}}/A_{K_{m}}A_{K^{\prime}_{m}}\to 0\;.$ (1)
Here we denote by $A_{M}$ the $p$-Sylow of the class group of any number field
$M$. If $M_{\infty}/M$ is a $\mathbb{Z}_{p}$-extension or a fake
$\mathbb{Z}_{p}$-extension, let $X_{M_{\infty}/M}$ (or $X_{M}$ if the (fake)
$\mathbb{Z}_{p}$-extension is clear) be the projective limit of $A_{M_{n}}$
with respect to the norm map. Moreover
$\iota_{m}\left(([I],\,[I^{\prime}])\right)=[II^{\prime}\mathcal{O}_{L_{m}}]$
if $I$ (resp. $I^{\prime}$) is an ideal of $K_{m}$ (resp. $K_{m}^{\prime}$).
Here we are identifying $A_{K_{m}}$ and $A_{K^{\prime}_{m}}$ with their
isomorphic images in $A_{L_{m}}$ (the extension maps $A_{K_{m}}\to A_{L_{m}}$
and $A_{K^{\prime}_{m}}\to A_{L_{m}}$ are injections since $L_{m}/K_{m}$ and
$L_{m}/K^{\prime}_{m}$ are of degree $2\neq p$). Passing to projective limit
with respect to norms we get
$0\to\mathrm{Ker}(\iota_{\infty})\longrightarrow X_{K}\oplus
X_{K^{\prime}}\stackrel{{\scriptstyle\iota_{\infty}}}{{\longrightarrow}}X_{L}\longrightarrow
X_{L}/X_{K}X_{K^{\prime}}\to 0\;.$ (2)
Then the main result of Section 5 is
###### Theorem.
The following holds
1. 1.
$\mathrm{Ker}(\iota_{\infty})$ is a $\mathbb{Z}_{p}$-module of rank $1$ if $p$
splits in $L$ and it is finite otherwise;
2. 2.
$X_{L}/X_{K}X_{K^{\prime}}$ is finite and its order divides
$h_{L}^{(p)}/p^{n_{0}}$.
where $n_{0}$ is the smallest natural number such that $L_{\infty}/L_{n_{0}}$
is totally ramified at every prime above $p$ and $h_{L}^{(p)}$ denotes the
order of the $p$-Sylow subgroup of the class group of $L$. In particular,
$X_{L}$ is finitely generated as $\mathbb{Z}_{p}$-module if and only if
$X_{K}$ is finitely generated as $\mathbb{Z}_{p}$-module and its rank is twice
the rank of $X_{K}$ if $p$ does not split and
$2\mathrm{rk}_{\mathbb{Z}_{p}}X_{K}+1$ if $p$ splits.
The tecniques involved in the proof give also an algebraic proof (only for odd
parts) of the formula proved in Section 2 and Section 3.
_Aknowledgements_ We would like to thank Ralph Greenberg for suggesting us to
work on this topic and for many useful comments. Moreover, we thank Jean-
François Jaulent for informing us that most of the results of this paper were
proved with different techniques in [Ja2].
## 2 Class numbers formula for dihedral extensions.
Let $q$ be an odd natural number. Let $K$ be a number field and let $F/K$ be a
Galois extension whose Galois group is isomorphic to the dihedral group
$D_{q}$ (we shall identify from now on $\mathrm{Gal}(F/K)$ with $D_{q}$).
Recall that $D_{q}$ is the group generated by $\rho$ and $\sigma$ with
relations
$\rho^{q}=\sigma^{2}=1,\quad\sigma\rho\sigma=\rho^{-1}\;.$
In particular $D_{q}$ contains the cyclic group $C_{q}$ of order $q$ generated
by $\rho$. Let $L$ be the subextension of $F/K$ fixed by $C_{q}$. Similarly,
let $k$ be the subextension of $F/K$ fixed by the subgroup generated by
$\sigma$.
Let $M$ be a subextension of $F/K$: for a complex representation of
$\mathrm{Gal}(F/M)$ with character $\chi$, we consider the attached Artin
$L$-function that we denote by $L(s,\,\chi,F/M)$ where $s\in\mathbb{C}$ has
real part bigger than $1$. We denote by $\chi_{0}^{M}$ the trivial character
of $\mathrm{Gal}(F/M)$: note that
$L(s,\,\chi_{0}^{M},\,F/M)=\zeta_{M}(s)$
where $\zeta_{M}$ is the Dedekind zeta function of $M$. We use here the
notation $\zeta^{*}_{M}(s)$ for the special value of $\zeta_{M}$ at
$s\in\mathbb{C}\smallsetminus\\{1\\}$: by definition, $\zeta^{*}_{M}(s)$ is
the first nontrivial coefficient in the Taylor expansion of $\zeta_{M}$ around
$s$. By Dirichlet’s theorem, we have
$\zeta_{M}^{*}(0)=-\frac{h_{M}}{w_{M}}R_{M}\;,$ (3)
where $w_{M}$ is the number of roots of unity contained in $M$ (this formula
comes from the formula for the residue at $1$ of $\zeta_{M}$ and the
functional equation, see [Na], chapter $7$). This notation will be used
throughout of the paper.
We briefly recall how the irreducible characters of $D_{q}$ are defined (for
everything concerning representation theory in the following see [Se1], I,
§5.3). There are two representations of degree $1$, namely
$\chi_{0}(\rho^{a}\sigma^{b})=1\quad\textrm{for each $0\leqslant a\leqslant
q-1,\,\,0\leqslant b\leqslant 1$}\;,$
$\chi_{1}(\rho^{a}\sigma^{b})=(-1)^{b}\quad\textrm{for each $0\leqslant
a\leqslant q-1,\,\,0\leqslant b\leqslant 1$}\;.$
Observe that $\chi_{0}^{k}=\chi_{0}$. Furthermore there are $q-1$
representations of degree $2$, namely $\psi_{1},\,\ldots,\,\psi_{(q-1)}$ which
are defined by
$\psi_{h}(\rho^{a})=\left(\begin{array}[]{cc}\zeta_{q}^{ha}&0\\\
0&\zeta_{q}^{-ha}\end{array}\right),\qquad\psi_{h}(\rho^{a}\sigma)=\left(\begin{array}[]{cc}0&\zeta_{q}^{-ha}\\\
\zeta_{q}^{ha}&0\end{array}\right)\;\forall\;0\leqslant a\leqslant q-1\;.$
for every $1\leqslant h\leqslant q-1$, where $\zeta_{q}$ is a primitive $q$-th
root of unity.
###### Proposition 2.1.
Let $r=(q-1)/2$. Then the representations
$\chi_{0},\,\chi_{1},\,\psi_{1},\,\psi_{2},\,\ldots,\,\psi_{r}$
are the irreducible representations of $D_{q}$.
###### Proof.
See [Se1], I, §5.3. ∎
In the following we shall denote by $\chi^{(h)}$ the character of $\psi_{h}$.
Furthermore, if $H$ is a subgroup of $D_{q}$ and $\chi$ is a character of $H$
whose corresponding representations is $\psi$, we denote by
$\mathrm{Ind}_{H}^{D_{q}}\chi$ the character of the representation of $D_{q}$
induced by $\psi$. Then we have
$\left(\mathrm{Ind}_{H}^{D_{q}}\chi\right)(u)=\sum_{\begin{subarray}{c}r\in
R\\\ r^{-1}ur\in H\end{subarray}}\chi(r^{-1}ur)\;,$ (4)
where $R$ is any system of representatives for $D_{q}/H$. The next lemma
describes the characters of some induced representations in terms of the
irreducible characters.
###### Lemma 2.2.
The following holds
$\mathrm{Ind}_{\\{1\\}}^{D_{q}}\chi_{0}^{L}=\chi_{0}+\chi_{1}+2\sum_{h=1}^{r}\chi^{(h)}\;,$
(5)
$\mathrm{Ind}_{\langle\sigma\rangle}^{D_{q}}\chi_{0}^{K}=\chi_{0}+\sum_{h=1}^{r}\chi^{(h)}\;,$
(6) $\mathrm{Ind}_{C_{q}}^{D_{q}}\chi_{0}^{F}=\chi_{0}+\chi_{1}\;.$ (7)
###### Proof.
Equality in (5) follows from the fact that both terms equal the character of
the regular representation of $D_{q}$.
In order to prove (6) we use (4) with $H=\langle\sigma\rangle$: choose
$R=C_{q}$. Then clearly
$\left(\mathrm{Ind}_{\langle\sigma\rangle}^{D_{q}}\chi_{0}^{K}\right)(\rho^{a})=\left\\{\begin{array}[]{ll}q&\textrm{if
$a=0$}\\\ 0&\textrm{if $0<a\leqslant q-1$}\end{array}\right.$
and
$\left(\mathrm{Ind}_{\langle\sigma\rangle}^{D_{q}}\chi_{0}^{K}\right)(\rho^{a}\sigma)=1$
for every $0\leqslant a\leqslant q-1$ (since
$\rho^{-c}\rho^{a}\sigma\rho^{c}\in\langle\sigma\rangle$ if and only if
$a\equiv 2c\pmod{q}$ and the latter has only one solution). On the other hand,
the right-hand side of (6) verifies
$\left(\chi_{0}+\sum_{h=1}^{r}\chi^{(h)}\right)(1)=q$
and, if $0<a\leqslant q-1$,
$\Big{(}\chi_{0}+\sum_{h=1}^{r}\chi^{(h)}\Big{)}(\rho^{a})=1+\sum_{h=1}^{r}\left(\zeta_{q}^{ha}+\zeta_{q}^{-ha}\right)=1+\sum_{h=1}^{q-1}\zeta_{q}^{ha}=1-1=0\;.$
Furthermore, if $0\leqslant a\leqslant q-1$,
$\Big{(}\chi_{0}+\sum_{h=1}^{r}\chi^{(h)}\Big{)}(\rho^{a}\sigma)=1+0=1$
which completes the proof of (6); (7) can be proven similarly. ∎
From now on, we let $\mu(M)$ denote the group of roots on unity of a number
field $M$.
###### Lemma 2.3.
The following holds
$\mu(k)=\mu(K),\qquad\mu(F)=\mu(L)\;.$
###### Proof.
Let $\zeta\in\mu(F)\setminus\mu(K)$ be a root of unity of $F$ which does not
lie in $K$, and set $M=K(\zeta)$. Then $M/K$ is a nontrivial abelian extension
of $K$ contained in $F$. In particular $\mathrm{Gal}(F/M)$ contains the
commutator subgroup of $D_{q}$ which is equal to $C_{q}$. Therefore, $M/K$
being nontrivial, $\mathrm{Gal}(F/M)=C_{q}$ and $M=L$. This shows at once that
$\mu(L)=\mu(F)$ and $\mu(k)=\mu(K)$ (since $F\cap k=K$). ∎
###### Theorem 2.4.
The following equality holds
$\zeta_{F}(s)=\zeta_{L}(s)\frac{\zeta_{k}(s)^{2}}{\zeta_{K}(s)^{2}}$
for each $s\in\mathbb{C}\smallsetminus\\{1\\}$. In particular
$h_{F}=h_{L}\frac{h_{k}^{2}}{h_{K}^{2}}\cdot\frac{R_{k}^{2}R_{L}}{R_{K}^{2}R_{F}}\;,$
and $R_{k}^{2}R_{L}/R_{K}^{2}R_{F}$ is a rational number.
###### Proof.
In the following we use various known properties of Artin $L$-functions: for
their proofs see [He], §3. First of all note that, for every $s\in\mathbb{C}$
such that $\mathrm{Re}\,s>1$,
$\zeta_{F}(s)=L(s,\,\chi_{0}^{F},\,F/F)=$
$=L(s,\,\mathrm{Ind}_{\\{1\\}}^{D_{q}}\chi_{0}^{F},\,F/K)=L(s,\,\chi_{0}+\chi_{1}+2\sum_{h=1}^{r}\chi^{(h)},\,F/K)=$
$=L(s,\,\chi_{0},\,F/K)L(s,\,\chi_{1},\,F/K)\prod_{h=1}^{r}L(s,\chi^{(h)},\,F/K)^{2}$
by Lemma 2.2.
Now we consider $\zeta_{k}$: we have
$\zeta_{k}(s)=L(s,\,\chi_{0}^{k},\,F/k)=$
$=L(s,\,\mathrm{Ind}_{\langle\sigma\rangle}^{D_{q}}\chi_{0}^{k},\,F/K)=L(s,\,\chi_{0}+\sum_{h=1}^{r}\chi^{(h)},\,F/K)=$
$=L(s,\,\chi_{0},\,F/K)\prod_{h=1}^{r}L(s,\chi^{(h)},\,F/K)$
by Lemma 2.2. Lastly, we consider $\zeta_{L}$: we have
$\zeta_{L}(s)=L(s,\,\chi_{0}^{L},\,F/L)=$
$=L(s,\,\mathrm{Ind}_{C_{q}}^{D_{q}}\chi_{0}^{k},\,F/K)=L(s,\,\chi_{0}+\chi_{1},\,F/K)=$
$=L(s,\,\chi_{0},\,F/K)L(s,\chi_{1},\,F/K)$
again by Lemma 2.2. Hence
$\zeta_{F}(s)=\zeta_{L}(s)\frac{\zeta_{k}(s)^{2}}{\zeta_{K}(s)^{2}}$ (8)
for $s\in\mathbb{C}$ with $\mathrm{Re}\,s>1$ because
$\zeta_{K}(s)=L(s,\,\chi_{0},\,F/K)\;.$
We deduce that (8) holds for every $s\in\mathbb{C}\smallsetminus\\{1\\}$. In
particular the left and the right terms have the same special value at $0$. We
then deduce from (3) that
$h_{F}=h_{L}\frac{h_{k}^{2}}{h_{K}^{2}}\cdot\frac{R_{k}^{2}R_{L}}{R_{K}^{2}R_{F}}\cdot\frac{w_{K}^{2}w_{F}}{w_{k}^{2}w_{L}}$
(9)
and the formula in our statement then comes from Lemma 2.3. ∎
## 3 Algebraic interpretation of regulators
We shall now prove an algebraic interpretation of the term
$(R_{k}^{2}R_{L})/(R_{K}^{2}R_{F})$ appearing in Theorem 2.4. An algebraic
proof of the formula resulting from (9) can also be found in [HK], [Ja1], [Le]
(see also the last section). The notation is the same as in Section 2, but we
fix the following convention for the rest of this section:
> $K$ is totally real of degree $n$ over $\mathbb{Q}$ while $F$ is totally
> imaginary (thus of degree $2qn$). Therefore $L$ is a CM-field and $L^{+}=K$.
As usual, $r_{1}(M)$ and $r_{2}(M)$ denote the number of real and imaginary
places, respectively, of a number field $M$, and we recall the notation
$r=(q-1)/2$ introduced in Proposition 2.1: we have
###### Lemma 3.1.
With the above convention, $r_{1}(k)=n$ and $r_{2}(k)=n(q-1)/2=nr$.
###### Proof.
Since $F$ is totally imaginary every infinite prime
$\vartheta_{i}^{\prime}:F\hookrightarrow\mathbb{C}$ of $F$ has a decomposition
subgroup of order $2$ inside $D_{q}$. On the other hand, the number of real
embeddings of $k$ coincides with the number of infinite primes of $k$ that
ramify in $F/k$, therefore such that
$\mathcal{I}(\vartheta_{i}^{\prime})\subseteq\mathrm{Gal}(F/k)$ where
$\mathcal{I}(\vartheta_{i}^{\prime})$ is the decomposition group of
$\vartheta_{i}^{\prime}$: this is equivalent to
$\mathcal{I}(\vartheta_{i}^{\prime})=\mathrm{Gal}(F/k)$. Inside $F$ there are
exactly $q$ fields of index $2$, and they are all isomorphic to $k$: therefore
the number of infinite primes of $F$ such that
$\mathcal{I}(\vartheta_{i}^{\prime})=\mathrm{Gal}(F/k)$ must coincide with the
number of infinite primes such that
$\mathcal{I}(\vartheta_{j}^{\prime})=\mathrm{Gal}(F/k^{\prime})$ for every
$k^{\prime}$ conjugate to $k$. Since there are exactly $nq$ infinite primes in
$F$, Dirichlet’s Box Principle tells us that exactly $n$ decomposition
subgroups coincide with $\mathrm{Gal}(F/k)$, as stated. ∎
Let now $1\neq\rho\in D_{q}$ be an automorphism of $F$ fixing $K$ of order $q$
and set $k^{\prime}=\rho(k)$. Since $\sigma$ and $\rho$ generate $D_{q}$ and
$k$ is fixed by $\sigma$, if $\rho(k)=k$ then $k$ would be a normal extension
of $K$: therefore $k^{\prime}\neq k$. We set
$E_{F}=\mathcal{O}_{F}^{\times}/\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{F}^{\times}$
and similarly for $k$, $k^{\prime}$, $L$ and $K$. Note that there are
canonical embeddings $E_{k}\hookrightarrow E_{F}$,
$E_{k^{\prime}}\hookrightarrow E_{F}$ and $E_{L}\hookrightarrow E_{F}$.
Moreover, it is not hard to see that
$E_{k}E_{k^{\prime}}=\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times}/\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times})$
(both groups are subgroups of $E_{F}$).
###### Lemma 3.2.
$\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times}$ is of finite index
in $\mathcal{O}_{F}^{\times}$ and $E_{k}E_{k^{\prime}}$ is of finite index in
$E_{F}$.
###### Proof.
Clearly it is enough to prove the first assertion, since we have an exact
sequence
$0\to\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{F}^{\times}/\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times})\to\mathcal{O}_{F}^{\times}/\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times}\to
E_{F}/E_{k}E_{k^{\prime}}\to 0.$
Thanks to Lemma 3.1, $\mathrm{rk}_{\mathbb{Z}}\mathcal{O}_{F}^{\times}=nq-1$
while
$\mathrm{rk}_{\mathbb{Z}}\mathcal{O}_{k}^{\times}=\mathrm{rk}_{\mathbb{Z}}\mathcal{O}_{k^{\prime}}^{\times}=n(r+1)-1$.
Therefore all we need to prove is that
$\mathcal{O}_{k}^{\times}\cap\mathcal{O}_{k^{\prime}}^{\times}\subseteq\mathcal{O}_{K}^{\times}$:
indeed, this would imply that
$\mathcal{O}_{k}^{\times}\cap\mathcal{O}^{\times}_{k^{\prime}}=\mathcal{O}_{K}^{\times}$,
and so
$\mathrm{rk}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times})=\mathrm{rk}_{\mathbb{Z}}\mathcal{O}_{k}^{\times}+\mathrm{rk}_{\mathbb{Z}}\mathcal{O}_{k^{\prime}}^{\times}-\mathrm{rk}_{\mathbb{Z}}\mathcal{O}_{K}^{\times}=nq-1$
which is precisely $\mathrm{rk}_{\mathbb{Z}}\mathcal{O}_{F}^{\times}$.
But the inclusion
$\mathcal{O}_{k}^{\times}\cap\mathcal{O}_{k^{\prime}}^{\times}\subseteq\mathcal{O}_{K}^{\times}$
is immediate once we know that $k\cap k^{\prime}=K$; and this is clear, for
$\mathrm{Gal}(F/k\cap k^{\prime})$ contains both $\sigma$ and $\rho\sigma$,
and thus both $\sigma$ and $\rho$. Hence $\mathrm{Gal}(F/k\cap
k^{\prime})=D_{q}=\mathrm{Gal}(F/K)$, from which $k\cap k^{\prime}=K$. ∎
Remark. The above proof shows, in particular, that
$\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times}\mathcal{O}_{L}^{\times}$
(resp. $E_{k}E_{k^{\prime}}E_{L}$) is of finite index in
$\mathcal{O}_{F}^{\times}$ (resp. $E_{F}$).
As in the proof of Lemma 3.2, the units of $k$ have $\mathbb{Z}$-rank equal to
$n(r+1)-1$, while those of $K$ and of $L$ have $\mathbb{Z}$-rank equal to
$n-1$; finally, then,
$\mathrm{rk}_{\mathbb{Z}}(\mathcal{O}_{F}^{\times})=nq-1$. By the elementary
divisors theorem and Lemma 2.3, we can choose subsets
$\\{\eta_{j}\\}_{j=1}^{n(r+1)}\subseteq\mathcal{O}_{k}^{\times}$ and
$\\{a_{j}\\}_{j=1}^{n}\subseteq\mathbb{N}$ such that
$\mathcal{O}_{k}^{\times}=\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{K}^{\times}\oplus\bigoplus_{j=1}^{n(r+1)}\eta_{j}^{\mathbb{Z}}\quad\textrm{and}\quad\mathcal{O}_{K}^{\times}=\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{K}^{\times}\oplus\bigoplus_{j=1}^{n}\eta_{j}^{a_{j}\mathbb{Z}}$
(10)
(recall that
$\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{K}^{\times}=\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{k}^{\times}$,
by Lemma 2.3). Then
$\mathcal{O}_{k^{\prime}}^{\times}=\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{K}^{\times}\oplus\bigoplus_{j=1}^{n(r+1)}\rho(\eta_{j})^{\mathbb{Z}}.$
Moreover we also have
$\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times}=\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times})\oplus\bigoplus_{j=1}^{n(r+1)}\eta_{j}^{\mathbb{Z}}\oplus\bigoplus_{j=n+1}^{n(r+1)}\rho(\eta_{j})^{\mathbb{Z}}.$
(11)
This can be seen as follows: first of all we show that
$\bigoplus_{j=1}^{n(r+1)}\eta_{j}^{\mathbb{Z}}\cap\bigoplus_{j=n+1}^{n(r+1)}\rho(\eta_{j})^{\mathbb{Z}}=\\{1\\}.$
Suppose that we have
$\prod_{j=1}^{n(r+1)}\eta_{j}^{b_{j}}\prod_{j=n+1}^{n(r+1)}\rho(\eta_{j})^{c_{j}}=1.$
Then
$\prod_{j=n+1}^{n(r+1)}\rho(\eta_{j})^{c_{j}}\in\mathcal{O}_{k}^{\times}\cap\mathcal{O}_{k^{\prime}}^{\times}=\mathcal{O}_{K}^{\times}=\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{K}^{\times}\oplus\bigoplus_{j=1}^{n}\eta_{j}^{a_{j}\mathbb{Z}}.$
Now note that $\rho(\eta_{j})=\eta_{j}\zeta_{j}$ where
$\zeta_{j}\in\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{F}^{\times}$ is an
$a_{j}$-th root of unity (this follows from
$\rho(\eta_{j}^{a_{j}})=\eta_{j}^{a_{j}}$ which holds because
$\eta_{j}^{a_{j}}\in K$). This means that
$\prod_{j=n+1}^{n(r+1)}\rho(\eta_{j})^{c_{j}}=\zeta\prod_{j=n+1}^{n(r+1)}\eta_{j}^{c_{j}}=\xi\prod_{j=1}^{n}\eta_{j}^{a_{j}d_{j}}$
for some $\xi\in\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{K}^{\times}$ and
$\zeta=\prod\zeta_{j}^{c_{j}}$. This equation can actually be seen in
$\mathcal{O}_{F}^{\times}$ and gives $\zeta=\xi$ and $c_{j}=0$ for any
$n+1\leqslant j\leqslant n(r+1)$ and $d_{j}=0$ for any $1\leqslant j\leqslant
n$ since
$(\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{F}^{\times}\cap\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{K}^{\times})\oplus\bigoplus_{j=1}^{n(r+1)}\eta_{j}^{\mathbb{Z}}=\\{1\\}.$
But then we also have $b_{j}=0$ for any $1\leqslant j\leqslant n(r+1)$.
Therefore (11) is proved and we have
$R_{F}[E_{k}E_{k^{\prime}}^{\times}]=R_{F}\left[\bigoplus_{j=1}^{n(r+1)}\eta_{j}^{\mathbb{Z}}\oplus\bigoplus_{j=n+1}^{n(r+1)}\rho(\eta_{j})^{\mathbb{Z}}\right]$
where, for a subgroup $A\subseteq E_{F}$, $R_{F}[A]$ denotes its regulator.
Remark. Before we prove the main result of this section we observe that
$\prod_{j=1}^{n}a_{j}=|\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}/\mathcal{O}_{K}^{\times})|$
(12)
(which is clear from (10)) and there is an isomorphism
$\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times})/\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{K}^{\times}\cong\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}/\mathcal{O}_{K}^{\times}).$
(13)
To see this, consider the map
$\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times})\stackrel{{\scriptstyle\phi}}{{\longrightarrow}}\mathcal{O}_{k}^{\times}/\mathcal{O}_{K}^{\times}$
defined by $\phi(xx^{\prime})=[x]$, where $x\in\mathcal{O}_{k}^{\times}$,
$x^{\prime}\in\mathcal{O}_{k^{\prime}}^{\times}$ and
$xx^{\prime}\in\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times})$.
First of all, this definition makes sense, since if $xx^{\prime}=yy^{\prime}$
with $y\in\mathcal{O}_{k}^{\times}$ and
$y^{\prime}\in\mathcal{O}_{k^{\prime}}^{\times}$, then
$xy^{-1}=(x^{\prime})^{-1}y^{\prime}\in\mathcal{O}_{k}\cap\mathcal{O}_{k^{\prime}}=\mathcal{O}_{K}$
(and therefore $\phi(yy^{\prime})=[y]=[x]$). Now clearly
$\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{K}^{\times})=\mathrm{ker}\phi$ (once
more because $\mathcal{O}_{k}\cap\mathcal{O}_{k^{\prime}}=\mathcal{O}_{K}$)
and of course
$\mathrm{Im}\phi\subseteq\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}/\mathcal{O}_{K}^{\times})$.
On the other hand, suppose that $x\in\mathcal{O}_{k}^{\times}$ and there
exists $n\in\mathbb{N}$ such that $x^{n}\in\mathcal{O}_{K}^{\times}$. Then
$(x\rho(x^{-1}))^{n}=1$ (recall that $\rho(k)=k^{\prime}$) which means that
$x\rho^{-1}(x^{-1})\in\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times})$
and $\phi(x\rho(x^{-1}))=[x]$. This proves
$\mathrm{Im}\phi=\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}/\mathcal{O}_{K}^{\times})$
and therefore $\phi$ gives an isomorphism as in (13). In particular using (12)
and (13), we get
$\prod_{j=1}^{n}a_{j}=(\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times}):\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{K}^{\times}).$
(14)
###### Proposition 3.3.
The following equality holds:
$(\mathcal{O}_{F}^{\times}:\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times}\mathcal{O}_{L}^{\times})=\frac{qR_{k}^{2}R_{L}}{R_{K}^{2}R_{F}}(\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times}\cap\mathcal{O}_{L}^{\times}:\mathcal{O}_{K}^{\times})\>.$
###### Proof.
Note that
$(\mathcal{O}^{\times}_{F}:\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}}\mathcal{O}^{\times}_{L})=\frac{(\mathcal{O}^{\times}_{F}:\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}})}{(\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}}\mathcal{O}^{\times}_{L}:\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}})}=$
$=\frac{(E_{F}:E_{k}E_{k^{\prime}})(\mathrm{tor}_{\mathbb{Z}}\mathcal{O}^{\times}_{F}:\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}}))}{(\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}}\mathcal{O}^{\times}_{L}:\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}})}.$
This follows from the fact that the natural map
$\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}}/\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}})\longrightarrow
E_{k}E_{k}^{\prime}=\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}}\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{F}^{\times})/\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{F}^{\times})$
is an isomorphism. Now
$\frac{(E_{F}:E_{k}E_{k^{\prime}})(\mathrm{tor}_{\mathbb{Z}}\mathcal{O}^{\times}_{F}:\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}}))}{(\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}}\mathcal{O}^{\times}_{L}:\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}})}=$
$=\frac{(E_{F}:E_{k}E_{k^{\prime}})(\mathrm{tor}_{\mathbb{Z}}\mathcal{O}^{\times}_{F}:\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}}))}{(\mathcal{O}^{\times}_{L}:\mathcal{O}^{\times}_{K})}(\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}}\cap\mathcal{O}^{\times}_{L}:\mathcal{O}^{\times}_{K}).$
Hence we need to prove that
$\frac{(E_{F}:E_{k}E_{k^{\prime}})(\mathrm{tor}_{\mathbb{Z}}\mathcal{O}^{\times}_{F}:\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}}))}{(\mathcal{O}^{\times}_{L}:\mathcal{O}^{\times}_{K})}=\frac{qR_{k}^{2}R_{L}}{R_{K}^{2}R_{F}}.$
(15)
We first prove that
$R_{L}[E_{k}E_{k^{\prime}}]=\frac{q2^{n-1}(R_{k})^{2}}{R_{K}}(\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}}):\mathrm{tor}_{\mathbb{Z}}\mathcal{O}^{\times}_{K}).$
(16)
Thanks to Lemma 3.1 we define $\gamma^{\prime}_{l}:k\hookrightarrow\mathbb{R}$
for $0\leqslant l\leqslant n-1$ to be the real embeddings of $k$ and
$\tau^{\prime}_{i}:k\hookrightarrow\mathbb{C}$ for $1\leqslant i\leqslant nr$
to be the non-equivalent imaginary embeddings111The reason for the primes will
appear shortly. of $k$. Analogously, let
$\vartheta^{\prime}_{i}:F\hookrightarrow\mathbb{C}$ for $0\leqslant i\leqslant
nq-1$ be the non-equivalent (imaginary) embeddings of $F$. We order them so
that $\vartheta^{\prime}_{lq}$ extends $\gamma_{l}^{\prime}$ for $0\leqslant
l\leqslant n-1$; while $\vartheta^{\prime}_{lq+i}$ and
$\vartheta^{\prime}_{lq+i+r}$ extend $\tau^{\prime}_{lr+i}$ for $0\leqslant
l\leqslant n-1$ and for $1\leqslant i\leqslant r$. Without loss of generality
(changing $\rho$ if necessary in another element of order $q$) we can also
assume that $\rho(\vartheta_{lq+i}^{\prime})=\vartheta_{lq+i+1}^{\prime}$ for
$0\leqslant i\leqslant r-1$ and $0\leqslant l\leqslant n-1$. The relation
$\rho\sigma=\sigma\rho^{-1}$ together with
$\sigma(\vartheta_{lq+i}^{\prime})=\vartheta_{lq+i+r}^{\prime}$ then gives
$\left\\{\begin{array}[]{rll}\rho(\vartheta^{\prime}_{lq+i})&=\vartheta^{\prime}_{lq+i+1}&0\leqslant
i\leqslant r-1,0\leqslant l\leqslant n-1\\\
\rho(\vartheta^{\prime}_{lq+r})&=\vartheta^{\prime}_{(l+1)q-1}&0\leqslant
l\leqslant n-1\\\
\rho(\vartheta^{\prime}_{lq+r+1})&=\vartheta^{\prime}_{lq}&0\leqslant
l\leqslant n-1\\\
\rho(\vartheta^{\prime}_{lq+r+j})&=\vartheta^{\prime}_{lq+r+j-1}&2\leqslant
j\leqslant r,0\leqslant l\leqslant n-1\;.\\\ \end{array}\right.$ (17)
By definition, setting
$\vartheta_{i}=2\log|\vartheta^{\prime}_{i}|,\tau_{i}=2\log|\tau_{i}^{\prime}|$
and $\gamma_{l}=\log|\gamma^{\prime}_{l}|$, the regulators take the form
$R_{F}[\langle\eta_{1},\ldots,\rho(\eta_{n(r+1)-1})\rangle]=\bigg{|}\det\left(\begin{array}[]{c}\vartheta_{i}(\eta_{j})_{\begin{subarray}{c}0\leqslant
i\leqslant nq-2\\\ 1\leqslant j\leqslant n-1\end{subarray}}\\\
\hline\cr\vartheta_{i}(\eta_{j})_{\begin{subarray}{c}0\leqslant i\leqslant
nq-2\\\ n\leqslant j\leqslant n(r+1)-1\end{subarray}}\\\
\hline\cr\vartheta_{i}(\rho(\eta_{j}))_{\begin{subarray}{c}0\leqslant
i\leqslant nq-2\\\ n\leqslant j\leqslant
n(r+1)-1\end{subarray}}\end{array}\right)\bigg{|}\;$
and222In the next and in the last formula we use a somehow non- standard
notation to write matrices. It should though be clear from the contest what we
mean: in the last formula the matrix naturally splits vertically in three
submatrices, each of one we describe explicitely. In the following, the
splitting is horizontal.
$R_{k}=\bigg{|}\det\left(\begin{array}[]{c|c}\gamma_{l}(\eta_{j})_{\begin{subarray}{c}0\leqslant
l\leqslant n-1\\\ 1\leqslant j\leqslant
n(r+1)-1\end{subarray}}&\tau_{i}(\eta_{j})_{\begin{subarray}{c}1\leqslant
i\leqslant nr-1\\\ 1\leqslant j\leqslant
n(r+1)-1\end{subarray}}\end{array}\right)\bigg{|}\;.$
Before rewriting $\vartheta_{i}(\eta_{j})$ in terms of the $\tau_{i}$’s, two
remarks are in order. First of all, the lowest part of the matrix defining the
first regulator can be rewritten in terms of the $\vartheta_{i}(\eta_{j})$
only, thanks to (17). Secondly, in the definition of a regulator in $F$ (resp.
in $k$), only $nq-1$ (resp. $n(r+1)-1$) embeddings play a role, since the
units lie in the subspace defined by
$\vartheta_{nq-1}=-\sum_{i=0}^{nq-2}\vartheta_{i}\quad\big{(}resp.\;\tau_{nr}=-\sum_{l=0}^{n-1}\gamma_{l}-\sum_{i=1}^{nr-1}\tau_{i}\big{)}\;.$
(18)
In the sequel this relations will be used: moreover, unlike (18) that holds
for all units in $F$, there is also the relation
$\vartheta_{(n-1)q}(\eta_{j}^{a_{j}})=-\sum_{l=0}^{n-2}\vartheta_{lq}(\eta_{j}^{a_{j}})\qquad\forall\;\;1\leqslant
j\leqslant n-1$
since $\eta_{j}^{a_{j}}\in E_{K}$ for $1\leqslant j\leqslant n-1$. But then of
course
$\vartheta_{(n-1)q}(\eta_{j})=-\sum_{l=0}^{n-2}\vartheta_{lq}(\eta_{j})\qquad\forall\;\;1\leqslant
j\leqslant n-1.$ (19)
Observe now that our ordering ensures us that for all $1\leqslant j\leqslant
n(r+1)-1$ we have $\vartheta_{lq+i}(\eta_{j})=\tau_{lr+i}(\eta_{j})$ if
$0\leqslant l\leqslant n-1$ and $1\leqslant i\leqslant r$; that we have
$\vartheta_{lq+i+r}(\eta_{j})=\tau_{lr+i}(\eta_{j})$ if $0\leqslant l\leqslant
n-1$ and $1\leqslant i\leqslant r$; and
$\vartheta_{lq}(\eta_{j})=2\gamma_{l}(\eta_{j})$ for $0\leqslant l\leqslant
n-1$. Putting all together, (16) has been reduced (use (17), (18), (10) and
(14)) to the equation
$|\det(\Xi)|=\frac{q|\det(\Psi)|^{2}2^{n-1}}{|\det(\Phi)|^{-1}}\prod_{j=1}^{n}a_{j}=\frac{q|\det(\Psi)|^{2}2^{n-1}}{|\det(\Phi)|^{-1}}(\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}^{\times}_{k}\mathcal{O}^{\times}_{k^{\prime}}):\mathrm{tor}_{\mathbb{Z}}\mathcal{O}^{\times}_{K})$
(20)
for the three matrices appearing below in (21), and whose determinants give
the regulators we are computing: we should thus introduce some notation in
order to define them.
* A
$A[l]\in\mathcal{M}_{(n-1)\times(r+1)}(\mathbb{R})$ is the matrix
$A[l]_{i,j}=\tau_{lr+i}(\eta_{j})$ for $0\leqslant l\leqslant n-1$,
$1\leqslant i\leqslant r$ and $1\leqslant j\leqslant n-1$; and
$A[l]_{0,j}=2\gamma_{l}(\eta_{j})$ for $0\leqslant l\leqslant n-1$ and
$1\leqslant j\leqslant n-1$. First of all, using (18), we have
$A[n-1]_{r,j}=\displaystyle-\sum_{i=0}^{n-1}\gamma_{i}(\eta_{j})-\sum_{i=1}^{nr-1}\tau_{i}(\eta_{j})=-\sum_{l=0}^{n-1}\frac{A[l]_{0,j}}{2}-\sum_{l=0}^{n-1}\sum_{i=1}^{r-1}A[l]_{i,j}-\sum_{l=1}^{n-1}A[l-1]_{r,j}$
for all $1\leqslant j\leqslant n-1$. Thanks to (19) we also have
$\displaystyle
A[n-1]_{0,j}=-\sum_{l=0}^{n-2}2\gamma_{l}(\eta_{j})=-\sum_{l=0}^{n-2}A[l]_{0,j}$
for every $1\leqslant j\leqslant n-1$, finally finding
$A[n-1]_{r,j}=\displaystyle-\sum_{l=0}^{n-1}\sum_{i=1}^{r-1}A[l]_{i,j}-\sum_{l=1}^{n-1}A[l-1]_{r,j}$
for all $1\leqslant j\leqslant n-1$.
* B
For $0\leqslant l\leqslant n-2$, $B[l]\in\mathcal{M}_{(n-1)\times
r}(\mathbb{R})$ is the matrix $B[l]_{i,j}=\tau_{lr+i}(\eta_{j})=A[l]_{i,j}$
for $1\leqslant j\leqslant n-1$ and $1\leqslant i\leqslant r$, while
$B[n-1]\in\mathcal{M}_{(n-1)\times(r-1)}(\mathbb{R})$ is the matrix
$B[n-1]_{i,j}=\tau_{(n-1)r+i}(\eta_{j})=A[n-1]_{i,j}$ for $1\leqslant
j\leqslant n-1$ and $1\leqslant i\leqslant r-1$ (this modification of the last
$B[l]$ comes from the fact that the $(nq-1)$-st embedding $\vartheta_{nq-1}$
does not show up in the regulator, thanks to (18): the same phenomenon will
appear below in D and in $\widetilde{\textbf{D}}$).
* C
We now define $C[l]\in\mathcal{M}_{(nr)\times(r+1)}(\mathbb{R})$ to be the
matrix $C[l]_{i,j}=\tau_{lr+i}(\eta_{j})$ for $0\leqslant l\leqslant n-1$,
$n\leqslant j\leqslant n(r+1)-1$, $1\leqslant i\leqslant r$ and
$C[l]_{0,j}=2\gamma_{l}(\eta_{j})$ for $0\leqslant l\leqslant n-1$,
$n\leqslant j\leqslant n(r+1)-1$. Here again, by (18), we find
$C[n-1]_{r,j}=\displaystyle-\sum_{l=0}^{n-1}\frac{C[l]_{0,j}}{2}-\sum_{l=0}^{n-1}\sum_{i=1}^{r-1}C[l]_{i,j}-\sum_{l=1}^{n-1}C[l-1]_{r,j}$
for all $n\leqslant j\leqslant n(r+1)-1$.
* D
For $0\leqslant l\leqslant n-2$ we define $D[l]\in\mathcal{M}_{(nr)\times
r}(\mathbb{R})$ to be the matrix $D[l]_{i,j}=\tau_{lr+i}(\eta_{j})=C[k]_{i,j}$
for $n\leqslant j\leqslant n(r+1)-1$ and $1\leqslant i\leqslant r$ while, as
before, $D[n-1]\in\mathcal{M}_{(nr)\times(r-1)}(\mathbb{R})$ is the matrix
$D[n-1]_{i,j}=\tau_{(n-1)r+i}(\eta_{j})=C[n-1]_{i,j}$ for $n\leqslant
j\leqslant n(r+1)-1$ and $1\leqslant i\leqslant r-1$.
Finally, we let $\rho$ act on these last two sets of matrices: but we use (17)
to write their elements as other embeddings of the same units.
* $\widetilde{\textbf{C}}$
We set $\tilde{C}[l]\in\mathcal{M}_{(nr)\times(r+1)}(\mathbb{C})$ to be the
matrix $\tilde{C}[l]_{i,j}=\tau_{lr+i+1}(\eta_{j})=C[l]_{i+1,j}$ for
$0\leqslant l\leqslant n-1$, $n\leqslant j\leqslant n(r+1)-1$, $0\leqslant
i\leqslant r-1$ and
$\tilde{C}[l]_{r,j}=\tau_{(l+1)r}(\eta_{j})=\tilde{C}[l]_{r-1,j}$ for
$0\leqslant l\leqslant n-1$, $n\leqslant j\leqslant n(r+1)-1$. Applying (18)
we find
$\tilde{C}[n-1]_{r,j}=\tilde{C}[n-1]_{r-1,j}=\displaystyle-\sum_{i=0}^{n-1}\frac{\tilde{C}[i]_{0,j}}{2}-\sum_{l=0}^{n-1}\sum_{i=1}^{r-1}\tilde{C}[l]_{i,j}-\sum_{l=1}^{n-1}\tilde{C}[l-1]_{r,j}$
for all $n\leqslant j\leqslant n(r+1)-1$.
* $\widetilde{\textbf{D}}$
For $0\leqslant l\leqslant n-2$, we define
$\tilde{D}[l]\in\mathcal{M}_{(nr)\times(r)}(\mathbb{C})$ to be the matrix
$\tilde{D}[l]_{i,j}=\tau_{lr+i-1}(\eta_{j})=C[l]_{i-1,j}$ for $n\leqslant
j\leqslant n(r+1)-1$ and $2\leqslant i\leqslant r$ and
$\tilde{D}[l]_{1,j}=2\gamma_{l}(\eta_{j})=C[l]_{0,j}$ for $n\leqslant
j\leqslant n(r+1)-1$; for $l=n-1$,
$\tilde{D}[n-1]\in\mathcal{M}_{(nr)\times(r-1)}(\mathbb{C})$ is the matrix
$\tilde{D}[n-1]_{i,j}=\tau_{(n-1)r+i-1}(\eta_{j})=C[n-1]_{i-1,j}$ for
$n\leqslant j\leqslant n(r+1)-1$ and $2\leqslant i\leqslant r-1$ while
$\tilde{D}[n-1]_{1,j}=2\gamma_{(n-1)}(\eta_{j})=C[n-1]_{0,j}$ for $n\leqslant
j\leqslant n(r+1)-1$.
Observe that our indexing of elements in the various submatrices might be
confusing: indeed, the _row_ index always starts from $1$, as well as the
column index for B and for D, $\widetilde{\textbf{D}}$ while the _column_
index for A and for C, $\widetilde{\textbf{C}}$ starts with $0$: this is
consistent with our indexing for the embeddings. We agree to denote with
$M^{i}$ the $i$-th column of a matrix $M$ and with $M_{i}$ its $i$-th row and,
finally, we introduce the notation ${{}_{{}_{\scriptscriptstyle{2}}}\\!M}$ to
denote the matrix such that
${{}_{{}_{\scriptscriptstyle{2}}}\\!M}^{0}=(1/2)M^{0}$ and
${{}_{{}_{\scriptscriptstyle{2}}}\\!M}^{i}=M^{i}$ for $i\geqslant 1$. Having
set all this up, we put
$\Xi=\left(\begin{array}[]{c|c|c|c|c|c|c}A[0]&B[0]&A[1]&B[1]&\cdots&A[n-1]&B[n-1]\\\
&&&&&&\\\ \hline\cr&&&&&&\\\ C[0]&D[0]&C[1]&D[1]&\cdots&C[n-1]&D[n-1]\\\
&&&&&&\\\ \hline\cr&&&&&&\\\
\tilde{C}[0]&\tilde{D}[0]&\tilde{C}[1]&\tilde{D}[1]&\cdots&\tilde{C}[n-1]&\tilde{D}[n-1]\end{array}\right)\;,$
(21)
$\Psi=\left(\begin{array}[]{c|c|c|c}{{}_{{}_{\scriptscriptstyle{2}}}\\!A[0]}&{{}_{{}_{\scriptscriptstyle{2}}}\\!A[1]}&\ldots&\big{(}\frac{A[n-1]^{0}}{2},A[n-1]^{1},\ldots,A[n-1]^{r-2}\big{)}\\\
&&&\\\ \hline\cr&&&\\\
{{}_{{}_{\scriptscriptstyle{2}}}\\!B[0]}&{{}_{{}_{\scriptscriptstyle{2}}}\\!B[1]}&\ldots&\big{(}\frac{B[n-1]^{0}}{2},B[n-1]^{1},\ldots,B[n-1]^{r-2}\big{)}\end{array}\right)$
(21)
and
$\qquad\Phi=\frac{1}{2}\Big{(}A[0]^{0},A[1]^{0},\ldots,A[n-2]^{0}\Big{)}\;.$
(21)
To check (20) is a straightforward but pretty cumbersome row-and-columns
operation. We give all the details in the case $n=1$ in the Appendix (the
general case being a similar but much lengthier and heavier computation).
Hence (16) holds and in particular (see [Wa], Lemma 4.15)
$(E_{F}:E_{k}E_{k^{\prime}})=\frac{R_{F}[E_{k}E_{k^{\prime}}]}{R_{F}}=\frac{2^{n-1}qR_{k}^{2}}{R_{K}R_{F}}(\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times}):\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{K}^{\times}).$
(22)
Now note that, using Lemma 2.3, we get
$\frac{(\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{F}^{\times}:\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times}))}{(\mathcal{O}_{L}^{\times}:\mathcal{O}_{K}^{\times})}=\frac{(\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{F}^{\times}:\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{K}^{\times})}{(\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times}):\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{K}^{\times})(E_{L}:E_{K})(\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{L}^{\times}:\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{K}^{\times})}$
$=\frac{1}{(\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times}):\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{K}^{\times})Q}$
where
$Q=(\mathcal{O}_{L}^{\times}:\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{L}^{\times})\mathcal{O}_{K}^{\times})=(E_{L}:E_{K})$
is equal to $1$ or $2$ (see Theorem 4.12 of [Wa]). Now by (22) we have
$(E_{F}:E_{k}E_{k^{\prime}})=\frac{2^{n-1}qR_{k}^{2}(\mathcal{O}_{L}^{\times}:\mathcal{O}_{K}^{\times})}{QR_{K}R_{F}(\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{F}^{\times}:\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times}))}$
which is exactly (15), thanks to Proposition 4.16 of [Wa]. ∎
###### Theorem 3.4.
[See [HK] Satz 5, [Ja1] Proposition 12, [Le] Theorem 2.2] Let $q$ be an odd
number and let $K$ be a totally real number field. For every totally imaginary
dihedral extension $F/K$ of degree $2q$ the equality
$h_{F}=\frac{(\mathcal{O}_{F}^{\times}:\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times}\mathcal{O}_{L}^{\times})}{(\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times}\cap\mathcal{O}_{L}^{\times}:\mathcal{O}_{K}^{\times})}\cdot\frac{h_{L}h_{k}^{2}}{qh_{K}^{2}}$
holds, where $k\subset F$ is a subfield of index $2$, $k^{\prime}=\rho(k)$ for
some element of order $q$ in $\mathrm{Gal}(F/K)$ and $L\subset F$ is the
subfield of index $q$.
###### Proof.
Just apply Theorem 2.4 together with Proposition 3.3. ∎
Remark. Note that, when $K=\mathbb{Q}$, then
${(\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times}\cap\mathcal{O}_{L}^{\times}:\mathcal{O}_{K}^{\times})}=1.$
This can be seen as follows: suppose $x\in\mathcal{O}_{k}^{\times}$,
$x^{\prime}\in\mathcal{O}_{k^{\prime}}^{\times}$ and
$xx^{\prime}\in\mathcal{O}_{L}^{\times}$. Then $(xx^{\prime})^{12}=1$, since
$\mathcal{O}_{L}^{\times}=\mathrm{tor}_{\mathbb{Z}}(\mathcal{O}_{L}^{\times})$
because $L$ is imaginary quadratic. This implies that
$x^{12}\in\mathcal{O}_{k}^{\times}\cap\mathcal{O}_{k^{\prime}}^{\times}=\mathcal{O}_{\mathbb{Q}}^{\times}=\\{\pm
1\\}.$
In particular $x\in\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{k}^{\times}$ and,
since $\mu(k)=\mu(\mathbb{Q})$, we must have $x=1$ or $x=-1$ (see also [HK],
Satz 5).
## 4 Iwasawa type class number formula.
In this section we consider the behaviour of the class number in a tower of
dihedral fields. First of all, recall the following classical definition:
###### Definition 4.1.
Let $K$ be a number field and let $p$ be a prime number. A Galois extension
$K_{\infty}/K$ such that $\mathrm{Gal}(K_{\infty}/K)\cong\mathbb{Z}_{p}$ is
called a $\mathbb{Z}_{p}$-extension. In this case,
$\bigcup_{n\in\mathbb{N}}K_{n}=K_{\infty}\supset\ldots\supset K_{n}\supset
K_{n-1}\supset\ldots K_{1}\supset K_{0}=K\;,$
where $K_{n}/K$ is a Galois extension such that
$\mathrm{Gal}(K_{n}/K)\cong\mathbb{Z}/p^{n}\mathbb{Z}$.
The behaviour of the class number in this tower is controlled by a celebrated
theorem of Kenkichi Iwasawa, namely
###### Theorem (K. Iwasawa, [Iw], Theorem 4.2).
Let $K_{\infty}/K$ be a $\mathbb{Z}_{p}$-extension and let $p^{e_{n}}$ be the
exact power of $p$ dividing the class number of $K_{n}$. Then there exist
three integers $\mu,\lambda$ and $\nu$ such that
$e_{n}=\mu p^{n}+\lambda n+\nu\qquad\mathrm{for}\;n\gg 0\,.$
We want now to investigate if the same holds in a more general setting, namely
dropping the Galois condition. We start with the following
###### Definition 4.2.
Let $p$ be a prime number, let $K$ be a number field and let $K_{\infty}/K$ be
a _non_ Galois extension. Suppose that there exists a Galois extension $L/K$
disjoint from $K_{\infty}/K$ such that $LK_{\infty}$ is a Galois closure of
$K_{\infty}/K$. If $LK_{\infty}/L$ is a $\mathbb{Z}_{p}$-extension, then
$K_{\infty}/K$ is called a fake $\mathbb{Z}_{p}$-extension.
Remark. If $K_{\infty}/K$ is a fake $\mathbb{Z}_{p}$-extension as in Definiton
4.2, then
$\mathrm{Gal}(LK_{\infty}/K)\cong\mathrm{Gal}(LK_{\infty}/L)\rtimes\mathrm{Gal}(LK_{\infty}/K_{\infty}).$
Indeed, one can also formulate the definition of fake
$\mathbb{Z}_{p}$-extensions in terms of structures of Galois groups. Moreover
$K_{\infty}$ is then the union
$K_{\infty}=\bigcup_{n\in\mathbb{N}}K_{n}$
where $K_{n}$ is the extension of $K$ fixed by
$\mathrm{Gal}(LK_{\infty}/L)^{p^{n}}\rtimes\mathrm{Gal}(LK_{\infty}/K_{\infty})$,
of degree $p^{n}$. Note, moreover, that $K_{n}/K$ is the only subextension of
$K_{\infty}/K$ of degree $p^{n}$ and every subextension of $K_{\infty}/K$ is
one of the $K_{n}$, a property which is also enjoyed by
$\mathbb{Z}_{p}$-extension. It would be interesting to know whether or not
this property characterize (fake)-$\mathbb{Z}_{p}$-extensions. We thank
Gabriele Dalla Torre for many fruitful discussions on this subject.
As an example of a fake $\mathbb{Z}_{p}$-extension, let $L$ be an imaginary
quadratic field and let $p$ be an odd prime: it is known (see, for instance,
[Wa], chapter 13) that the compositum of its $\mathbb{Z}_{p}$-extensions has
Galois group isomorphic to $\mathbb{Z}_{p}^{2}$. Since
$\mathrm{Gal}(L/\mathbb{Q})$ acts semisimply on this Galois group, it
decomposes $\mathbb{Z}_{p}^{2}$ accordingly to its characters, giving two
independent $\mathbb{Z}_{p}$-extensions, both Galois over $\mathbb{Q}$: the
cyclotomic $\mathbb{Z}_{p}$-extension $L_{cyc}$ and the anticyclotomic one
$L_{\infty}$. The first one is cyclic over $\mathbb{Q}$, the second one is
pro-dihedral, namely,
$\mathrm{Gal}(L_{cyc}/\mathbb{Q})\cong\varprojlim\mathbb{Z}/p^{n}\mathbb{Z}\times\mathbb{Z}/2\mathbb{Z}\quad\mathrm{while}\quad\mathrm{Gal}(L_{\infty}/\mathbb{Q})\cong\varprojlim
D_{p^{n}}\;.$
Let $D_{p^{\infty}}$ be the pro-dihedral group isomorphic to
$\mathrm{Gal}(L_{\infty}/\mathbb{Q})$: it admits two topological generators,
which we consider fixed from now on, $\sigma$ and $\rho_{\infty}$ such that
$\sigma^{2}=1\qquad\sigma\rho_{\infty}=\rho_{\infty}^{-1}\sigma\;.$ (23)
If $L_{n}$ is the $n$-th layer of the anticyclotomic extension of $L$, then
$L_{n}=L_{\infty}^{\langle\rho_{\infty}^{p^{n}}\rangle}$ (where
$\langle\rho_{\infty}^{p^{n}}\rangle$ denotes the closed subgroup of
$D_{p^{\infty}}$ generated by $\rho_{\infty}^{p^{n}}$): we define,
accordingly,
$K_{n}=L_{\infty}^{\langle\rho_{\infty}^{p^{n}},\sigma\rangle}\subset L_{n}$
(where $\langle\rho_{\infty}^{p^{n}},\sigma\rangle$ denotes the closed
subgroup of $D_{p^{\infty}}$ generated by $\rho_{\infty}^{p^{n}}$ and
$\sigma$). Then $K_{n}$ is a (non normal) extension of $K=\mathbb{Q}$ of
degree $p^{n}$ and we set $K_{\infty}=L_{\infty}^{\langle\sigma\rangle}$ to
find a diagram of fields as in the introduction. Therefore
$K_{\infty}/\mathbb{Q}$ is a fake $\mathbb{Z}_{p}$-extension.
Another example may be the following: let $K=\mathbb{Q}(\zeta_{p})$ where $p$
is a primitive $p$-th root of unity and let $a\in K^{\times}$,
$a\notin\mu_{p}$. Then $K(\sqrt[p^{\infty}]{a})/K$ is a fake
$\mathbb{Z}_{p}$-extension, as it can easily be seen by taking
$L=F(\zeta_{p^{\infty}})$: this case would fit in a much more general setting,
as the one introduced in [VV], and we hope to investigate it in a future work.
In the sequel we study the pro-dihedral case over $\mathbb{Q}$, with notations
introduced in the above example. The main result of this section is then
Theorem 4.7 below. The strategy for the study of the growth of the class
number in this setting is given by Theorem 3.4. In the following, we shall
always make the following
* [H]
Hypothesis: If $p=3$, then $L\neq\mathbb{Q}(\sqrt{-3})$
This is not a real restriction, since in that case the class groups of
$K_{n}$, $L_{n}$ and $L$ have trivial $3$-Sylow subgroups and any of the
stated result trivially holds. For $n\geqslant 1$, we define
$P_{n}=\mathcal{O}_{K_{n}}^{\times}\mathcal{O}_{K_{n}^{\prime}}^{\times}\otimes\mathbb{Z}_{p},\quad
U_{n}=\mathcal{O}_{L_{n}}\otimes\mathbb{Z}_{p},\quad R_{n}=U_{n}/P_{n}$
where the inclusion $P_{n}\hookrightarrow U_{n}$ is induced by the injection
$\mathcal{O}_{K_{n}}^{\times}\mathcal{O}_{K_{n}^{\prime}}^{\times}\hookrightarrow
E_{L_{n}}$: therefore $|R_{n}|$ is the $p$-part of the quotient of units
appearing in Theorem 3.4 (note that $E_{L}$ is trivial thanks to assumption
[H]). Morever, we can also write
$P_{n}=E_{K_{n}}E_{K_{n}^{\prime}}\otimes\mathbb{Z}_{p}\quad\text{ and }\quad
U_{n}=E_{L_{n}}\otimes\mathbb{Z}_{p}$
because $\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{K_{n}}^{\times}$ and
$\mathrm{tor}_{\mathbb{Z}}\mathcal{O}_{L_{n}}^{\times}$ are always of order
coprime to $p$, again by [H]. Note that $P_{n}$, $U_{n}$ and $R_{n}$ are
$D_{p^{n}}$-modules (see for example [HK], Lemma 1). We let
$\Gamma=\mathrm{Gal}(L_{\infty}/L)$ and we write
$G_{n}=\Gamma/\Gamma^{p^{n}}=\mathrm{Gal}(L_{n}/L)$ (or, more generally,
$G_{m,n}=\mathrm{Gal}(L_{m}/L_{n})$ for $m\geqslant n\geqslant 0$), while we
henceforth call $\Delta$ the group $\mathrm{Gal}(L/\mathbb{Q})$ and we fix a
subgroup of $\mathrm{Gal}(L_{\infty}/\mathbb{Q})$, also called $\Delta$,
mapping isomorphically onto $\mathrm{Gal}(L/\mathbb{Q})$ via restriction.
Finally, $\Lambda=\mathbb{Z}_{p}[[\Gamma]]$ will be the completed group
algebra of $\Gamma$, isomorphic to $\mathbb{Z}_{p}[[T]]$ by the choice of a
topological generator $\gamma_{0}$ of $\Gamma$.
What we want to do is to compare the action of $G_{n}$ with that of
$D_{p^{n}}$, by comparing their cohomology. We start with the following
cohomological result which is certainly well-know but difficult to find in
print. As a matter of notation, recall that if $H$ is a group and $B$ an
$H$-module, then $B^{H}$ denotes invariants, $B_{H}$ coinvariants,
$N_{H}=\sum_{h\in H}h\in\mathbb{Z}[H]$ the norm, $B[N_{H}]$ the kernel of
multiplication by the norm and $I_{H}$ the augmentation ideal.
###### Proposition 4.3.
Let $A$ be a $2$-divisible abelian $D_{p^{n}}$-module: then, for every
$i\geqslant 0$, there are canonical isomorphisms induced by restriction (or
corestriction)
$\begin{split}H^{i}(D_{p^{n}},A)\cong H^{i}(G_{n},A)^{\Delta}\;,\\\
H_{i}(D_{p^{n}},A)\cong H_{i}(G_{n},A)_{\Delta}\;.\end{split}$
Moreover,
$\widehat{H}^{0}(D_{p^{n}},A)\cong\widehat{H}^{0}(G_{n},A)^{\Delta}$
and
$\widehat{H}^{-1}(D_{p^{n}},A)\cong\widehat{H}^{-1}(G_{n},A)_{\Delta}\cong\widehat{H}^{-1}(G_{n},A)^{\Delta}\;.$
Finally, the Tate isomophism
$\widehat{H}^{i}(G_{n},A)\cong\widehat{H}^{i+2}(G_{n},A)$ is
$\Delta$-antiequivariant, so that (as $D_{p^{n}}$-modules with trivial action)
$\widehat{H}^{-1}(D_{p^{n}},A)=\widehat{H}^{-1}(G_{n},A)^{\Delta}=\widehat{H}^{1}(G_{n},A)/\widehat{H}^{1}(D_{p^{n}},A)$
and
$\widehat{H}^{1}(D_{p^{n}},A)=\widehat{H}^{1}(G_{n},A)^{\Delta}=\widehat{H}^{-1}(G_{n},A)/\widehat{H}^{-1}(D_{p^{n}},A)$
###### Proof.
The first isomorphism is an immediate application of the Hochschild-Serre
Spectral Sequence (see, for instance, [We], 6.8). We now consider Tate
cohomology: taking $\Delta$-invariants in the tautological sequence
$0\rightarrow N_{G_{n}}A\rightarrow
H^{0}(G_{n},A)\rightarrow\widehat{H}^{0}(G_{n},A)\rightarrow 0$
we find (use, as before, that $N_{G_{n}}A$ is $2$-divisible, thus its
$\Delta$-cohomology is trivial)
$\widehat{H}^{0}(G_{n},A)^{\Delta}=H^{0}(G_{n},A)^{\Delta}/(N_{G_{n}}A)^{\Delta}$.
Since, as observed, $N_{G_{n}}A$ has trivial $\Delta$-cohomology and $\Delta$
is a cyclic group,
$\widehat{H}^{0}(\Delta,N_{G_{n}}A)\cong H^{2}(\Delta,N_{G_{n}}A)=0$
so that $(N_{G_{n}}A)^{\Delta}=N_{\Delta}N_{G_{n}}A=N_{D_{p^{n}}}A$ and
finally (using the first isomorphism in our statement)
$\widehat{H}^{0}(G_{n},A)^{\Delta}=H^{0}(G_{n},A)^{\Delta}/(N_{G_{n}}A)^{\Delta}=H^{0}(D_{p^{n}},A)/N_{D_{p^{n}}}A=\widehat{H}^{0}(D_{p^{n}},A)$
as claimed. In degree $-1$, take $\Delta$-coinvariants of the tautological
exact sequence defining the Tate group to get the sequence
$0\rightarrow\widehat{H}^{-1}(G_{n},A)_{\Delta}\rightarrow
H_{0}(G_{n},A)_{\Delta}\rightarrow(A/A[N_{G_{n}}])_{\Delta}\rightarrow 0\;.$
(24)
where, as before, we have
$(A/A[N_{G_{n}}])_{\Delta}=A_{\Delta}/A[N_{G_{n}}]_{\Delta}$. Take now
$\Delta$-coninvariants in the exact sequence
$0\rightarrow A[N_{G_{n}}]\rightarrow A\rightarrow N_{G_{n}}A\rightarrow 0$
(25)
to identify the quotient $A_{\Delta}/A[N_{G_{n}}]_{\Delta}$ with
$(N_{G_{n}}A)_{\Delta}$. We claim that
$A/A[N_{D_{p^{n}}}]\stackrel{{\scriptstyle
N_{G_{n}}}}{{\longrightarrow}}(N_{G_{n}}A)_{\Delta}=N_{G_{n}}A/I_{\Delta}(N_{G_{n}}A)$
(26)
is an isomorphism: first of all, the map is well defined, since for every
$x\in A[N_{D_{p^{n}}}]$, we have $N_{G_{n}}(x)\in
N_{G_{n}}(A)[N_{\Delta}]=(I_{\Delta})N_{G_{n}}(A)$ because
$\widehat{H}^{-1}(\Delta,N_{G_{n}})=H^{1}(\Delta,N_{G_{n}})=0$. The same
argument shows injectivity, since for every $a\in A$ such that
$N_{\Delta}(N_{G_{n}}(a))=0$, we have $N_{D_{p^{n}}}(a)=0$, while surjectivity
is obvious. Plugging now the isomorphism of (26) in (24) through the
identification induced by (25) we find
$0\rightarrow\widehat{H}^{-1}(G_{n},A)_{\Delta}\rightarrow
H_{0}(D_{p^{n}},A)\rightarrow A/A[N_{D_{p^{n}}}]\rightarrow 0\;.$
showing our claim. The fact now that
$\widehat{H}^{-1}(G_{n},A)^{\Delta}=\widehat{H}^{-1}(G_{n},A)_{\Delta}$ comes
from splitting any $2$-divisible $\Delta$-module $M$ as $M=M^{+}\oplus M^{-}$
_canonically_ , writing $m=(m+\delta m)/2+(m-\delta m)/2$: here we denote by
$M^{+}$ the eigenspace on which $\Delta$ acts trivially and by $M^{-}$ the
eigenspace on which it acts as $-1$. Then
$M^{\Delta}=M^{+}=M/M^{-}=M_{\Delta}$.
Finally, we discuss the $\Delta$-antiequivariance of Tate isomorphisms. Recall
that the isomorphism is given by the cup product with a fixed generator $\chi$
of $H^{2}(G_{n},\mathbb{Z})$:
$\begin{array}[]{ccc}\widehat{H}^{i}(G_{n},A)&\longrightarrow&\widehat{H}^{i+2}(G_{n},A)\\\
x&\longmapsto&x\cup\chi\end{array}$
The action of $\delta\in\Delta$ on $\widehat{H}^{i}(G_{n},A)$ is $\delta_{*}$
in the notation of [NSW], I.5 and this action is $-1$ on
$H^{2}(G_{n},\mathbb{Z})$ as can immediately be seen through the isomorphism
$H^{2}(G_{n},\mathbb{Z})\cong\mathrm{Hom}(G_{n},\mathbb{Q}/\mathbb{Z})$ (see
[We], example 6.7.10). Then, by Proposition 1.5.3 of [NSW],
$\delta_{*}(x\cup\chi)=-(\delta_{*}x)\cup\chi$ which gives the result. ∎
The key tool for studying the growth of the $p$-part of $h_{K_{n}}$ along the
fake $\mathbb{Z}_{p}$-extension is to interpret the quotient $R_{n}$ as a
cohomology group. We have the following
###### Proposition 4.4.
With notations as above, for every $n\geqslant 0$ there is an isomorphisms of
abelian groups
$R_{n}\cong H^{1}(G_{n},U_{n})^{\Delta}\cong H^{1}(D_{p^{n}},U_{n})\;.$
###### Proof.
Along the proof, set $V_{n}=E_{K_{n}}\otimes\mathbb{Z}_{p}$,
$V_{n}^{\prime}=E_{K^{\prime}_{n}}\otimes\mathbb{Z}_{p}$.
$U_{n}/I_{G_{n}}U_{n}$ is a $2$-divisible $\Delta$-module. Hence
$U_{n}/I_{G_{n}}U_{n}=(U_{n}/I_{G_{n}}U_{n})^{+}\oplus(U_{n}/I_{G_{n}}U_{n})^{-}\;,$
as in the proof of Proposition 4.3. Moreover, we claim that
$(U_{n}/I_{G_{n}}U_{n})^{+}=(U_{n}/I_{G_{n}}U_{n})^{\Delta}=V_{n}V_{n}^{\prime}/I_{G_{n}}U_{n}\;:$
(27)
this is quite clear by definition of the action of $\Delta$ since
$V_{n}I_{G_{n}}U_{n}/I_{G_{n}}U_{n}=(U_{n}/I_{G_{n}}U_{n})^{\langle\sigma\rangle}=(U_{n}/I_{G_{n}}U_{n})^{\Delta}$
but also
$V_{n}^{\prime}I_{G_{n}}U_{n}/I_{G_{n}}U_{n}=(U_{n}/I_{G_{n}}U_{n})^{{}^{\langle\rho^{2}\sigma\rangle}}=(U_{n}/I_{G_{n}}U_{n})^{\Delta}\;.$
We deduce that
$V_{n}V_{n}^{\prime}I_{G_{n}}U_{n}/I_{G_{n}}U_{n}=(U_{n}/I_{G_{n}}U_{n})^{\Delta}\;,$
and since
$I_{G_{n}}U_{n}\subseteq V_{n}V_{n}^{\prime}$
(see for example [Le], Lemma 3.3) we get (27). Then
$R_{n}=U_{n}/I_{G_{n}}U_{n}\Big{/}V_{n}V_{n}^{\prime}/I_{G_{n}}U_{n}\cong(U_{n}/I_{G_{n}}U_{n})^{-}$
as $\Delta$-modules. Since
$U_{n}/I_{G_{n}}U_{n}=\widehat{H}^{-1}(G_{n},U_{n})$ we find, by Proposition
4.3, that
$R_{n}=\widehat{H}^{-1}(G_{n},U_{n})/\widehat{H}^{-1}(G_{n},U_{n})^{\Delta}=\widehat{H}^{1}(D_{p^{n}},U_{n})\;.$
(28)
∎
###### Corollary 4.5.
$R_{n}$ is a $G_{n}$-module with trivial action and the injection
$i_{n}:U_{n}\hookrightarrow U_{n+1}$ induces an injective map
$i_{n}:R_{n}\hookrightarrow R_{n+1}\;.$
###### Proof.
By the first equality in (28), the action of $G_{n}$ on $R_{n}$ is trivial
since $R_{n}$ is a quotient of a trivial $G_{n}$-module. Now the induced map
$i_{n}:R_{n}\rightarrow R_{n+1}$ corresponds to the restriction on minus parts
of
$i_{n}:U_{n}/I_{G_{n}}U_{n}\rightarrow U_{n+1}/I_{G_{n+1}}U_{n+1}\;.$
The commutativity of the diagram
|
---|---
$\textstyle{\widehat{H}^{-1}(G_{n},U_{n})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{i_{n}}$$\scriptstyle{\cup\chi}$$\scriptstyle{\cong}$$\textstyle{\widehat{H}^{-1}(G_{n+1},U_{n+1})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\cup\chi}$$\scriptstyle{\cong}$$\textstyle{\widehat{H}^{1}(G_{n},U_{n})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\mathrm{inf}}$$\textstyle{\widehat{H}^{1}(G_{n+1},U_{n+1})}$
is immediate to check and proves our statement. ∎
Remark. For $m\geqslant n\geqslant 0$, let $N_{m,n}:L_{m}\rightarrow L_{n}$ be
the usual “arithmetic” norm. With the same notation we will also indicate the
induced maps on $E_{L_{n}}$, or on $E_{K_{n}}$, as well as on $P_{n}$ and
$U_{n}$. One can check that this induces a well-defined map
$N_{m,n}:R_{m}\rightarrow R_{n}$, so that we can form the projective limits of
the tautological exact sequence
$0\longrightarrow P_{n}\longrightarrow U_{n}\longrightarrow
R_{n}\longrightarrow 0$
to get an exact sequence
$0\longrightarrow\varprojlim P_{n}:=P_{\infty}\longrightarrow\varprojlim
U_{n}:=U_{\infty}\longrightarrow\varprojlim R_{n}:=R_{\infty}\longrightarrow
0$ (29)
that is exact on the right since $\varprojlim^{1}P_{n}=0$ as all the $P_{n}$’s
are compact modules (see, for instance, [We], Proposition 3.5.7): in
particular, $R_{\infty}\cong U_{\infty}/P_{\infty}$ as $\Lambda$-modules. In
Iwasawa theory, one classically tries to get information at finite levels from
the study of some $\Lambda$-module, via the so-called _co-descent_ maps:
indeed, if $Z=\varprojlim Z_{n}$ is a $\Lambda$-module one has a co-descent
map $k_{n}:(Z)_{\Gamma_{n}}\rightarrow Z_{n}$. Since the size of
$(Z)_{\Gamma_{n}}$ is well-behaved with respect to $n$, if one can bound the
orders of $\mathrm{ker}(k_{n})$ and of $\mathrm{coker}(k_{n})$ indipendently
of $n$, then one can also control the growth of $Z_{n}$. Unfortunately, we
cannot apply this strategy to study the order of $R_{n}$, since Corollary 4.5
shows, by passing to the limit, that the $\Lambda$-module $R_{\infty}$ has a
trivial action of $\Gamma$ and this is precisely the obstruction for the
boundness of $\mathrm{ker}(k_{n})$ and of $\mathrm{coker}(k_{n})$.
Before stating our main result, we need a general lemma. In the following, for
a number field $M$, denote by $A_{M}$ the $p$-Sylow subgroup of the class
group of $M$ (isomorphic to the maximal $p$-quotient of the class group).
###### Lemma 4.6.
Let $M_{\infty}/M$ be a $\mathbb{Z}_{p}$-extension in which all primes above
$p$ are ramified. For every sufficiently large $n$ (_i. e._ large enough that
all primes above $p$ are totally ramified in $M_{\infty}/M_{n}$) let
$\mathfrak{p}_{1,n},\ldots,\mathfrak{p}_{s,n}$ be the primes in $M_{n}$ above
$p$ and let $\mathfrak{P}_{n}=\prod_{i=1}^{s}\mathfrak{p}_{i,n}$ be their
product. Then there exist two integers
$\lambda_{\mathfrak{P}},\nu_{\mathfrak{P}}$ independent of $n$ such that the
order of the the projection of the class of $\mathfrak{P}_{n}$ in $A_{M_{n}}$
is $n\lambda_{\mathfrak{P}}+\nu_{\mathfrak{P}}$.
###### Proof.
For every $n\in\mathbb{N}$, let $H_{n}$ be the cyclic subgroup of $A_{M_{n}}$
generated by the projection of $\mathfrak{P}_{n}$. Clearly, the $H_{n}$’s form
a projective system and we set $Y=\varprojlim H_{n}$. Setting $X=\varprojlim
A_{M_{n}}$, then $Y\subseteq X$ is a $\Lambda$-module and $X/Y$ is a
noetherian $\Lambda$-module (it corresponds to the maximal unramified
extension of $M_{\infty}$ in which the product of all Frobenius automorphisms
of primes above $p$ is trivial): let $\mu,\lambda,\nu$ be the Iwasawa
invariants of $X$. Then $X/Y$ also admits three Iwasawa invariants
$\tilde{\lambda},\tilde{\mu},\tilde{\nu}$: moreover, $Y$ is clearly finitely
generated over $\mathbb{Z}_{p}$, so $\tilde{\mu}=\mu$. Setting
$\lambda_{\mathfrak{P}}=\lambda-\tilde{\lambda}$ and
$\nu_{\mathfrak{P}}=\nu-\tilde{\nu}$ we establish the Lemma. ∎
Remark. Observe that the proof itself shows that $Y$ is procyclic, so it is
either finite or free of rank $1$ over $\mathbb{Z}_{p}$ and, accordingly,
$\lambda_{\mathfrak{P}}\leqslant 1$. We will come later on this. Now we go
back to our anticyclotomic setting.
###### Theorem 4.7.
Let $p^{\varepsilon_{n}}$ be the order of the $p$-Sylow class group of
$K_{n}$. Then there exist integers $\mu_{K},\lambda_{K},\nu_{K}$ such that
$2\varepsilon_{n}=\mu_{K}p^{n}+\lambda_{K}n+\nu_{K}\qquad\mathrm{for}\quad
n\gg 0\;.$
###### Proof.
As it is well-known (see, for instance, [Wa], Lemma 13.3) only primes above
$p$ may ramify in $L_{\infty}/L$ and at least one of those must eventually
ramify, while the fact that $L_{n}/\mathbb{Q}$ is Galois for every $n$ shows
that, if one is ramified in $L_{n}$ so is the other (if it exists) and with
the same ramification index. Let thus $n_{0}$ be the smallest integer such
that they are totally ramified in $L_{\infty}/L_{n_{0}}$ and assume
$n\geqslant\max\\{n_{0},\tilde{n}\\}$ where $\tilde{n}$ is the smallest
integer such that the formula in Iwasawa’s theorem (see the beginning of this
section) for $L_{\infty}/L$ applies. Then by Theorem 3.4 applied with
$k=K_{n}$ and $F=L_{n}$ we have
$2\varepsilon_{n}=e_{n}-r_{n}+n-f=\mu_{L}p^{n}+(\lambda_{L}+1)n+\nu^{\prime}-r_{n}\;,$
(30)
where $|R_{n}|=p^{r_{n}}$, $|A_{L}|=p^{f}$, $\nu^{\prime}=\nu_{L}-f$ and
$\mu_{L},\lambda_{L},\nu_{L}$ are the Iwasawa invariants of $L_{\infty}/L$. We
thus want to control the growth of $r_{n}$ along the tower. To do this, we
apply Proposition 4.4 studying explicitly $H^{1}(D_{p^{n}},U_{n})$, since
$r_{n}=v_{p}(|H^{1}(D_{p^{n}},U_{n})|)\;.$ (31)
To analyze $H^{1}(D_{p^{n}},U_{n})$, set
$B^{\diamond}:=B\otimes_{\mathbb{Z}}\mathbb{Z}_{p}$ for any abelian group $B$:
this is an exact functor so we have the exact sequence
$0\rightarrow U_{n}\rightarrow(L_{n}^{\times})^{\diamond}\rightarrow
Pr_{n}^{\diamond}\rightarrow 0$
where $Pr_{n}$ is the group of principal ideals of $L_{n}$. Taking
$D_{p^{n}}$-cohomology we get, by Hilbert 90, an isomorphism333We use, here
and in what follows, that for every $G$-module $A$, there is an isomorphism
$H^{q}(G,A)\otimes\mathbb{Z}_{p}\cong H^{q}(G,A\otimes\mathbb{Z}_{p})$ holding
for every $q\geqslant 0$. This is an easy exercise about the Grothendieck
spectral sequence for the functors $(-)\otimes\mathbb{Z}_{p}$ and $(-)^{G}$.
To verify that tensoring with $\mathbb{Z}_{p}$ sends injective $G$-modules to
$G$-acyclic, use the explicit description in [Se2], chapitre VII. For the
equivalent result in Tate cohomology apply Proposition 4.3 together with the
above remark. René Schoof pointed out to us that one can also prove directly
the isomorphism
$\widehat{H}^{q}(G,A)\otimes\mathbb{Z}_{p}\cong\widehat{H}^{q}(G,A\otimes\mathbb{Z}_{p})$
by tensoring the complex giving raise to Tate cohomology with
$\mathbb{Z}_{p}$.
$(Pr_{n}^{\diamond})^{D_{p^{n}}}/(\mathbb{Q}^{\times})^{\diamond}\cong
H^{1}(D_{p^{n}},U_{n})\;.$ (32)
On the other hand, the exact sequence
$0\rightarrow Pr_{n}^{\diamond}\rightarrow Id_{n}^{\diamond}\rightarrow
A_{L_{n}}\rightarrow 0$ (33)
defining the class group (so $Id_{n}$ is the group of fractional ideals of
$L_{n}$) induces an inclusion
$(Pr_{n}^{\diamond})^{D_{p^{n}}}/(\mathbb{Q}^{\times})^{\diamond}\hookrightarrow(Id_{n}^{\diamond})^{D_{p^{n}}}/(\mathbb{Q}^{\times})^{\diamond}\;.$
This last quotient is fairly explicit: indeed, an ideal
$I=\prod\mathfrak{q}_{i}^{a_{i}}$ is fixed by $D_{p^{n}}$ if and only if every
prime appears with the same exponent with all its $D_{p^{n}}$-conjugates: for
a prime $\mathfrak{l}$ call the product of all this conjugates
$Orb(\mathfrak{l})$. Then clearly (recall that all modules here are
$\mathbb{Z}_{p}$-modules, so primes ramified only in $L/\mathbb{Q}$ generate
the same module as the rational prime below them)
$Orb(\mathfrak{l})^{\mathbb{Z}_{p}}=\left\\{\begin{array}[]{llr}\ell^{\mathbb{Z}_{p}}&\mathrm{where\;}\mathfrak{l}\mid\ell\in\mathbb{Q}&\mathrm{if\;}\mathfrak{l}\nmid
p\\\ \\\ \mathfrak{P}_{n}^{\mathbb{Z}_{p}}=(\prod_{\mathfrak{p}_{n}\mid
p\,\,\textrm{in
$L_{n}$}}\mathfrak{p}_{n})^{\mathbb{Z}_{p}}&\mathrm{where\;}\mathfrak{P}_{n}^{p^{n-n_{0}}}=p\in\mathbb{Q}&\mathrm{if\;}\mathfrak{l}\mid
p\end{array}\right.\;.$
The fact that these are the only possibilities for ramification follows from
the definition of $n_{0}$: moding now out by
$(\mathbb{Q}^{\times})^{\diamond}$ we find
$(Id_{n}^{\diamond})^{D_{p^{n}}}/(\mathbb{Q}^{\times})^{\diamond}=\mathfrak{P}_{n}^{\mathbb{Z}_{p}}/p^{\mathbb{Z}_{p}}\cong\mathbb{Z}/p^{n-n_{0}}\mathbb{Z}\;,$
and accordingly
$(Pr_{n}^{\diamond})^{D_{p^{n}}}/(\mathbb{Q}^{\times})^{\diamond}=\mathfrak{P}_{n}^{p^{h_{n}}\mathbb{Z}_{p}}/p^{\mathbb{Z}_{p}}\cong\mathbb{Z}/(p^{n-n_{0}-h_{n}})\mathbb{Z}\;$
where $p^{h_{n}}$ is the order of the class of $\mathfrak{P}_{n}$ in
$A_{L_{n}}$. Applying Lemma 4.6 we find
$h_{n}=\lambda_{\mathfrak{P}}n+\nu_{\mathfrak{P}}$ and this, together with
(32), shows that
$H^{1}(D_{p^{n}},U_{n})\cong\mathbb{Z}/p^{(1-\lambda_{\mathfrak{P}})n-n_{0}-\nu_{\mathfrak{P}}}\mathbb{Z}\;.$
We now achieve the proof of the theorem plugging this information in (31) in
order to find the existence of suitable invariants $\lambda_{r}$ and $\nu_{r}$
such that $r_{n}=\lambda_{r}n+\nu_{r}$, so that equation (30) becomes our
statement. ∎
Remark. Following explicitly the proof, one finds that
$\lambda_{r}=1-\lambda_{\mathfrak{P}}$, which is at most $1$ by the above
remark, and $\nu_{r}=-n_{0}-\nu_{\mathfrak{P}}$. Accordingly,
$\mu_{K}=\mu_{L}\;,\qquad\lambda_{K}=\lambda_{L}+1-\lambda_{r}=\lambda_{L}+\lambda_{\mathfrak{P}}\;,\qquad\nu_{K}=\nu_{L}-f+n_{0}+\nu_{\mathfrak{P}}\;:$
in particular, $\lambda_{K}$ is either $\lambda_{L}$ or $\lambda_{L}+1$ and it
is _even_ : to see this, just use our formula to write explicitly
$2(\varepsilon_{n+1}-\varepsilon_{n})$. Analogously one can prove that
$\mu_{K}\equiv\nu_{K}\pmod{2}$.
###### Definition 4.8.
Denote by $p^{h_{n}}$ the order of the class of $\mathfrak{P}_{n}$ in
$A_{L_{n}}$, where $\mathfrak{P}_{n}$ is the product of all primes above $p$
in $L_{n}$. Moreover, let $n_{0}$ be the smallest integer such that
$L_{\infty}/L_{n_{0}}$ is totally ramified.
For later use, we extract the following result from the proof of the theorem.
###### Proposition 4.9.
There is a short exact sequence
$0\to H^{1}(D_{p^{n}},U_{n})\to\mathbb{Z}/p^{n-n_{0}}\mathbb{Z}\to
H^{0}(D_{p^{n}},A_{L_{n}})\to 0$
and isomorphisms
$H^{1}(D_{p^{n}},U_{n})\cong\mathbb{Z}/p^{n-n_{0}-h_{n}}\mathbb{Z},\quad
H^{0}(D_{p^{n}},A_{L_{n}})\cong\mathbb{Z}/p^{h_{n}}\mathbb{Z}\;.$
###### Proof.
In the long exact $D_{p^{n}}$-cohomology sequence of
$0\to U_{n}\to(L_{n}^{\times})^{\diamond}\to Pr_{n}^{\diamond}\to 0$
one has $H^{1}(D_{p^{n}},(L_{n}^{\times})^{\diamond})=0$ by Hilbert 90
together with
$H^{2}(D_{p^{n}},U_{n})=H^{2}(G_{n},U_{n})^{\Delta}=\widehat{H}^{0}(G_{n},U_{n})^{-}=0$
as $U_{0}=0$; thus $H^{1}(D_{p^{n}},Pr_{n}^{\diamond})=0$. Taking
$D_{p^{n}}$-cohomology in (33) one finds
$0\to H^{0}(D_{p^{n}},Pr_{n}^{\diamond})\to
H^{0}(D_{p^{n}},Id_{n}^{\diamond})\to H^{0}(D_{p^{n}},A_{L_{n}})\to 0\;$
and, moding out by $(\mathbb{Q}^{\times})^{\diamond}$,
$H^{0}(D_{p^{n}},Pr_{n}^{\diamond})/(\mathbb{Q}^{\times})^{\diamond}\hookrightarrow
H^{0}(D_{p^{n}},Id_{n}^{\diamond})/(\mathbb{Q}^{\times})^{\diamond}\twoheadrightarrow
H^{0}(D_{p^{n}},A_{L_{n}})\;;$ (34)
Since in the proof of Theorem 4.7 we found isomorphisms
$\begin{split}&H^{0}(D_{p^{n}},Pr_{n}^{\diamond})/(\mathbb{Q}^{\times})^{\diamond}\cong
H^{1}(D_{p^{n}},U_{n})\cong\mathbb{Z}/p^{n-n_{0}-h_{n}}\mathbb{Z},\\\ \\\
&H^{0}(D_{p^{n}},Id_{n}^{\diamond})/(\mathbb{Q}^{\times})^{\diamond}\cong\mathbb{Z}/p^{n-n_{0}}\mathbb{Z}\;,\end{split}$
the exact sequence (34) becomes that of our statement. ∎
###### Corollary 4.10.
$R_{n}$ is a cyclic group of order $p^{n-n_{0}-h_{n}}$ and $R_{\infty}$ is
procyclic.
###### Proof.
Combine Proposition 4.4 with Proposition 4.9. ∎
We stress on the fact that the preceding result gives also a way to compute
$R_{n}$ directly (_i. e._ without Theorem 3.4).
Applying the Snake Lemma to multiplication-by-$(\gamma_{n}-1)$ (where
$\gamma_{n}$ is a topological generator of $\Gamma_{n}$) to the sequence in
(29) gives the fundamental exact sequence
$\begin{split}U_{\infty}^{\Gamma_{n}}\longrightarrow
R_{\infty}\longrightarrow(P_{\infty})_{\Gamma_{n}}\longrightarrow(U_{\infty})_{\Gamma_{n}}\longrightarrow
R_{\infty}\longrightarrow 0\;.\end{split}$ (35)
where $(R_{\infty})^{\Gamma_{n}}=R_{\infty}=(R_{\infty})_{\Gamma_{n}}$ since
$R_{\infty}$ has trivial $\Gamma$ action by Corollary 4.5. The next
proposition shows that actually $(U_{\infty})^{\Gamma_{n}}=0$; it crucially
depends on a result of Jean-Robert Belliard (see [Be]).
###### Proposition 4.11.
$P_{\infty}$ and $U_{\infty}$ are free $\Lambda$-modules of rank $1$.
###### Proof.
By [Gre], Proposition 1, we know that the projective limit
$U_{\infty}^{\prime}$ of the $p$-units along the anticyclotomic extension is
$\Lambda$-free of rank $1$. Now, Proposition 1.3 of [Be] gives a sufficient
condition for a projective limit to be free. Namely, suppose that a projective
system of $\mathbb{Z}_{p}[G_{n}]$-modules $(Z_{n})_{n\in\mathbb{N}}$, equipped
with norm maps $N_{m,n}:Z_{m}\rightarrow Z_{n}$ and extension maps
$i_{n,m}:Z_{n}\rightarrow Z_{m}$ (both for $m\geqslant n\geqslant 0$)
verifying the obvious relations, satisfies the following conditions:
1. 1.
There exists another projective system $W_{n}\supseteq Z_{n}$ with norm and
injection maps inducing the above maps on $Z_{n}$ by restriction such that
$W_{\infty}=\varprojlim W_{n}$ is $\Lambda$-free;
2. 2.
Extension maps $i_{n,m}:W_{n}\hookrightarrow W_{m}^{G_{m,n}}$ are injective
for $m\geqslant n\geqslant 0$;
3. 3.
$Z_{m}^{G_{m,n}}=i_{n,m}(Z_{n})$ (at least for $m\geqslant n\gg 0$) ,
then $Z_{\infty}=\varprojlim Z_{n}$ is also $\Lambda$-free.
First of all we apply this result to $U_{\infty}\subseteq
U_{\infty}^{\prime}$, finding that it is $\Lambda$-free, and of $\Lambda$-rank
equal to $1$ thanks to the exact sequence
$0\longrightarrow U_{\infty}\longrightarrow
U_{\infty}^{\prime}\stackrel{{\scriptstyle\prod
v_{\mathfrak{p}}}}{{\longrightarrow}}\mathbb{Z}_{p}[S_{\infty}]\longrightarrow
0$
where $S_{\infty}$ is the (finite) set of $p$-places in $L_{\infty}$ (the
right arrow is the product of all $\mathfrak{p}$-valuations for
$\mathfrak{p}\mid p$): in particular, $U_{\infty}^{\Gamma_{n}}=0$. Then we
apply the proposition again with $Z_{n}=P_{n}$ and $W_{n}=U_{n}$: in fact,
only the third condition needs to be checked and that comes from the Snake
Lemma applied to the diagram
$\begin{CD}0@>{}>{}>P_{n}@>{}>{}>U_{n}@>{}>{}>R_{n}@>{}>{}>0\\\
@V{i_{n,m}}V{}V@V{i_{n,m}}V{}V@V{i_{n,m}}V{}V\\\
0@>{}>{}>P_{n}^{G_{n,m}}@>{}>{}>U_{n}^{G_{n,m}}@>{}>{}>R_{m}^{G_{n,m}}\end{CD}$
noting that the vertical arrow in the middle is an isomorphism and the right-
hand vertical arrow is injective by Corollary 4.5. Thus we get that
$P_{\infty}$ is $\Lambda$-free: its $\Lambda$-rank is equal to the
$\Lambda$-rank of $U_{\infty}$ by Lemme 1.1 of [Be] together with (35), since
we have already proved that $U_{\infty}^{\Gamma_{n}}=0$. ∎
This Proposition already shows that either $R_{\infty}=0$ or $R_{\infty}$ is
free of rank $1$ over $\mathbb{Z}_{p}$: indeed, (29) shows that $R_{\infty}$
injects in $(P_{\infty})_{\Gamma_{n}}$ for all $n$, as
$(R_{\infty})_{\Gamma_{n}}=R_{\infty}$. On the other hand, $P_{\infty}$ being
free, the $\mathbb{Z}_{p}$-rank of $(P_{\infty})_{\Gamma}$ coincides with the
$\Lambda$-rank of $P_{\infty}$ which is $1$ by the above Proposition. As
$\mathbb{Z}_{p}$ does not admit any finite non-trivial submodules, the only
possibilities for $R_{\infty}$ are $0$ or $\mathbb{Z}_{p}$.
Otherwise we can argue as follows: by Corollary 4.5 we know that
$R_{n}\cong\mathbb{Z}/p^{(1-\lambda_{\mathfrak{P}})n-c}\mathbb{Z}$ for some
constant $c$. If $\lambda_{\mathfrak{P}}=0$ then
$R_{\infty}\cong\mathbb{Z}_{p}$. If $\lambda_{\mathfrak{P}}=1$ then the
$R_{n}$’s have bounded orders: since transition maps are induced by norms (as
$R_{n}\hookrightarrow R_{n+1}$ by Corollary 4.5, we need not to distinguish
between _algebraic_ and _arithmetic_ norm) and Proposition 4.4 shows that
$G_{n}$ acts trivially on $R_{n}$, $R_{\infty}$ is the projective limit of
cyclic groups of bounded order with respect to multiplication by $p$, so it is
$0$. We have thus proved
###### Corollary 4.12.
With notations as in Theorem 4.7, if $\lambda_{\mathfrak{P}}=1$ then
$R_{\infty}=0$ and if $\lambda_{\mathfrak{P}}=0$ then $R_{\infty}$ is free of
rank $1$ over $\mathbb{Z}_{p}$.
Remark. In the proof of Theorem 5.9 below we will show that
$\lambda_{\mathfrak{P}}=1$ if $p$ splits in $F$ and $\lambda_{\mathfrak{P}}=0$
if $p$ does not.
## 5 Structure of $X_{K}$
We now want to connect the study of $X_{L}$ and $X_{K}$: we recall that
$X_{L}:=\varprojlim A_{L_{n}}\quad\textrm{and}\quad X_{K}:=\varprojlim
A_{K_{n}},$
projective limits being taken with respect to norms. If $L_{\infty}/L$ were
the cyclotomic $\mathbb{Z}_{p}$-extension of $L$, then $X_{L}$ would be known
to be $\mathbb{Z}_{p}$-finitely generated by a celebrated result of B. Ferrero
and L. Washington (see [FW]), but for the anticyclotomic extension this is no
more the case (see [Gi] and [Ja2]). We are interested in giving conditions for
$X_{K}$ to be finitely generated as $\mathbb{Z}_{p}$-module. Our strategy is
to study the quotient $X_{L}/X_{K}X_{K^{\prime}}$. The following exact
sequence is then useful
$0\to\mathrm{Ker}(\iota_{n})\longrightarrow A_{K_{n}}\oplus
A_{K^{\prime}_{n}}\stackrel{{\scriptstyle\iota_{n}}}{{\longrightarrow}}A_{L_{n}}\longrightarrow
A_{L_{n}}/A_{K_{n}}A_{K^{\prime}_{n}}\to 0$ (36)
where
$\iota_{n}\left(([I],\,[I^{\prime}])\right)=[II^{\prime}\mathcal{O}_{L_{n}}]$
if $I\subset\mathcal{O}_{K_{n}}$ and
$I^{\prime}\subseteq\mathcal{O}_{K^{\prime}_{n}}$ are ideals. Passing to
projective limit we get
$0\to\mathrm{Ker}(\iota_{\infty})\longrightarrow X_{K}\oplus
X_{K^{\prime}}\stackrel{{\scriptstyle\iota_{\infty}}}{{\longrightarrow}}X_{L}\longrightarrow
X_{F}/X_{K}X_{K^{\prime}}\to 0$ (37)
and
$\mathrm{Ker}(\iota_{\infty})=\lim_{\longleftarrow}\mathrm{Ker}(\iota_{n})\;.$
We will describe $\mathrm{Ker}(\iota_{n})$ and
$A_{L_{n}}/A_{K_{n}}A_{K^{\prime}_{n}}$ in terms of cohomology groups. The
following diagram will be useful:
$\begin{CD}000\\\ @V{}V{}V@V{}V{}V@V{}V{}V\\\
0@>{}>{}>U_{n}@>{}>{}>(L_{n}^{\times})^{\diamond}@>{}>{}>Pr_{n}^{\diamond}@>{}>{}>0\\\
@V{}V{}V@V{}V{}V@V{}V{}V\\\
0@>{}>{}>\mathcal{U}_{n}^{\diamond}@>{}>{}>\mathcal{J}_{n}^{\diamond}@>{}>{}>Id_{n}^{\diamond}@>{}>{}>0\\\
@V{}V{}V@V{}V{}V@V{}V{}V\\\
0@>{}>{}>Q_{n}@>{}>{}>\mathcal{C}_{n}^{\diamond}@>{}>{}>A_{L_{n}}@>{}>{}>0\\\
@V{}V{}V@V{}V{}V@V{}V{}V\\\ 000\\\ \end{CD}$ (38)
Here $\mathcal{J}_{n}$ is the idèles group, $\mathcal{C}_{n}$ is the idèles
class group, $\mathcal{U}_{n}$ is the group of idèles units, $Id_{n}$ is the
group of ideals and $Pr_{n}$ is the group of principal ideals of $L_{n}$.
Remember that for an abelian group $B$ we set
$B^{\diamond}:=B\otimes\mathbb{Z}_{p}$.
###### Proposition 5.1.
We have
$\mathrm{Ker}(\iota_{n})\cong H^{0}(D_{p^{n}},A_{L_{n}})\cong
H^{1}(D_{p^{n}},Q_{n})$
and
$\widehat{H}^{-1}(D_{p^{n}},A_{L_{n}})\cong\widehat{H}^{0}(D_{p^{n}},Q_{n})\;.$
###### Proof.
Let $\varphi$ be the map
$\varphi:\mathrm{Ker}(\iota_{n})\longrightarrow H^{0}(D_{p^{n}},A_{L_{n}})$
defined by $\varphi([I],[I^{\prime}]))=\iota_{K_{n}}([I])$. Since
$\iota_{K_{n}}([I^{\prime}])=\iota_{K_{n}^{\prime}}([I])^{-1}$, both $\sigma$
and $\rho\sigma$ fix $\iota_{K_{n}}([I])$, so the map takes indeed value in
$H^{0}(D_{p^{n}},A_{L_{n}})$. We claim that $\varphi$ is an isomorphism. It is
clearly injective: to check surjectivity, just observe that
$A_{L_{n}}^{\langle\sigma\rangle}=\iota_{K_{n}}(A_{K_{n}})$ (and analogously
for $\sigma\rho$), so that, for every $[J]\in H^{0}(D_{p^{n}},A_{L_{n}})$, we
can write $[J]=\iota_{K_{n}}([I])=\iota_{K_{n}^{\prime}}([I^{\prime}])$ and
$[J]=\varphi([I],[I^{\prime}])$ and the claim is established.
Now consider the exact sequence
$0\to Q_{n}\to\mathcal{C}_{n}^{\diamond}\to A_{L_{n}}\to 0$
as in diagram (38). We take $D_{p^{n}}$-Tate cohomology, making constant use
of Proposition 4.3, and we get
$\begin{split}0\to\widehat{H}^{-1}(D_{p^{n}},A_{L_{n}})&\to\widehat{H}^{0}(D_{p^{n}},Q_{n})\to\widehat{H}^{0}(D_{p^{n}},\mathcal{C}_{n}^{\diamond})\\\
&\to\widehat{H}^{0}(D_{p^{n}},A_{L_{n}})\to\widehat{H}^{1}(D_{p^{n}},Q_{n})\to
0\;,\end{split}$ (39)
since by class field theory
$\widehat{H}^{i}(G_{n},\mathcal{C}_{n}^{\diamond})=0$ if $i\equiv 1\pmod{2}$
and
$|\widehat{H}^{i}(D_{p^{n}},\mathcal{C}_{n}^{\diamond})|=|\widehat{H}^{i}(G_{n},\mathcal{C}_{n}^{\diamond})^{\Delta}|\;.$
However, also the middle term in the exact sequence (39) is trivial, since
class field theory gives a $\Delta$-modules isomorphism
$\widehat{H}^{0}(G_{n},\mathcal{C}_{n}^{\diamond})\stackrel{{\scriptstyle\cong}}{{\longrightarrow}}\mathrm{Gal}(L_{n}/L)$
and the latter has no $\Delta$-invariants (since
$\mathrm{Gal}(L_{n}/\mathbb{Q})$ is dihedral). Note that
$\widehat{H}^{0}(D_{p^{n}},A_{L_{n}})=H^{0}(D_{p^{n}},A_{L_{n}})$
since $A_{L_{n}}[N_{D_{p^{n}}}]=A_{L_{n}}$ because $A_{\mathbb{Q}}=0$. ∎
###### Lemma 5.2.
Let $g$ denote the number of primes above $p$ in $L$ (hence $g\in\\{1,\,2\\}$)
and let $n_{0}$ be as in Definition 4.8. Then
$\widehat{H}^{0}(D_{p^{n}},\mathcal{U}_{n}^{\diamond})$ is a cyclic group of
order $p^{(g-1)(n-n_{0})}$ and
$\widehat{H}^{1}(D_{p^{n}},\mathcal{U}_{n}^{\diamond})$ is a cyclic group of
order $p^{n-n_{0}}$.
###### Proof.
We start by studying $G_{n}$-cohomology. Local class field theory gives
$\Delta$-equivariant identifications
$\widehat{H}^{0}(G_{n},\mathcal{U}_{n}^{\diamond})\cong
I(\mathfrak{p_{1}})\times I(\mathfrak{p_{2}})\quad\text{resp.
}\widehat{H}^{0}(G_{n},\mathcal{U}_{n}^{\diamond})\cong I(\mathfrak{p_{1}})$
where $I(\mathfrak{p}_{i})$ is the inertia subgroup inside $G_{n}$ of the
prime $\mathfrak{p}_{i}$ of $L$ above $p$, accordingly as $p$ splits or not in
$L$ (here and in the rest of the proof, we let $i=1,2$ if $p$ splits, while
$i=1$ if $p$ does not split). Analogously,
$\begin{split}H^{1}(G_{n},\mathcal{U}_{n}^{\diamond})&\cong\widehat{H}^{1}(G_{n},\mathcal{O}_{n,\mathfrak{P}_{1}}^{\times})\times\widehat{H}^{1}(G_{n},\mathcal{O}_{n,\mathfrak{P}_{2}}^{\times})\\\
\text{resp.}\quad
H^{1}(G_{n},\mathcal{U}_{n}^{\diamond})&\cong\widehat{H}^{1}(G_{n},\mathcal{O}_{n,\mathfrak{P}_{1}}^{\times})\end{split}$
where $\mathcal{O}_{n,\mathfrak{P}_{i}}^{\times}$ are the local units at the
prime $\mathfrak{P}_{i}$ of $L_{n}$ above $p$, accordingly again as $p$ splits
or not in $L$.
Concerning $\widehat{H}^{0}$, it is clear how $\Delta$ acts on the cohomology
group, since if there is only one inertia group it acts on it as $-1$; and if
there are two of them it acts on $-1$ on each subgroup, and permutes them.
Since the inertia subgroups are cyclic of order $p^{n-n_{0}}$, we get our
claim, using that
$\widehat{H}^{0}(D_{p^{n}},\mathcal{U}_{n}^{\diamond})=\widehat{H}^{0}(G_{n},{\mathcal{U}_{n}^{\diamond}})^{\Delta}$.
Passing now to $\widehat{H}^{1}$, we observe that
$\widehat{H}^{-1}(G_{n},\mathcal{O}_{n,\mathfrak{P}_{i}}^{\times})$ is
generated by $\rho\pi/\pi$ for some chosen uniformizer $\pi$ of a fixed
completion $L_{n,\mathfrak{P}_{i}}$ of $L_{n}$ at $\mathfrak{P}_{i}$. The
action of $\Delta$ is
$\delta(\rho\pi/\pi)=(\rho^{-1}\delta\pi)/\delta\pi\equiv\rho^{-1}\pi/\pi\pmod{I_{G_{n}}\mathcal{O}_{n,\mathfrak{P}_{i}}^{\times})}$:
starting from
$\rho^{2}\pi/\rho\pi\equiv\rho\pi/\pi\pmod{I_{G_{n}}\mathcal{O}_{n,\mathfrak{P}_{i}}^{\times})}$
one finds
$\rho^{2}\pi/\pi\equiv(\rho\pi/\pi)^{2}\pmod{I_{G_{n}}\mathcal{O}_{n,\mathfrak{P}_{i}}^{\times})}$
and, inductively,
$\rho^{-1}\pi/\pi\equiv(\rho\pi/\pi)^{-1}\pmod{I_{G_{n}}\mathcal{O}_{n,\mathfrak{P}_{i}}^{\times})}$
so that $\Delta$ acts as $-1$ on
$\widehat{H}^{-1}(G_{n},\mathcal{O}_{n,\mathfrak{P}_{i}}^{\times})$. Again,
the fact that this group is cyclic of order $p^{n-n_{0}}$ by local class field
theory, together with
$\widehat{H}^{1}(D_{p^{n}},\mathcal{U}_{n}^{\diamond})=\widehat{H}^{1}(G_{n},{\mathcal{U}_{n}^{\diamond}})^{-}$
shows our result. ∎
###### Proposition 5.3.
There is an exact sequence
$0\to\widehat{H}^{0}(D_{p^{n}},\mathcal{U}_{n}^{\diamond})\to\widehat{H}^{0}(D_{p^{n}},\,Q_{n})\stackrel{{\scriptstyle
0}}{{\rightarrow}}R_{n}\stackrel{{\scriptstyle\alpha}}{{\to}}\widehat{H}^{1}(D_{p^{n}},\mathcal{U}_{n}^{\diamond})\to\mathrm{Ker}(\iota_{n})\to
0\;.$
Hence we get
$\widehat{H}^{0}(D_{p^{n}},\mathcal{U}_{n}^{\diamond})\cong\widehat{H}^{0}(D_{p^{n}},\,Q_{n})$
and the short exact sequence
$0\to
R_{n}\to\widehat{H}^{1}(D_{p^{n}},\,\mathcal{U}_{n}^{\diamond})\to\mathrm{Ker}(\iota_{n})\to
0\;.$
In particular,
$\mathrm{Ker}(\iota_{n})\cong\mathbb{Z}/p^{h_{n}}\mathbb{Z}$
where $h_{n}$ is as in Definition 4.8.
###### Proof.
The exact sequence is (a short piece of) the long exact sequence of
$D_{p^{n}}$-Tate cohomology of the righthand column of diagram (38): here
$R_{n}$ and $\mathrm{Ker}(\iota_{n})$ appear thanks to Proposition 4.4 and
Proposition 5.1. Now note that
$\widehat{H}^{1}(D_{p^{n}},\mathcal{U}_{n}^{\diamond})\cong\mathbb{Z}/p^{n-n_{0}}\mathbb{Z}$
by Lemma 5.2. Moreover,
$\mathrm{Ker}(\iota_{n})=H^{0}(D_{p^{n}},A_{L_{n}})$
by Proposition 5.1 and
$R_{n}=H^{1}(D_{p^{n}},U_{n})$
by Proposition 4.4. Hence, using Proposition 4.9, we deduce that $\alpha$ is
necessarily injective, so the third map is necessarily $0$. ∎
###### Lemma 5.4.
The following inclusions hold
$I_{G_{n}}A_{L_{n}}\subseteq\iota_{K_{n}}(A_{K_{n}})\iota_{K^{\prime}_{n}}(A_{K_{n}^{\prime}})\subseteq
A_{L_{n}}[N_{G_{n}}].$
###### Proof.
For the inclusion
$I_{G_{n}}A_{L_{n}}\subseteq\iota_{K_{n}}(A_{K_{n}})\iota_{K^{\prime}_{n}}(A_{K_{n}^{\prime}})$
see for example Lemma 3.3 of [Le], using that
$(1+\sigma)A_{L_{n}}=\iota_{K_{n}}(A_{K_{n}})$ and
$(1+\rho^{2}\sigma)A_{L_{n}}=\iota_{K^{\prime}_{n}}(A_{K_{n}^{\prime}})$. Then
note that the norm element $N_{D_{p^{n}}}\in\mathbb{Z}_{p}[D_{p^{n}}]$ is the
zero map on $A_{L_{n}}$ since $\mathbb{Z}$ is principal, hence
$N_{G_{n}}(\iota_{K_{n}}(A_{K_{n}}))=N_{G_{n}}((1+\sigma)A_{L_{n}})=N_{D_{p_{n}}}(A_{L_{n}})=0$
and the analogous result holds for $A_{K^{\prime}_{n}}$, thereby proving the
claimed inclusion. ∎
###### Lemma 5.5.
For every $n\geqslant 0$ there is an isomorphism
$A_{L_{n}}[N_{G_{n}}]/A_{K_{n}}A_{K^{\prime}_{n}}\cong
H^{1}(D_{p^{n}},A_{L_{n}})\;.$
###### Proof.
The proof goes exactly in the same way as in Proposition 4.4, except for the
fact that here we cannot replace $A_{L_{n}}[N_{G_{n}}]$ with $A_{L_{n}}$ (but
we can use Lemma 5.4 above). ∎
Collecting together these results we can give an algebraic proof of a version
of the formula in Theorem 3.4. We need to recall a well known result. If
$M_{1}/M_{0}$ is any finite Galois extension, we shall denote by
$Ram(M_{1}/M_{0})$ the product of the ramification indexes in $M_{1}/M_{0}$ of
the (finite) primes of $M_{0}$. Then we have the following formula, coming
from a computation with Herbrand quotients:
###### Fact 5.6 (Ambiguous Class Number Formula).
Let $M_{1}/M_{0}$ be a finite Galois extension of odd degree and set
$G=\mathrm{Gal}(M_{1}/M_{0})$. Then
$|Cl_{M_{1}}^{G}|=\frac{|Cl_{M_{0}}|Ram(M_{1}/M_{0})}{[M_{1}:\,M_{0}](E_{M_{0}}:\,E_{M_{0}}\cap
N_{M_{1}/M_{0}}(M_{1}^{\times}))}\;.$
###### Proof.
See [Gra], II 6.2.3. ∎
###### Proposition 5.7.
The following formula holds (compare with Theorem 3.4)
$h_{L_{n}}^{(p)}=\frac{(h_{K_{n}}^{(p)})^{2}h_{L}^{(p)}|R_{n}|}{p^{n}}\;.$
###### Proof.
From the exact sequence (36) we deduce that
$h_{L_{n}}=\frac{h_{K_{n}}^{2}|A_{L_{n}}/A_{K_{n}}A_{K^{\prime}_{n}}|}{|\mathrm{Ker}(\iota_{n})|}\;.$
Note that
$|A_{L_{n}}/A_{K_{n}}A_{K^{\prime}_{n}}|=|A_{L_{n}}[N_{G_{n}}]/A_{K_{n}}A_{K^{\prime}_{n}}||A_{L_{n}}/A_{L_{n}}[N_{G_{n}}]|=$
$=|H^{1}(D_{p^{n}},A_{L_{n}})||N_{G_{n}}(A_{L_{n}})|\;,$
using Lemma 5.5. Moreover, by the Ambiguous Class Number Formula,
$|N_{G_{n}}(A_{L_{n}})|=\frac{|A_{L_{n}}^{G_{n}}|}{|\widehat{H}^{0}(G_{n},A_{L_{n}})|}=\frac{h_{L}^{(p)}Ram(L_{n}/L)}{p^{n}|H^{1}(G_{n},A_{L_{n}})|}$
(we use the fact that
$|\widehat{H}^{0}(G_{n},A_{L_{n}})|=|H^{1}(G_{n},A_{L_{n}})|$). Now observe
that
$\frac{|H^{1}(D_{p^{n}},A_{L_{n}})|}{|H^{1}(G_{n},A_{L_{n}})|}=\frac{1}{|\widehat{H}^{-1}(D_{p^{n}},A_{L_{n}})|}=\frac{1}{|\widehat{H}^{0}(D_{p^{n}},Q_{n})|}$
by Proposition 5.1 and Proposition 4.3. Furthermore
$\frac{1}{|\widehat{H}^{0}(D_{p^{n}},Q_{n})|}=\frac{1}{|\widehat{H}^{0}(D_{p^{n}},\mathcal{U}_{n}^{\diamond})|}\;,$
by Proposition 5.3. Hence we get
$|A_{L_{n}}/A_{K_{n}}A_{K^{\prime}_{n}}|=\frac{h_{L}^{(p)}Ram(L_{n}/L)}{p^{n}|\widehat{H}^{0}(D_{p^{n}},\mathcal{U}_{n}^{\diamond})|}\;.$
Now $b_{n}=n-n_{0}$. Then
$Ram(L_{n}/L)=p^{gb_{n}}$
(where $g$ is the same as in Lemma 5.2) and
$|\widehat{H}^{0}(D_{p^{n}},\mathcal{U}_{n}^{\diamond})|=p^{(g-1)b_{n}}\;.$
by Lemma 5.2. Therefore,
$|A_{L_{n}}/A_{K_{n}}A_{K^{\prime}_{n}}|=\frac{h_{L}^{(p)}p^{gb_{n}}}{p^{(g-1)b_{n}+n}}=\frac{h_{L}^{(p)}}{p^{n_{0}}}$
(40)
and
$h_{L_{n}}^{(p)}=\frac{(h_{K_{n}}^{(p)})^{2}h_{L}^{(p)}p^{b_{n}}}{p^{n}|\mathrm{Ker}(\iota_{n})|}\;.$
Using once more Proposition 5.3 we deduce that
$|R_{n}||\mathrm{Ker}(\iota_{n})|=p^{b_{n}}$
which gives the formula of the proposition. ∎
Remark. We want to stress here that our proof works as well in a more general
setting. Indeed, we assumed that $L_{n}/L$ is part of the anticyclotomic
extension because this is the context for our further application. But since
both Proposition 5.3, 5.1 and Lemma 5.2 continue to hold true, _mutatis
mutandis_ , for any dihedral extension, the above proposition can be proven in
the same way for any such a dihedral extension. Moreover, as pointed out for
example by Lemmermeyer in [Le], Theorem 2.2, the formula above is trivially
true for any odd prime number $\ell\neq p$: summarizing, our proof can be
generalized to show (algebraically) that for any dihedral extension
$F/\mathbb{Q}$ of degree $p^{n}$, the relation
$h_{F}=\frac{h_{k}^{2}h_{L}[\mathcal{O}_{L}^{\times}:\mathcal{O}_{k}^{\times}\mathcal{O}_{k^{\prime}}^{\times}\mathcal{O}_{L}^{\times}]}{p^{n}}$
holds, up to powers of $2$.
Now we make some remarks about the structure of $X_{K}$. First we need a
lemma:
###### Lemma 5.8.
Suppose that $p$ splits in $L$: then for any prime $\mathfrak{p}$ over $p$ in
$L$, the union $(L_{cyc}L_{\infty})_{\mathfrak{p}}$ of all completions at $p$
of the finite layers of $L_{cyc}L_{\infty}/L$ is a $\mathbb{Z}_{p}$-extension
over $(L_{cyc})_{\mathfrak{p}}$.
###### Proof.
Let $\mathfrak{p}_{1}$ and $\mathfrak{p}_{2}$ be the two primes of $F$ above
$p$. Let $H$ be the maximal subextension of $F_{cyc}F_{\infty}/F$ such that
$\mathfrak{p}_{1}$ is totally split in $H/F$. By global class field theory,
explicitely writing down the normic subgroups corresponding to $H$ and to
$F_{cyc}F_{\infty}$, one sees that $H/F$ must be finite. The lemma now follows
since the decomposition group of $\mathfrak{p}_{1}$ (and therefore also that
of $\mathfrak{p}_{2}$) in $\mathrm{Gal}(F_{cyc}F_{\infty}/F)$ is of
$\mathbb{Z}_{p}$-rank $2$. ∎
###### Theorem 5.9.
$\mathrm{Ker}(\iota_{\infty})$ is a $\mathbb{Z}_{p}$-module of rank $1$ if $p$
splits in $L$ and it is finite otherwise. Moreover,
$X_{L}/X_{K}X_{K^{\prime}}$ is finite and its order divides
$h_{L}^{(p)}/p^{n_{0}}$. In particular $X_{L}$ is finitely generated as
$\mathbb{Z}_{p}$-module if and only if $X_{K}$ is.
###### Proof.
Before starting the proof, observe that $\mathrm{Ker}(\iota_{\infty})$ is a
$\mathbb{Z}_{p}$-module of rank at most $1$ (use the fact that each
$\mathrm{Ker}(\iota_{n})$ is cyclic, see Proposition 5.3).
Suppose now that $p$ splits in $L$, say
$p\mathcal{O}_{L}=\mathfrak{p}_{1}\mathfrak{p}_{2}$. Let
$\mathfrak{P}_{i,1},\ldots,\mathfrak{P}_{i,s}$ be the prime ideals of
$L_{\infty}$ which lie above $\mathfrak{p}_{i}$ (clearly $s=p^{a}$ for some
$a\in\mathbb{N}$) for $i=1,\,2$. Let, as before, $n_{0}$ be the smallest
natural number such that $L_{\infty}/L_{n_{0}}$ is totally ramified. Set
$L^{\prime}=L_{cyc}L_{\infty}$ and note that $L^{\prime}/L_{\infty}$ is
unramified everywhere: indeed, $L^{\prime}/L_{\infty}$ is clearly unramified
at every prime which does not lie above $p$. On the other hand,
$L^{\prime}_{\mathfrak{P}_{i,j}}/L_{\infty\;\mathfrak{P}_{i,j}}$ must be
unramified because $\mathbb{Q}_{p}=L_{\mathfrak{p}_{i}}$ admits only two
independent $\mathbb{Z}_{p}$-extension, one being the unramified one. Let now
$M_{\infty}$ be the maximal unramified abelian pro-$p$-extension of
$L_{\infty}$ (hence $L^{\prime}\subseteq M_{\infty}$). For each $i=1,\,2$ and
$j=1,\,\ldots,\,s$, consider the Frobenius
$\mathrm{Frob}(\mathfrak{P}_{i,j},M_{\infty}/L_{\infty})$ of
$\mathfrak{P}_{i,j}$ in $M_{\infty}/L_{\infty}$ which is an element of
infinite order since its restriction
$\mathrm{Frob}(\mathfrak{P}_{i,j},L^{\prime}/L_{\infty})$ to $L^{\prime}$ is
of infinite order by Lemma 5.8. Furthermore
$\mathrm{Gal}(L_{\infty}/\mathbb{Q})$ acts by conjugation on the set
$\\{\mathrm{Frob}(\mathfrak{P}_{i,j},M_{\infty}/L_{\infty})\\}_{i=1,2;\;j=1,\ldots,s}$:
in other words if $\tau\in\mathrm{Gal}(L_{\infty}/\mathbb{Q})$ we have
$\tau\cdot\mathrm{Frob}(\mathfrak{P}_{i,j},M_{\infty}/L_{\infty})=\widetilde{\tau}\mathrm{Frob}(\mathfrak{P}_{i,j},M_{\infty}/L_{\infty})\widetilde{\tau}^{-1}=\mathrm{Frob}(\widetilde{\tau}(\mathfrak{P}_{i,j}))\;.$
where $\widetilde{\tau}$ is an extension of $\tau$ to $M_{\infty}$. On the
other hand we must have
$\widetilde{\tau}\mathrm{Frob}(\mathfrak{P}_{i,j},L^{\prime}/L_{\infty})\widetilde{\tau}^{-1}=\mathrm{Frob}(\mathfrak{P}_{i,j},L^{\prime}/L_{\infty})$
because $L^{\prime}/\mathbb{Q}$ is a Galois extension whose Galois group is
isomorphic to
$\mathrm{Gal}(L^{\prime}/L_{\infty})\times\mathrm{Gal}(L_{\infty}/\mathbb{Q})\;.$
In particular we deduce that
$\prod_{i=1}^{2}\prod_{j=1}^{s}\mathrm{Frob}(\mathfrak{P}_{i,j},L^{\prime}/L_{\infty})=\mathrm{Frob}(\mathfrak{P}_{1,1},L^{\prime}/L_{\infty})^{2s}.$
Hence this product is an element of infinite order in
$\mathrm{Gal}(L^{\prime}/L_{\infty})$ and the same holds for the products of
$\mathrm{Frob}(\mathfrak{P}_{i,j},M_{\infty}/L_{\infty})$. It corresponds by
class field theory to
$\lim_{\longleftarrow}[\mathfrak{P}_{n}]\in X_{L}$
where $\mathfrak{P}_{n}$ is the product of all primes above $p$ in $L_{n}$.
Now note that
$\mathrm{Ker}(\iota_{n})=H^{0}(D_{p^{n}},\,A_{L_{n}})=\langle[\mathfrak{P}_{n}]\rangle\cong\mathbb{Z}/p^{h_{n}}\mathbb{Z}.$
In fact, clearly
$H^{0}(D_{p^{n}},\,A_{L_{n}})\supseteq\langle[\mathfrak{P}_{n}]\rangle$
and both groups have order $p^{h_{n}}$ (use Proposition 4.9). Hence
$\mathrm{Ker}(\iota_{\infty})$ is infinite since it contains an element of
infinite order and it has $\mathbb{Z}_{p}$-rank $1$).
Now suppose that $p$ does not split in $L$ and let again
$M_{\infty}/L_{\infty}$ be the maximal pro-$p$ abelian extension of
$L_{\infty}$ everywhere unramified, so that we have an isomorphism
$\mathrm{Gal}(M_{\infty}/L_{\infty})\cong X_{L}$. Let $M_{0}/L_{\infty}$ be
the fixed field by $TX_{L}\subset X_{L}$, viewing $X_{L}$ as a
$\Lambda$-module: then $M_{0}$ is the maximal unramified extension of
$L_{\infty}$ which is pro-$p$ abelian over $L$ (see [Wa], chapter 13) and we
let $\mathscr{G}=\mathrm{Gal}(M_{0}/L)$. Since $\mathscr{G}$ is abelian, we
can speak of the inertia subgroup $\mathscr{I}\vartriangleleft\mathscr{G}$ of
$\mathfrak{p}$ (the unique prime in $L$ above $p$): then $M^{\mathscr{I}}/L$
is an abelian extension everywhere unramified, thus finite. This shows that
$p$ is finitely split and has finite inertia degree in $M_{0}/L$. Therefore
$M_{0}/L_{\infty}$ is finite, being unramified everywhere, and then its Galois
group (which is isomorphic to $X_{L}/TX_{L}$) is finite. The exact sequence
$0\rightarrow\mathrm{Ker}(\cdot T)\rightarrow
X_{L}\stackrel{{\scriptstyle\cdot T}}{{\longrightarrow}}X_{L}\rightarrow
X_{L}/TX_{L}\rightarrow 0$
shows that $\mathrm{Ker}(\cdot T)=\mathrm{Ker}(\gamma_{0}-1)=X_{L}^{\Gamma}$
is finite (recall that the isomorphism $\mathbb{Z}_{p}[[\Gamma]]\cong\Lambda$
is induced by $\gamma_{0}-1\mapsto T$, where $\gamma_{0}$ is a fixed
topological generator of $\Gamma$). Since $\mathfrak{P}_{n}$ is clearly fixed
by $\Gamma$, we see that their projective limit $Y$ is in $X_{L}^{\Gamma}$ and
is therefore finite. In particular, their order $p^{h_{n}}$ is bounded, and
since $\mathrm{Ker}(\iota_{n})$ has order $p^{h_{n}}$ by Proposition 5.3, we
immediately see that $\mathrm{Ker}(\iota_{\infty})$ is finite (actually
$Y=\mathrm{Ker}(\iota_{\infty})$).
The second assertion of the theorem is exactly (40) and then the last one
easily follows from (37). ∎
Remark. Suppose that $\mu_{L}=0$: therefore $X_{K}$ is a finitely generated
$\mathbb{Z}_{p}$-module and $\lambda_{X_{K}}$ is its rank. From (37) we deduce
that
$\lambda_{K}=2\lambda_{X_{K}}$
thanks to the remark after Theorem 4.7.
Examples. Suppose that the $p$-Hilbert class field of $L$ is cyclic and that
it is contained in the compositum of the $\mathbb{Z}_{p}$-extensions of $L$.
Then it must be in $L_{\infty}$ ($A_{L}^{\Delta}$ is trivial since
$A_{\mathbb{Q}}=0$). With this in mind we give the following examples:
* •
Take $L=\mathbb{Q}(\sqrt{-191})$. Then the $13$-Hilbert class field is cyclic
of order $13$ and is contained in the compositum of the
$\mathbb{Z}_{13}$-extensions of $L$ (see [Gra], Examples 2.6.3). Then
$n_{0}=1$ and we have $h^{(13)}_{L}=13^{n_{0}}=13$ and this gives
$X_{L}=X_{K}X_{K^{\prime}}$ by Theorem 5.9.
* •
Take $L=\mathbb{Q}(\sqrt{-383})$. Then the $17$-Hilbert class field is cyclic
of order $17$ but linearly disjoint from the compositum of the
$\mathbb{Z}_{17}$-extensions of $L$ (see [Gra], Examples 2.6.3). In particular
$n_{0}=0$ and $X_{L}/X_{K}X_{K^{\prime}}$ is cyclic of order dividing $17$ by
Theorem 5.9. Actually $X_{L}/X_{K}X_{K^{\prime}}$ is of order $17$: for,
$L_{\infty}/L$ is totally ramified and this implies that the arithmetic norms
$A_{L_{m}}\to A_{L_{n}}$ are surjective for every $m\geqslant n\geqslant 0$.
Then it is easy to see that the arithmetic norms
$A_{L_{m}}/A_{K_{m}}A_{K^{\prime}_{m}}\to
A_{L_{n}}/A_{K_{n}}A_{K^{\prime}_{n}}$ are surjective too, for every
$m\geqslant n\geqslant 0$. Since each $A_{L_{n}}/A_{K_{n}}A_{K^{\prime}_{n}}$
is of order $17$ by (40) we are done.
## Appendix
In this appendix we perform the computations needed in Proposition 3.3.
###### Lemma.
With notations introduced in section 3, $|\det(M)|=q|\det(A)|^{2}$, where
$M=\left(\begin{array}[]{c|c}A^{i}_{\hskip 7.0pt0\leqslant i\leqslant
r-1}&B^{i}_{\hskip 7.0pt0\leqslant i\leqslant r-1}\\\ \hline\cr C^{i}_{\hskip
7.0pt0\leqslant i\leqslant r-1}&D^{i}_{\hskip 7.0pt0\leqslant i\leqslant
r-1}\end{array}\right),\;A=\left((1/2)A^{0},A^{1},\ldots,A^{r-1}\right)$
are the matrices appearing in Proposition 3.3. Therefore they are defined as
follows:
$A^{i}=\left\\{\begin{array}[]{lr}\left(\begin{array}[]{c}2\gamma(\eta_{1})\\\
\vdots\\\ 2\gamma(\eta_{r})\end{array}\right)&\mathrm{for}\;i=0\\\ \\\
\left(\begin{array}[]{c}\tau_{i}(\eta_{1})\\\ \vdots\\\
\tau_{i}(\eta_{r})\end{array}\right)&\mathrm{for}\;1\leqslant i\leqslant
r-1\\\ \end{array}\right.\;,$
$B^{i}=\left\\{\begin{array}[]{lr}\left(\begin{array}[]{c}-\displaystyle{\sum_{j=1}^{r-1}\tau_{j}(\eta_{1})-\gamma(\eta_{1})}\\\
\vdots\\\
-\displaystyle{\sum_{j=1}^{r-1}\tau_{j}(\eta_{r})-\gamma(\eta_{r})}\end{array}\right)&\mathrm{for}\;i=0\\\
\\\ \left(\begin{array}[]{c}\tau_{i}(\eta_{1})\\\ \vdots\\\
\tau_{i}(\eta_{r})\end{array}\right)=A^{i}&\mathrm{for}\;1\leqslant i\leqslant
r-1\\\ \end{array}\right.\;,$
$C^{i}=\left\\{\begin{array}[]{lr}\left(\begin{array}[]{c}\tau_{i+1}(\eta_{1})\\\
\vdots\\\
\tau_{i+1}(\eta_{r})\end{array}\right)=A^{i+1}&\mathrm{for}\;0\leqslant
i\leqslant r-2\\\ \\\ B^{0}&\mathrm{for}\;i=r-1\\\ \end{array}\right.$
and
$D^{i}=\left\\{\begin{array}[]{lr}B^{0}&\mathrm{for}\;i=0\\\
A^{i-1}&\mathrm{for}\;1\leqslant i\leqslant r-1\\\ \end{array}\right.\;.$
###### Proof.
In what follows we will transform $M$ in another matrix $N$ (appearing below)
with trivial upper-right and lower-left blocks, and we do this by elementary
operations that don’t change the (absolute value of the) determinant.
Writing $M^{i}$ for the $i$-th column ($1\leqslant i\leqslant 2r$) of $M$,
let’s perform the substitution $M^{i}\mapsto M^{i}-M^{i-r}$ for $r+2\leqslant
i\leqslant q-1$ and $M^{r+1}\mapsto M^{r+1}+\sum_{i=1}^{r}M^{i}$. $M$ then
becomes
$M^{\prime}=\left(\begin{array}[]{c|c}A^{i}_{\hskip 7.0pt0\leqslant i\leqslant
r-1}&E^{i}_{\hskip 7.0pt0\leqslant i\leqslant r-1}\\\ \hline\cr C^{i}_{\hskip
7.0pt0\leqslant i\leqslant r-1}&F^{i}_{\hskip 7.0pt0\leqslant i\leqslant
r-1}\end{array}\right)$
where $A^{i}$ and $C^{i}$ are as above, while $E^{0}=A^{0}/2$,
$E^{i}=\underline{0}$ for $1\leqslant i\leqslant r-1$, $F^{0}=D^{0}+\sum
C^{i}$ and $F^{i}=D^{i}-C^{i}$ for $i>0$, _i. e._
$F^{i}=\left\\{\begin{array}[]{lr}B^{0}+\displaystyle{\sum_{j=0}^{r-1}C^{j}=2B^{0}+\sum_{j=1}^{r-1}A^{j}}&\mathrm{for}\;i=0\\\
A^{i-1}-A^{i+1}&\mathrm{for}\;1\leqslant i\leqslant r-2\\\
A^{r-2}-B^{0}&\mathrm{for}\;i=r-1\\\ \end{array}\right.\;.$
Before going further, let’s kill the first column in the upper right block: it
is enough to subtract from this column (it is the $r+1$-th) a half of the
first one, namely $M^{\prime r+1}\mapsto M^{\prime r+1}-(1/2)M^{\prime 1}$,
thus finding
$M^{\prime\prime}=\left(\begin{array}[]{c|c}A^{i}_{\hskip 7.0pt0\leqslant
i\leqslant r-1}&0\\\ \hline\cr C^{i}_{\hskip 7.0pt0\leqslant i\leqslant
r-1}&G^{i}_{\hskip 7.0pt0\leqslant i\leqslant r-1}\end{array}\right)$
where
$G^{i}=\left\\{\begin{array}[]{lr}B^{0}-(1/2)A^{0}-(1/2)A^{1}&\mathrm{for}\;i=0\\\
A^{i-1}-A^{i+1}&\mathrm{for}\;1\leqslant i\leqslant r-2\\\
A^{r-2}-B^{0}&\mathrm{for}\;i=r-1\\\ \end{array}\right.\;.$
We can now use all the $G^{i}$’s freely without changing the other blocks. In
particular, we will in the sequel operate in the submatrix formed by the
$G^{i}$’s. Observe, first of all, that
$\sum_{i=1}^{r-2}G^{i}=A^{0}+A^{1}-A^{r-2}-A^{r-1}.\>$
Using this, let’s substitute $G^{0}\mapsto G^{0}+(1/2)\sum G^{i}$, finding
$B^{0}-(1/2)A^{r-2}-(1/2)A^{r-1}:=X$ in the first column. Another step is now
to change this in $X\mapsto X+(1/2)G^{r-1}=(1/2)B^{0}-(1/2)A^{r-1}:$ $M$ has
now been reduced to
$M^{\prime\prime\prime}=\left(\begin{array}[]{c|c}A^{i}_{\hskip
7.0pt0\leqslant i\leqslant r-1}&0\\\ \hline\cr C^{i}_{\hskip 7.0pt0\leqslant
i\leqslant r-1}&H^{i}_{\hskip 7.0pt0\leqslant i\leqslant
r-1}\end{array}\right)$
where
$H^{i}=\left\\{\begin{array}[]{lr}(1/2)B^{0}-(1/2)A^{r-1}&\mathrm{for}\;i=0\\\
A^{i-1}-A^{i+1}&\mathrm{for}\;1\leqslant i\leqslant r-2\\\
A^{r-2}-B^{0}&\mathrm{for}\;i=r-1\\\ \end{array}\right.\;.$
Now we should transform $H^{r-1}\mapsto
H^{r-1}+2H^{0}=A^{r-2}-A^{r-1}:=H^{\prime r-1}$. Keeping on setting
$H^{i}\mapsto H^{i}-H^{\prime i+1}$ for $1\leqslant i\leqslant r-2$ we
inductively build a matrix $N$ whose determinant satisfies
$\det(N)=(1/2)\det(M)$, namely
$N=\left(\begin{array}[]{c|c}A^{i}_{\hskip 7.0pt0\leqslant i\leqslant
r-1}&0\\\ \hline\cr C^{i}_{\hskip 7.0pt0\leqslant i\leqslant r-1}&H^{\prime
i}_{\hskip 7.0pt0\leqslant i\leqslant r-1}\end{array}\right)$
where
$H^{\prime i}=\left\\{\begin{array}[]{lr}B^{0}-A^{r-1}&\mathrm{for}\;i=0\\\
A^{i-1}-A^{i}&\mathrm{for}\;1\leqslant i\leqslant r-1\\\
\end{array}\right.\;.$
Now we can finally perform our last transformations: the idea is to reduce the
right bottom block of $N$ to a matrix having $qA^{1}$ as second column. First
of all, we substitute $H^{\prime 1}\mapsto H^{\prime 1}+2H^{\prime 0}:=T_{1}$:
clearly every other column is unchanged, while this second column becomes
$A^{0}-A^{1}+2B^{0}-2A^{r-1}=\left(\begin{array}[]{c}-3\tau_{1}(\eta_{1})-\displaystyle{\sum_{i=2}^{r-2}2\tau_{i}(\eta_{1})-4\tau_{r-1}(\eta_{1})}\\\
\vdots\\\
-3\tau_{1}(\eta_{r})-\displaystyle{\sum_{i=2}^{r-2}2\tau_{i}(\eta_{r})-4\tau_{r-1}(\eta_{r})}\end{array}\right).\\\
$
Now we can repeatedly subtract to this column suitable multiples of the
$H^{\prime i}$’s in order to be left only with $-qA^{1}$: in fact, we define
inductively the matrix (as before, we perform these substitution only in the
submatrix formed by the $H^{\prime i}$’s)
$\Pi_{j}:=\left(H^{\prime 0},\underbrace{T_{j-1}-2j\cdot H^{\prime
r+1-j}}_{T_{j}},H^{\prime 2},\ldots,H^{\prime r-1}\right)\;,\;2\leqslant
j\leqslant r-1\;.$
The definition of the $H^{\prime i}$’s implies that
$T_{j}=T_{j-1}-2jA^{r-j}+2jA^{r+1-j}$ so that $T_{j}$ verifies
$T_{j}=\left(\begin{array}[]{c}-3\tau_{1}(\eta_{1})-\displaystyle{\sum_{i=2}^{r-1-j}2\tau_{i}(\eta_{1})-2(j+1)\tau_{r-j}(\eta_{1})}\\\
\vdots\\\
-3\tau_{1}(\eta_{r})-\displaystyle{\sum_{i=2}^{r-1-j}2\tau_{i}(\eta_{r})-2(j+1)\tau_{r-j}(\eta_{r})}\end{array}\right)\\\
\;,2\leqslant j\leqslant r-3$
and, for the remaining cases (as degenerate versions of the same formula),
$T_{r-2}=\left(\begin{array}[]{c}-3\tau_{1}(\eta_{1})-(q-3)\tau_{2}(\eta_{1})\\\
\vdots\\\
-3\tau_{1}(\eta_{r})-(q-3)\tau_{2}(\eta_{r})\end{array}\right),\;T_{r-1}=\left(\begin{array}[]{c}-q\tau_{1}(\eta_{1})\\\
\vdots\\\ -q\tau_{1}(\eta_{r})\end{array}\right)=-qA^{1}\;.$
Recalling now the definition of $N$ and that the performed transformations do
not change the left hand blocks of it, we find
$\begin{split}|\det(M)|=&\frac{1}{2}|\det(N)|=\\\
&=\frac{q}{2}\Big{|}\det\left(\begin{array}[]{c|c}A^{i}_{\hskip
7.0pt0\leqslant i\leqslant r-1}&0\\\ \hline\cr C^{i}_{\hskip 7.0pt0\leqslant
i\leqslant r-1}&\Pi^{\prime}_{r-1}\end{array}\right)\Big{|}\;,\end{split}$
(41)
where $\Pi^{\prime}_{r-1}$ is as $\Pi_{r-1}$ but with the second column
divided by $-q$. Looking now at the definition of $H^{\prime i}$’s shows that
we can still transform
$\Pi^{\prime}_{r-1}\mapsto\left(B^{0},A^{1},A^{2},\ldots,A^{r}\right)\;,$
since $A^{2}=A^{1}-H^{\prime 2}$, $A^{3}=A^{2}-H^{\prime 3}$ and so on. For
exactly the same reason, the first column $B^{0}$ may safely be substituted by
$(1/2)A^{0}$ so that finally
$\Pi^{\prime}_{r-1}\mapsto\left((1/2)A^{0},A^{1},\ldots,A^{r-1}\right)$
and (41) shows that
$|\det(M)|=\frac{q}{4}\big{|}\det\left(\begin{array}[]{c|c}A^{i}_{\hskip
7.0pt0\leqslant i\leqslant r-1}&0\\\ \hline\cr C^{i}_{\hskip 7.0pt0\leqslant
i\leqslant r-1}&A^{i}_{\hskip 7.0pt0\leqslant i\leqslant
r-1}\end{array}\right)\Big{|}\;.$
At last, one can use the bottom right block to kill the bottom left one
without changing the upper left block, simply by the definition of the
$C^{i}$’s: thus
$|\det(M)|=\frac{q}{4}\Big{|}\det\left(\begin{array}[]{c|c}A^{i}_{\hskip
7.0pt0\leqslant i\leqslant r-1}&0\\\ \hline\cr 0&A^{i}_{\hskip 7.0pt0\leqslant
i\leqslant r-1}\end{array}\right)\Big{|}=q|\det(A)|^{2}\;,$
as we wanted. ∎
## References
* [Be] J.-R. Belliard, _Sous-modules d’unités en théorie d’Iwasawa_ , Publ. Math. UFR Sci. Tech. Besançon, Univ. Franche-Comté, (2002).
* [FW] B. Ferrero and L. C. Washington, _The Iwasawa invariant $\mu_{p}$ vanishes for abelian number fields_, Ann. of Math. CIX (1979), 377-395.
* [Gi] R. Gillard, _Remarques sur certaines extensions prodiédrales de corps de nombres_ , C. R. Acad. Sci. Paris Sér. A-B 282 n. 1 (1976), A13-A15.
* [Gra] G. Gras, _Class field theory_ , Springer Monographs in Mathematics, Springer-Verlag, Berlin, 2005.
* [Gre] C. Greither, _Sur les normes universelles dans les $\mathbb{Z}_{p}$-extensions_, J. Théor. Nombres Bordeaux. 6 (1994), 205-220.
* [HK] F. Halter-Koch, _Einheiten und Divisorenklassen in Galois’schen algebraischen Zahlkörpern mit Diedergruppe der Ordnung $2\ell$ für eine ungerade Primzahl $\ell$_, Acta Arith. XXXIII (1977), 355-364.
* [He] H. A. Heilbronn, _Zeta-functions and $L$-functions_ in _Algebraic Number Theory, Proceedings of an istructional conference organized by the London Mathematical Society_ , edited by J.W.S. Cassels and A. Frölich, Academic Press, 1967.
* [Iw] K. Iwasawa, _On $\mathbb{Z}_{\ell}$-extensions of Number Fields_, Ann. of Math. LXLVIII, (1973), 246-326.
* [Ja1] J.-F. Jaulent, _Unités et classes dans les extensions métabeliénnes de degré $n\ell^{s}$ sur un corps de nombres algébriques_, Ann. Inst. Fourier (Grenoble) 31 n. 1 (1981), 39-62.
* [Ja2] J.-F. Jaulent, _Sur la théorie des genres dans les tours métabeliénnes_ , Séminaire de Theorie de Nombres 1981/82, Univ. Bordeaux 1, exposé n. 24 (1982), 18 pp.
* [Le] F. Lemmermeyer, _Class groups of dihedral extensions_ , Math. Nachr. CCLXXVIII (2005), 679-691.
* [Na] W. Narkiewicz, _Elementary and analytic Theory of Algebraic Numbers_ , Springer Monographs in Mathematics, Springer-Verlag, Berlin, 2004.
* [NSW] J. Neukirch, A. Schmidt and K. Wingberg, _Cohomology of Number Fields_ , Grundlehren der matematischen Wissenschaften 323, Springer-Verlag, Berlin, 2000.
* [Se1] J.-P. Serre, _Représentations linéaires des groupes finis_ , Collection Méthodes, Hermann, Paris, 1967.
* [Se2] J.-P. Serre, _Corps Locaux_ , Hermann, Paris, 1967.
* [Ta] J. T. Tate, _Global Class Field Theory_ in _Algebraic Number Theory, Proceedings of an istructional conference organized by the London Mathematical Society_ , edited by J.W.S. Cassels and A. Frölich, Academic Press, 1967.
* [VV] O. Venjakob and D. Vogel, _A non-commutative Weierstrass preparation theorem and applications to Iwasawa theory_ , J. Reine Angew. Math. 559 (2003), 153-191.
* [Wa] L. C. Washington, _Introduction to Cyclotomic Fields_ , GTM 83, Springer-Verlag, Berlin, 1997.
* [We] C. A. Weibel, _An introduction to homological algebra_ , Cambridge studies in advanced mathematics 38, Cambridge University Press, Cambridge, 1997.
Luca Caputo
Dipartimento di Matematica
Università di Pisa
Largo Bruno Pontecorvo, 5
56127 - Pisa - ITALY
caputo@mail.dm.unipi.it
Filippo A. E. Nuccio
Dipartimento di Matematica
Università “La Sapienza”
Piazzale Aldo Moro, 5
00185 - Rome - ITALY
nuccio@mat.uniroma1.it
Corresponding author
| arxiv-papers | 2008-07-07T22:07:13 | 2024-09-04T02:48:56.606416 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/",
"authors": "Luca Caputo, Filippo A. E. Nuccio",
"submitter": "Filippo Alberto Edoardo Nuccio Mortarino Majno di Capriglio",
"url": "https://arxiv.org/abs/0807.1135"
} |
0807.1183 | # Multiple D2-Brane Action from M2-Branes
Tianjun Li Institute of Theoretical Physics, Chinese Academy of Sciences,
Beijing 100080, P. R. China George P. and Cynthia W. Mitchell Institute for
Fundamental Physics, Texas A$\&$M University, College Station, TX 77843, USA
Yan Liu Institute of Theoretical Physics, Chinese Academy of Sciences,
Beijing 100080, P. R. China Dan Xie George P. and Cynthia W. Mitchell
Institute for Fundamental Physics, Texas A$\&$M University, College Station,
TX 77843, USA
###### Abstract
We study the detail derivation of the multiple D2-brane effective action from
multiple M2-branes in the Bagger-Lambert-Gustavsson (BLG) theory and the
Aharony-Bergman-Jafferis-Maldacena (ABJM) theory by employing the novel Higgs
mechanism. We show explicitly that the high-order $F^{3}$ and $F^{4}$ terms
are commutator terms, and conjecture that all the high-order terms are
commutator terms. Because the commutator terms can be treated as the covariant
derivative terms, these high-order terms do not contribute to the multiple
D2-brane effective action. Inspired by the derivation of a single D2-brane
from a M2-brane, we consider the curved M2-branes and introduce an auxiliary
field. Integrating out the auxiliary field, we indeed obtain the correct high-
order $F^{4}$ terms in the D2-brane effective action from the BLG theory and
the ABJM theory with $SU(2)\times SU(2)$ gauge symmetry, but we can not obtain
the correct high-order $F^{4}$ terms from the ABJM theory with $U(N)\times
U(N)$ and $SU(N)\times SU(N)$ gauge symmetries for $N>2$. We also briefly
comment on the (gauged) BF membrane theory.
###### pacs:
04.65.+e, 04.50.-h, 11.25.Hf
††preprint: MIFP-08-16
## I Introduction
Inspired by the ideas that the Chern-Simons gauge theories without Yang-Mills
kinetic terms may be used to describe ${\cal N}=8$ superconformal M2-brane
world-volume theory 0411077 ; 0412310 , Barger and Lambert 0611108 ; 0711.0955
; 0712.3738 , as well as Gustavasson (BLG) 0709.1260 ; 0802.3456 have
successfully constructed three-dimensional ${\mathcal{N}}=8$ superconformal
Chern-Simons gauge theory with manifest $SO(8)$ R-symmetry based on three
algebra. And then there is intensive research on the world-volume action of
multiple coincident M2-branes Mukhi:2008ux ; 0803.3242 ; 0803.3611 ; 0803.3803
; 0804.0913 ; 0804.1114 ; 0804.1256 ; 0804.1784 ; 0804.2110 ; 0804.2186 ;
0804.2201 ; 0804.2519 ; 0804.2662 ; 0804.3078 ; 0804.3567 ; 0804.3629 ;
0805.1012 ; 0805.1087 ; 0805.1202 ; 0805.1703 ; 0805.1895 ; 0805.1997 ;
0805.2898 ; 0805.3125 ; 0805.3193 ; 0805.3236 ; 0805.3427 ; 0805.3662 ;
0805.3930 ; 0805.4003 ; 0805.4363 ; 0805.4443 ; 0806.0054 ; 0806.0335 ;
0806.0363 ; 0806.0738 ; 0806.1218 ; 0806.1420 ; 0806.1519 ; 0806.1639 ;
0806.1990 ; 0806.2270 ; 0806.2584 ; 0806.3242 ; 0806.3253 ; 0806.3391 ;
0806.3498 ; 0806.3520 ; 0806.3727 ; Minahan:2008hf ; Larsson:2008ke ;
Furuuchi:2008ki ; Armoni:2008kr ; Agarwal:2008rr ; Gaiotto:2008cg ;
Bandos:2008fr ; 0806.4807 ; Bedford:2008hn ; Arutyunov:2008if ;
Stefanski:2008ik ; Grignani:2008is ; Hosomichi:2008jb ; Fre:2008qc ;
Okuyama:2008qd ; Bagger:2008se ; Terashima:2008sy ; Grignani:2008te ;
Chakrabortty:2008zk ; Terashima:2008ba ; Gromov:2008bz ; Ahn:2008gd ;
0807.0777 ; 0807.0802 ; 0807.0808 ; 0807.0812 ; 0807.0880 ; 0807.0890 ;
0807.1074 . Although the BLG theory is expected to describe any number of
M2-branes, there is one and only one known example with gauge group $SO(4)$
for the positive definite metric 0804.2110 ; 0804.2662 ; 0804.3078 . At the
level one of the Chern-Simons gauge theory, the BLG SO(4) gauge theory
describes two M2-branes on a $R^{8}/Z_{2}$ orbifold 0804.1114 ; 0804.1256 .
Thus, it is very important to generalize the BLG theory so that it can
describe an arbitrary number of M2-branes.
By relaxing the requirement of the positive definite metric on three algebra,
three groups 0805.1012 ; 0805.1087 ; 0805.1202 proposed the so called BF
membrane theory with arbitrary semi-simple Lie groups. However, the BF
membrane theory has ghost fields and then the unitarity problem in the
classical theory due to the Lorenzian three algebra. To solve these problems,
the global shift symmetries for the bosonic and fermionic ghost fields with
wrong-sign kinetic terms are gauged, which ensures the absence of the negative
norm states in the physical Hilbert space 0806.0054 ; 0806.0738 . However,
this gauged BF membrane theory might be equivalent to three-dimensional ${\cal
N}=8$ supersymmetric Yang-Mills theory 0806.1639 via a duality transformation
due to de Wit, Nicolai and Samtleben deWit:2004yr .
Very recently, Aharony, Bergman, Jafferis and Maldacena (ABJM) have
constructed three-dimensional Chern-Simons theories with gauge groups
$U(N)\times U(N)$ and $SU(N)\times SU(N)$ which have explicit ${\cal N}=6$
superconformal symmetry 0806.1218 (For Chern-Simons gauge theories with
${\cal N}=3$ and $4$ supersymmetries, see Refs. 0704.3740 ; Gaiotto:2008sd ).
Using brane constructions they argued that the $U(N)\times U(N)$ theory at
Chern-Simons level $k$ describes the low-energy limit of $N$ M2-branes on a
$C^{4}/Z_{k}$ orbifold. In particular, for $k=1$ and $2$, ABJM conjectured
that their theory describes the $N$ M2-branes respectively in the flat space
and on a $R^{8}/Z_{2}$ orbifold, and then might have ${\cal N}=8$
supersymmetry. For $N=2$, this theory has extra symmetries and is the same as
the BLG theory 0806.1218 .
On the other hand, D-branes are the hypersurfaces on which the open strings
can end, and their dynamics is described by open string field theory
Polchinski:1995mt . The low-energy world-volume action for D-branes can be
obtained by calculating the string scattering amplitudes Tseytlin:1997csa or
by using the T-duality Myers:1999ps . As usual in string theory, there are
high-order $\alpha^{\prime}=\ell_{s}^{2}$ corrections, where $\ell_{s}$ is the
string length scale. For a single D-brane, the D-brane action, which includes
all order corrections in the gauge field strength but not its derivatives,
takes the Dirac-Born-Infeld (DBI) form Leigh:1989jq . For multiple coincident
D-branes, Tseytlin assumed that all the commutator terms should be treated as
covariant derivative terms for gauge field strength, and thus should not be
included in the effective action Tseytlin:1997csa . And he proposed that the
action is the symmetrized trace of the direct non-Abelian generalization of
the DBI action Tseytlin:1997csa . This non-Abelian DBI action gives the
correct terms up to the order $F^{4}$ that were completely determined
previously Tseytlin:1986ti ; Gross:1986iv . But it fails for the higher order
terms Hashimoto:1997gm ; Schwarz:2001ps . Because the $F^{3}$ terms can always
be written as the commutator terms, they are not interesting in the
discussions of the D-brane effective action.
With the multiple M2-brane and D2-brane theories, we can study the deep
relation between them. As we know, the full effective action of a D2-brane can
be obtained by the reduction of the eleven-dimensional supermembrane action
Bergshoeff:1996tu . So, whether we can obtain the effective non-Abelian action
for multiple D2-branes from the reduction of the BLG and ABJM theories is an
interesting open question. Mukhi and Papageorgakis proposed a novel Higgs
mechanism by giving vacuum expectation value (VEV) to a scalar field, which
can promote the topological Chern-Simons gauge fields to dynamical gauge
fields Mukhi:2008ux . And they indeed obtained the maximally supersymmetric
Yang-Mills theory for two D2-branes from the BLG theory at the leading order.
Also, there exists a series of high-order corrections Mukhi:2008ux .
In this paper, we consider the derivation of the multiple D2-brane effective
action from the multiple M2-branes in the BLG and ABJM theories in details.
Concentrating on pure Yang-Mills fields, we show that the high-order $F^{3}$
and $F^{4}$ terms are commutator terms, and argue that all the high-order
terms are also commutator terms. Thus, these high-order terms are irrelevant
to the multiple D2-brane effective action. Note that the (gauged) BF membrane
theory does not have high-order terms, the BLG theory, the (gauged) BF
membrane theory, and the ABJM theory give the same D2-brane effective action.
In order to generate the non-trivial high-order $F^{4}$ terms, inspired by the
derivation of a single D2-brane from a M2-brane Bergshoeff:1996tu , we
consider the curved M2-branes and introduce an auxiliary field. In particular,
the VEV of the scalar field in the novel Higgs mechanism depends on the
auxiliary field. After we integrate out the massive gauge fields and auxiliary
field, we indeed obtain the high-order $F^{4}$ terms in the D2-brane effective
action from the BLG theory and the ABJM theory with $SU(2)\times SU(2)$ gauge
group. However, we still can not obtain the correct $F^{4}$ terms in the
generic ABJM theories with gauge groups $U(N)\times U(N)$ and $SU(N)\times
SU(N)$ for $N>2$. The reason might be that the $SU(2)\times SU(2)$ gauge
theory has three-dimensional ${\cal N}=8$ superconformal symmetry while the
$U(N)\times U(N)$ and $SU(N)\times SU(N)$ gauge theories with $N>2$ may only
have three-dimensional ${\cal N}=6$ superconformal symmetry Bagger:2008se . We
also briefly comment on the (gauged) BF membrane theory.
This paper is organized as follows. In Section II, we briefly review the novel
Higgs mechanism in the BLG theory and (gauged) BF membrane theory, and study
the novel Higgs mechanism in the ABJM theory. In Section III, we calculate the
effective D2-brane action with the leading order $F^{2}$, and high-order
$F^{3}$ and $F^{4}$ terms from M2-branes. In Section IV, we generate the high-
order $F^{4}$ terms by considering the curved M2-branes and introducing an
auxiliary field. Our discussion and conclusions are given in Section V.
## II Novel Higgs Mechanism
In this Section, we briefly review the novel Higgs mechanism from M2-branes to
D2-branes in the BLG theory and (gauged) BF membrane theory, and study it in
the ABJM theory.
### II.1 The BLG Theory and BF Membrane Theory
In the Lagrangian for the BLG theory with gauge group $SO(4)$ 0711.0955 , we
define
$\displaystyle
f^{abcd}~{}\equiv~{}f\epsilon^{abcd}~{},~{}~{}~{}f=\frac{2\pi}{k}~{},~{}\,$
(1)
where $k$ is the level of the Chern-Simons terms. We also make the following
transformation on the Yang-Mills fields
$\displaystyle A_{\mu AB}\longrightarrow{1\over f}A_{\mu AB}~{}.~{}\,$ (2)
Then the Lagrangian for the BLG theory with gauge group $SO(4)$ becomes
$\displaystyle{\mathcal{L}}$ $\displaystyle=$ $\displaystyle-{1\over
2}D^{\mu}X^{AI}D_{\mu}X^{I}_{A}+{i\over
2}\bar{\Psi}^{A}\Gamma^{\mu}D_{\mu}\Psi_{A}+{{if}\over
4}\bar{\Psi}_{B}\Gamma_{IJ}X_{C}^{I}X_{D}^{J}\Psi_{A}\epsilon^{ABCD}$ (3)
$\displaystyle-V(X)+{1\over{2f}}\epsilon^{\mu\nu\lambda}(\epsilon^{ABCD}A_{\mu
AB}\partial_{\nu}A_{\lambda CD}+{2\over
3}{\epsilon^{CDA}}_{G}\epsilon^{EFGB}A_{\mu AB}A_{\nu CD}A_{\lambda
EF})~{},~{}$
where $A=1,2,3,4$, $I=1,2,...,8$, and
$V(X)={{f^{2}}\over
12}\epsilon_{ABCD}{\epsilon_{EFG}}^{D}X^{A(I)}X^{B(J)}X^{C(K)}X^{E(I)}X^{F(J)}X^{G(K)}~{}.~{}\,$
(4)
As we know, the strong coupling limit of Type IIA theory is M-theory, and the
coupling constant in Type IIA theory is related to the radius of the circle of
the eleventh dimension in M-theory. Thus, for D2-branes, the gauge coupling
constant is also related to the radius of the circle of the eleventh
dimension. And at the strong coupling limit the D2-branes become M2-branes. To
derive the D2-branes from M2-branes via the novel Higgs mechanism, we
compactify the M-theory on the circle of the eleventh dimension by giving VEV
to a linear combination of the scalar fields $X^{A(I)}$ Mukhi:2008ux . Because
we have the $SO(8)$ R-symmetry and SO(4) gauge symmetry, we can always make
the rotation so that only the component $\langle X^{8(\phi)}\rangle$ develops
a VEV
$\langle X^{8(\phi)}\rangle~{}=~{}v_{0}~{}=~{}{v\over{\sqrt{f}}}~{},~{}\,$ (5)
where we split the index $A$ into two sets $a=1,2,3$ and $\phi=4$. In
addition, the gauge fields are splitted into $A_{\mu}^{a}$ and $B_{\mu}^{a}$
$A_{\mu}^{a}\equiv A_{\mu}^{a\phi}~{},~{}~{}~{}B_{\mu}^{a}\equiv{1\over
2}{\epsilon^{a}}_{bc}A_{\mu}^{bc}~{}.~{}\,$ (6)
And then the Chern-Simons terms can be rewritten as
${1\over 2}\epsilon^{\mu\nu\lambda}\epsilon^{ABCD}A_{\mu
AB}\partial_{\nu}A_{\lambda
CD}=4\epsilon^{\mu\nu\lambda}B_{\mu}^{a}\partial_{\nu}A_{\lambda a}~{},~{}\,$
(7) ${1\over
3}\epsilon^{\mu\nu\lambda}{\epsilon_{CDA}}^{G}\epsilon_{EFGB}A_{\mu}^{AB}A_{\nu}^{CD}A_{\lambda}^{EF}=-4\epsilon^{\mu\nu\lambda}\epsilon_{abc}B_{\mu}^{a}A_{\nu}^{b}A_{\lambda}^{c}-{4\over
3}\epsilon^{\mu\nu\lambda}\epsilon_{abc}B_{\mu}^{a}B_{\nu}^{b}B_{\lambda}^{c}~{},~{}\,$
(8)
where we neglect the total derivative term. Combining these two terms, the
Chern-Simons action becomes
${\mathcal{L}}_{\rm CS}={1\over
f}\left(2\epsilon^{\mu\nu\lambda}B_{\mu}^{a}F_{\nu\lambda a}-{4\over
3}\epsilon^{\mu\nu\lambda}\epsilon_{abc}B_{\mu}^{a}B_{\nu}^{b}B_{\lambda}^{c}\right)~{},~{}\,$
(9)
where $F_{\nu\lambda a}=\partial_{\nu}A_{\lambda a}-\partial_{\lambda}A_{\nu
a}-2\epsilon_{abc}A_{\nu}^{b}A_{\lambda}^{c}$ is the field strength for the
gauge field $A_{\mu}^{a}$. Similarly, the kinetic terms for the scalar fields
are
$D_{\mu}X^{a(I)}=\partial_{\mu}X^{a(I)}+\epsilon^{a}_{~{}BCD}A_{\mu}^{CD}X^{B(I)}=\partial_{\mu}X^{a(I)}-2{\epsilon^{a}}_{cb}A_{\mu}^{c}X^{b(I)}+2B_{\mu}^{a}X^{\phi(I)},$
(10) $D_{\mu}X^{\phi(I)}=\partial_{\mu}X^{\phi(I)}-2B_{\mu a}X^{a(I)}.$ (11)
Substituting these back into the action and setting $X^{\phi(8)}\rightarrow
X^{\phi(8)}+v$, we obtain the terms involving $B_{\mu}^{a}$ from the scalar
kinetic terms
$\displaystyle{\mathcal{L}}=-{{2v^{2}}\over
f}B_{\mu}^{a}B_{a}^{\mu}-{{4v}\over{\sqrt{f}}}X^{8\phi}B_{\mu}^{a}B_{a}^{\mu}-2X^{8\phi}X^{8\phi}B_{\mu}^{a}B_{a}^{\mu}-2B_{\mu}^{a}B_{a}^{\mu}X^{\phi(i)}X_{\phi(i)}$
$\displaystyle-2B_{\mu}^{a}X^{\phi(i)}D^{\mu}X_{a}^{(i)}-{{2v}\over{\sqrt{f}}}B_{\mu}^{a}D^{\mu}X_{a}^{(8)}-2X^{8\phi}B_{\mu}^{a}D^{\mu}X_{a}^{(8)}$
$\displaystyle-2B_{\mu
a}X^{a(I)}B_{b}^{\mu}X^{b(I)}+2B_{a}^{\mu}X^{a(I)}\partial_{\mu}X^{\phi(I)},$
(12)
where $i=1,2,...,7$, and the new defined covariant derivative is
$D_{\mu}X^{a(I)}=\partial_{\mu}X^{a(I)}-2{\epsilon^{a}}_{bc}A_{\mu}^{b}X^{c(I)}$.
Therefore, the relevant Lagrangian for pure Yang-Mills fields is
$\displaystyle{\mathcal{L}}_{\rm YM}={1\over
f}\left(-2v^{2}B_{\mu}^{a}B_{a}^{\mu}+2\epsilon^{\mu\nu\lambda}B_{\mu}^{a}F_{\nu\lambda
a}-{4\over
3}\epsilon^{\mu\nu\lambda}\epsilon_{abc}B_{\mu}^{a}B_{\nu}^{b}B_{\lambda}^{c}\right)~{}.~{}\,$
(13)
Next, we would like to briefly review the result of the novel Higgs mechanism
in the BF membrane theory 0805.1012 ; 0805.1087 ; 0805.1202 . Here, we follow
the convention in Ref. 0805.1087 except that we choose
$\displaystyle(B_{\mu})_{a}~{}\equiv~{}{1\over
2}(A_{\mu})_{bc}f^{~{}~{}bc}_{a}~{}.~{}\,$ (14)
In this theory, the equation of motion for ghost field $X_{-}^{I}$ gives the
constraint $\partial^{2}X^{I}_{+}=0$. So, we can give a constant VEV to
$X^{8}_{+}$, i.e., $X^{8}_{+}=v$. And then we obtain the relevant Lagrangian
for pure gauge fields
${\mathcal{L}}=-2v^{2}B^{a}_{\mu}B^{\mu}_{a}+2\epsilon^{\mu\nu\lambda}B^{a}_{\mu}F^{a}_{\nu\lambda}~{}.~{}\,$
(15)
It should be noted that unlike the Lagrangian in Eq. (13) in the BLG theory,
there is no cubic term for $B_{\mu}^{a}$ in above Lagrangian. And this is one
of the motivations of the work 0806.1639 which showed that the gauged BF
membrane theory might be equivalent to the maximally supersymmetric three-
dimensional Yang-Mills theory via a duality transformation due to de Wit,
Nicolai and Samtleben deWit:2004yr .
After gauging the shift symmetries for the ghost fields $X^{I}_{-}$ and
$\Psi_{-}$ in the BF membrane theory 0806.0054 ; 0806.0738 by introducing new
gauge fields, we could make the gauge choice to decouple the ghost states. And
the equation of motion for the new gauge fields gives the constraint
$\partial_{\mu}X^{I}_{+}=0$, which indicates that $X^{I}_{+}$ must be a
constant. We emphasize that in this case the relevant Lagrangian for pure
Yang-Mills fields is still given by Eq. (15).
### II.2 The ABJM Theory
Very recently, Aharony, Bergman, Jafferis and Maldacena (ABJM) have
constructed three-dimensional $U(N)\times U(N)$ and $SU(N)\times SU(N)$ Chern-
Simons gauge theories with ${\cal N}=6$ superconformal symmetry. From the
brane constructions, they argued that the $U(N)\times U(N)$ theory at Chern-
Simons level $k$ describes the low-energy limit of $N$ M2-branes probing a
$C^{4}/Z_{k}$ singularity. It was conjectured that for $k=1$ and $2$, the ABJM
theory respectively describes $N$ M2-branes in flat space and on a
$R^{8}/Z_{2}$ orbifold, and then may have ${\cal N}=8$ supersymmetry. For
$N=2$, this theory has additional symmetries and becomes identical to the BLG
theory. In this subsection, we will study the novel Higgs mechanism in the
ABJM theory.
Following the convention in Ref. 0806.1519 , we can write the explicit
Lagrangian in ABJM theory as follows
$\displaystyle{\mathcal{L}}$ $\displaystyle=$ $\displaystyle
2K\epsilon^{\mu\nu\lambda}{\rm
Tr}\left(A^{\prime}_{\mu}\partial_{\nu}A^{\prime}_{\lambda}+\frac{2i}{3}A^{\prime}_{\mu}A^{\prime}_{\nu}A^{\prime}_{\lambda}-\hat{A}_{\mu}\partial_{\nu}\hat{A}_{\lambda}-\frac{2i}{3}\hat{A}_{\mu}\hat{A}_{\nu}\hat{A}_{\lambda}\right)$
(16) $\displaystyle~{}~{}~{}~{}-{\rm
Tr}\left(({\mathcal{D}}_{\mu}Z)^{\dagger}{\mathcal{D}}^{\mu}Z+({\mathcal{D}}_{\mu}W)^{\dagger}{\mathcal{D}}^{\mu}W-i\zeta^{\dagger}\gamma^{\mu}{\mathcal{D}}_{\mu}\zeta-i\omega^{\dagger}\gamma^{\mu}{\mathcal{D}}_{\mu}\omega\right)$
$\displaystyle~{}~{}~{}~{}-V_{\mathrm{ferm}}-V_{\mathrm{bos}}~{},~{}\,$
where
$\displaystyle K~{}=~{}{k\over{8\pi}}~{},~{}\,$ (17) $\displaystyle
Z^{1}=X^{1}+iX^{5}~{},~{}~{}~{}Z^{2}=X^{2}+iX^{6}~{},~{}~{}~{}W_{1}=X^{3\dagger}+iX^{7\dagger}~{},~{}~{}~{}W_{2}=X^{4\dagger}+iX^{8\dagger}~{},~{}\,$
(18)
where $X^{i}$ belongs to the bifundamental representation of $U(N)\times U(N)$
or $SU(N)\times SU(N)$, and here we do not present the potential
$V_{\mathrm{ferm}}$ and $V_{\mathrm{bos}}$ since they are irrelevant in the
following discussions. For our convention, we choose
$\displaystyle{\rm Tr}(T^{a}T^{b})~{}=~{}{1\over
2}\delta_{ab}~{},~{}~{}~{}[T^{a},~{}T^{b}]~{}=~{}if_{abc}T^{c}~{},~{}\,$ (19)
where $T^{a,b,c}$ are the generators of the corresponding gauge group.
Similar to the novel Higgs mechanism in the BLG theory, we give the diagonal
VEV to $X^{8}$ as follows
$\langle X^{8}\rangle~{}=~{}v_{0}I_{N\times N}~{}=~{}v{\sqrt{K}}I_{N\times
N}~{},~{}\,$ (20)
where $I_{N\times N}$ is the $N$ by $N$ indentity matrix. Also, we define
$\displaystyle A_{\mu}={1\over
2}(A^{\prime}_{\mu}+\hat{A}_{\mu})~{},~{}~{}~{}B_{\mu}={1\over
2}(A^{\prime}_{\mu}-\hat{A}_{\mu})~{}.~{}\,$ (21)
So we have
$\displaystyle
A^{\prime}_{\mu}=A_{\mu}+B_{\mu}~{},~{}~{}~{}\hat{A}_{\mu}=A_{\mu}-B_{\mu}~{}.~{}\,$
(22)
From the kinetic term for $W_{2}$ and the Chern-Simons terms, we obtain the
relevant Lagrangian for pure Yang-Mills fields
${\mathcal{L}}_{\rm
YM}=K\left(-2v^{2}B^{a}_{\mu}B^{\mu}_{a}+2\epsilon^{\mu\nu\lambda}B^{a}_{\mu}F_{a\nu\lambda}-\frac{2}{3}\epsilon^{\mu\nu\lambda}f_{abc}B^{a}_{\mu}B^{b}_{\nu}B^{c}_{\lambda}\right)~{},~{}$
(23)
where
$F_{\mu\nu}=\partial_{\mu}A_{\nu}-\partial_{\nu}A_{\mu}+i[A_{\mu},A_{\nu}]$.
Note that the BLG theory with $SO(4)$ gauge group is the same as the ABJM
theory with $SU(2)\times SU(2)$ gauge group, so we can obtain the Lagrangian
in Eq. (13) from that in the above Eq. (23) by rescaling $f_{abc}$.
## III Effective Action for the Pure Gauge Fields
Because $B_{\mu}^{a}$ is massive, we will calculate the effective action for
pure Yang-Mills fields by integrating it out. Due to the absence of the cubic
term for $B_{\mu}^{a}$ in the (gauged) BF membrane theory, we do not have the
high-order corrections in the effective action of gauge fields. Thus, we will
concentrate on the BLG theory and ABJM theory. The relevant Lagrangians for
pure gauge fields are the same for the BLG theory and the ABJM theory with
$SU(2)\times SU(2)$ gauge symmetry, and the ABJM theory is more general. Thus,
we will use the Lagrangian in Eq. (23) in the following discussions.
From the Lagrangian in Eq. (23), we get the equation of motion for
$B^{a}_{\mu}$
$\displaystyle
B^{\mu}_{a}~{}=~{}\frac{1}{2v^{2}}\epsilon^{\mu\nu\lambda}F_{a\nu\lambda}-\frac{1}{2v^{2}}\epsilon^{\mu\nu\lambda}f_{abc}B^{b}_{\nu}B^{c}_{\lambda}~{}.~{}\,$
(24)
We can solve the above equation by parametrizing the solution in ${1/{v^{2}}}$
expansion
$B^{\mu}_{a}=\sum_{n}{1\over{v^{2n}}}(C_{2n})_{a}^{\mu}~{}.~{}\,$ (25)
Substituting it back into Eq. (24), we obtain
$\sum_{n}{1\over{v^{2n}}}(C_{2n})^{\mu}_{a}~{}=~{}{1\over{2v^{2}}}\epsilon^{\mu\nu\lambda}F_{a\nu\lambda}-{1\over{2v^{2}}}\sum_{n,m}{1\over
v^{2n+2m}}{\epsilon^{\mu\nu\lambda}}f_{abc}(C_{2n})_{\nu}^{b}(C_{2m})_{\lambda}^{c}+...~{}.~{}\,$
(26)
Because we only know for sure the high-order terms up to the order of $F^{4}$
in D2-brane effective action Tseytlin:1997csa ; Hashimoto:1997gm ;
Schwarz:2001ps , we only need to calculate the solution to Eq. (24) up to the
order of $1/v^{10}$ or $(C_{10})^{\mu}_{a}$. And the non-vanishing terms in
the solution are
$\displaystyle(C_{2})^{\mu}_{a}~{}=~{}{1\over
2}{\epsilon^{\mu\nu\lambda}}F_{a\nu\lambda}~{},~{}~{}~{}(C_{6})^{\mu}_{a}~{}=~{}-{1\over
2}{\epsilon^{\mu\nu\lambda}}f_{abc}(C_{2})_{\nu}^{b}(C_{2})_{\lambda}^{c}~{},~{}\,$
(27)
$\displaystyle(C_{10})^{\mu}_{a}=-{\epsilon^{\mu\nu\lambda}}f_{abc}(C_{2})_{\nu}^{b}(C_{6})_{\lambda}^{c}~{}.~{}\,$
(28)
Integrating $B^{a}_{\mu}$ out, we get the Lagrangian for pure Yang-Mills
fields
$\displaystyle{\mathcal{L}_{\rm YM}}$ $\displaystyle=$
$\displaystyle{\mathcal{L}^{(2)}_{\rm YM}}+{\mathcal{L}^{(3)}_{\rm
YM}}+{\mathcal{L}^{(4)}_{\rm YM}}+...~{},~{}\,$ (29)
where
$\displaystyle{\mathcal{L}^{(2)}_{\rm
YM}}~{}=~{}\frac{2K}{v^{2}}(C_{2})_{\mu}^{a}(C_{2})^{\mu}_{a}~{},~{}\,$ (30)
$\displaystyle{\mathcal{L}^{(3)}_{\rm
YM}}~{}=~{}-\frac{2K}{3v^{6}}\epsilon^{\mu\nu\lambda}f_{abc}(C_{2})_{\mu}^{a}(C_{2})_{\nu}^{b}(C_{2})_{\lambda}^{c}~{},~{}\,$
(31) $\displaystyle{\mathcal{L}^{(4)}_{\rm
YM}}~{}=~{}-\frac{2K}{v^{10}}(C_{6})_{\mu}^{a}(C_{6})^{\mu}_{a}-\frac{2K}{v^{10}}\epsilon^{\mu\nu\lambda}f_{abc}(C_{2})_{\mu}^{a}(C_{2})_{\nu}^{b}(C_{6})_{\lambda}^{c}~{}.~{}\,$
(32)
Using Eqs. (27) and (28), and the useful identities in the Appendix A, we
obtain
$\displaystyle{\mathcal{L}^{(2)}_{\rm YM}}~{}=~{}-\frac{2K}{v^{2}}{\rm
Tr}\left(F^{2}\right)~{},~{}\,$ (33) $\displaystyle{\mathcal{L}^{(3)}_{\rm
YM}}~{}=~{}\frac{i4K}{3v^{6}}{\rm
Tr}\left(F_{\alpha_{1}\beta_{1}}[F^{\beta_{1}\beta_{3}},F^{~{}~{}\alpha_{1}}_{\beta_{3}}]\right)~{},~{}\,$
(34) $\displaystyle{\mathcal{L}^{(4)}_{\rm YM}}~{}=~{}\frac{K}{2v^{10}}{\rm
Tr}\left([F^{\rho\sigma},F^{\eta\delta}][F_{\eta\delta},F_{\rho\sigma}]\right)~{}.~{}\,$
(35)
Thus, ${\mathcal{L}^{(2)}_{\rm YM}}$ is the kinetic term for the gauge fields
$A^{\mu}_{a}$ and is the leading order of the supersymmetric Yang-Mills
effective action. Moreover, the gauge coupling in the BLG theory is
$\displaystyle g_{YM}^{2}~{}=~{}{{fv^{2}}\over 4}~{}=~{}{{f^{2}v_{0}^{2}}\over
4}~{}\propto~{}{{v_{0}^{2}}\over{k^{2}}}~{},~{}\,$ (36)
and the gauge coupling in the ABJM theory is
$\displaystyle
g_{YM}^{2}~{}=~{}{{v^{2}}\over{4K}}~{}=~{}{{v_{0}^{2}}\over{4K^{2}}}~{}\propto~{}{{v_{0}^{2}}\over{k^{2}}}~{}.~{}\,$
(37)
So for very large $v_{0}$ and $k$, we can still keep the gauge coupling as a
fixed constant. For D2-branes, the gauge coupling is related to the string
coupling and the string length as follows
$g_{YM}=({g_{s}\over\ell_{s}})^{1\over 2}~{}.~{}\,$ (38)
And then for the fixed string coupling, we have
$g_{YM}^{2}\propto\alpha^{-1/2}$. Therefore, ${1/v}$ is proportional to
${\alpha}^{\prime{1/4}}$, ${\mathcal{L}^{(3)}_{\rm YM}}$ and
${\mathcal{L}^{(4)}_{\rm YM}}$ are proportional to
$g_{YM}^{-2}\alpha^{\prime}$ and $g_{YM}^{-2}\alpha^{\prime 2}$, respectively.
In short, they are at the correct orders according to the $\alpha^{\prime}$
expansion.
Because ${\mathcal{L}^{(3)}_{\rm YM}}$ and ${\mathcal{L}^{(4)}_{\rm YM}}$ only
have commutator terms, these high-order terms are covariant derivative terms
and then do not contribute to the effective action for the D2-branes
Tseytlin:1997csa . We conjecture that all the high-order terms obtained by
this approach are the commutator terms. The point is that the equation of
motion for $B^{a}_{\mu}$ in Eq. (24) can be rewritten as follows
$\displaystyle
B^{a\mu}~{}=~{}\frac{1}{2v^{2}}\epsilon^{\mu\nu\lambda}F^{a}_{\nu\lambda}+\frac{i}{v^{2}}\epsilon^{\mu\nu\lambda}{\rm
Tr}\left(T^{a}[B_{\nu},B_{\lambda}]\right)~{}.~{}\,$ (39)
Because all the high-order terms originally come from the last term in the
above equation which is a commutator term, all the high-order terms should be
the commutator terms and then the covariant terms. Thus, moduloing the
commutator terms or covariant derivative terms, we only have the kinetic term
for the gauge fields $A^{\mu}_{a}$ from the BLG and ABJM theories, which is
the leading order in the D2-brane effective action. And then the effective
action for pure Yang-Mills fields from the BLG and ABJM theories is the same
as that from the (gauged) BF membrane theory after we integrate $B^{\mu}_{a}$
out. Therefore, how to obtain the non-trivial $F^{4}$ terms in the D2-brane
effective action from the BLG theory, the (gauged) BF membrane theory, and the
ABJM theory is still a big problem.
## IV D2-Branes from the Curved M2-Branes
In spired by the derivation of a single D2-brane from a M2-brane
Bergshoeff:1996tu , we would like to consider the multiple curved M2-branes.
To employ the trick in Ref. Bergshoeff:1996tu , we only need to introduce
gravity. For simplicity, we do not consider the dilaton, the vector and scalar
fields in the eleven-dimensional metric due to compactification, and RR
fields, etc. And our ansatz for the Lagrangian of the curved M2-branes is
${\mathcal{L}}_{\rm Curved}=-\beta_{0}\sqrt{-{\rm det}(g)}+\sqrt{-{\rm
det}(g)}~{}{\mathcal{L}}_{\rm M2s}~{},~{}\,$ (40)
where $\beta_{0}$ is a positive constant like membrane tension, $g_{\mu\nu}$
is the induced metric on the world-volume of multiple M2-branes, and
${\mathcal{L}}_{\rm M2s}$ is formally given in Eq. (3) for the BLG theory or
in Eq. (16) for the ABJM theory. In ${\mathcal{L}}_{\rm M2s}$, we need to
replace $\eta_{\mu\nu}$ and $\partial_{\alpha}$ by $g_{\mu\nu}$ and
$\nabla_{\alpha}$, respectively. Also, we replace $\epsilon_{\mu\nu\lambda}$
by $\varepsilon_{\mu\nu\lambda}=\sqrt{-g}\epsilon_{\mu\nu\lambda}$ which will
be covariant under coordinate transformation. This is a natural action for the
multiple M2-branes in the curved space-time since it can come back to flat
theory after we decouple the gravity.
Similar to the discussions in Ref. Bergshoeff:1996tu , we introduce an
auxiliary filed $u$ and rewrite the above Lagrangian as follows
${\mathcal{L}}_{\rm Curved}=\frac{\beta^{2}_{0}}{2u}{\rm
det}~{}(g)-\frac{u}{2}+\sqrt{-{\rm det}(g)}~{}{\mathcal{L}}_{M2s}~{}.~{}\,$
(41)
We can obtain the Lagrangian in Eq. (40) from Eq. (41) by integrating out the
auxiliary filed $u$.
To match the convention in Tseytlin:1997csa , we give the following VEV to the
scalar field $\phi$
$<\phi>=(\frac{8u}{\sqrt{-{\rm
det}(g)}})^{1/2}~{}{{K^{\prime}}\over{\beta_{0}}}~{}{1\over{2\pi\alpha^{\prime}}}~{}I_{N\times
N}~{},~{}\,$ (42)
where we can take $\phi=X^{8(\phi)}$, $K^{\prime}=1/f$, and $N=1$ in the BLG
theory, take $\phi=X_{+}^{8}$, $K^{\prime}=1$ and $N=1$ in the (gauged) BF
membrane theory, and take $\phi=X^{8}$ and $K^{\prime}=K$ in the ABJM theory.
Thus, the relevant Lagrangian is
$\displaystyle{\mathcal{L}}_{\rm Curved}$ $\displaystyle=$
$\displaystyle\frac{\beta^{2}_{0}}{2u}{\rm det}~{}(g)-\frac{u}{2}+\sqrt{-{\rm
det}(g)}\left(-2\langle\phi^{2}\rangle
B^{a}_{\mu}B^{\mu}_{a}+2K^{\prime}\varepsilon^{\mu\nu\lambda}B_{\mu}^{a}F_{\nu\lambda}^{a}\right.$
(43)
$\displaystyle\left.-\frac{2}{3}K^{\prime}\varepsilon^{\mu\nu\lambda}f_{abc}B_{\mu}^{a}B_{\nu}^{b}B_{\lambda}^{c}\right)~{}.~{}\,$
Using the results of the novel Higgs mechanism in the Section III and
neglecting the commutator terms for the $A_{\mu}^{a}$ field strength, we
obtain
${\mathcal{L}}=\frac{\beta^{2}_{0}}{2u}{\rm
det}(g)\left(1+\frac{(2\pi\alpha^{\prime})^{2}}{4}F^{a\mu\nu}F^{a}_{\mu\nu}\right)-\frac{u}{2}~{}.~{}\,$
(44)
Moreover, we use the following identity for the $3\times 3$ matrices that is
proved in the Appendix A
${\rm Str}~{}{\rm det}~{}(g+2\pi\alpha^{\prime}{F})~{}=~{}{\rm
det}(g)~{}\left(1+\frac{(2\pi\alpha^{\prime})^{2}}{4}F^{a\mu\nu}F^{a}_{\mu\nu}\right)~{},~{}\,$
(45)
where “Str” is the symmetrized trace that acts on the gauge group indices, and
“det” acts on the world-volume coordinate indices. Integrating out the
auxiliary field $u$, we obtain the Lagrangian for multiple D2-brane effective
action
$\displaystyle{\mathcal{L}}=-\beta_{0}{\sqrt{-{\rm Str}~{}{\rm
det}~{}(g+2\pi\alpha^{\prime}{F})}}~{}.~{}\,$ (46)
However, the well-known Lagrangian for the multiple D2-brane DBI action is
Tseytlin:1997csa
$\displaystyle{\mathcal{L}}=-c_{0}{\rm Str}\left[\sqrt{-~{}{\rm
det}~{}(g+2\pi\alpha^{\prime}{F})}\right]~{},~{}\,$ (47)
where $c_{0}$ is a constant. Because in general the Lagrangian in Eq. (46) is
not equivalent to that in Eq. (47), we still can not get the correct $F^{4}$
terms for generic case.
Interestingly, for gauge symmetry $SU(2)\times SU(2)$ in the BLG theory, or
the (gauged) BF membrane theory, or the ABJM theory, we indeed can get the
correct $F^{4}$ terms. Let us prove it in the following. From the Lagrangian
in Eq. (46), we obtain
$\displaystyle{\mathcal{L}}=-\beta_{0}{\sqrt{-{\rm
det}(g)\left(1+\frac{(2\pi\alpha^{\prime})^{2}}{4}F^{a\mu\nu}F^{a}_{\mu\nu}\right)}}~{}.~{}\,$
(48)
Expandind the above Lagrangian, we have the relevant Lagrangian for pure Yang-
Mills fields at the Minkowski space-time limit
$\displaystyle{\mathcal{L}}=-{{\beta_{0}(2\pi\alpha^{\prime})^{2}}\over{4}}{\rm
Tr}\Biggl{[}F^{\mu\nu}F_{\mu\nu}-{{(2\pi\alpha^{\prime})^{2}}\over{4}}(F^{\mu\nu}F_{\mu\nu})^{2}+...\Biggr{]}~{}.~{}\,$
(49)
From the known effective action for multiple D2-branes, the relevant
Lagrangian for pure Yang-Mills fields up to the $F^{4}$ terms is
Tseytlin:1997csa
$\displaystyle{\mathcal{L}}_{DBI}$ $\displaystyle=$ $\displaystyle c_{1}{\rm
Tr}\Biggl{\\{}F^{\mu\nu}F_{\mu\nu}-{1\over
3}(2\pi\alpha^{\prime})^{2}\left(F^{\mu\nu}F_{\rho\nu}F_{\mu\sigma}F^{\rho\sigma}+\frac{1}{2}F^{\mu\nu}F_{\rho\nu}F^{\rho\sigma}F_{\mu\sigma}\right.$
(50)
$\displaystyle\left.-\frac{1}{4}F^{\mu\nu}F_{\mu\nu}F^{\rho\sigma}F_{\rho\sigma}-\frac{1}{8}F^{\mu\nu}F^{\rho\sigma}F_{\mu\nu}F_{\rho\sigma}\right)+...\Biggr{\\}}~{},~{}\,$
where $c_{1}=\pi^{2}\alpha^{\prime 2}c_{0}$. For gauge group $SU(2)$, we
obtain
$\displaystyle{\mathcal{L}}_{DBI}$ $\displaystyle=$ $\displaystyle c_{1}{\rm
Tr}\Biggl{\\{}F^{\mu\nu}F_{\mu\nu}-(2\pi\alpha^{\prime})^{2}\left(\frac{1}{8}(F^{\mu\nu}F_{\mu\nu})^{2}+\frac{1}{24}F^{\mu\nu}F^{\rho\sigma}[F_{\mu\nu},F_{\rho\sigma}]\right)+...\Biggr{\\}}~{}.~{}\,$
(51)
Therefore, neglecting the commutator terms and rescaling the gauge fields, we
can show that the correct $F^{4}$ terms in the effective D2-brane action in
Eq. (49) from the two M2-branes in the BLG and ABJM theories are equivalent to
these in the known DBI action in Eq. (51).
In short, we can generate the correct $F^{4}$ terms in the effective D2-brane
action from the BLG theory and the ABJM theory with gauge group $SU(2)\times
SU(2)$. However, we can not get the correct $F^{4}$ terms from the ABJM theory
with $U(N)\times U(N)$ and $SU(N)\times SU(N)$ gauge symmetries for $N>2$. It
seems to us that the reasons are the following: the BLG theory and the ABJM
theory with gauge group $SU(2)\times SU(2)$ have three-dimensional ${\cal
N}=8$ superconformal symmetry while the ABJM theory with $U(N)\times U(N)$ and
$SU(N)\times SU(N)$ gauge symmetries for $N>2$ might only have three-
dimensional ${\cal N}=6$ superconformal symmetry Bagger:2008se . However, for
the (gauged) BF membrane theory, although the constraint
$\nabla_{\mu}X_{+}^{8}=0$ is still satisfied, it might be equivalent to three-
dimensional ${\cal N}=8$ supersymmetric Yang-Mills theory. In particular, for
the (gauged) BF membrane theory with $SU(2)\times SU(2)$ gauge symmetry, we
can generate the correct $F^{4}$ terms since it is similar to the
corresponding BLG and ABJM theories.
## V Discussion and Conclusions
Using the novel Higgs mechanism, we considered the derivation of the multiple
D2-brane effective action for pure Yang-Mills fields from the multiple
M2-branes in the BLG theory and the ABJM theory. We showed that the high-order
$F^{3}$ and $F^{4}$ terms are commutator terms, and we argued that all the
high-order terms are commutator terms as well. Thus, these high-order terms do
not contribute to the multiple D2-brane effective action. In order to generate
the non-trivial high-order $F^{4}$ terms and inspired by the derivation of one
D2-brane from one M2-brane, we considered the curved M2-branes and introduce
an auxiliary field. In particular, the VEV of the scalar field in the novel
Higgs mechanism depends on the auxiliary field. After we integrate out the
massive gauge fields and auxiliary field, we obtain the correct high-order
$F^{4}$ terms in the D2-brane effective action from the BLG theory and the
ABJM theory with $SU(2)\times SU(2)$ gauge group. However, we still can not
obtain the correct $F^{4}$ terms in the generic ABJM theory with gauge groups
$U(N)\times U(N)$ and $SU(N)\times SU(N)$ for $N>2$. This might be related to
the possible fact that the $SU(2)\times SU(2)$ gauge theory has three-
dimensional ${\cal N}=8$ superconformal symmetry while the $U(N)\times U(N)$
and $SU(N)\times SU(N)$ gauge theories for $N>2$ might only have three-
dimensional ${\cal N}=6$ superconformal symmetry. We also briefly comment on
the (gauged) BF membrane theory.
###### Acknowledgements.
This research was supported in part by the Cambridge-Mitchell Collaboration in
Theoretical Cosmology (TL).
## Appendix A Mathematical Identifies
In this appendix we collect or prove the useful identities in this paper:
(1) Two useful identities about $\epsilon$ in three-dimensional Minkowski
space-time
$\displaystyle\epsilon^{\mu\nu\lambda}{\epsilon_{\lambda}}^{\rho\sigma}=(-\eta^{\mu\rho}\eta^{\nu\sigma}+\eta^{\mu\sigma}\eta^{\nu\rho})~{},~{}\,$
(52)
$\displaystyle\epsilon^{\mu\gamma\lambda}\epsilon^{\nu\rho\sigma}=-\eta^{\mu\nu}(\eta^{\gamma\rho}\eta^{\lambda\sigma}-\eta^{\gamma\sigma}\eta^{\lambda\rho})+\eta^{\mu\rho}(\eta^{\gamma\nu}\eta^{\lambda\sigma}-\eta^{\gamma\sigma}\eta^{\lambda\nu})-\eta^{\mu\sigma}(\eta^{\gamma\nu}\eta^{\lambda\rho}-\eta^{\gamma\rho}\eta^{\lambda\nu})~{}.~{}$
(53)
(2) Let us prove the identity in Eq. (45) which is right for the Abelian and
non-Abelian cases
$\displaystyle{\rm Str}~{}{\rm det}(g+a{F})$ $\displaystyle=$
$\displaystyle{\rm det}(g_{\mu\nu})~{}{\rm Str}~{}{\rm
det}(g^{\nu}_{~{}~{}\lambda}+a{F}^{\nu}_{~{}~{}\lambda})$ (54)
$\displaystyle=$ $\displaystyle({\rm det}~{}g)~{}{\rm
Str}~{}\epsilon_{\alpha_{1}\alpha_{2}\alpha_{3}}(g^{1}_{~{}~{}\alpha_{1}}+a{F}^{1}_{~{}~{}\alpha_{1}})(g^{2}_{~{}~{}\alpha_{2}}+a{F}^{2}_{~{}~{}\alpha_{2}})(g^{3}_{~{}~{}\alpha_{3}}+a{F}^{3}_{~{}~{}\alpha_{3}})$
$\displaystyle=$ $\displaystyle({\rm det}~{}g)~{}{\rm
Str}~{}\Biggl{[}1+a{F}^{\alpha}_{~{}~{}\alpha}+a^{2}\left(\epsilon_{1\alpha_{2}\alpha_{3}}{F}^{2}_{~{}~{}\alpha_{2}}{F}^{3}_{~{}~{}\alpha_{3}}+\epsilon_{\alpha_{1}2\alpha_{3}}{F}^{1}_{~{}~{}\alpha_{1}}{F}^{3}_{~{}~{}\alpha_{3}}\right.$
$\displaystyle\left.~{}~{}+\epsilon_{\alpha_{1}\alpha_{2}3}{F}^{1}_{~{}~{}\alpha_{1}}{F}^{2}_{~{}~{}\alpha_{2}}\right)+a^{3}\epsilon_{\alpha_{1}\alpha_{2}\alpha_{3}}{F}^{1}_{~{}~{}\alpha_{1}}{F}^{2}_{~{}~{}\alpha_{2}}{F}^{3}_{~{}~{}\alpha_{3}}\Biggr{]}$
$\displaystyle=$ $\displaystyle({\rm
det}~{}g)~{}(1+\frac{a^{2}}{4}F^{a\mu\nu}F^{a}_{\mu\nu})~{},~{}\,$
where $a=2\pi\alpha^{\prime}$.
## References
* (1) J. H. Schwarz, JHEP 0411, 078 (2004).
* (2) A. Basu and J. A. Harvey, Nucl. Phys. B 713, 136 (2005).
* (3) J. Bagger and N. Lambert, Phys. Rev. D 75, 045020 (2007).
* (4) J. Bagger and N. Lambert, Phys. Rev. D 77, 065008 (2008).
* (5) J. Bagger and N. Lambert, JHEP 0802, 105 (2008).
* (6) A. Gustavsson, arXiv:0709.1260 [hep-th].
* (7) A. Gustavsson, JHEP 0804, 083 (2008).
* (8) S. Mukhi and C. Papageorgakis, JHEP 0805, 085 (2008).
* (9) M. A. Bandres, A. E. Lipstein and J. H. Schwarz, JHEP 0805, 025 (2008).
* (10) D. S. Berman, L. C. Tadrowski and D. C. Thompson, arXiv:0803.3611 [hep-th].
* (11) M. Van Raamsdonk, arXiv:0803.3803 [hep-th].
* (12) A. Morozov, arXiv:0804.0913 [hep-th].
* (13) N. Lambert and D. Tong, arXiv:0804.1114 [hep-th].
* (14) J. Distler, S. Mukhi, C. Papageorgakis and M. Van Raamsdonk, JHEP 0805, 038 (2008).
* (15) U. Gran, B. E. W. Nilsson and C. Petersson, arXiv:0804.1784 [hep-th].
* (16) P. M. Ho, R. C. Hou and Y. Matsuo, arXiv:0804.2110 [hep-th].
* (17) J. Gomis, A. J. Salim and F. Passerini, arXiv:0804.2186 [hep-th].
* (18) E. A. Bergshoeff, M. de Roo and O. Hohm, arXiv:0804.2201 [hep-th].
* (19) K. Hosomichi, K. M. Lee and S. Lee, arXiv:0804.2519 [hep-th].
* (20) G. Papadopoulos, JHEP 0805, 054 (2008).
* (21) J. P. Gauntlett and J. B. Gutowski, arXiv:0804.3078 [hep-th].
* (22) G. Papadopoulos, arXiv:0804.3567 [hep-th].
* (23) P. M. Ho and Y. Matsuo, arXiv:0804.3629 [hep-th].
* (24) J. Gomis, G. Milanesi and J. G. Russo, arXiv:0805.1012 [hep-th].
* (25) S. Benvenuti, D. Rodriguez-Gomez, E. Tonni and H. Verlinde, arXiv:0805.1087 [hep-th].
* (26) P. M. Ho, Y. Imamura and Y. Matsuo, arXiv:0805.1202 [hep-th].
* (27) A. Morozov, arXiv:0805.1703 [hep-th].
* (28) Y. Honma, S. Iso, Y. Sumitomo and S. Zhang, arXiv:0805.1895 [hep-th].
* (29) H. Fuji, S. Terashima and M. Yamazaki, arXiv:0805.1997 [hep-th].
* (30) P. M. Ho, Y. Imamura, Y. Matsuo and S. Shiba, arXiv:0805.2898 [hep-th].
* (31) C. Krishnan and C. Maccaferri, arXiv:0805.3125 [hep-th].
* (32) Y. Song, arXiv:0805.3193 [hep-th].
* (33) I. Jeon, J. Kim, N. Kim, S. W. Kim and J. H. Park, arXiv:0805.3236 [hep-th].
* (34) M. Li and T. Wang, arXiv:0805.3427 [hep-th].
* (35) K. Hosomichi, K. M. Lee, S. Lee, S. Lee and J. Park, arXiv:0805.3662 [hep-th].
* (36) S. Banerjee and A. Sen, arXiv:0805.3930 [hep-th].
* (37) H. Lin, arXiv:0805.4003 [hep-th].
* (38) J. Figueroa-O’Farrill, P. de Medeiros and E. Mendez-Escobar, arXiv:0805.4363 [hep-th].
* (39) A. Gustavsson, arXiv:0805.4443 [hep-th].
* (40) M. A. Bandres, A. E. Lipstein and J. H. Schwarz, arXiv:0806.0054 [hep-th].
* (41) J. H. Park and C. Sochichiu, arXiv:0806.0335 [hep-th].
* (42) F. Passerini, arXiv:0806.0363 [hep-th].
* (43) J. Gomis, D. Rodriguez-Gomez, M. Van Raamsdonk and H. Verlinde, arXiv:0806.0738 [hep-th].
* (44) O. Aharony, O. Bergman, D. L. Jafferis and J. Maldacena, arXiv:0806.1218 [hep-th].
* (45) C. Ahn, arXiv:0806.1420 [hep-th].
* (46) M. Benna, I. Klebanov, T. Klose and M. Smedback, arXiv:0806.1519 [hep-th].
* (47) B. Ezhuthachan, S. Mukhi and C. Papageorgakis, arXiv:0806.1639 [hep-th].
* (48) S. Cocetti and A. Sen arXiv:0806.1990 [hep-th].
* (49) A. Mauri and A. C. Petkou, arXiv:0806.2270 [hep-th].
* (50) E. A. Bergshoeff, M. de Roo, O. Hohm and D. Roest, arXiv:0806.2584 [hep-th].
* (51) P. de Medeiros, J. Figueroa-O’Farrill and E. Mendez-Escobar, arXiv:0806.3242 [hep-th].
* (52) M. Blau and M. O’Loughlin, arXiv:0806.3253 [hep-th].
* (53) T. Nishioka and T. Takayanagi, arXiv:0806.3391 [hep-th].
* (54) Y. Honma, S. Iso, Y. Sumitomo and S. Zhang, arXiv:0806.3498 [hep-th].
* (55) C. Sochichiu, arXiv:0806.3520 [hep-th].
* (56) Y. Imamura and K. Kimura, arXiv:0806.3727 [hep-th].
* (57) J. A. Minahan and K. Zarembo, arXiv:0806.3951 [hep-th].
* (58) T. a. Larsson, arXiv:0806.4039 [hep-th].
* (59) K. Furuuchi, S. Y. Shih and T. Takimi, arXiv:0806.4044 [hep-th].
* (60) A. Armoni and A. Naqvi, arXiv:0806.4068 [hep-th].
* (61) A. Agarwal, arXiv:0806.4292 [hep-th].
* (62) D. Gaiotto, S. Giombi and X. Yin, arXiv:0806.4589 [hep-th].
* (63) I. A. Bandos and P. K. Townsend, arXiv:0806.4777 [hep-th].
* (64) C. Ahn, arXiv:0806.4807 [hep-th].
* (65) J. Bedford and D. Berman, arXiv:0806.4900 [hep-th].
* (66) G. Arutyunov and S. Frolov, arXiv:0806.4940 [hep-th].
* (67) B. j. Stefanski, arXiv:0806.4948 [hep-th].
* (68) G. Grignani, T. Harmark and M. Orselli, arXiv:0806.4959 [hep-th].
* (69) K. Hosomichi, K. M. Lee, S. Lee, S. Lee and J. Park, arXiv:0806.4977 [hep-th].
* (70) P. Fre and P. A. Grassi, arXiv:0807.0044 [hep-th].
* (71) K. Okuyama, arXiv:0807.0047 [hep-th].
* (72) J. Bagger and N. Lambert, arXiv:0807.0163 [hep-th].
* (73) S. Terashima, arXiv:0807.0197 [hep-th].
* (74) G. Grignani, T. Harmark, M. Orselli and G. W. Semenoff, arXiv:0807.0205 [hep-th].
* (75) S. Chakrabortty, A. Kumar and S. Jain, arXiv:0807.0284 [hep-th].
* (76) S. Terashima and F. Yagi, arXiv:0807.0368 [hep-th].
* (77) N. Gromov and P. Vieira, arXiv:0807.0437 [hep-th].
* (78) C. Ahn and P. Bozhilov, arXiv:0807.0566 [hep-th].
* (79) N. Gromov and P. Vieira, arXiv:0807.0777 [hep-th].
* (80) B. Chen and J. B. Wu, arXiv:0807.0802 [hep-th].
* (81) S. Cherkis and C. Saemann, arXiv:0807.0808 [hep-th].
* (82) C. Chu, P. M. Ho, Y. Matsuo and S. Shiba, arXiv:0807.0812 [hep-th].
* (83) M. A. Bandres, A. E. Lipstein and J. H. Schwarz, arXiv:0807.0880 [hep-th].
* (84) Y. Zhou, arXiv:0807.0890 [hep-th].
* (85) J. Gomis, D. Rodriguez-Gomez, M. Van Raamsdonk and H. Verlinde, arXiv:0807.1074 [hep-th].
* (86) B. de Wit, H. Nicolai and H. Samtleben, arXiv:hep-th/0403014.
* (87) D. Gaiotto and X. Yin, JHEP 0708, 056 (2007).
* (88) D. Gaiotto and E. Witten, arXiv:0804.2907 [hep-th].
* (89) J. Polchinski, Phys. Rev. Lett. 75, 4724 (1995).
* (90) A. A. Tseytlin, Nucl. Phys. B 501, 41 (1997).
* (91) R. C. Myers, JHEP 9912, 022 (1999).
* (92) R. G. Leigh, Mod. Phys. Lett. A 4, 2767 (1989).
* (93) A. A. Tseytlin, Nucl. Phys. B 276 (1986) 391 [Erratum-ibid. B 291 (1987) 876].
* (94) D. J. Gross and E. Witten, Nucl. Phys. B 277, 1 (1986).
* (95) A. Hashimoto and W. Taylor, Nucl. Phys. B 503, 193 (1997).
* (96) J. H. Schwarz, arXiv:hep-th/0103165, and the references therein.
* (97) E. Bergshoeff and P. K. Townsend, Nucl. Phys. B 490, 145 (1997).
| arxiv-papers | 2008-07-08T07:04:26 | 2024-09-04T02:48:56.615751 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Tianjun Li, Yan Liu, Dan Xie",
"submitter": "Tianjun Li",
"url": "https://arxiv.org/abs/0807.1183"
} |
0807.1193 | # Quantum phonon transport of molecular junctions amide-linked with carbon
nanotubes: a first-principle study
J. T. Lü tower.lu@gmail.com Jian-Sheng Wang Center for Computational Science
and Engineering and Department of Physics, National University of Singapore,
Singapore 117542, Republic of Singapore
(8 July 2008)
###### Abstract
Quantum phonon transport through benzene and alkane chains amide-linked with
single wall carbon nanotubes (SWCNTs) is studied within the level of density
functional theory. The force constant matrices are obtained from standard
quantum chemistry software. The phonon transmission and thermal conductance
are from the nonequilibrium Green’s function and the mode-matching method. We
find that the ballistic thermal conductance is not sensitive to the
compression or stretching of the molecular junction. The terminating groups of
the SWCNTs at the cutting edges only influence the thermal conductance
quantitatively. The conductance of the benzene and alkane chains shows large
difference. Analysis of the transmission spectrum shows that (i) the low
temperature thermal conductance is mainly contributed by the SWCNT transverse
acoustic modes, (ii) the degenerate phonon modes show different transmission
probability due to the presence of molecular junction, (iii) the SWCNT
twisting mode can hardly be transmitted by the alkane chain. As a result, the
ballistic thermal conductance of alkane chains is larger than that of benzene
chains below $38$ K, while it is smaller at higher temperature.
###### pacs:
05.60.Gg, 44.10.+i, 63.22.-m, 65.80.+n
## I Introduction
The electronic transport properties of a molecular device depend much on the
underlying nuclear configuration and the electron-phonon couplingGalperin et
al. (2007). The mutual interaction between them induces Joule heating in the
molecule, which may have important consequences on the functionality and
reliability of a molecular device. Study of the heat transport by phonons in
molecular junctions is crucial for a better understanding of their electronic
transport properties. Phonon thermal transport itself is also interesting.
Detailed understanding of the underlying transport mechanism is especially
useful for the design of novel phononic devicesLi et al. (2004); Chang et al.
(2006). Furthermore, combined study of the electronic and phononic transport
in molecular junctions is the first step toward the design of molecular
thermoelectric devicesReddy et al. (2007). Great experimental progress has
been made in these directions recently, which enables one to measure the
thermal and thermoelectric transport properties of molecular junctionsWang et
al. (2007a); Reddy et al. (2007). Advances in experimental technique call for
new theoretical method to predict the thermal conductance of molecular
junctions. Although semiempirical or minimal model calculation is helpful to
understand the underlying physicsSegal et al. (2003); A. Buldum and Ciraci
(1999). For a detailed quantitative study, a parameter-free, first-principle
method is highly desirableMingo et al. (2008). Furthermore, for many molecular
structures there exists no empirical inter-atom potential.
In this paper we introduce such a method based on a standard quantum chemistry
software, the Gaussian03 parckageFrisch and et al. . Given the molecule
structure, we can obtain the force constant matrices after performing the
energy minimization. The thermal conductance can be calculated via available
theoretical methods. Among them are the nonequilibrium Green’s function (NEGF)
methodWang et al. (2006, 2007b, 2008); Mingo (2006); Yamamoto and Watanabe
(2006), which has been successfully used to predict the electronic conductance
of molecular junctionsTaylor et al. (2001); Brandbyge et al. (2002); Damle et
al. (2001). To study the transport of each phonon mode, we will also use the
mode-matching methodAndo (1991), which is equivalent to the NEGF method in the
ballistic limitKhomyakov et al. (2005). Using these methods, we first compare
the thermal conductance of a benzene ring amide-linked with two $(6,0)$
single-walled carbon nanotubes (SWCNTs) under different compression or
stretching conditions. Then we study the effect of SWCNT terminating group at
the cutting edges. Finally, we compare the phonon transmission probability and
the thermal conductance of benzene and alkane chains with the same leads.
Although the electron contribution to the thermal conductance may be
comparable with or larger than that of phonons, inclusion of this effect is
out of the scope of present study.
## II Molecular structure and theoretical method
In this section we first introduce the system Hamiltonian and the procedure to
obtain it from the Gaussian03 package. Then we briefly outline the NEGF, the
mode-matching method, and their relationship. For a detailed discussion, we
refer the reader to Ref. Wang et al., 2008. The system we are interested in is
a molecular junction connected with two periodic semi-infinite leads at both
sides. It is a standard treatment to divide the whole structure into three
parts: the center junction and two leads acting as thermal baths. The
boundaries between the center and the baths may be at arbitrary positions, and
not correspond to any physical interface. But it is desirable to include part
of the leads into the center region, since we need to make sure that there is
no direct interaction between the two baths, which is required by the NEGF
formalism. By doing this, we can also include the charge transfer effect
between the leads and the molecule junction. In this setup, the system
Hamiltonian can be written as
${\cal
H}=\\!\\!\\!\\!\\!\sum_{\alpha=L,C,R}\\!\\!\\!\\!\\!H_{\alpha}+(u^{L})^{T}V^{LC}u^{C}+(u^{C})^{T}V^{CR}u^{R}+V_{n},$
(1)
where
$H_{\alpha}=\frac{1}{2}{(\dot{u}^{\alpha})}^{T}\dot{u}^{\alpha}+\frac{1}{2}{(u^{\alpha})}^{T}K^{\alpha}u^{\alpha}$
represents harmonic oscillators, $u^{\alpha}$ is a column vector consisting of
all the mass normalised displacement variables in region $\alpha$, and
$\dot{u}^{\alpha}$ is the corresponding conjugate momentum. $K^{\alpha}$ is
the spring constant matrix in the tight-binding form and $V^{LC}=(V^{CL})^{T}$
is the coupling matrix of the left lead to the central region; similarly for
$V^{CR}$. $V_{n}$ is the nonlinear interaction in the center, which could be
$V_{n}=\frac{1}{3}\sum_{ijk}t_{ijk}u_{i}u_{j}u_{k}$ for cubic nonlinearality.
We ignore the nonlinear interaction in this paper and only briefly discuss its
effect in Sec. III.
We study the phonon thermal conductance of benzene and alkane chains
covalently bonded with two $(6,0)$ SWCNTs via the amide group. This is
relevant to a recent experimental setupGuo et al. (2006), where the SWCNT is
oxidatively cutted, and the cutting edges are covalently bonded with molecular
chains via the amide group. To get the force constants of the system, we do
two separate runs for the center and the leads using Gaussian03. For the
center, we include extra one and a half periods of SWCNT at each side, which
is terminated by hydrogen atoms at the outer boundaries. The cutting edges may
be terminated by H or COOH group. We optimize the center at the level of b3lyp
density functional method using the 6-31G basis set. We first relax the
structure by constraining the leads to be coaxial to get an optimized distance
between the two leads. Then we fix the outer half period of the SWCNTs and the
H atoms at both sides and allow all other atoms to move freely. The optimized
structure is used to get the dynamical matrix. The inset of Fig. 2 shows the
optimized benzene structure terminated by the H and COOH groups at the cutting
edges. For the lead, it is desirable to use periodic boundary condition to
compute the dynamical matrix, while this is out of the ability of Gaussian03.
So we optimise $7$ SWCNT periods and extract the force constant from the
central period to minimize the finite size effect. To connect the molecule
with the two leads, we remove the outer fixed atoms in the molecule and
connect the remaining period of SWCNT with one semi-infinite SWCNT at each
side. For the coupling matrix between the center and the leads, we only
include coupling between one period of SWCNT atoms in the center and one
period in the leads. This is a good approximation so long as we include enough
SWCNT atoms into the center molecule.
In the NEGF method as described in Refs. Wang et al., 2006, 2007b, 2008;
Mingo, 2006; Yamamoto and Watanabe, 2006, thermal conductance of the molecular
junction can be calculated from the Landauer formula ($\hbar=1$ throughout the
formula)
$\sigma=\frac{1}{2\pi}\int_{0}^{\infty}\\!\\!\\!d\omega\,\omega\,T[\omega]\frac{\partial
f(\omega)}{\partial T}$ (2)
with the transmission coefficient
$\displaystyle T[\omega]$ $\displaystyle=$ $\displaystyle{\rm
Tr}\bigl{\\{}G^{r}\Gamma_{L}G^{a}\Gamma_{R}\bigr{\\}}.$ (3)
$f(\omega)$ is the Bose-Einstein distribution function. The retarded Green’s
function $G^{r}$ is obtained from
$G^{r}[\omega]=\left((\omega+i0^{+})^{2}-K^{C}-\Sigma^{r}_{L}[\omega]-\Sigma^{r}_{R}[\omega]\right)^{-1},$
(4)
where the retarded self-energy of lead $\alpha$ is
$\Sigma^{r}_{\alpha}[\omega]=V^{C\alpha}g^{r}_{\alpha}[\omega]V^{\alpha C},$
(5)
and the lead surface Green’s function $g^{r}_{\alpha}[\omega]$ can be
calculated from the generalized eigenvalue methodWang et al. (2008), e.g.,
$g^{r}_{R}[\omega]=\left((\omega+i0^{+})^{2}-k^{R}_{11}-k^{R}_{01}F^{+}_{R}(1)\right)^{-1}.$
(6)
$k^{R}_{11}$ and $k^{R}_{01}$ are the diagonal and off-diagonal parts of the
right lead dynamical matrix. Their sizes are determined by the degrees of
freedom $M$ in each period of the lead. The matrix $F^{+}_{R}(s)$ translates
to the right the displacement in the $n$th period to the $(n+s)$th period
$u^{+}_{R}(n+s)=F^{+}_{R}(s)u^{+}_{R}(n)$. It is constructed from the eigen
values and vectors of the generalized eigen value problem
$\left(\begin{array}[]{cc}(\omega+i0^{+})^{2}I\\!\\!-\\!\\!k^{R}_{11}&-I\\\
k^{R}_{10}&0\end{array}\right)\left(\begin{array}[]{c}\epsilon\\\
\zeta\end{array}\right)=\lambda\left(\begin{array}[]{cc}k^{R}_{01}&0\\\
0&I\end{array}\right)\left(\begin{array}[]{c}\epsilon\\\
\zeta\end{array}\right)$ (7)
as $F^{+}_{R}(s)=E^{+}_{R}\Lambda_{+}^{s}(E^{+}_{R})^{-1}$. Here $I$ is an
$M\\!\times\\!M$ identity matrix. The diagonal matrix $\Lambda_{+}^{s}$
consists of all the eigen values $|\lambda_{+}^{s}|<1$, and $E^{+}_{R}$ the
corresponding eigen vectors
$E^{+}_{R}=(\varepsilon^{+}_{1},\varepsilon^{+}_{2},\cdots,\varepsilon^{+}_{M^{\prime}})$.
Note that $M^{\prime}$ may be less than $M$, in which case the matrix inverse
becomes pseudo-inverse. A similar left-translation matrix $F^{-}_{R}(s)$ can
be constructed from $\Lambda_{-}^{s}$ and $E_{R}^{-}$, which include all the
eigen values $|\lambda_{-}^{s}|>1$ (excluding infinity) and the corresponding
eigen vectors.
While the NEGF method is systematic and suitable to take into account the
nonlinear interaction, the transmission coefficient from it is the sum of all
the eigen modes from the leads. Thus it is difficult to analyse the
contribution from each mode. The mode-matching method provides another way to
calculate the transmission coefficient in the ballistic limitAndo (1991);
Khomyakov et al. (2005); Wang et al. (2008). Single mode transmission and mode
mixing effect can be studied by this method. The transmission matrix is given
by
$t^{RL}_{mn}=\sqrt{\frac{v_{R,m}^{+}a_{L}}{v_{L,n}^{+}a_{R}}}\tau^{RL}_{mn},$
(8)
and
$\tau^{RL}=(E^{+}_{R})^{-1}g^{r}_{R}V^{RC}G^{r}V^{CL}g^{r}_{L}k_{10}^{L}(F^{+}_{L}(-1)-F^{-}_{L}(-1))E^{+}_{L},$
(9)
where $v^{+}_{\alpha,m}$ is the group velocity of the $m$th right propagating
mode for the lead $\alpha$. While the mode indices $m$ and $n$ are only for
propagating modes, the matrices $E^{\pm}_{\alpha}$ include all the propagating
and evanescent modes. The total transmission coefficient as in the Landauer
formula is $T[\omega]={\rm Tr}\bigl{\\{}(t^{RL})^{\dagger}t^{RL}\bigr{\\}}$.
Similar relations hold for waves incidented from the right lead and
transmitted to the left $t^{LR}_{nm}$.
The NEGF and the mode-matching method are exactly equivalent in the ballistic
case as show in Ref. Khomyakov et al., 2005. All the matrices needed by Eqs.
(8–9) can be obtained by solving the generalized eigen value problem Eq. (7).
It is interesting to note that by doing a singular value decomposition on the
transmission matrix $t$ we can get the transmission eigenchannel information
without any other efforts, for which different methods have been developed in
the electronic transport literatureInglesfield et al. (2005).
## III Numerical results and discussion
We now present our numerical results. We begin with the simplest case where
there is only one benzene ring at the center. We compare the thermal
conductance of the molecular junction at different compression or stretching
configurations by changing the distance between the two leads. The purpose of
this study is twofold. Firstly, we want to study the effect of the compression
or stretching on the thermal conductance. Secondly, our optimization process
discussed in Sec. II can not ensure that we have reached the lowest energy
configuration while keeping the two SWCNTs to be coaxial. This is because we
have fixed the position of the outer C and H atoms at both sides. If the
thermal conductance is not sensitive to the distance between the two leads,
our results make sense even if we do not find the lowest energy configuration.
As we can see in Fig. 1, this is indeed the case. In Fig. 1, from $1$ to $5$
the distance between the two carbon atoms connecting to the amide groups is
$10.07$, $10.21$, $10.38$, $10.51$, $10.64$ Å, respectively. The inset shows
the atom configuration of the five cases. In the compression states ($10.07$
and $10.21$ Å), the relative position between the benzene ring and the SWCNT
leads changes compared to the full relaxed ($10.38$ Å) and stretching ($10.51$
and $10.64$ Å) states (inset of Fig. 1). Although the atom configurations
change a lot, the ballistic thermal conductance of the molecular junction only
changes slightly in all the temperature range studied here. So we can conclude
that the ballistic thermal conductance of the molecular junction is not
sensitive to small compression or stretching of the molecule.
Figure 1: Ballistic thermal conductance of the benzene junction terminated by
H atoms at different configurations. $3$ is the full relaxed structure. $1-2$
are in the compression state, and $3-4$ are in the stretching state.
In a related experimentGuo et al. (2006), the cutting edge carbon atoms are
expected to be saturated by the COOH group, not the H group shown in Fig. 1.
In Ref. Ren et al., 2007, the authors show that the electron transmission is
largely affected by the terminating groups. It is also interesting to know how
the terminating group affects the quantum thermal conductance. We still use
the single benzene structure to study this problem. Figure 2 shows the thermal
conductance of the two configurations. The full relaxed structures are shown
in the inset. The electrical conductance is mainly determined by the energy
channel near the chemical potential, while the thermal conductance is jointly
contributed by many phonon modes. The terminating group only has a large
influence on the high energy (short wavelength) optical phonon modes. The low
energy (long wavelength) phonon modes are not sensitive to the local
environment at the cutting edges. As a result, the thermal conductance only
shows quantitative difference, which becomes larger at high temperatures.
Figure 2: Ballistic thermal conductance of the benzene junctions terminated
by H and COOH group.
We now compare the conductance of the benzene rings and that of the alkane
chains, with the SWCNTs terminated by H atoms. We expect that the COOH
terminating structures show similar behavior. To ensure the two molecules have
comparable length, we include two benzene rings for the benzene structure and
$8$ CH2 groups for the alkane chain. The total transmission probability and
the thermal conductance are shown in Figs. 3 and 4, respectively. Since the
nonlinear interaction is not considered here, phonon modes with different
energies are independent of each other. The transmission probability is
nonzero only in the overlapped energy range of the SWCNT leads and the center
molecule. Strong coupling with the leads gives rise to wide broadening and
strong shifting of the discrete molecule phonon energy. It is hard to make a
one-to-one correspondence between the transmission peaks and the isolated
molecule phonon eigen frequencies. This is especially true for low energy
phonon modes, which have relatively large spatial extent. The high energy
modes are highly localized and show sharp peaks in the transmission spectrum.
Due to the highly localized nature and low Bose-Einstein weighting, these
modes can hardly transfer energy. There is a wide zero transmission gap around
$0.07$ eV for the alkane chain, which is a characteristic of the alkane chain
phonon spectrumSegal et al. (2003).
Figure 3: Phonon transmission probability as a function of energy for the
benzene and alkane chains. Figure 4: Ballistic thermal conductance as a
function of temperature for the benzene and alkane chains. The inset shows the
crossing point of the thermal conductance at about $38$ K. Figure 5:
Transmission probability of different phonon branches: (a) TA modes, (b) LA
mode, (c) twisting mode, (d) lowest optical modes.
To gain further insight into the transmission spectrum, we also study the
single mode transmission using the mode-matching method. Figure 5 shows the
transmission of several important branches in SWCNT phonon spectrum: Two
transverse acoustic (TA) phonon modes, one longitudinal acoustic (LA) mode,
one twisting mode and the lowest two degenerate optical modes. These low
frequency modes are expected to contribute much to the thermal conductance.
The degenerate phonon modes of SWCNTs show different transmission in both
cases. This is consistent with the fact that the junction structure destroys
the degeneracy in transverse directions. At the energy range below $0.01$ eV,
the main contribution comes from the TA modes of the SWCNT. The alkane chain
shows larger transmission than the benzene chain in this energy range. Above
$0.01$ eV, the benzene chain shows larger transmission in most cases. The
twisting mode shows the largest difference in two structures. While it can
hardly be transmitted by the alkane chain, it has a large contribution in the
benzene chain.
The thermal conductance of the molecular junction depends not only on the
transmission coefficient, but also on the phonon occupation number, which is
reflected as the Bose-Einstein distribution in the Landauer formula. According
the analysis of the transmission spectrum, we may expect that the alkane chain
has larger thermal conductance at low temperatures, while at higher
temperatures the order reverses. This is confirmed in Fig. 4. The crossing
temperature is about $38$ K. The room temperature thermal conductance is about
$0.075$ nW/K for alkane chains and $0.125$ nW/K for benzene chains. In a
recent experimentWang et al. (2007a), the thermal conductance of alkane chain
is found to be smaller than our theoretical value. This difference comes from
the effect of the leads. We are using SWCNTs here, while in the experiment it
is the bulk gold connected with the alkane chain via the sufur atom. At least
two factors from the leads may account for this decrease. The first is the
smaller phonon spectrum overlap between gold and alkane chain, and the second
is the weaker coupling between them.
Some comments are worthwhile. The nonlinear interaction may change the
transmission spectrum, and lead to a decrease of the thermal conductance. A
perturbative analysis of the single benzene structure shows that the room
temperature thermal conductance drops about $30\%$ of the ballistic value if
we include the cubic force constant calculated from Gaussion03. Mean-filed
approximation in the NEGF formalism only works well for simple structures with
relatively weak nonlinear interactionWang et al. (2008). It fails to converge
in the self-consistent iteration in present case. So it is still a challenge
to find a good approximation for the nonlinear self-energies in NEGF method.
## IV Conclusions
In this paper, we introduce a straightforward method to calculate the phononic
thermal conductance of molecular junctions in the ballistic regime from first-
principles. The force constant matrices are obtained from Gaussian03 quantum
chemistry software. The phonon transmission and thermal conductance are
calculated using the NEGF or mode-matching method. Further information can be
obtained from the transmission spectrum of each single mode. Using this method
we show that the ballistic thermal conductance of the benzene ring amide
linked with SWCNTs is not sensitive to the distance between the two SWCNTs.
The benzene rings show larger thermal conductance than the alkane chains. This
method is general, and can be easily applied to other material systems.
###### Acknowledgements.
The authors thank Nan Zeng for discussions. JTL is grateful to the hospitality
of Prof. J. C. Cao at Shanghai Institute of Microsystem and Information
Technology, where this paper was finished. This work was supported in part by
a Faculty Research Grant (R-144-000-173-101/112) of National University of
Singapore.
## References
* Galperin et al. (2007) M. Galperin, M. A. Ratner, and A. Nitzan, J. Phys.: Condens. Matter 19, 103201 (2007).
* Li et al. (2004) B. Li, L. Wang, and G. Casati, Phys. Rev. Lett. 93, 184301 (2004).
* Chang et al. (2006) C. W. Chang, D. Okawa, A. Majumdar, and A. Zettl, Science 314, 1121 (2006).
* Reddy et al. (2007) P. Reddy, S.-Y. Jang, R. Segalman, and A. Majumdar, Science 315, 1568 (2007).
* Wang et al. (2007a) Z. Wang, J. A. Carter, A. Lagutchev, Y. K. Koh, N.-H. Seong, D. G. Cahill, and D. D. Dlott, Science 317, 787 (2007a).
* Segal et al. (2003) D. Segal, A. Nitzan, and P. Hänggi, J. Chem. Phys. 119, 6840 (2003).
* A. Buldum and Ciraci (1999) A. Buldum, D. M. Leitner and S. Ciraci, Europhys. Lett. 47, 208 (1999).
* Mingo et al. (2008) N. Mingo, D. A. Stewart, D. A. Broido, and D. Srivastava, Phys. Rev. B 77, 033418 (2008).
* (9) M. J. Frisch and et al., gaussian 03, Revision C.02 (Gaussian Inc., Wallingford, 2004).
* Wang et al. (2006) J.-S. Wang, J. Wang, and N. Zeng, Phys. Rev. B 74, 033408 (2006).
* Wang et al. (2007b) J.-S. Wang, N. Zeng, J. Wang, and C. K. Gan, Phys. Rev. E 75, 061128 (2007b).
* Mingo (2006) N. Mingo, Phys. Rev. B 74, 125402 (2006).
* Yamamoto and Watanabe (2006) T. Yamamoto and K. Watanabe, Phys. Rev. Lett. 96, 255503 (2006).
* Wang et al. (2008) J.-S. Wang, J. Wang, and J. T. Lü, Eur. Phys. J. B 62, 381 (2008).
* Taylor et al. (2001) J. Taylor, H. Guo, and J. Wang, Phys. Rev. B 63, 245407 (2001).
* Brandbyge et al. (2002) M. Brandbyge, J.-L. Mozos, P. Ordejón, J. Taylor, and K. Stokbro, Phys. Rev. B 65, 165401 (2002).
* Damle et al. (2001) P. S. Damle, A. W. Ghosh, and S. Datta, Phys. Rev. B 64, 201403 (2001).
* Ando (1991) T. Ando, Phys. Rev. B 44, 8017 (1991).
* Khomyakov et al. (2005) P. A. Khomyakov, G. Brocks, V. Karpan, M. Zwierzycki, and P. J. Kelly, Phys. Rev. B 72, 035450 (2005).
* Guo et al. (2006) X. Guo, J. P. Small, J. E. Klare, Y. Wang, M. S. Purewal, I. W. Tam, B. H. Hong, R. Caldwell, L. Huang, S. O’Brien, et al., Science 311, 356 (2006).
* Inglesfield et al. (2005) J. E. Inglesfield, S. Crampin, and H. Ishida, Phys. Rev. B 71, 155120 (2005), A. Bagrets, N. Papanikolaou, and I. Mertig, Phys. Rev. B 75, 235448 (2007), M. Paulsson and M. Brandbyge, Phys. Rev. B 76, 115117 (2007).
* Ren et al. (2007) W. Ren, J. Reimers, N. Hush, Y. Zhu, J. Wang, and H. Guo, J. Phys. Chem. C 111, 3700 (2007).
| arxiv-papers | 2008-07-08T08:57:42 | 2024-09-04T02:48:56.621422 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "J. T. L\\\"u and Jian-Sheng Wang",
"submitter": "L\\\"u Jing Tao",
"url": "https://arxiv.org/abs/0807.1193"
} |
0807.1214 | # The Rank of the Endomorphism Monoid of a Partition
João Araújo and Csaba Schneider Informatics Research Laboratory
Computer and Automation Research Institute
1518 Budapest Pf. 63
Hungary Centro de Álgebra da Universidade de Lisboa
Av. Prof. Gama Pinto 2
1649-003 Lisboa
Portugal Email: csaba.schneider@gmail.com WWW: www.sztaki.hu/$\sim$schneider
Universidade Aberta
R. Escola Politécnica 147
1269-001 Lisboa
Portugal Centro de Álgebra
Universidade de Lisboa
1649-003 Lisboa
Portugal Email: jaraujo@lmc.fc.ul.pt
###### Abstract.
The rank of a semigroup is the cardinality of a smallest generating set. In
this paper we compute the rank of the endomorphism monoid of a non-trivial
uniform partition of a finite set, that is, the semigroup of those
transformations of a finite set that leave a non-trivial uniform partition
invariant. That involves proving that the rank of a wreath product of two
symmetric groups is two and then use the fact that the endomorphism monoid of
a partition is isomorphic to a wreath product of two full transformation
semigroups. The calculation of the rank of these semigroups solves an open
question.
###### Key words and phrases:
transformation semigroups, rank, relative rank, wreath product, symmetric
groups
###### 1991 Mathematics Subject Classification:
20M20, 20B30, 20E22
## 1\. Introduction
If $S$ is a semigroup and $U$ is a subset of $S$ then we say that $U$
generates $S$ if every element of $S$ is expressible as a word in the elements
of $U$. We use the convention that the empty word represents the identity
element. The rank of a semigroup $S$, denoted by $\mbox{\emph{\mbox{\rm
rank}\,}}S$, is the minimum among the cardinalities of its generating sets. It
is well-known that a finite full transformation semigroup has rank 3, while a
finite full partial transformation semigroup has rank 4 (see [14, Exercises
1.9.7 and 1.9.13]). Similar results were proved for many different classes of
transformation semigroups (such as total, partial, partial one-to-one, order
preserving) and their ideals; see [4, 10, 15, 20, 22].
Some generalizations of the notion of rank, for instance the idempotent rank
and the nilpotent rank, also attracted a great deal of attention (see [5, 9,
19], among others).
Finally, in recent years, the notion of relative rank has been subjected to
extensive research (see for example [7, 13, 11, 12, 16]). Relative rank is a
useful notion when dealing with finite semigroups (see Lemma 3.1), and it is
crucial when dealing with uncountable semigroups. In fact, in such semigroups,
the notion of rank is not very informative as the rank and the cardinality
coincide. To a large extent, this line of research was prompted by two old
papers by Sierpiński [24] and by Banach [3] (see also [2]).
In this paper we deal with the endomorphism monoid of a non-trivial uniform
partition. We prove that the rank of such semigroup is 4, thus settling a
problem posed in [17]. We also calculate the ranks of some related
transformation semigroups.
If $X$ is a finite set then the set of transformations on $X$ form the full
transformation monoid on $X$ and is denoted by ${\sf T}(X)$. We assume that
transformations act on the right; that is, if $x\in X$ and $f\in{\sf T}(X)$,
then $xf$ denotes the image of $x$ under $f$. Let $\mathcal{P}$ be a partition
of $X$; that is, $\mathcal{P}=\\{P_{1},\ldots,P_{m}\\}$ where
$P_{1},\ldots,P_{m}\subseteq X$, $P_{i}\cap P_{j}=\emptyset$ whenever $i\neq
j$, and $X=P_{1}\cup\cdots\cup P_{m}$. The equivalence relation that
corresponds to a partition $\mathcal{P}$ is denoted $\sim_{\mathcal{P}}$. The
elements of ${\sf T}(X)$ that preserve the partition $\mathcal{P}$ form a
semigroup and is denoted by ${\sf T}(X,\mathcal{P})$. Using symbols,
$\begin{array}[]{rcl}{\sf T}(X,\mathcal{P})&=&\\{f\in{\sf
T}(X)\mid(\forall{P_{i}\in\mathcal{P}})(\exists{P_{j}\in\mathcal{P}})\
P_{i}f\subseteq P_{j}\\}\\\ &=&\\{f\in{\sf T}(X)\mid\mbox{if
}x\sim_{\mathcal{P}}y\mbox{ then }xf\sim_{\mathcal{P}}yf\\}.\end{array}$
The partition $\mathcal{P}$ is called uniform if $|P_{i}|=|P_{j}|$ for all
$i,\ j\in\\{1,\ldots,m\\}$. The main result of Huisheng’s paper [17] is that
for a uniform partition $\mathcal{P}$ the rank of ${\sf T}(X,\mathcal{P})$ is
at most 6, or smaller in some degenerate cases. Huisheng’s proof relied on the
observation that there is a strong relationship between $\mbox{\emph{\mbox{\rm
rank}\,}}{\sf T}(X,\mathcal{P})$ and the rank of the group $G$ of invertible
transformations in ${\sf T}(X,\mathcal{P})$. He proved that the rank of $G$ is
at most 4 (or smaller in some degenerate cases). In the present paper, we are
able to show, for $|X|\geqslant 3$, that $\mbox{\emph{\mbox{\rm rank}\,}}G=2$
(see Theorem 4.1). In order to facilitate the proof of our results, we use the
concept of wreath products of transformation semigroups; see Section 2.
We can also consider the rank of some further interesting semigroups related
to a partition $\mathcal{P}$ of a set $X$. Let us define
$\begin{array}[]{rcl}\Sigma(X,\mathcal{P})&=&\\{f\in{\sf T}(X)\ |\
x\sim_{\mathcal{P}}y\mbox{ if and only if }xf\sim_{\mathcal{P}}yf\\},\mbox{
and}\\\ \Gamma(X,\mathcal{P})&=&\\{f\in{\sf
T}(X,\mathcal{P})\mid(\forall{P_{i},\ P_{j}\in\mathcal{P}})\ \mbox{ either
}P_{i}f\cap P_{j}\neq\emptyset\mbox{ or }P_{i}f=P_{j}\\}.\\\ \end{array}$
It is routine to check that $\Sigma(X,\mathcal{P}),\
\Gamma(X,\mathcal{P})\leqslant{\sf T}(X,\mathcal{P})$.
The following theorem improves Huisheng’s result and settles the problem of
determining the rank of ${\sf T}(X,\mathcal{P})$. It also gives the rank of
$\Sigma(X,\mathcal{P})$ and $\Gamma(X,\mathcal{P})$. A partition of $X$ is
said to be trivial if it has $1$ or $|X|$ parts (that is, the trivial
partitions are the identity and the universal partitions).
###### Theorem 1.1.
If $X$ is a finite set such that $|X|\geqslant 3$, and $\mathcal{P}$ is a non-
trivial uniform partition of $X$, then $\mbox{\emph{\mbox{\rm rank}\,}}{\sf
T}(X,\mathcal{P})=4$ and $\mbox{\emph{\mbox{\rm
rank}\,}}\Sigma(X,\mathcal{P})=\mbox{\emph{\mbox{\rm
rank}\,}}\Gamma(X,\mathcal{P})=3$.
## 2\. Wreath products of transformation semigroups
We define the wreath product of two transformation semigroups following [25]
(see also [21, Chapter 10]). The material of this section is well-known,
however, we felt that it was necessary to present it in order to make the
paper self-contained and also to set our system of notation. Let $S$ and $R$
be two transformation semigroups acting on the sets $Y$ and $Z$, respectively,
and let $B$ denote the set of functions $f:Z\rightarrow S$. The underlying set
of the wreath product $S\,{\sf wr}\,R$ is the Cartesian product $B\times R$
and to each element $(f,r)\in B\times R$, we assign a transformation of the
set $Y\times Z$ defined by
(1) $(y,z)(f,r)=(yf(z),zr)\quad\mbox{for all}\quad y\in Y,\ z\in Z,\ f\in B,\
r\in R.$
It is easy to see that this assignment is injective and that the image of this
assignment is closed under composition; that is, the image is a subsemigroup
of ${\sf T}(Y\times Z)$.
From now on we assume that the sets $Y$ and $Z$ are finite and that $S$ and
$R$ are monoids, that is, they contain the identity transformation ${\rm id}$.
Let us assume without loss of generality that $Y=\\{1,\ldots,n\\}$ and
$Z=\\{1,\ldots,m\\}$. In this case, a function $f:Z\rightarrow S$ can be
represented as the $m$-tuple $(s_{1},\ldots,s_{m})$ where $f(i)=s_{i}$ for all
$i\in\\{1,\ldots,m\\}$. This defines a bijection between $B\times R$ and
$S^{m}\times R$, and, in turn, between $S\,{\sf wr}\,R$ and $S^{m}\times R$.
Therefore we may view an element of $S\,{\sf wr}\,R$ as a pair
$((s_{1},\ldots,s_{m}),r)$ where $s_{1},\ldots,s_{m}\in S$ and $r\in R$. The
element $((s_{1},\ldots,s_{m}),r)$ will more briefly be denoted by
$(s_{1},\ldots,s_{m})r$. Setting $s=(s_{1},\ldots,s_{m})$ the same element can
be written as $sr$. By (1), the action of $(s_{1},\ldots,s_{m})r$ on a pair
$(y,z)\in Y\times Z$ is given by
(2) $(y,z)(s_{1},\ldots,s_{m})r=(ys_{z},zr).$
Let $S\leqslant{\sf T}(Y)$ and $R\leqslant{\sf T}(Z)$ be transformation
monoids as above. Then $S$ and $R$ can naturally be embedded into the wreath
product $S\,{\sf wr}\,R$. Indeed, we may consider the following submonoids of
$S\,{\sf wr}\,R$:
(3) $\overline{S}_{i}=\\{({\rm id},\ldots,{\rm
id},\stackrel{{\scriptstyle\mbox{\footnotesize$i$-th component}}}{{s}},{\rm
id}\ldots,{\rm id}){\rm id}\in S\,{\sf wr}\,R\ |\ s\in S\\},$
and
(4) $\overline{R}=\\{({\rm id},\ldots,{\rm id})r\in S\,{\sf wr}\,R\ |\ r\in
R\\}.$
Set, for $i=\\{1,\ldots,m\\}$, $\overline{Y}_{i}=\\{(y,i)\ |\ y\in Y\\}$. Easy
computation shows that the elements of $\overline{S}_{i}$ leave the set
$\overline{Y}_{i}$ invariant, and $\overline{S}_{i}$, considered as a
transformation monoid on $\overline{Y}_{i}$, is isomorphic to the
transformation monoid $S$. Let us further define
$\overline{Z}=\\{\overline{Y}_{1},\ldots,\overline{Y}_{m}\\}$. Then
$\overline{R}$ leaves $\overline{Z}$ invariant, and $\overline{R}$, considered
as a transformation monoid on $\overline{Z}$, is isomorphic to $R$. Since
$\overline{S}_{1}\times\cdots\times\overline{S}_{m}\cong S^{m}$ and
$\overline{R}\cong R$, we may consider $S^{m}$ and $R$ as submonoids of
$S\,{\sf wr}\,R$.
Since $R$ is assumed to be a transformation monoid on the set
$Z=\\{1,\ldots,m\\}$, we may define a homomorphism $\vartheta:R\rightarrow
S^{m}$:
$(r\vartheta)(s_{1},\ldots,s_{m})=(s_{1r},\ldots,s_{mr}).$
We note that the action of $\mbox{End}(S^{m})$ on $S^{m}$ is a left-action.
The homomorphism $\vartheta$ is useful for expressing the operation in
$S\,{\sf wr}\,R$. Indeed, let $s_{1},\ s_{2}\in S^{m}$ and $r_{1},\ r_{2}\in
R$. Then, viewing $s_{1}r_{1}$ and $s_{2}r_{2}$ as elements of $S\,{\sf
wr}\,R$, easy computation shows that
(5) $(s_{1}r_{1})(s_{2}r_{2})=(s_{1}(r_{1}\vartheta)s_{2})(r_{1}r_{2}).$
Therefore the wreath product $S\,{\sf wr}\,R$ can also be viewed as the
semidirect product $S^{m}\rtimes R$ of $S^{m}$ and $R$ with respect to the
homomorphism $\vartheta:R\rightarrow\mbox{End}(S^{m})$ (see [21, page 186]).
In particular, if $r$ is an invertible element of $R$ and $s\in S^{m}$, then,
considering $s$ and $r$ as elements of $S\,{\sf wr}\,R$, equation (5) implies
that
(6) $rsr^{-1}=(r\vartheta)s.$
Therefore conjugation by an invertible element of $R$ leaves $S^{m}$
invariant, and the conjugation action of $R$ is given by the homomorphism
$\vartheta$. Indeed, from the last displayed line we obtain that
$s^{r}=r^{-1}sr=(r^{-1}\vartheta)s$. Note that the conjugation action of the
element $r$ is actually given by the endomorphism $r^{-1}\vartheta$. The
reason for this is that we assumed that endomorphisms act on the left, while
the usual definition makes the conjugation action a right-action. Further, if
$r$ is an invertible element of $R$ and $i,\ j\in\\{1,\ldots,m\\}$ such that
$ir=j$ then (6) implies that
$\overline{S}_{i}^{r}=(r^{-1}\vartheta)\overline{S}_{i}=\overline{S}_{j}$.
For a finite set $X$, the invertible elements of ${\sf T}(X)$ form the
symmetric group ${\sf S}{(X)}$. If $\mathcal{P}$ is a partition of $X$ then
set ${\sf S}{(X,\mathcal{P})}={\sf T}(X,\mathcal{P})\cap{\sf S}{(X)}$. The
next lemma connects the semigroups related to a partition to wreath products.
###### Lemma 2.1.
Let $Y=\\{1,\ldots,n\\}$ and $Z=\\{1,\ldots,m\\}$, set $X=Y\times Z$, and let
$\mathcal{P}$ denote the partition
$\\{\\{(1,1),\ldots,(n,1)\\},\ldots,\\{(1,m),\ldots,(n,m)\\}\\}$. Then
1. (i)
${\sf T}(X,\mathcal{P})={\sf T}(Y)\,{\sf wr}\,{\sf T}(Z)$;
2. (ii)
$\Sigma(X,\mathcal{P})={\sf T}(Y)\,{\sf wr}\,{\sf S}{(Z)}$;
3. (iii)
$\Gamma(X,\mathcal{P})={\sf S}{(Y)}\,{\sf wr}\,{\sf T}(Z)$;
4. (iv)
${\sf S}{(X,\mathcal{P})}={\sf S}{(Y)}\,{\sf wr}\,{\sf S}{(Z)}$.
###### Proof.
Since the proofs of these statements are very similar to each other, we only
verify assertion (ii). Let $(y_{1},z_{1}),\ (y_{2},z_{2})\in Y\times Z$ and
assume that $t_{1},\ldots,t_{m}\in{\sf T}(Y)$ $s\in{\sf S}{(Z)}$ so that
$(t_{1},\ldots,t_{m})s\in{\sf T}(Y)\,{\sf wr}\,{\sf S}{(Z)}$. Set
$w=(t_{1},\ldots,t_{m})s$. We have, for $i=1,\ 2$, that
$(y_{i},z_{i})w=(y_{i},z_{i})(t_{1},\ldots,t_{m})s=(y_{i}t_{z_{i}},z_{i}s).$
Thus, if $(y_{1},z_{1})\sim_{\mathcal{P}}(y_{2},z_{2})$, then $z_{1}=z_{2}$,
and then $z_{1}s=z_{2}s$; hence, in this case,
$(y_{1},z_{1})w\sim_{\mathcal{P}}(y_{2},z_{2})w$. Conversely, if
$(y_{1},z_{1})w\sim_{\mathcal{P}}(y_{2},z_{2})w$, then $z_{1}s=z_{2}s$, which,
using that $s$ is a invertible, gives that $z_{1}=z_{2}$; therefore
$(y_{1},z_{1})\sim_{\mathcal{P}}(y_{2},z_{2})$. This shows that
$w\in\Sigma(X,\mathcal{P})$, and so ${\sf T}(Y)\,{\sf wr}\,{\sf
S}{(Z)}\leqslant\Sigma(X,\mathcal{P})$.
Suppose now that $x\in\Sigma(X,\mathcal{P})$. Then the defining property of
$\Sigma(X,\mathcal{P})$ implies that $x$ induces a permutation on the set
$\mathcal{P}$. Since there is a one-to-one correspondence between
$\mathcal{P}$ and $Z=\\{1,\ldots,m\\}$, we obtain that $x$ induces a
permutation on $Z$. Let this permutation be $s$. For $i=1,\ldots,m$, let us
define a transformation $t_{i}\in{\sf T}(Y)$. Let $j\in\\{1,\ldots,m\\}$ such
that $ix=j$. Then, for all $k\in\\{1,\ldots,n\\}$ there is some
$l_{k}\in\\{1,\ldots,n\\}$ such that $(k,i)x=(l_{k},j)$. Let $t_{i}$ be the
transformation that maps $k$ to $l_{k}$ for all $k\in\\{1,\ldots,n\\}$. Then
routine computation shows that $x=(t_{1},\ldots,t_{m})s$, and so $x\in{\sf
T}(Y)\,{\sf wr}\,{\sf S}{(Z)}$. Thus, $\Sigma(X,\mathcal{P})\leqslant{\sf
T}(Y)\,{\sf wr}\,{\sf S}{(Z)}$, and so ${\sf T}(Y)\,{\sf wr}\,{\sf
S}{(Z)}=\Sigma(X,\mathcal{P})$ ∎
## 3\. Relative rank of semigroups
Let $U\subseteq S$ be a subset of a semigroup $S$. The relative rank of $S$
modulo $U$, denoted by $\mbox{\emph{\mbox{\rm rank}\,}}{(S:U)}$, is the
minimum among the cardinalities of subsets $V$ of $S$ such that $S=\langle
V\cup U\rangle$. The relative rank was introduced in [13]. The next lemma
shows that the rank of a transformation semigroup is related to its relative
rank modulo the unit group.
###### Lemma 3.1.
Let $S$ be a finite transformation semigroup and let $G$ be the group of units
in $S$. If $U\subseteq S$ such that $\bigl{<}U\bigr{>}=S$, then $\bigl{<}U\cap
G\bigr{>}=G$. In particular, $\mbox{\emph{\mbox{\rm
rank}\,}}S=\mbox{\emph{\mbox{\rm rank}\,}}{(S:G)}+\mbox{\emph{\mbox{\rm
rank}\,}}G$.
###### Proof.
It suffices to show that $G\leqslant\bigl{<}U\cap G\bigr{>}$, and so suppose
that $g\in G$. Since $U$ is a generating set of $S$, we obtain that
$g=u_{1}u_{2}\cdots u_{r}$ with some $u_{1},\ldots,u_{r}\in U$. Since $g$ is
invertible, we obtain that $u_{1},\ldots,u_{r}$ must also be invertible, and
so $u_{1},\ldots,u_{r}\in G$. Thus $g\in\bigl{<}U\cap G\bigr{>}$, and hence
the assertion follows. ∎
Next we determine the relative ranks of ${\sf T}(X,\mathcal{P})$,
$\Sigma(X,\mathcal{P})$ and $\Gamma(X,\mathcal{P})$ modulo their unit groups.
###### Lemma 3.2.
If $X$ is a finite set and $\mathcal{P}$ is a uniform partition of $X$, then
$\mbox{\emph{\mbox{\rm rank}\,}}{({\sf T}(X,\mathcal{P}):{\sf
S}{(X,\mathcal{P})})}=2,$
and
$\mbox{\emph{\mbox{\rm rank}\,}}{(\Gamma(X,\mathcal{P}):{\sf
S}{(X,\mathcal{P})})}=\mbox{\emph{\mbox{\rm
rank}\,}}{(\Sigma(X,\mathcal{P}):{\sf S}{(X,\mathcal{P})})}=1.$
###### Proof.
We may suppose without loss of generality that $X=Y\times Z$ where
$Y=\\{1,\ldots,n\\}$, $Z=\\{1,\ldots,m\\}$ and $\mathcal{P}$ is the partition
$\\{\\{(1,1),\ldots,(n,1)\\},\ldots,\\{(1,m),\ldots,(n,m)\\}\\}.$
By Lemma 2.1, ${\sf T}(X,\mathcal{P})={\sf T}(Y)\,{\sf wr}\,{\sf T}(Z)$,
$\Sigma(X,\mathcal{P})={\sf T}(Y)\,{\sf wr}\,{\sf S}{(Z)}$,
$\Gamma(X,\mathcal{P})={\sf S}{(Y)}\,{\sf wr}\,{\sf T}(Z)$, and ${\sf
S}{(X,\mathcal{P})}={\sf S}{(Y)}\,{\sf wr}\,{\sf S}{(Z)}$, and so it suffices
to show that
(7) $\mbox{\emph{\mbox{\rm rank}\,}}{({\sf T}(Y)\,{\sf wr}\,{\sf T}(Z):{\sf
S}{(Y)}\,{\sf wr}\,{\sf S}{(Z)})}=2$
and
(8) $\mbox{\emph{\mbox{\rm rank}\,}}{({\sf T}(Y)\,{\sf wr}\,{\sf S}{(Z)}:{\sf
S}{(Y)}\,{\sf wr}\,{\sf S}{(Z)})}=\mbox{\emph{\mbox{\rm rank}\,}}{({\sf
S}{(Y)}\,{\sf wr}\,{\sf T}(Z):{\sf S}{(Y)}\,{\sf wr}\,{\sf S}{(Z)})}=1.$
Let $\overline{\alpha}$ denote the transformation in ${\sf T}(Y)$ such that
$1\overline{\alpha}=2$ and $i\overline{\alpha}=i$ for all
$i\in\\{2,\ldots,n\\}$. Then ${\sf T}(Y)=\bigl{<}{\sf
S}{(Y)}\cup\\{\overline{\alpha}\\}\bigr{>}$ (see [14, Exercise 1.9.7]). Set
$\alpha=(\overline{\alpha},{\rm id},\ldots,{\rm id}){\rm id}$ and $\beta=({\rm
id},\ldots,{\rm id})\overline{\alpha}$. We claim that
(9) $\displaystyle{\sf T}(Y)\,{\sf wr}\,{\sf S}{(Z)}$ $\displaystyle=$
$\displaystyle\bigl{<}{\sf S}{(Y)}\,{\sf wr}\,{\sf
S}{(Z)}\cup\\{\alpha\\}\bigr{>},$ (10) $\displaystyle{\sf S}{(Y)}\,{\sf
wr}\,{\sf T}(Z)$ $\displaystyle=$ $\displaystyle\bigl{<}{\sf S}{(Y)}\,{\sf
wr}\,{\sf S}{(Z)}\cup\\{\beta\\}\bigr{>},$ (11) $\displaystyle{\sf T}(Y)\,{\sf
wr}\,{\sf T}(Z)$ $\displaystyle=$ $\displaystyle\bigl{<}{\sf S}{(Y)}\,{\sf
wr}\,{\sf S}{(Z)}\cup\\{\alpha,\beta\\}\bigr{>}.$
For $i\in\\{1,\ldots,m\\}$, let us define the following submonoids of ${\sf
T}(Y)\,{\sf wr}\,{\sf T}(Z)$:
$\displaystyle\overline{T}_{i}$ $\displaystyle=$ $\displaystyle\\{({\rm
id},\ldots,{\rm id},\stackrel{{\scriptstyle\mbox{\footnotesize$i$-th
component}}}{{t}},{\rm id},\ldots,{\rm id}){\rm id}\in{\sf T}(Y)\,{\sf
wr}\,{\sf T}(Z)\ |\ t\in{\sf T}(Y)\\},$ $\displaystyle\overline{S}_{i}$
$\displaystyle=$ $\displaystyle\\{({\rm id},\ldots,{\rm
id},\stackrel{{\scriptstyle\mbox{\footnotesize$i$-th component}}}{{s}},{\rm
id}\ldots,{\rm id}){\rm id}\in{\sf T}(Y)\,{\sf wr}\,{\sf T}(Z)\ |\ s\in{\sf
S}{(Y)}\\},$ $\displaystyle\overline{{\sf T}(Z)}$ $\displaystyle=$
$\displaystyle\\{({\rm id},\ldots,{\rm id})t\in{\sf T}(Y)\,{\sf wr}\,{\sf
T}(Z)\ |\ t\in{\sf T}(Z)\\},$ $\displaystyle\overline{{\sf S}{(Z)}}$
$\displaystyle=$ $\displaystyle\\{({\rm id},\ldots,{\rm id})s\in{\sf
T}(Y)\,{\sf wr}\,{\sf T}(Z)\ |\ s\in{\sf S}{(Z)}\\}.$
Let us first prove (9). As $\alpha\in{\sf T}(Y)\,{\sf wr}\,{\sf S}{(Z)}$, we
find that $\bigl{<}{\sf S}{(Y)}\,{\sf wr}\,{\sf
S}{(Z)}\cup\\{\alpha\\}\bigr{>}\leqslant{\sf T}(Y)\,{\sf wr}\,{\sf S}{(Z)}$.
Since ${\sf T}(Y)=\bigl{<}{\sf S}{(Y)}\cup\\{\overline{\alpha}\\}\bigr{>}$, we
obtain that
$\overline{T}_{1}=\bigl{<}\overline{S}_{1}\cup\\{\alpha\\}\bigr{>}$, and hence
$\overline{T}_{1}\leqslant\bigl{<}{\sf S}{(Y)}\,{\sf wr}\,{\sf
S}{(Z)}\cup\\{\alpha\\}\bigr{>}$. For all $i\in\\{1,\ldots,m\\}$, there is
some $r\in{\sf S}{(Z)}$ such that $1r=i$, and, as discussed before Lemma 2.1,
we obtain that $r^{-1}\overline{T}_{1}r=\overline{T}_{i}$. Therefore
$\overline{T}_{1},\ldots,\overline{T}_{m}\leqslant\bigl{<}{\sf S}{(Y)}\,{\sf
wr}\,{\sf S}{(Z)}\cup\\{\alpha\\}\bigr{>}$. As $\overline{{\sf
S}{(Z)}}\leqslant\bigl{<}{\sf S}{(Y)}\,{\sf wr}\,{\sf
S}{(Z)}\cup\\{\alpha\\}\bigr{>}$ and ${\sf T}(Y)\,{\sf wr}\,{\sf
S}{(Z)}=(\overline{T}_{1}\times\cdots\times\overline{T}_{m})\rtimes\overline{{\sf
S}{(Z)}}$, we obtain that ${\sf T}(Y)\,{\sf wr}\,{\sf
S}{(Z)}\leqslant\bigl{<}{\sf S}{(Y)}\,{\sf wr}\,{\sf
S}{(Z)}\cup\\{\alpha\\}\bigr{>}$, and so the required equality holds.
Now we show that (10). As $\beta\in{\sf S}{(Y)}\,{\sf wr}\,{\sf T}(Z)$, we
have that $\bigl{<}{\sf S}{(Y)}\,{\sf wr}\,{\sf
S}{(Z)}\cup\\{\beta\\}\bigr{>}\leqslant{\sf S}{(Y)}\,{\sf wr}\,{\sf T}(Z)$. As
${\sf T}(Y)=\bigl{<}{\sf S}{(Y)}\cup\\{\overline{\alpha}\\}\bigr{>}$, we
obtain that $\overline{{\sf T}(Z)}=\bigl{<}\overline{{\sf
S}{(Z)}}\cup\\{\beta\\}\bigr{>}$. As ${\sf S}{(Y)}\,{\sf wr}\,{\sf
T}(Z)=(\overline{S}_{1}\times\cdots\times\overline{S}_{m})\rtimes\overline{{\sf
T}(Z)}$, we obtain that ${\sf S}{(Y)}\,{\sf wr}\,{\sf
T}(Z)\leqslant\bigl{<}{\sf S}{(Y)}\,{\sf wr}\,{\sf
S}{(Z)}\cup\\{\beta\\}\bigr{>}$, and so the claim is proved.
As ${\sf T}(Y)\,{\sf wr}\,{\sf
T}(Z)=(\overline{T}_{1}\times\cdots\times\overline{T}_{m})\rtimes\overline{{\sf
T}(Z)}$, the arguments in the previous two paragraphs show that ${\sf
T}(Y)\,{\sf wr}\,{\sf T}(Z)\leqslant\bigl{<}{\sf S}{(Y)}\,{\sf wr}\,{\sf
S}{(Z)}\cup\\{\alpha,\beta\\}\bigr{>}$. As $\alpha,\ \beta\in{\sf T}(Y)\,{\sf
wr}\,{\sf T}(Z)$, we obtain (11).
As ${\sf S}{(Y)}\,{\sf wr}\,{\sf S}{(Z)}$ is a proper submonoid of each of the
monoids ${\sf T}(Y)\,{\sf wr}\,{\sf T}(Z)$, ${\sf T}(Y)\,{\sf wr}\,{\sf
S}{(Z)}$ and ${\sf S}{(Y)}\,{\sf wr}\,{\sf T}(Z)$, equation (8) must be valid.
In order to show (7), it suffices to prove that $\mbox{\emph{\mbox{\rm
rank}\,}}{({\sf T}(Y)\,{\sf wr}\,{\sf T}(Z):{\sf S}{(Y)}\,{\sf wr}\,{\sf
S}{(Z)})}>1$. Suppose that $\gamma\in{\sf T}(Y)\,{\sf wr}\,{\sf T}(Z)$ such
that $\bigl{<}{\sf S}{(Y)}\,{\sf wr}\,{\sf
S}{(Z)}\cup\\{\gamma\\}\bigr{>}={\sf T}(Y)\,{\sf wr}\,{\sf T}(Z)$. Then there
are $g,\ g_{1},\ldots,g_{k}\in{\sf S}{(Y)}\,{\sf wr}\,{\sf S}{(Z)}$, such that
$g\gamma g_{1}\gamma\cdots\gamma g_{k}=\alpha$. Thus $\gamma
g_{1}\gamma\cdots\gamma g_{k}=g^{-1}\alpha$ and hence $\ker\gamma\subseteq\ker
g^{-1}\alpha=\\{((1,1)g,(2,1)g)\\}\cup\Delta$, where $\Delta=\\{(x,x)\mid x\in
X\\}$. Since $\gamma\not\in{\sf S}{(Y)}\,{\sf wr}\,{\sf S}{(Z)}$, we obtain
that $\ker\gamma=\\{((1,1)g,(2,1)g)\\}\cup\Delta$. Similarly, there exist $h,\
h_{1},\ldots,h_{k}\in{\sf S}{(Y)}\,{\sf wr}\,{\sf S}{(Z)}$ such that $h\gamma
h_{1}\gamma\cdots\gamma h_{k}=\beta$, Thus $\gamma h_{1}\gamma\cdots\gamma
h_{k}=h^{-1}\beta$ and hence
$\ker\gamma\subseteq\ker
h^{-1}\beta=\\{((1,1)h,(1,2)h),\ldots,((n,1)h,(n,2)h)\\}\cup\Delta.$
Hence, for some $i\in\\{1,\ldots,n\\}$, we have
$((1,1)g,(2,1)g)=((i,1)h,(i,2)h)$, that is,
$((1,1),(2,1))gh^{-1}=((i,1),(i,2))$. This, however, is a contradiction,
because $(1,1)\sim_{\mathcal{P}}(2,1)$, but $(i,1)\not\sim_{\mathcal{P}}(i,2)$
and the transformation $gh^{-1}$ preserves the equivalence relation
$\sim_{\mathcal{P}}$.
Therefore we verified that $\bigl{<}{\sf S}{(Y)}\,{\sf wr}\,{\sf
S}{(Z)}\cup\\{\gamma\\}\bigr{>}$ is a proper submonoid of the wreath product
${\sf T}(Y)\,{\sf wr}\,{\sf T}(Z)$, which shows that equation (7) must hold. ∎
## 4\. The Rank of ${\sf S}{(X,\mathcal{P})}$
In this section we prove the following theorem.
###### Theorem 4.1.
If $X$ is a finite set such that $|X|\geqslant 3$ and $\mathcal{P}$ is a
uniform partition of $X$ then ${\sf S}{(X,\mathcal{P})}$ is generated by two
elements.
In our terminology, the previous theorem gives that $\mbox{\emph{\mbox{\rm
rank}\,}}{\sf S}{(X,\mathcal{P})}=2$. We note that the rank of a transitive
permutation group, such as ${\sf S}{(X,\mathcal{P})}$, is defined in
permutation group theory as the number of orbits of a point-stabilizer. Thus,
in order to avoid possible confusion, we decided to state the theorem above
without using the notation $\mbox{\emph{\mbox{\rm rank}\,}}{\sf
S}{(X,\mathcal{P})}$.
Theorem 1.1 will follow from Lemmas 3.1 and 3.2 and Theorem 4.1.
Using the fact that ${\sf S}{(X,\mathcal{P})}$ is isomorphic to a wreath
product ${\sf S}{(Y)}\,{\sf wr}\,{\sf S}{(Z)}$, it is not difficult to see
that ${\sf S}{(X,\mathcal{P})}$ is generated by four elements. Indeed,
consider the subgroups $\overline{{\sf S}{(Y)}}_{i}$ and $\overline{{\sf
S}{(Z)}}$ defined in (3) and (4). Since ${\sf S}{(Y)}\,{\sf wr}\,{\sf
S}{(Z)}=(\overline{{\sf S}{(Y)}}_{1}\times\cdots\times\overline{{\sf
S}{(Y)}}_{m})\rtimes{\sf S}{(Z)}$ and ${\sf S}{(Z)}$ is transitive by
conjugation on the subgroups $\overline{{\sf S}{(Y)}}_{i}$, we obtain that
${\sf S}{(Y)}\,{\sf wr}\,{\sf S}{(Z)}=\bigl{<}\overline{{\sf
S}{(Y)}}_{1},\overline{{\sf S}{(Z)}}\bigr{>}$. Since $\overline{{\sf
S}{(Y)}}_{1}$ and $\overline{{\sf S}{(Z)}}$ are full symmetric groups, they
are generated by two elements, and so we obtain that ${\sf S}{(Y)}\,{\sf
wr}\,{\sf S}{(Z)}$ is generated by at most four elements. Essentially this is
proved in [17, Theorem 2.6].
Before proving Theorem 4.1, we state two simple lemmas.
Let $G$ be a permutation group acting on the set $\\{1,\ldots,n\\}$ and let
$\mathbb{F}$ be a field. Let $V$ denote the $n$-dimensional vector space over
$\mathbb{F}$ with basis $\\{e_{1},\ldots,e_{n}\\}$. The group $G$ can be
thought of as a permutation group on the set $\\{e_{1},\ldots,e_{n}\\}$, and
this defines an $\mathbb{F}G$-module structure on $V$ as follows:
$e_{i}g=e_{ig}\quad\mbox{for}\quad i\in\\{1,\ldots,n\\}\mbox{ and }g\in G.$
The module $V$ is called the permutation module for $G$ over $\mathbb{F}$.
The following lemma is well-known; see, for instance, [18, Lemma 5.3.4].
###### Lemma 4.1.
If $X=\\{1,\ldots,n\\}$, then the permutation module for ${\sf S}{(X)}$ over a
field $\mathbb{F}$ of characteristic $p$ has precisely two proper non-trivial
submodules:
$\displaystyle U_{1}$ $\displaystyle=$ $\displaystyle\left\\{(a,a,\ldots,a)\
|\ a\in\mathbb{F}\right\\}\quad\mbox{and}$ $\displaystyle U_{2}$
$\displaystyle=$ $\displaystyle\left\\{(a_{1},\ldots,a_{n})\ |\
a_{1}+\cdots+a_{n}=0\right\\}.$
Further, if $p\mid n$ then $U_{1}\leqslant U_{2}$; otherwise $V=U_{1}\oplus
U_{2}$.
Suppose that $G=G_{1}\times\cdots\times G_{k}$ where the $G_{i}$ are finite
groups. For $I\subseteq\\{G_{1},\ldots,G_{k}\\}$ the function
$\varrho_{I}:G\rightarrow\prod_{G_{i}\in I}G_{i}$ is the natural projection
map. We also write $\varrho_{i}$ for $\varrho_{\\{G_{i}\\}}$. A subgroup $X$
of $G$ is said to be a strip if for each $i=1,\ldots,k$ either
$X\varrho_{i}=1$ or $X\varrho_{i}\cong X$. The set of $G_{i}$ such that
$X\varrho_{i}\neq 1$ is called the support of $X$ and is denoted ${\sf
Supp}\,X$. Two strips $X_{1}$ and $X_{2}$ are disjoint if ${\sf
Supp}\,X_{1}\cap{\sf Supp}\,X_{2}=\emptyset$. A strip $X$ is said to be full
if $X\varrho_{i}=G_{i}$ for all $G_{i}\in{\sf Supp}\,X$, and $X$ is called
non-trivial if $|{\sf Supp}\,X|\geqslant 2$. A subgroup $K$ of $G$ is said to
be subdirect if $K\varrho_{i}=G_{i}$ for all $i$.
We recall a well-known lemma on finite simple groups which can be found in
[23, page 328]. The proof of the lemma is elementary and does not use the
finite simple group classification.
###### Lemma 4.2.
Let $M$ be a direct product of finitely many non-abelian, finite simple groups
and let $H$ be a subdirect subgroup of $M$. Then $H$ is the direct product of
pairwise disjoint full strips of $M$.
The wreath product of two transformation semigroups $S$ and $R$ was defined in
Section 2, and the definition can also be used to construct the wreath product
of two permutation groups $G$ and $H$. Recall that the wreath product $G\,{\sf
wr}\,H$ is isomorphic to $G^{m}\rtimes H$ and a typical element of $G\,{\sf
wr}\,H$ is denoted by $(\pi_{1},\ldots,\pi_{m})\sigma$ where $\pi_{i}\in G$
and $\sigma\in H$. Setting $\pi=(\pi_{1},\ldots,\pi_{m})$, the same element
can also be written as $\pi\sigma$. The following lemma facilitates the
calculations in $G\,{\sf wr}\,H$.
###### Lemma 4.3.
Let $\pi\sigma,\ \pi_{1}\sigma_{1},\ \pi_{2}\sigma_{2}\in G\,{\sf wr}\,H$
where $G$ and $H$ are as above. Then
1. (i)
$\pi_{1}\sigma_{1}\pi_{2}\sigma_{2}=\pi_{1}(\pi_{2})^{\sigma_{1}^{-1}}\sigma_{1}\sigma_{2}$;
2. (ii)
$(\pi\sigma)^{-1}=(\pi^{-1})^{\sigma}\sigma^{-1}$;
3. (iii)
$(\pi\sigma)^{n}=\pi\pi^{\sigma^{-1}}\pi^{\sigma^{-2}}\cdots\pi^{\sigma^{-n+1}}\sigma^{n}$.
In particular the projection map $\varrho:G\,{\sf wr}\,H\rightarrow H$ defined
by $\pi\sigma\mapsto\sigma$ is a homomorphism.
###### Proof.
The assertion that $\varrho$ is a homomorphism follows from (i). The rest can
be verified using (5) and (6). ∎
Now we are ready to prove Theorem 4.1. Permutations of a finite set will be
written as products of disjoint cycles.
###### The proof of Theorem 4.1.
By Lemma 2.1(iv), it suffices to show, for $Y=\\{1,\ldots,n\\}$ and
$Z=\\{1,\ldots,m\\}$, that the group $W={\sf S}{(Y)}\,{\sf wr}\,{\sf S}{(Z)}$
is generated by two elements whenever $nm\geqslant 3$. Since, for a finite set
$Y$, the group ${\sf S}{(Y)}$ is generated by two elements, we may assume that
$n\geqslant 2$ and $m\geqslant 2$. Let ${{\sf A}{(Y)}}$ denote the group of
even permutations of $Y$. Then ${{\sf A}{(Y)}}$ is a normal subgroup of index
two of ${\sf S}{(Y)}$. As $W={\sf S}{(Y)}^{m}\rtimes{\sf S}{(Z)}$, we may
consider the subgroups ${{\sf A}{(Y)}}^{m}$ and ${\sf S}{(Y)}^{m}$ of $W$ and
let $A$ and $S$ denote these subgroups respectively.
Let us define
$\displaystyle x$ $\displaystyle=$
$\displaystyle\left\\{\begin{array}[]{ll}({\rm id},(1,2),{\rm id},\ldots,{\rm
id})(1,2,\ldots,m)&\mbox{if either $n$ or $m$ is odd}\\\ ({\rm id},(1,2),{\rm
id},\ldots,{\rm id})(2,3,\ldots,m)&\mbox{otherwise}\end{array}\right.$
$\displaystyle y$ $\displaystyle=$ $\displaystyle((1,2,\ldots,n),{\rm
id},\ldots,{\rm id})(1,2).$
Set $M=\left<x,y\right>$ and we claim that $M=W$.
Let $\varrho:W\rightarrow{\sf S}{(Z)}$ denote the natural projection map in
Lemma 4.3. If either $n$ or $m$ is odd then
$\left<(1,2),(1,2,\ldots,m)\right>\leqslant M\varrho$; otherwise
$\left<(1,2),(2,3,\ldots,m)\right>\leqslant M\varrho$. As
$\left<(1,2),(1,2,\ldots,m)\right>=\left<(1,2),(2,3,\ldots,m)\right>={\sf
S}{(Z)},$
we obtain that $M\varrho={\sf S}{(Z)}$. As $S=\ker\varrho$, in order to prove
that $M=W$, it suffices to show that $S\leqslant M$.
Next we claim that $A\leqslant M$. If $n=2$, then $A=1$, and so we may assume
that $n\geqslant 3$. First we suppose that $n=3$. In this case
$A\cong(C_{3})^{m}$ and so $A$ can be viewed as a $W$-module over
$\mathbb{F}_{3}$, and, in particular, it can be viewed as a ${\sf
S}{(Z)}$-module over the same field. In fact, $A$ is the natural permutation
module for ${\sf S}{(Z)}$. Since $M\varrho={\sf S}{(Z)}$, the intersection
$A\cap M$ is an ${\sf S}{(Z)}$-submodule of $A$. Now
$y^{2}=((1,2,3),(1,2,3),{\rm id},\ldots,{\rm id})$ and so $y^{2}\in A\cap M$,
but, if $m\geqslant 3$, then $y^{2}$ is not an element of either of the two
proper submodules listed in Lemma 4.1. Therefore $A\cap M=A$, and so
$A\leqslant M$ follows when $n=3$ and $m\geqslant 3$. The case $(n,m)=(3,2)$
can be checked using a computer algebra package such as GAP [8] or Magma [6].
Let us assume that $n=4$, that is, $Y=\\{1,\ldots,4\\}$, and, as above, we may
also assume without loss of generality that $m\geqslant 3$. Note that ${{\sf
A}{(Y)}}$ admits the decomposition ${{\sf A}{(Y)}}=U\rtimes V$ where
$U=\left<(1,2)(3,4),(1,3)(2,4)\right>$ and $V=\left<(1,2,3)\right>$. Further,
$U$ can be considered as an irreducible $V$-module over $\mathbb{F}_{2}$.
Let $m$ be odd and set $z_{1}=(x^{m}y^{2})^{2}$. As
$x^{m}y^{2}=((1,3,4),(1,3,4),(1,2),\ldots,(1,2)),$
we obtain that $z_{1}=(x^{m}y^{2})^{2}=((1,4,3),(1,4,3),{\rm id},\ldots,{\rm
id})$. As $M\varrho={\sf S}{(Z)}$, the subgroup $M$ has an element of the form
$\pi(2,3)$ where $\pi\in S$. Then set $z=z_{1}^{\pi(2,3)}$ and compute that
$z=z_{1}^{\pi(2,3)}=(\sigma_{1},{\rm id},\sigma_{3},{\rm id},\ldots,{\rm id})$
where $\sigma_{1}$ and $\sigma_{3}$ are three-cycles in ${\sf S}{(Y)}$. Now
set
$w_{1}=(x^{m})^{y^{4}}x^{m}=((1,2)(3,4),(1,2)(3,4),{\rm id},\ldots,{\rm id})$
and consider the element $w=w_{1}^{z}w_{1}$. Now, as the first component of
$z\in S$ is a non-trivial three cycle, the element $w$ is of the form
$w=(\sigma,{\rm id},\ldots,{\rm id})$ where
$\sigma\in\\{(1,2)(3,4),(1,3)(4,2),(1,4)(2,3)\\}$.
Assume now that $m$ is even and set $z=(x^{m-1}y^{2})^{4}$. Easy computation
yields that
$z=(x^{m-1}y^{2})^{4}=({\rm id},(1,3,4),{\rm id},\ldots,{\rm id})$
and that
$w=(x^{m-1})^{y^{4}}x^{m-1}=({\rm id},(1,2)(3,4),{\rm id},\ldots,{\rm id}).$
As $V$ is irreducible on $U$, the computation above shows that the
$\left<z\right>$-submodule generated by $w$ coincides with $U\times
1\times\cdots\times 1$ of $m$ is odd, and $1\times U\times 1\times\cdots\times
1$ if $m$ is even. Thus $U\times 1\times\cdots\times 1\leqslant M$ in the
former case, and $1\times U\times 1\times\cdots\times 1\leqslant M$ in the
latter. As $M\varrho={\sf S}{(Z)}$, we also obtain that $U^{m}\leqslant M$.
Now the quotient $A/U^{m}\cong V^{m}\cong(C_{3})^{m}$ can be considered as a
permutation module for ${\sf S}{(Z)}$ over $\mathbb{F}_{3}$, and as $(M\cap
A)/U^{m}\unlhd M/U^{m}$ and $M\varrho={\sf S}{(Z)}$, we obtain that $(M\cap
A)/U^{m}$ is a ${\sf S}{(Z)}$-submodule. However, the image of the element $z$
above is not in either of the proper submodules listed in Lemma 4.1, and so we
obtain that $(M\cap A)/U^{m}=A/U^{m}$ which shows that $A\leqslant M$.
Hence we have shown that $A\leqslant M$ if $n\leqslant 4$. Assume now that
$n\geqslant 5$. In this case ${{\sf A}{(Y)}}$ is a non-abelian finite simple
group and so $A$ is a non-abelian characteristically simple group.
Set $Q=M\cap S$. Clearly, $Q\leqslant S$ and $Q\unlhd M$. Let us show that $Q$
is a subdirect subgroup of $S={\sf S}{(Y)}^{m}$. If $n$ or $m$ is odd then set
$z=x^{m}=((1,2),(1,2),\ldots,(1,2))$; otherwise set $z=x^{m-1}=({\rm
id},(1,2),(1,2),\ldots,(1,2))$. Further,
$y^{2}=((1,2,\ldots,n),(1,2,\ldots,n),{\rm id},\ldots,{\rm id}).$
Therefore $z,\ y^{2}\in Q$. Let $\varrho_{2}$ denote the second coordinate
projection $\varrho_{2}:S\rightarrow{\sf S}{(Y)}$. Since
$z\varrho_{2}=(1,2)\in Q\varrho_{2}$ and $y^{2}\varrho_{2}=(1,2,\ldots,n)\in
Q\varrho_{2}$, we obtain that $Q\varrho_{2}={\sf S}{(Y)}$. Simple computation
shows that if $\tau\in Q$ and $\pi\sigma\in M$ then
$\tau^{\pi\sigma}\varrho_{2\sigma}=\tau^{\pi}\varrho_{2}$. Therefore
${\sf S}{(Y)}=Q\varrho_{2}=Q^{\pi\sigma}\varrho_{2\sigma}=Q\varrho_{2\sigma}.$
As $Q\varrho_{2}={\sf S}{(Y)}$, it follows that $Q\varrho_{i}={\sf S}{(Y)}$
for all $i$, and so $Q$ is a subdirect subgroup of $S={\sf S}{(Y)}^{m}$.
Consider the commutator subgroup $Q^{\prime}$. Since $S^{\prime}=A$, we obtain
that $Q^{\prime}\leqslant A$. Further, as $Q\varrho_{i}={\sf S}{(Y)}$, we also
obtain, for all $i$, that $Q^{\prime}\varrho_{i}={{\sf A}{(Y)}}$. Therefore
$Q^{\prime}$ is a subdirect subgroup of $A={{\sf A}{(Y)}}^{m}$. Set $R=M\cap
A$. Since $Q^{\prime}\leqslant R$ and $Q^{\prime}$ is a subdirect subgroup, we
obtain that so is $R$; that is, by Lemma 4.2, $R$ is a direct product of
disjoint strips.
Let $S$ be a strip in $R$ and let $\mathcal{S}\subseteq\\{1,\ldots,m\\}$ be
the support of $S$. We claim that $\mathcal{S}$ is a block for the action of
${\sf S}{(Z)}$. Indeed, if $\sigma\in{\sf S}{(Z)}$ then there is some $\pi\in
S$ such that $\pi\sigma\in M$. Then $S^{\pi\sigma}$ is strip in $R$ and so
either $S=S^{\pi\sigma}$ or $S\cap S^{\pi\sigma}=1$. The support of
$S^{\pi\sigma}$ is $\mathcal{S}^{\sigma}$. Thus either
$\mathcal{S}^{\sigma}=\mathcal{S}$ or
$\mathcal{S}^{\sigma}\cap\mathcal{S}=\emptyset$, which shows that
$\mathcal{S}$ is a block, as required. Since ${\sf S}{(Z)}$ is primitive on
$\\{1,\ldots,m\\}$ either $|\mathcal{S}|=1$ or $\mathcal{S}=\\{1,\ldots,m\\}$.
If the latter holds, then $Q$ is a strip. This, however, is impossible.
Indeed, if $m=2$ and $n$ is odd, then set $z=xy=({\rm
id},(1,2)(1,2,\ldots,n))$; if $m=2$ and $n$ is even then set
$z=xy^{2}=((1,2,\ldots,n),(1,2)(1,2,\ldots,n))$; if $m\geqslant 3$, then set
$z=y^{2}=((1,2,\ldots,n),(1,2,\ldots,n),{\rm id},\ldots,{\rm id})$. Then in
all cases $z^{2}\in R$, but $z^{2}$ is not in a full strip (if $m=2$ and $n$
is even then note that the first component of $z^{2}$ is of order $n/2$ and
the second is of order $n-1$). Thus $|\mathcal{S}|=1$, and so $A\leqslant M$.
This completes the proof of the claim that that $A\leqslant M$ for all $n$ and
$m$.
Note that $A$ is a normal subgroup of $W$ and let $x\mapsto\widehat{x}$ denote
the natural homomorphism $W\rightarrow W/A$. If $H\leqslant W$, then
$\widehat{H}$ denotes the image $HA/A$ of $H$. Then $\widehat{W}\cong
C_{2}\,{\sf wr}\,{\sf S}{(Z)}$ and $\widehat{S}\cong(C_{2})^{m}$. We claim
that $\widehat{S}\leqslant\widehat{M}$. Note that
$\widehat{M}\cap\widehat{S}\unlhd\widehat{M}$. If $\pi\sigma\in\widehat{M}$
and $u\in\widehat{M}\cap\widehat{S}$ then $u^{\pi\sigma}=u^{\sigma}$. As
$u^{\pi\sigma}\in\widehat{M}\cap\widehat{S}$ we obtain that
$u^{\sigma}\in\widehat{M}\cap\widehat{S}$ which shows that
$\widehat{M}\cap\widehat{S}$ is a ${\sf S}{(Z)}$-submodule of
$\widehat{S}\cong(C_{2})^{m}$. It is clear that $\widehat{S}$ is the natural
permutation module for ${\sf S}{(Z)}$ over $\mathbb{F}_{2}$. Lemma 4.1 lists
the non-trivial proper submodules $U_{1}$ and $U_{2}$ of $\widehat{S}$.
If both $n$ and $m$ are even then $\widehat{x}^{m-1}=(0,1,\ldots,1)$ and this
element is not in either $U_{1}$ or $U_{2}$. Hence
$\widehat{S}\leqslant\widehat{M}$. If this is not the case, then
$\widehat{x}^{m}=(1,1,\ldots,1)$ which shows that $U_{1}\leqslant\widehat{M}$.
If $n$ is even and $m$ is odd then $\widehat{y}^{2}=(1,1,0,\ldots,0)$ and so
$U_{2}\leqslant\widehat{M}$. Therefore in this case $U_{1}\oplus
U_{2}\leqslant\widehat{M}$ follows.
Suppose that $n$ is odd. If $m=2$ then $xy=({\rm id},(1,2)(1,2,\ldots,n))$.
Thus $\widehat{xy}$ is not in either of the proper submodules listed in Lemma
4.1. Hence $\widehat{S}\leqslant\widehat{M}$ follows in this case. If
$m\geqslant 3$, then
$xy=({\rm id},(1,2),{\rm id},\ldots,{\rm id},(1,2,\ldots,n))(2,3\ldots,m).$
Let $\pi=({\rm id},(1,2),{\rm id},\ldots,{\rm id},(1,2,\ldots,n))$ and
$\sigma=(2,3,\ldots,m)$ so that $xy=\pi\sigma$. As $\sigma^{m-1}=1$, it
follows from Lemma 4.3 that
$(xy)^{m-1}=\pi\pi^{\sigma^{-1}}\ldots\pi^{\sigma^{-m+2}}.$
Thus $(xy)^{m-1}$ is of the form $({\rm id},\pi_{2},\ldots,\pi_{m})$ where,
for $i=2,\ldots,m$, the permutation $\pi_{i}$ is either $(1,2)(1,2,\ldots,n)$
or $(1,2,\ldots,n)(1,2)$, that is, $\pi_{i}$ is a cycle with length $n-1$. As
$n$ is odd, $\pi_{i}\not\in{{\sf A}{(Y)}}$, and so
$(\widehat{xy})^{m-1}=(0,1,1,\ldots,1)$. Now if $m$ is even then
$(\widehat{xy})^{m-1}$ is not an element of $U_{1}$ or $U_{2}$, and so
$\widehat{S}\leqslant\widehat{M}$ follows also in this case. If $m$ is odd
then this shows that $U_{2}\leqslant\widehat{M}$, and as we proved above that
$U_{1}\leqslant\widehat{M}$, it follows that
$\widehat{S}\leqslant\widehat{M}$.
Thus we have shown that $S\leqslant M$ as required. As explained above, $M=W$
now follows. ∎
The main result of the paper can now be proved.
###### The proof of Theorem 1.1.
As usual, we assume without loss of generality that $X=Y\times Z$ where
$Y=\\{1,\ldots,n\\}$, $Z=\\{1,\ldots,m\\}$ and $\mathcal{P}$ is the partition
$\\{\\{(1,1),\ldots,(n,1)\\},\ldots,\\{(1,m),\ldots,(n,m)\\}\\}.$
By Lemmas 2.1, 3.1, 3.2 and Theorem 4.1,
$\mbox{\emph{\mbox{\rm rank}\,}}{{\sf T}(X,\mathcal{P})}=\mbox{\emph{\mbox{\rm
rank}\,}}{({\sf T}(X,\mathcal{P}):{\sf
S}{(X,\mathcal{P})})}+\mbox{\emph{\mbox{\rm rank}\,}}{{\sf
S}{(X,\mathcal{P})}}\\\ =\mbox{\emph{\mbox{\rm rank}\,}}{({\sf T}(Y)\,{\sf
wr}\,{\sf T}(Z):{\sf S}{(Y)}\,{\sf wr}\,{\sf S}{(Z)})}+\mbox{\emph{\mbox{\rm
rank}\,}}{{\sf S}{(Y)}\,{\sf wr}\,{\sf S}{(Z)}}=4.$
The assertions concerning $\mbox{\emph{\mbox{\rm
rank}\,}}\Gamma(X,\mathcal{P})$ and $\mbox{\emph{\mbox{\rm
rank}\,}}\Sigma(X,\mathcal{P})$ can be proved very similarly. ∎
## Acknowledgment
We acknowledge with gratitude some conversations with John D. Dixon, Peter M.
Neumann and Joseph J. Rotman.
The first author was partially supported by FCT and FEDER, Project POCTI-
ISFL-1-143 of Centro de Algebra da Universidade de Lisboa, and by FCT and
PIDDAC through the project PTDC/MAT/69514/2006.
The second author is grateful to the Centro de Álgebra da Universidade de
Lisboa for the invitation and the hospitality; he was also supported by the
Hungarian Scientific Research Fund (OTKA) grant F049040.
## References
* [1] J. Araújo and J. D. Mitchell. Relative ranks in the monoid of endomorphisms of an independence algebra. Monatsh. Math., 151(1):1–10, 2007.
* [2] J. Araújo, J. D. Mitchell, and N. Silva. On generating countable sets of endomorphisms. Algebra Universalis, 50(1):61–67, 2003.
* [3] S. Banach. Sur un theorème de M. Sierpiński. Fund. Math., 25:5–6, 1935.
* [4] George Barnes and Inessa Levi. Ranks of semigroups generated by order-preserving transformations with a fixed partition type. Comm. Algebra, 31(4):1753–1763, 2003.
* [5] George Barnes and Inessa Levi. On idempotent ranks of semigroups of partial transformations. Semigroup Forum, 70(1):81–96, 2005.
* [6] Wieb Bosma, John Cannon, and Catherine Playoust. The Magma algebra system I: The user language. J. Symbolic Comput., 24(3-4):235–265, 1997. Computational algebra and number theory (London, 1993).
* [7] J. Cichoń, J. D. Mitchell, and M. Morayne. Generating continuous mappings with Lipschitz mappings. Trans. Amer. Math. Soc., 359(5):2059–2074 (electronic), 2007.
* [8] The GAP Group. GAP – Groups, Algorithms, and Programming, Version 4.4.10, 2007\.
* [9] G. U. Garba. On the nilpotent ranks of certain semigroups of transformations. Glasgow Math. J., 36(1):1–9, 1994.
* [10] Gracinda M. S. Gomes and John M. Howie. On the ranks of certain semigroups of order-preserving transformations. Semigroup Forum, 45(3):272–282, 1992.
* [11] P. M. Higgins, J. M. Howie, J. D. Mitchell, and N. Ruškuc. Countable versus uncountable ranks in infinite semigroups of transformations and relations. Proc. Edinb. Math. Soc. (2), 46(3):531–544, 2003.
* [12] P. M. Higgins, J. D. Mitchell, M. Morayne, and N. Ruškuc. Rank properties of endomorphisms of infinite partially ordered sets. Bull. London Math. Soc., 38(2):177–191, 2006.
* [13] Peter M. Higgins, John M. Howie, and Nikola Ruškuc. Generators and factorisations of transformation semigroups. Proc. Roy. Soc. Edinburgh Sect. A, 128(6):1355–1369, 1998.
* [14] John M. Howie. Fundamentals of semigroup theory, volume 12 of London Mathematical Society Monographs. New Series. The Clarendon Press Oxford University Press, New York, 1995. Oxford Science Publications.
* [15] John M. Howie and Robert B. McFadden. Idempotent rank in finite full transformation semigroups. Proc. Roy. Soc. Edinburgh Sect. A, 114(3-4):161–167, 1990.
* [16] John M. Howie, N. Ruškuc, and P. M. Higgins. On relative ranks of full transformation semigroups. Comm. Algebra, 26(3):733–748, 1998.
* [17] Pei Huisheng. On the rank of the semigroup $T_{E}(X)$. Semigroup Forum, 70(1):107–117, 2005.
* [18] Peter Kleidman and Martin Liebeck. The subgroup structure of the finite classical groups, volume 129 of London Mathematical Society Lecture Note Series. Cambridge University Press, Cambridge, 1990.
* [19] Inessa Levi. Nilpotent ranks of semigroups of partial transformations. Semigroup Forum, 72(3):459–476, 2006.
* [20] Inessa Levi and Steve Seif. Combinatorial techniques for determining rank and idempotent rank of certain finite semigroups. Proc. Edinb. Math. Soc. (2), 45(3):617–630, 2002.
* [21] J. D. P. Meldrum. Wreath products of groups and semigroups, volume 74 of Pitman Monographs and Surveys in Pure and Applied Mathematics. Longman, Harlow, 1995.
* [22] Maria Isabel Marques Ribeiro. Rank properties in finite inverse semigroups. Proc. Edinburgh Math. Soc. (2), 43(3):559–568, 2000.
* [23] Leonard L. Scott. Representations in characteristic $p$. In The Santa Cruz Conference on Finite Groups (Univ. California, Santa Cruz, Calif., 1979), volume 37 of Proc. Sympos. Pure Math., pages 319–331. Amer. Math. Soc., Providence, R.I., 1980.
* [24] Wacł aw Sierpiński. Sur les suites infinies de fonctions définies dans les ensembles quelconques. Fund. Math., 24:209–212, 1935.
* [25] Howard Straubing. The wreath product and its applications. In Formal properties of finite automata and applications (Ramatuelle, 1988), volume 386 of Lecture Notes in Comput. Sci., pages 15–24. Springer, Berlin, 1989.
| arxiv-papers | 2008-07-08T11:33:16 | 2024-09-04T02:48:56.626014 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Joao Araujo and Csaba Schneider",
"submitter": "Csaba Schneider",
"url": "https://arxiv.org/abs/0807.1214"
} |
0807.1247 | # MEROMORPHIC FUNCTIONS WITH SEVERAL ESSENTIAL SINGULARITIES
A. A. Kondratyuk
Lviv National University,
Universitetska street 1, 79000, Lviv, Ukraine
July 8, 2008
Key words:
Annulus, meromorphic function, Nevanlinna characteristic, index, value
distribution theory.
Mathematics Subject Classification (2000)
30D35
Preprint
Lviv National University,
Faculty of Mathematics and Mechanics
Abstract
A two-parameter characteristic of functions meromorphic on annuli is
introduced and an extension of the Nevanlinna value distribution theory for
such functions is proposed.
## Introduction
Meromorphic functions with $m+1$ possible essential singularities
$\\{c_{j}\\},$ $c_{j}\in\mathbb{C},$ $j=1,2,...,m,$ $c_{m+1}=\infty,$ are
considered. In other words, we consider meromorphic functions in a
$m$-punctured plane. The example of such a function can be given by the
composition $f\circ\mathcal{R}$ of a function $f,$ transcendental meromorphic
in $\mathbb{C},$ and a rational function $\mathcal{R}$ with $m+1$ distinct
poles in $\overline{\mathbb{C}}.$
We introduce a $m+1$-parameter characteristic of meromorphic functions in a
$m$-punctured plane and investigate the distribution of their values. The
introduced characteristic possesses the properties similar to these ones of
the classical Nevanlinna characteristic.
To begin consider meromorphic functions with two possible essential
singularities. Up to a linear fractional transformation we have $0$ and
$\infty.$ That is, we consider meromorphic functions in the plane punctured at
the origin. We will approach to both singularities by the annuli.
$A_{sr}=\\{z:\,s<|z|<r\\}.$ In other words,
$\mathbb{C}\setminus\\{0\\}=\bigcup\limits_{0<s<r<+\infty}A_{sr}.$
It seems that consideration of annuli $A_{\frac{1}{\tau}r},\,1\leq\tau,\,1\leq
r,$ will be more convenient.
Note, that there are essential differences between disks and annuli in the
topoligal sence which are reflected in the theory of meromorphic functions.
Firstly, the fundamental (Poincré) group of a disk is trivial, while for an
annulus, we have a group isomorphic to the additive group $\mathbb{Z}.$
Secondly, the group of automorphisms of the unit disk is rich. It consists of
the Möbius transformations. Therefore, the Poisson integral formula is an
invariant form of the Gauss mean theorem and the Poisson-Jensen formula is an
invariant form of Jensen’s formula. More over, all disks are conformally
equivalent. The situation with annuli is another. The group of automorphisms
of an annulus is poor. If $s\neq\frac{1}{r},$ it consists of rotations only.
Annuli $A_{1r}$ and $A_{1\rho}$ is conformally equivalent iff $r=\rho.$
Therefore, the theory of meromorphic functions on annuli is more complicated
that this one in disks. But many results for meromorphic functions in
$\mathbb{C}$ have their counterparts for these ones in a punctured plane
$\mathbb{C}^{*}.$ For example, the Picard theorem may be reformulated in the
form. ”For every non-constant meromorphic function in $\mathbb{C}^{*}$ there
is a linear fractional transformation $\omega(\mathbb{C}^{*})\subset
f(\mathbb{C}^{*}).$” If both, $\omega$ and $f,$ are identical maps, we have
the coincidence.
In order to introduce a two-parameter characteristic we use the notion of
index of a meromorphic function along a circle.
## $\bf 1^{0}.$ Index of $f$ along a circle
###### Lemma 1.
Let $f$ be a function meromorphic on $\\{z:\,|z|=t\\},$ non identical zero.
Then
$\displaystyle{}\displaystyle\nu(t,f)=\frac{1}{\pi}\int\limits_{|z|=t}{\rm
Im}\left(\frac{f^{\prime}(z)}{f(z)}\,dz\right)$ (1)
is an integer.
###### Proof.
Denote $\gamma(\theta)=te^{i\theta},$ $0\leq\theta\leq 2\pi.$ If $f$ is
holomorphic on the circle $\\{z:\,|z|=t\\}$ without zeroes on this circle then
$\Gamma(\theta)=(f\circ\gamma)(\theta),$ $0\leq\theta\leq 2\pi,$ is a closed
path, and
$\displaystyle\displaystyle\frac{1}{2\pi
i}\int\limits_{|z|=t}\frac{f^{\prime}(z)}{f(z)}\,dz=\frac{1}{2\pi
i}\int\limits_{0}^{2\pi}\frac{f^{\prime}(\gamma(\theta)))}{f(\gamma(\theta)))}\,\gamma^{\prime}(\theta)\,d\theta=$
$\displaystyle\displaystyle=\frac{1}{2\pi
i}\int\limits_{0}^{2\pi}\frac{\Gamma^{\prime}(\theta)}{\Gamma(\theta)}\,d\theta=\frac{1}{2\pi
i}\int\limits_{\Gamma}\frac{d\zeta}{\zeta}={\rm Ind}_{\Gamma}(0).$
Taking the real parts of both sides we have
$\displaystyle\displaystyle\nu(t,f)=2\,{\rm Ind}_{\Gamma}(0).$
As index of $\zeta=0$ with respect to $\Gamma$ is an integer we obtain the
needful conclusion in this case.
If $f$ has a unique simple zero $a=te^{i\alpha},$ then $f(z)=g(z)(z-a),$
$\displaystyle\frac{f^{\prime}}{f}=\frac{g^{\prime}}{g}+\frac{1}{z-a},$ and
$\displaystyle\displaystyle{\rm Im}\left(\frac{1}{z-a}\,dz\right)={\rm
Im}\left(\frac{ite^{i\theta}d\theta}{t(e^{i\theta}-e^{i\alpha})}\right)=$
$\displaystyle\displaystyle={\rm
Re}\frac{1}{1-e^{i(\alpha-\theta)}}\,d\theta=\frac{1}{2}\,d\theta,$
because the transformation $\displaystyle w=\frac{1}{1-z}$ maps the unit
circle on the straight line ${\rm Re}\,w=1/2.$ Hence, in this case
$\displaystyle\displaystyle\nu(t,f)=\nu(t,g)+1.$
If $f$ has a simple pole on the considered circle, then we obtain the similar
equality with ”-” instead of ”+”. The final conclusion in the general case now
follows by induction.
The value $\nu(t,f)$ is said to be index of $f$ along the circle
$\\{z:\,|z|=t\\}.$
If a branch of $\log f$ may by determined on the circle, then $d\log
f=\frac{f^{\prime}}{f}\,dz.$ Thus, ${\rm
Im}\left(\frac{f^{\prime}}{f}\,dz\right)=d\arg f(z)$ and $\nu(t,f)$ is the
increment of $\frac{1}{\pi}\,\arg f$ along the circle.
∎
## $\bf 2^{0}.$ Version 1 of Jensen’s theorem
###### Lemma 2.
Let $f$ be a function meromorphic on the closure of the annulus
$A_{sr}=\\{z:\,s<|z|<r\\},$ and non identical zero. Then
$\displaystyle{}\displaystyle\int\limits_{s}^{r}\frac{\nu(t,f)}{t}\,dt=\frac{1}{\pi}\int\limits_{0}^{2\pi}\log|f(re^{i\theta})|\,d\theta-\frac{1}{\pi}\int\limits_{0}^{2\pi}\log|f(se^{i\theta})|\,d\theta.$
(2)
###### Proof.
Assume that there are neither zeroes nor poles of $f$ on the interval
$\\{z=te^{i\theta},\,s\leq t\leq r\\}.$ Then a branch of $\log f$ can by
determined on this interval, and we have
$\displaystyle{}\displaystyle\log f(re^{i\theta})-\log
f(se^{i\theta})=\int\limits_{s}^{r}\frac{f^{\prime}(z)}{f(z)}\,dz.$
This implies
$\displaystyle{}\displaystyle\log|f(re^{i\theta})|-\log|f(se^{i\theta})|=\int\limits_{s}^{r}{\rm
Re}\left(\frac{f^{\prime}(te^{i\theta})}{f(te^{i\theta})}\,e^{i\theta}\right)\,dt.$
The last equality is valid for all $\theta$ from $[0,2\pi]$ except for a
finite number of $\theta.$ The integration over $\theta$ yields
$\displaystyle\displaystyle\frac{1}{\pi}\int\limits_{0}^{2\pi}\log|f(re^{i\theta})|\,d\theta-\frac{1}{\pi}\int\limits_{0}^{2\pi}\log|f(se^{i\theta})|\,d\theta=$
$\displaystyle\displaystyle=\frac{1}{\pi}\int\limits_{0}^{2\pi}d\theta\int\limits_{s}^{r}\frac{1}{t}{\rm
Im}\left(\frac{f^{\prime}}{f}\,ite^{i\theta}\right)dt.$
Using the Fubini theorem and changing the order of integration we obtain
(LABEL:eq2). ∎
## $\bf 3^{0}.$ Version 2 of Jensen’s theorem
Suppose $s\leq 1\leq r$ and $f$ meromorphic on the closure of the annulus
$A_{sr}$ and non identical zero. Put in (LABEL:eq2) $r=1,$ $\tau=\frac{1}{s}$
and $t=\frac{1}{u}.$ Then
$\displaystyle\displaystyle-\int\limits_{1/s}^{1}\frac{\nu(\frac{1}{u},f)}{u}\,du=\int\limits_{1}^{\tau}\frac{\nu(\frac{1}{u},f)}{u}\,du=$
$\displaystyle\displaystyle=\frac{1}{\pi}\int\limits_{0}^{2\pi}\log|f(e^{i\theta})|\,d\theta-\frac{1}{\pi}\int\limits_{0}^{2\pi}\log\left|f\left(\frac{1}{\tau}\,e^{i\theta}\right)\right|\,d\theta.$
(3)
Putting now in (LABEL:eq2) $s=1$ and subtracting (LABEL:eq10) we have
$\displaystyle\displaystyle\int\limits_{1}^{r}\frac{\nu(t,f)}{t}\,dt-\int\limits_{1}^{\tau}\frac{\nu(\frac{1}{t},f)}{t}\,dt=\frac{1}{\pi}\int\limits_{0}^{2\pi}\log|f(re^{i\theta})|\,d\theta+$
$\displaystyle\displaystyle+\frac{1}{\pi}\int\limits_{0}^{2\pi}\log\left|f\left(\frac{1}{\tau}\,e^{i\theta}\right)\right|\,d\theta-\frac{2}{\pi}\int\limits_{0}^{2\pi}\log\left|f\left(e^{i\theta}\right)\right|\,d\theta.$
(4)
In order to obtain Version 2 of Jensen’s theorem we are going to connect the
notion of index $\nu(t,f)$ with the counting functions of zeroes and poles of
$f.$
Let $\mathbb{T}$ be the unit circle and $n(s,r;f)$ be the number of poles of
$f$ in $A_{sr}.$ It follows from the argument principle that
$\displaystyle\displaystyle\nu(t,f)-\nu(1,f)=$
$\displaystyle\displaystyle=2\,n(1,t;\frac{1}{f})+n(\mathbb{T},\frac{1}{f})-2\,n(1,t;f)-n(\mathbb{T},f),$
(5)
and
$\displaystyle\displaystyle\nu(1,f)-\nu(\frac{1}{t},f)=$
$\displaystyle\displaystyle=2\,n(\frac{1}{t},1;\frac{1}{f})+n(\mathbb{T},\frac{1}{f})-2\,n(\frac{1}{t},1;f)-n(\mathbb{T},f)$
(6)
under the assumption that neither zeroes nor poles of $f$ lie on the circles
$\\{z:\,|z|=t\\}$ and $\\{z:\,z=1/t\\}.$
Define
$\displaystyle\displaystyle
N(\tau,r;f)=\int\limits_{1}^{\tau}\frac{n(\frac{1}{t},1;f)}{t}\,dt+\int\limits_{1}^{r}\frac{n(1,t;f)}{t}\,dt+n(\mathbb{T},f)\log\sqrt{\tau
r}.$
Relations (LABEL:eq11) – (LABEL:eq13) yield the following lemma.
###### Lemma 3.
Let $f$ be a meromorphic function on the closure of $A_{\frac{1}{\tau}r}$ and
non identical zero. Then
$\displaystyle\displaystyle
N(\tau,r;f)-N(\tau,r;f)=\frac{1}{2\pi}\int\limits_{0}^{2\pi}\log\left|f\left(\frac{e^{i\theta}}{\tau}\right)\right|\,d\theta+$
$\displaystyle\displaystyle+\frac{1}{2\pi}\int\limits_{0}^{2\pi}\log|f\left(r{e^{i\theta}}\right)|\,d\theta-\frac{1}{\pi}\int\limits_{0}^{2\pi}\log|f\left({e^{i\theta}}\right)|\,d\theta+$
(7) $\displaystyle\displaystyle+\nu(1,f)\log\sqrt{\frac{\tau}{r}}.$
## $\bf 4^{0}.$ Characteristic of $f$
Now we are able to introduce a two-parameter characteristic of meromorphic
functions in an annulus which possesses the properties like to its classical
counterpart. We follow the Cartan idea. Applying (LABEL:eq15) to the function
$f(z)-e^{i\varphi},$ and integrating over $\varphi$ on $[0,2\pi]$ we obtain
$\displaystyle\displaystyle\frac{1}{2\pi}\int\limits_{0}^{2\pi}N\left(\tau,r;\frac{1}{f-e^{i\varphi}}\right)\,d\varphi-N(\tau,r;f)=$
$\displaystyle\displaystyle=\frac{1}{2\pi}\int\limits_{0}^{2\pi}\log^{+}\left|f\left(\frac{e^{i\theta}}{\tau}\right)\right|\,d\theta+\frac{1}{2\pi}\int\limits_{0}^{2\pi}\log^{+}|f\left(r{e^{i\theta}}\right)|\,d\theta-$
(8)
$\displaystyle\displaystyle-\frac{1}{\pi}\int\limits_{0}^{2\pi}\log^{+}|f\left({e^{i\theta}}\right)|\,d\theta+\frac{1}{4\pi}\int\limits_{0}^{2\pi}\nu(1,f-e^{i\varphi})\,d\varphi\log{\frac{\tau}{r}},$
of course, if the last integral exists.
We will show that and evaluate the integral.
Let $f$ be a meromorphic function on $\mathbb{T}.$
Denote $E_{f}^{+}=\left\\{\mathbb{T}\ni z:\,|f(z)|>1\right\\},$
$E_{f}^{0}=\left\\{\mathbb{T}\ni z:\,|f(z)|=1\right\\}$ and
$E_{f}^{-}=\left\\{\mathbb{T}\ni z:\,|f(z)|<1\right\\}.$
###### Lemma 4.
Let $f$ be a meromorphic function on the unit circle $\mathbb{T},$
$f(z)\not\equiv 0.$ Then
$\displaystyle\displaystyle\frac{1}{4\pi}\int\limits_{0}^{2\pi}\nu(1,f-e^{i\varphi})\,d\varphi=\frac{1}{2\pi}\int\limits_{E_{f}^{+}}{\rm
Im}\left(\frac{f^{\prime}}{f}\,dz\right)+\frac{1}{4\pi}\int\limits_{E_{f}^{0}}{\rm
Im}\left(\frac{f^{\prime}}{f}\,dz\right).$
The proof of Lemma 4 needs some auxiliary results.
###### Lemma 5.
For each function $f$ meromorphic on $\\{z:\,|z|=t\\},$ $f\not\equiv 0,$ and
each $\zeta\in\mathbb{C}$ the relation
$\displaystyle\displaystyle\nu(t,f-\zeta)=\nu(t,f)-\frac{1}{\pi}\int\limits_{|z|=t}{\rm
Im}\left(\frac{\zeta f^{\prime}}{f(\zeta-f)}\,dz\right)$ (9)
holds.
###### Proof.
Relation (LABEL:eq18) follows immediately from (1) and the idenity
$\displaystyle\displaystyle\frac{f^{\prime}}{f-\zeta}=\frac{f^{\prime}}{f}\left(1-\frac{\zeta}{\zeta-f}\right).$
∎
###### Lemma 6.
Let $f$ be a meromorphic function on the unit circle $\mathbb{T},$
$f(z)\not\equiv 0.$ Then
$\displaystyle\displaystyle\frac{1}{4\pi}\iint\limits_{\mathbb{T}\times\mathbb{T}}{\rm
Re}\left(\frac{f^{\prime}}{f(\zeta-f)}\,dz\,d\zeta\right)=$
$\displaystyle\displaystyle=-\frac{1}{2\pi}\int\limits_{E_{f}^{-}}{\rm
Im}\left(\frac{f^{\prime}}{f}\,dz\right)-\frac{1}{4\pi}\int\limits_{E_{f}^{0}}{\rm
Im}\left(\frac{f^{\prime}}{f}\,dz\right).$ (10)
###### Proof.
If $z\in\mathbb{T}\setminus E_{f}^{0},$ i.e. $|f(z)|\neq 1,$ we have
$\displaystyle\displaystyle\frac{1}{2\pi
i}\int\limits_{\mathbb{T}}\frac{d\zeta}{\zeta-f(z)}=\begin{cases}0,&{\text{i}f}\
\ |f(z)|>1,\\\ 1,&{\text{i}f}\ \ |f(z)|<1.\end{cases}$
This is true for each $z$ from $\mathbb{T}\setminus E_{f}^{0}$ except for a
finite number of $z.$ Multiplying the last equality by
$i\frac{f^{\prime}dz}{f}$ and taking the real parts we obtain
$\displaystyle\displaystyle\frac{1}{2\pi
i}\int\limits_{\mathbb{T}_{\zeta}}{\rm
Re}\left(\frac{f^{\prime}}{f(\zeta-f)}\,d\zeta\,dz\right)=\begin{cases}0,&{\text{i}f}\
\ z\in E_{f}^{+},\\\ -{\rm
Im}\left(\frac{f^{\prime}}{f}\,dz\right),&{\text{i}f}\ \ z\in
E_{f}^{-}.\end{cases}$
for each $z$ from $\mathbb{T}\setminus E_{f}^{0}$ except for a finite number
of $z.$ Therefore,
$\displaystyle\displaystyle\frac{1}{4\pi^{2}}\int\limits_{\mathbb{T}_{\zeta}}\int\limits_{\mathbb{T}_{z}\setminus
E_{f}^{0}}{\rm Re}\left(\frac{f^{\prime}}{f(\zeta-f)}\,dz\,d\zeta\right)=$
(11)
$\displaystyle\displaystyle=\frac{1}{2\pi}\int\limits_{\mathbb{T}_{z}\setminus
E_{f}^{0}}\frac{1}{2\pi}\int\limits_{\mathbb{T}_{\zeta}}{\rm
Re}\left(\frac{f^{\prime}}{f(\zeta-f)}\,dz\,d\zeta\right)=-\frac{1}{2\pi}\int\limits_{E_{f}^{-}}{\rm
Im}\left(\frac{f^{\prime}}{f}\,dz\right).$
If $|f(z)|=1,$ i.e. $z\in E_{f}^{0},$ then $f(z)=e^{i\alpha(\theta)}.$ We will
use the identity
$\displaystyle\displaystyle{\rm
Re}\left(\frac{f^{\prime}}{f(\zeta-f)}\,dz\,d\zeta\right)={\rm
Re}\left(\frac{f^{\prime}}{f}\,dz\right){\rm
Re}\left(\frac{d\zeta}{\zeta-f}\right)-$ $\displaystyle\displaystyle-{\rm
Im}\left(\frac{f^{\prime}}{f}\,dz\right){\rm
Im}\left(\frac{d\zeta}{\zeta-f}\right)$ (12)
If $E_{f}^{0}$ contains an open arc then $d\log|f(z)|={\rm
Re}\left(\frac{f^{\prime}}{f}\,dz\right)=0$ on this arc, because
$\log|f(z)|=0.$ By continuity
$\displaystyle\displaystyle{\rm Re}\left(\frac{f^{\prime}}{f}\,dz\right)=0$
on the closure of the arc. Thus, the intersection of $E_{f}^{0}$ with the set
$\left\\{z:\,{\rm Re}\left(\frac{f^{\prime}}{f}\,dz\right)\neq 0\right\\}$ is
empty or consists of isolated points. Besides this,
$\displaystyle\displaystyle{\rm Re}\frac{d\zeta}{\zeta-f}={\rm
Im}\frac{ie^{i\varphi}}{e^{i\varphi}-e^{i\alpha(\theta)}}={\rm
Re}\frac{1}{1-e^{i(\alpha(\theta)-\varphi)}}=\frac{1}{2}.$
Hence, relation (LABEL:eq24) implies
$\displaystyle\displaystyle\frac{1}{4\pi^{2}}\int\limits_{\mathbb{T}\times
E_{f}^{0}}{\rm
Re}\left(\frac{f^{\prime}}{f(\zeta-f)}\,dz\,d\zeta\right)=-\frac{1}{4\pi}\int\limits_{E_{f}^{0}}{\rm
Im}\left(\frac{f^{\prime}}{f}\,dz\right).$ (13)
Relations (LABEL:eq23) and (LABEL:eq27) yield (LABEL:eq20). ∎
###### Proof of Lemma 4.
Applying Lemma 5 we have
$\displaystyle\displaystyle\frac{1}{4\pi}\int\limits_{0}^{2\pi}\nu(1,f-e^{i\varphi})\,d\varphi=\frac{1}{2}\nu(1,f)-$
$\displaystyle\displaystyle-\frac{1}{4\pi^{2}}\int\limits_{0}^{2\pi}\int\limits_{\mathbb{T}}{\rm
Im}\left(\frac{f^{\prime}e^{i\varphi}dz}{f(e^{i\varphi}-f)}\right)\,d\varphi.$
(14)
As
$\displaystyle\displaystyle
d\varphi=\frac{-id\zeta}{\zeta},\qquad\zeta=e^{i\varphi},$
then
$\displaystyle\displaystyle-\frac{1}{4\pi^{2}}\int\limits_{0}^{2\pi}\int\limits_{\mathbb{T}}{\rm
Im}\left(\frac{\zeta f^{\prime}}{f(\zeta-f)}\,dz\,d\varphi\right)=$
$\displaystyle\displaystyle=\frac{1}{4\pi^{2}}\iint\limits_{\mathbb{T}\times\mathbb{T}}{\rm
Re}\left(\frac{f^{\prime}}{f(\zeta-f)}\,dz\,d\zeta\right).$
Applying Lemma 6, which evaluates the last integral, and (1) we obtain the
conclusion of Lemma 4 from (LABEL:eq28). ∎
Denote
$\displaystyle\displaystyle
m(\tau,r;f)=m\left(\frac{1}{\tau},f\right)+m(r,f)-2\,m(1,f),$
where $m(r,f)$ is usual Nevanlinna’s notation.
###### Definition 1.
Let $f$ be a meromorphic function on the closure of $A_{\frac{1}{\tau}r},$
$f\not\equiv 0.$ The function
$\displaystyle\displaystyle T(\tau,r;f)=N(\tau,r;f)+m(\tau,r;f)+$
$\displaystyle\displaystyle+c_{f}\log\frac{\tau}{r}\quad\tau\geq 1,\quad r\geq
1,.$ (15)
where
$\displaystyle\displaystyle c_{f}=\frac{1}{2\pi}\int\limits_{E_{f}^{+}}{\rm
Im}\left(\frac{f^{\prime}}{f}\,dz\right)+\frac{1}{4\pi}\int\limits_{E_{f}^{0}}{\rm
Im}\left(\frac{f^{\prime}}{f}\,dz\right)$ (16)
is called the characteristic of $f.$
A strange example. Let $m\in\mathbb{N}.$ Then
$\displaystyle\displaystyle
T(\tau,r;z^{m})=T(\tau,r;z^{-m})=m\,\frac{\log\tau+\log r}{2}.$
Note, that the introduced characteristic fits not only for meromorphic
functions in the punctured plane but in arbitrary annuli including punctured
disks.
###### Theorem 1.
Let $f,$ $f\not\equiv 0,$ be a meromorphic function on the annulus
$A_{s_{0}r_{0}},$ $s_{0}<1<r_{0}.$ Then the function $T(\tau,r;f)$ is non-
negative, continuous, non-decreasing and convex with respect to logarithm of
each variable for $1\leq\tau<1/s_{0},$ $1\leq r<r_{0},$
$\displaystyle\displaystyle T(1,1;f)=0\ \ {\mbox{a}nd}\ \
T(\tau,r;\frac{1}{f})=T(\tau,r;f).$
If $A_{s_{0}r_{0}}=\mathbb{C}\setminus\\{0\\},$ the function $f$ has the
meromorphic continuation at the origin and $T(r,f)$ is its Nevanlinna
characteristic, then
$\displaystyle\displaystyle T(r,f)-2\,T(1,f)\leq T(r,r;f)\leq T(r,f),\quad
r\geq 1.$ (17)
###### Proof.
The indicated properties follow immediately from (LABEL:eq16), Lemma 4 and
Lemma 3. Moreover, relation (LABEL:eq16) implies
$\displaystyle\displaystyle\tau\,\frac{\partial
T}{\partial\tau}=\frac{1}{2\pi}\int\limits_{0}^{2\pi}n\left(\frac{1}{\tau},1;\frac{1}{f-e^{i\varphi}}\right)\,d\varphi+\frac{1}{4\pi}\int\limits_{0}^{2\pi}n\left(\mathbb{T},\frac{1}{f-e^{i\varphi}}\right)\,d\varphi$
and
$\displaystyle\displaystyle r\,\frac{\partial T}{\partial
r}=\frac{1}{2\pi}\int\limits_{0}^{2\pi}n\left(1,r;\frac{1}{f-e^{i\varphi}}\right)\,d\varphi+\frac{1}{4\pi}\int\limits_{0}^{2\pi}n\left(\mathbb{T},\frac{1}{f-e^{i\varphi}}\right)\,d\varphi$
at the points of continuity of $n.$ Thus,
$\displaystyle\displaystyle\tau\,\frac{\partial
T}{\partial\tau}+r\,\frac{\partial T}{\partial
r}=\frac{1}{2\pi}\int\limits_{0}^{2\pi}n\left(\frac{1}{\tau},r;\frac{1}{f-e^{i\varphi}}\right)\,d\varphi.$
It seems, that the last identity with some boundary condition can be a
definition of $T(\tau,r;f).$
Verify also the identity $T(\tau,r;1/f)=T(\tau,r;f).$ It follows from Version
2 of Jensen’s theorem (see (LABEL:eq15)) and the relation
$\displaystyle\displaystyle\frac{1}{2}\,\nu(1,f)=c_{f}-c_{\frac{1}{f}}.$ (18)
Relation (LABEL:eq35) is in [1]. ∎
## $\bf 5^{0}.$ First Fundamental Theorem (FFT)
FFT. Let $f$ be a non-constant meromorphic function in
$\mathbb{C}\setminus\\{0\\}.$ Then for each $a\in\mathbb{C}$
$\displaystyle\displaystyle
N\left(\tau,r;\frac{1}{f-a}\right)+m\left(\tau,r;\frac{1}{f-a}\right)=$
$\displaystyle\displaystyle=T(\tau,r;f)+\varepsilon_{1}(\tau,r,a)+\varepsilon_{2}(a)\log\frac{\tau}{r},\quad\tau\geq
1,\quad r\geq 1$ (19)
where
$\displaystyle\displaystyle\left|\varepsilon_{1}(\tau,r,a)\right|\leq
4\,\log^{+}|a|+4\log 2,$
and
$\displaystyle\displaystyle\left|\varepsilon_{2}(a)\right|\leq C,\qquad C\
{\mbox{is a constant.}}$ (20)
###### Proof.
As in classical Nevanlinna theory we apply Version 2 of Jensen’s theorem (see
(LABEL:eq15)) for $f(z)-a.$ Then
$\displaystyle\displaystyle
N\left(\tau,r;\frac{1}{f-a}\right)+m\left(\tau,r;\frac{1}{f-a}\right)=N\left(\tau,r,f\right)+$
$\displaystyle\displaystyle+m\left(\tau,r;f-a\right)+\frac{1}{2}\,\nu(1,f-a)\log\frac{\tau}{r}=$
$\displaystyle\displaystyle=T\left(\tau,r;f\right)+m\left(\tau,r;f-a\right)-m\left(\tau,r;f\right)+$
$\displaystyle\displaystyle+\left(\frac{1}{2}\,\nu(1,f-a)-c_{f}\right)\log\frac{\tau}{r}.$
Set
$\displaystyle\displaystyle\varepsilon_{1}(\tau,r;a)=m(\tau,r;f-a)-m(\tau,r;f)$
and
$\displaystyle\displaystyle\varepsilon_{2}(a)=\frac{1}{2}\,\nu(1,f-a)-c_{f}.$
We have, as usually
$\displaystyle\displaystyle\left|\varepsilon_{1}(\tau,r,a)\right|\leq
4\log^{+}|a|+4\log 2.$
Applying (LABEL:eq18) for $\zeta=a$ we obtain
$\displaystyle\displaystyle\varepsilon_{2}(a)=\frac{1}{2\pi}\int\limits_{E_{f}^{-}}{\rm
Im}\left(\frac{f^{\prime}}{f}\,dz\right)+\frac{1}{4\pi}\int\limits_{E_{f}^{0}}{\rm
Im}\left(\frac{f^{\prime}}{f}\,dz\right)-$
$\displaystyle\displaystyle-\frac{1}{2\pi}\int\limits_{\mathbb{T}}{\rm
Im}\left(\frac{af^{\prime}}{f(a-f)}\,dz\right).$
Since the last integral is bounded as a function of $a$ then (LABEL:eq43) is
valid. ∎
## References
* [1] Andriy Kondratyuk and and Ilpo Laine, Meromorphic functions in multiply connected domains, Fourier series methods in complex analysis (Mekrijärvi, 2005). Univ. Joensuu Dept. Math. Rep. Ser. No 10 (2006), pp. 9–111.
| arxiv-papers | 2008-07-08T13:51:22 | 2024-09-04T02:48:56.631789 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Andriy Kondratyuk",
"submitter": "Andriy Kondratyuk Andriyovych",
"url": "https://arxiv.org/abs/0807.1247"
} |
0807.1413 | # Almost Sure Stabilization for Adaptive Controls of Regime-switching LQ
Systems with A Hidden Markov Chain
Bernard Bercu Université Bordeaux 1, Institut de Mathématiques de Bordeaux,
UMR 5251 and INRIA Bordeaux Sud-Ouest, Team CQFD, 351 cours de la Libération,
33405 Talence cedex, France. Email: Bernard.Bercu@math.u-bordeaux1.fr François
Dufour Université Bordeaux 1, Institut de Mathématiques de Bordeaux, UMR 5251
and INRIA Bordeaux Sud-Ouest, Team CQFD, 351 cours de la Libération, 33405
Talence cedex, France. Email: Francois.Dufour@math.u-bordeaux1.fr G. George
Yin Department of Mathematics, Wayne State University, Detroit, Michigan
48202. Email: gyin@math.wayne.edu
###### Abstract
This work is devoted to the almost sure stabilization of adaptive control
systems that involve an unknown Markov chain. The control system displays
continuous dynamics represented by differential equations and discrete events
given by a hidden Markov chain. Different from previous work on stabilization
of adaptive controlled systems with a hidden Markov chain, where average
criteria were considered, this work focuses on the almost sure stabilization
or sample path stabilization of the underlying processes. Under simple
conditions, it is shown that as long as the feedback controls have linear
growth in the continuous component, the resulting process is regular.
Moreover, by appropriate choice of the Lyapunov functions, it is shown that
the adaptive system is stabilizable almost surely. As a by-product, it is also
established that the controlled process is positive recurrent.
Key words. Adaptive control, hidden Markov chain, almost sure stabilization.
Abbreviated title. Almost Sure Stabilization of Adaptive Controls with
Switching
## 1 Introduction
This work deals with almost sure stabilization of adaptive control systems in
continuous-time with an unknown parameter process that is a hidden Markov
chain. The systems belong to the class of partially observed control systems.
Naturally, one estimates the parameter process by using nonlinear filtering
techniques and then uses the estimator in the systems in order to design
adaptive control strategies. The motivation of our study stems from
consideration of the following problem. Let us begin with a hybrid linear
quadratic (LQ) problem
$\dot{X}(t)=A_{\alpha(t)}X(t)+B_{\alpha(t)}U(t)$
where $\alpha(t)$ is a continuous-time Markov chain taking values in a finite
set ${\mathcal{M}}=\\{1,\ldots,m\\}$, $A_{i}$ and $B_{i}$ for
$i\in{\mathcal{M}}$ are matrices with compatible dimensions, and $U(t)$ is the
control process. One can observe that different from the traditional setup of
LQ problems, the system matrices $A_{i}$ and $B_{i}$ are both subject to
random switching influence. At any given instance, these coefficient matrices
are chosen from a set ${\mathcal{M}}$ with a finite number of candidates. The
selection rule is dictated by the modulating switching process $\alpha(t)$
that jump changes from one state to another at random times. Such systems have
enjoyed numerous applications in emerging application areas as financial
engineering, wireless communications, as well as in existing applications. A
particular important problem concerns the asymptotic behavior of such systems
when they are in operations for a long time. Our interest lies in finding
admissible controls so that the resulting system will be almost surely
stabilized. An added difficulty is that the process $X(t)$ can only be
observed with an additive noise
$dX(t)=[A_{\alpha(t)}X(t)+B_{\alpha(t)}U(t)]dt+dW(t).$
For such partially observed systems, it is natural to use nonlinear filtering
techniques. The associated filter is known as the Wonham filter [17], which is
one of a handful of finite dimensional filters in existence.
Linear quadratic (LQ) regulators appear to present rather simple structures.
Meanwhile, there are so many applications that can be described by such
processes. We refer the reader to [1, 2, 6, 16] for some recent work on the
associated control, estimation, and optimization problems for hybrid systems.
Emerging applications have also been found in manufacturing systems, in which
a Markov chain is used to represent the capacity of an unreliable machine, in
wireless communication, in which a Markov chain is used to depict randomly
time varying signals or channels. In financial engineering, a geometric
Brownian motion model for a stock is frequently used. The traditional setup
can be described by a linear stochastic differential equation, where both the
appreciate rate and volatility are constant. However, it has been recognized
that such a formulation is far from realistic. Very often, there are
additional randomness due to the variation of interest rates and other random
environment factors. For example, the well-known Markowitz’s mean-variance
portfolio selection is one of the LQ control problems. Some recent effort for
mean-variance control problems has been on obtaining optimal portfolio
selections when both the appreciation rate and volatility depend on a Markov
chain. For all of the applications mentioned above, practical considerations
often lead to deal with unobservable Markov chains. In many situation, the
Markov chain is used to model random environment. Thus, treat adaptive
controls, stability, and stabilization of such systems will have significant
impact to many applications.
There have been continued interest in dealing with hybrid systems under a
Markov switching. In [16], stabilization for robust controls of jump LQ
control problems was investigated. In [6], both controllability and
stabilizability of jump linear LQ systems were considered. Stability under
random perturbations of Markov chain type can be traced back to the work [8].
This line of work has been substantially expanded to diffusion systems in [9,
11]. Recently, renewed interests have been shown to deal with switching
diffusions; see for example [10, 14, 15, 20] among others.
In the literature, stabilization of continuous-time, adaptive control systems
with hidden Markov chains were considered in [3, 5]. In both of these
references, averaging criteria were used for the purpose of stabilization. To
be more precise, adaptive control strategies were developed in [5] to make
both the system and the control have bounded second moment in the sense
$\limsup_{t\to\infty}{\mathbb{E}}[|X(t)|^{2}+|U(t)|^{2}]<\infty,$
whereas adaptive controls were obtained in [3] to have the second moments of
the averages of both the system and control bounded in the sense
$\limsup_{t\to\infty}{1\over
t}{\mathbb{E}}\left[\int^{t}_{0}[|X(s)|^{2}+|U(s)|^{2}]ds\right]<\infty.$
In comparisons with the aforementioned references, it is a worthwhile effort
to examine the pathwise stabilization of the associated LQ problems under
partial observations. First, to be of any practical use in applications, the
system resulting from an adaptive control law should not allow wild behavior
in the sample paths. Secondly, owing to the use of adaptive control
strategies, known results in stability and stabilization in Markov-modulated
stochastic systems cannot be applied directly. As will be seen in later
section, the feedback adaptive controls render difficulty in analyzing the
underlying systems. Certain functions associated with the diffusion matrix in
fact grow faster than normally is allowed in the standard analysis. When
averaged criteria are used, this kind of difficulty will not show up since by
taking expectation, we can easily average out the Brownian motion term.
However, when pathwise criteria are the used, we can no longer use the
argument based on using expectations. Thus the consideration of pathwise
stabilization is both practically necessary and theoretically interesting.
To begin our quest of finding admissible controls that stabilize the systems
almost surely, we answer the question if the controlled process is regular. By
a process being regular we mean that it does not have finite explosion time
with probability one. We establish regularity under feedback controls under
linear growth conditions for the feedback controls. Then, we develop
sufficient conditions and admissible adaptive controls under which the system
is stabilizable. Moreover, as a by-product, we also establish positive
recurrence of the underlying processes as a corollary of our stabilization
result. For a deterministic system given by a differential equation, if the
solutions are ultimately uniformly bounded, then it is Lagrange stable. For
stochastic systems, almost sure boundedness excludes many cases (e.g., any
systems perturbed by a white noise). Thus, in lieu of such a boundedness, one
seeks stability in certain weak sense. So a process is recurrent if it starts
from a point outside a compact set, the process will return to the bounded set
with probability one. We say the process is positive recurrent if the expected
return time is finite. In fact, positive recurrence is termed weak stability
for diffusion processes in [18]. For a practical system, no finite explosion
time is a must. In addition, starting from a point outside of a bounded set,
the system should be able to return to the set infinitely often with
probability one. Moreover, the average return time cannot be infinitely long
otherwise the controlled system is useless. Thus, regularity and recurrence of
adaptive control systems can be viewed as “practical” stability conditions.
The rest of the paper is organised as follows. Section 2 presents the
formulation and preliminaries. Section 3 investigates the regularity of the
underlying process. Our conclusion is that, as long as the feedback controls
have linear growth, the resulting systems will be regular. Section 4 proceeds
with the study of stabilization. We conclude the paper with some additional
remarks in Section 5. In order to preserve the flow of presentation, proofs of
a couple of technical results are postponed to two appendices to facilitate
the reading.
## 2 Formulation and Preliminary
### 2.1 Problem Setup
Denote by $(\Omega,{\mathcal{A}},{\mathbb{P}})$ a probability space with an
associated nondecreasing family of $\sigma$-algebras $({\mathcal{F}}_{t})$.
Let $\alpha(t)$ be a continuous-time Markov chain with a finite state space
${\mathcal{M}}=\\{1,\ldots,m\\}$ and transition rate matrix
$\Pi=(\pi_{ij})\in{\mathbb{R}}^{m\times m}$, and $W(t)$ be a standard
${\mathbb{R}}^{n}$-valued Brownian motion. In the above and hereafter,
$A^{\prime}$ denotes the transpose of a matrix $A$,
$|A|=\sqrt{\hbox{tr}(AA^{\prime})}$ is the trace norm of $A$, and
$|v|=\sqrt{v^{\prime}v}$ is the usual Euclidean norm of a vector $v$.
Assume throughout the paper that $W(t)$ and $\alpha(t)$ are independent. Let
$X(t)\in{\mathbb{R}}^{n}$ and $U(t)\in{\mathbb{R}}^{d}$ be the state and
control processes, respectively. For $i\in{\mathcal{M}}$,
$A_{i}\in{\mathbb{R}}^{n\times n}$ and $B_{i}\in{\mathbb{R}}^{n\times d}$ are
matrices with appropriate dimensions. Our main interest focuses on the
following regime-switching stochastic system
$\begin{array}[]{ll}&\\!\\!\\!\displaystyle
dX(t)=[A_{\alpha(t)}X(t)+B_{\alpha(t)}U(t)]dt+dW(t)\end{array}$ (2.1)
with square integrable initial condition $X(0)=x$ As in [3, 5], denoting the
column vector of ${\mathbb{R}}^{m}$ of indicator functions by
$\Phi(t)=(\mbox{1}\kern-2.5pt\mbox{I}_{\\{\alpha(t)=1)\\}},\ldots,\mbox{1}\kern-2.5pt\mbox{I}_{\\{\alpha(t)=m\\}})^{\prime}$
where $\mbox{1}\kern-2.5pt\mbox{I}_{E}$ stands for the usual indicator
function of the event $E$, we may present the dynamics of the Markov chain by
$d\Phi(t)=\Pi^{\prime}\Phi(t)dt+dM(t).$
The process $M(t)$ is an ${\mathbb{R}}^{m}$-valued square integrable
martingale with right continuous trajectories. The independence of $\alpha(t)$
and $W(t)$ implies that of $\Phi(t)$ and $W(t)$. In all the sequel, we also
assume that $x$, $\Phi(t)$, and $W(t)$ are mutually independent. Consider the
quadratic cost criterion
$J_{T}(x,\Phi,U)={\mathbb{E}}_{x,\alpha}\Bigg{[}{1\over
2}\int^{T}_{0}[X^{\prime}(t)Q_{\alpha(t)}X(t)+U^{\prime}(t)R_{\alpha(t)}U(t))]dt\Bigg{]},$
where ${\mathbb{E}}_{x,\alpha}$ denotes the expectation with initial
conditions $X(0)=x$, $\alpha(0)=\alpha$, and for each $i\in{\mathcal{M}}$,
$Q_{i}$ is a symmetric positive semi-definite matrix, and $R_{i}$ is a
symmetric positive definite matrix.
One of the main features of the system considered here is that the Markov
chain under consideration is a hidden one. As treated in [3, 5], the essence
is that we are dealing with a system (2.1) with unknown mode that switches
back and forth among a finite set at random times. But different from previous
consideration, we wish to establish the regularity of the process and to find
conditions ensuring almost sure stabilization. The almost sure stabilization
poses new challenges and difficulties since we cannot average out the
martingale term by means of taking expectations. Compared with the
aforementioned papers, different techniques are needed. Here the keystone is
to find a suitable Lyapunov function.
Throughout the paper, the process $X(t)$ is assumed to be observable, but this
is not the case for the switching process $\alpha(t)$. The problem belongs to
the category of controls with partial observations. Observing $\alpha(t)$
through the adaptive control process with Gaussian white noise brings us to
the framework of the setup of Wonham filtering problems [17]. Denote by
${\mathcal{F}}^{X}_{t}$ the $\sigma$-algebra generated by
${\mathcal{F}}^{X}_{t}=\sigma\\{X(s),s\leq t\\}$. For the problem of interest,
a control is said to be admissible if for each $t\geq 0$, $U(t)$ is
${\mathcal{F}}^{X}_{t}$-measurable. We are now in position to state precisely
the problem we wish to study.
Problem statement. Under the setup presented so far, we aim to solve the
following problem.
1. 1.
We analyze (2.1) and obtain conditions under which the system will be regular.
Hence, our goal is to propose sufficient conditions ensuring the process will
not have finite explosion time. We show that, as long as the feedback control
$U$ (as a function of $x$) has linear growth in $x$, the resulting adaptive
control system will be regular.
2. 2.
We design admissible adaptive controls and provide sufficient conditions that
stabilize the closed-loop system almost surely (a.s.). Loosely, the sufficient
condition ensures that for almost all sample points $\omega$ (except a null
set), the corresponding system will be stabilizable. The precise definition of
almost sure stabilization will be provided in the next section.
### 2.2 Preliminary
As in [3, 5], we convert this partially observed system to a control process
with complete observation. It entails to replace the hidden state $\Phi(t)$ by
its estimator, namely the well-known Wonham filter $\widehat{\Phi}(t)$. Using
feedback control $U(t)=U(X(t),\widehat{\Phi}(t))$, we shall need the following
notation
$\begin{array}[]{ll}&\\!\\!\\!\displaystyle\widehat{\Phi}_{i}(t)={\mathbb{E}}[\mbox{1}\kern-2.5pt\mbox{I}_{\\{\alpha(t)=i\\}}|{\mathcal{F}}^{X}_{t}],\\\
&\\!\\!\\!\displaystyle\widehat{\Phi}(t)=(\widehat{\Phi}_{1}(t),\ldots,\widehat{\Phi}_{m}(t))^{\prime}\in{\mathbb{R}}^{m},\\\
&\\!\\!\\!\displaystyle
C(X(t))=(A_{1}X(t)+B_{1}U(t),\ldots,A_{m}X(t)+B_{m}U(t))\in{\mathbb{R}}^{n\times
m},\\\ &\\!\\!\\!\displaystyle D(\varphi)=({\rm
diag}(\varphi)-\varphi\varphi^{\prime})\ \hbox{ for }\
\varphi\in{\mathbb{R}}^{m},\\\ &\\!\\!\\!\displaystyle{\rm diag}(\varphi)={\rm
diag}(\varphi_{1},\ldots,\varphi_{m}).\end{array}$ (2.2)
Denote also the innovation process by
$dV(t)=dX(t)-C(X(t))\widehat{\Phi}(t)dt.$
Using the above notation, we can rewrite the converted completely observable
system as
$d\pmatrix{X(t)\cr\widehat{\Phi}(t)}=\pmatrix{C(X(t))\widehat{\Phi}(t)\cr\Pi^{\prime}\widehat{\Phi}(t)}dt+\pmatrix{I_{n}\cr
D(\widehat{\Phi}(t))C(X(t))^{\prime}}dV(t),$ (2.3)
where $I_{n}$ stands for the identity matrix of order $n$.
###### Remark 2.1
. Before proceeding further, we shall make a few remarks.
* •
The form $C(X(t))$ indicates the $X(t)$-dependence. When the feedback control
$U(t)$ is of linear form, it depends on $X(t)$ linearly. This point will be
used in what follows.
* •
The equivalent and completely observable system can be viewed as a controlled
diffusion, in which the usual diffusion term is replaced by
$\pmatrix{I_{n}\cr D(\widehat{\Phi}(t))C(X(t))^{\prime}}$
and the driven Brownian motion is given by $V(t)$.
* •
When linear feedback control is used, both the drift and diffusion grow at
most linearly, which is a useful observation.
* •
Since $\widehat{\Phi}(t)$ is the probability conditioned on the observation,
for each $t\geq 0$ and each $i\in{\mathcal{M}}$, $\widehat{\Phi}_{i}(t)\geq 0$
with
$\sum^{m}_{i=1}\widehat{\Phi}_{i}(t)=1.$
Denote the joint vector by
$Y(t)=(X(t),\widehat{\Phi}(t))^{\prime}\in{\mathbb{R}}^{n+m}$. In what
follows, we often consider $|Y(t)|\geq r$ for some $r>0$, where $|Y|$ is the
usual Euclidean norm. Denote by $N(0;r)\in{\mathbb{R}}^{n+m}$ the neighborhood
centered at 0 with radius $r$. Using the notation defined in (2.2) associated
with the stochastic differential equation (2.3), we define the following
operator. For each sufficiently smooth real-valued function
$h:{\mathbb{R}}^{n+m}/N(0;r)\mapsto{\mathbb{R}}$, define
$\begin{array}[]{ll}{\cal L}h(y)&\\!\\!\\!\displaystyle={\cal
L}h(x,\varphi)\\\ &\\!\\!\\!\displaystyle=\Bigl{(}\nabla
h(x,\varphi)\Bigr{)}^{\prime}\pmatrix{C(x)\varphi\cr\Pi^{\prime}\varphi}\\\
&\displaystyle\ +{1\over
2}\hbox{tr}\bigg{(}\pmatrix{I_{n}&C(x)D(\varphi)^{\prime}}\nabla^{2}h(x,\varphi)\pmatrix{I_{n}&C(x)D(\varphi)^{\prime}}^{\prime}\bigg{)}\end{array}$
(2.4)
where $\nabla h$ and $\nabla^{2}h$ are the gradient and Hessian of $h$,
respectively.
## 3 Regularity
First, let us recall the definition of regularity. According to [9], the
Markov process
$Y(t)=\pmatrix{X(t)\cr\widehat{\Phi}(t)}$
is regular, if for any $0<T<\infty$,
${\mathbb{P}}\left(\sup_{0\leq t\leq T}|Y(t)|=\infty\right)=0.$
Roughly speaking, regularity ensures the process under consideration will not
have finite explosion time. For our adaptive control systems, we proceed to
show that under linear feedback control, the systems is regular.
###### Theorem 3.1
. Assume that the feedback control $U(t)=U(X(t),\widehat{\Phi}(t))$ is
admissible and that it grows at most linearly in $X(t)$. Then, the feedback
control system (2.3) is regular.
###### Remark 3.2
. In fact, for our problem, we are mainly interested in linear (in $x$
variable) feedback controls. In this case, the linear growth condition is
clearly satisfied.
Proof. Let $\Theta$ be an open set in ${\mathbb{R}}^{n+m}$ and denote
$O=\Bigl{\\{}y=(x,\varphi)^{\prime}\in\Theta,\varphi=(\varphi_{1},\ldots,\varphi_{m})\hbox{
satisfying }\varphi_{i}\geq 0\hbox{ for }i\in{\mathcal{M}},\hbox{ and
}\sum^{m}_{i=1}\varphi_{i}=1\Bigr{\\}}.$
We first observe that both the drift and the diffusion coefficient given in
(2.3) satisfy the linear growth and Lipschitz condition in every open set in
$O\subset{\mathbb{R}}^{n+m}$. Thus, to prove the regularity, using the result
in [9], we only need to show that there is a nonnegative function
$\mathcal{U}$ which is twice continuously differentiable in $O_{r}=\\{y\in
O,|y|>r\\}$ for some $r>0$ with $y=(x,\varphi)^{\prime}$ such that
$\inf_{|y|>R}\mathcal{U}(y)\to\infty\ \hbox{ as }\ R\to\infty,$ (3.1)
and that there is an $\gamma>0$ satisfying
${\cal L}\mathcal{U}(y)\leq\gamma\mathcal{U}(y).$ (3.2)
Thus, to verify the regularity of the process $Y(t)$, all needed is to
construct an appropriate Lyapunov function $\mathcal{U}$. Note that we only
need a Lyapunov function that is smooth and defined in the complement of a
sphere. Equivalently, we only need the smoothness of the Lyapunov function to
be in a deleted neighborhood of the origin. To this end, take $r=1$ and denote
by $O_{1}$ the set
$\begin{array}[]{ll}O_{1}&\\!\\!\\!\displaystyle=\Bigl{\\{}y=(x,\varphi)^{\prime}\in{\mathbb{R}}^{n+m},\
|y|>1\hbox{ and }\varphi=(\varphi_{1},\ldots,\varphi_{m})\hbox{ satisfying
}\\\ &\displaystyle\qquad\qquad\varphi_{i}\geq 0\hbox{ for
}i\in{\mathcal{M}},\hbox{ and
}\sum^{m}_{i=1}\varphi_{i}=1\Bigr{\\}}.\end{array}$ (3.3)
Define $\mathcal{U}:O_{1}\mapsto{\mathbb{R}}$ as $\mathcal{U}(y)=|y|$. It is
easily checked that condition (3.1) holds. Moreover, we have
$\nabla\Big{|}\pmatrix{x\cr\varphi}\Big{|}={\pmatrix{x\cr\varphi}\over\Big{|}\pmatrix{x\cr\varphi}\Big{|}},$
and
$\nabla^{2}\Big{|}\pmatrix{x\cr\varphi}\Big{|}={I_{n+m}\over\Big{|}\pmatrix{x\cr\varphi}\Big{|}}-{\pmatrix{xx^{\prime}&x\varphi^{\prime}\cr\varphi
x^{\prime}&\varphi\varphi^{\prime}}\over\Big{|}\pmatrix{x\cr\varphi}\Big{|}^{3}}.$
Consequently, it follows from (2.4) that
$\begin{array}[]{ll}{\cal
L}\mathcal{U}(x,\varphi)&\\!\\!\\!\displaystyle={1\over\Big{|}\pmatrix{x\cr\varphi}\Big{|}}(x^{\prime}C(x)\varphi+\varphi^{\prime}\Pi^{\prime}\varphi)\\\
&\displaystyle\ +{1\over
2\Big{|}\pmatrix{x\cr\varphi}\Big{|}}(n+\hbox{tr}(D(\varphi)C^{\prime}(x)C(x)D^{\prime}(\varphi)))\\\
&\displaystyle\ -{1\over
2\Big{|}\pmatrix{x\cr\varphi}\Big{|}^{3}}\hbox{tr}\bigg{(}\pmatrix{I_{n}&C(x)D(\varphi)^{\prime}}\pmatrix{xx^{\prime}&x\varphi^{\prime}\cr\varphi
x^{\prime}&\varphi\varphi^{\prime}}\pmatrix{I_{n}&C(x)D(\varphi)^{\prime}}^{\prime}\bigg{)}.\end{array}$
(3.4)
Note that the set that we are working with is $O_{1}$ defined in (3.3). In
particular, the use of $O_{1}$ yields that for any $y\in O_{1}$, $|\varphi|$
is always bounded. We also note that owing to the definition of $C(x)$ and the
linear growth feedback controls used, $C(x)$ is a function grows at mostly
linearly in $x$. To proceed, henceforth, use $\gamma$ as a generic positive
constant with the convention that $\gamma+\gamma=\gamma$ and
$\gamma\gamma=\gamma$ in an appropriate sense. It follows that for the terms
on the third line from bottom of (3.4), for $|(x,\varphi)^{\prime}|$ large
enough,
$\begin{array}[]{rl}&\\!\\!\\!\displaystyle{1\over\Big{|}\pmatrix{x\cr\varphi}\Big{|}}\Bigl{|}x^{\prime}C(x)\varphi+\varphi^{\prime}\Pi^{\prime}\varphi\Bigr{|}\leq{\gamma\over\Big{|}\pmatrix{x\cr\varphi}\Big{|}}\Big{|}\pmatrix{x\cr\varphi}\Big{|}^{2}\leq\gamma\Big{|}\pmatrix{x\cr\varphi}\Big{|}.\end{array}$
Likewise, for the next two term, we have
$\begin{array}[]{rl}&\\!\\!\\!\displaystyle{1\over
2\Big{|}\pmatrix{x\cr\varphi}\Big{|}}\Bigl{|}n+\hbox{tr}(D(\varphi)C^{\prime}(x)C(x)D^{\prime}(\varphi))\Bigr{|}\leq\gamma\Big{|}\pmatrix{x\cr\varphi}\Big{|}.\end{array}$
Combining the above estimates, we can deduce that
${\cal
L}\mathcal{U}(x,\varphi)\leq\gamma\Big{|}\pmatrix{x\cr\varphi}\Big{|}=\gamma\mathcal{U}(x,\varphi)$
for some $\gamma>0$. Consequently, the second condition (3.2) is satisfied.
Thus the regularity of the feedback control is obtained. $\qquad\Box$
## 4 Stabilization
In this section, we establish conditions under which the system of interest is
stabilizable in the almost sure sense. We first present the definition and
then proceed to find sufficient conditions for stabilization.
###### Definition 4.1
. System (2.1) or equivalently (2.3) is said to be almost surely stabilizable
if there is a feedback control law $U(t)$ such that the resulting trajectories
satisfy
$\limsup_{t\to\infty}{1\over t}\log|X(t)|\leq 0\ \hbox{ almost surely.}$ (4.1)
Note that the definition given in (4.1) is natural. When studying stability of
stochastic differential equations, especially for pathwise stability, one uses
the so-called $q$th-moment Lyapunov exponent
$\limsup_{t\to\infty}{1\over t}\log|X(t)|^{q}$
for some $q>0$. Here, roughly, we require that under the control law, the
first-moment Lyapunov exponent is non-positive.
### 4.1 Auxiliary Results
Before proceeding further, let us first recall a lemma, which is concerned
with the existence of the associated system of Riccati equations when
quadratic cost criteria are used. The proof of the lemma is given in [7].
###### Lemma 4.2
. Consider the system of Riccati equations
$A^{\prime}_{i}P_{i}+P_{i}A_{i}-P_{i}B_{i}R^{-1}B^{\prime}_{i}P_{i}+\sum^{m}_{j=1}\pi_{ij}P_{j}+Q=0,\
i\in{\mathcal{M}},$ (4.2)
where $Q\in{\mathbb{R}}^{n\times n}$ is symmetric and positive semi-definite,
and $R\in{\mathbb{R}}^{m\times m}$ is symmetric and positive definite. The
system (4.2) has a solution if and only if for each $i\in{\mathcal{M}}$, there
is a matrix $\overline{P}_{i}$ satisfying
$A^{\prime}_{i}\overline{P}_{i}+\overline{P}_{i}A_{i}-\overline{P}_{i}A_{i}-\overline{P}_{i}B_{i}R^{-1}B^{\prime}_{i}\overline{P}_{i}+\sum^{m}_{j=1}\pi_{ij}\overline{P}_{j}+Q\leq
0.$ (4.3)
Furthermore, if $Q$ is positive definite, so are $P_{i}$ for
$i\in{\mathcal{M}}$.
To carry out the analysis, we need some auxiliary results on the bounds of the
quadratic variation process. Before getting the almost sure bounds, we examine
the moment bounds for certain related martingales, which turn out to be
interesting in their own right. The main ingredient is the use of properties
of the associated Markov chain.
#### Moment Bounds
###### Proposition 4.3
. Consider the stochastic differential equation
$d\widehat{\Phi}(t)=\Pi^{\prime}\widehat{\Phi}(t)dt+D(\widehat{\Phi}(t))C(X(t))^{\prime}dV(t)$
(4.4)
and define the associate martingale
$N(t)=\int^{t}_{0}D(\widehat{\Phi}(s))C(X(s))^{\prime}dV(s).$ (4.5)
Suppose that the Markov chain $\alpha(t)$ is irreducible. Then, for some
positive constant $K$ independent of $t$,
${\mathbb{E}}\left[{1\over t}|N(t)|^{2}\right]\leq K.$ (4.6)
Proof. The proof is given in Appendix A. $\qquad\Box$
###### Remark 4.4
. It follows from the proof of Proposition 4.3 that the limit of the matrix
$S=\lim_{t\to\infty}{1\over
t}\int^{t}_{0}\int^{t}_{0}\Pi^{\prime}[\widehat{\Phi}(u)-\nu][\widehat{\Phi}(s)-\nu]^{\prime}\Pi
duds$
is finite. Clearly, this matrix is symmetric and positive semi definite. A
moment of reflect reveals that we can further prove the asymptotic normality.
That is
${1\over\sqrt{t}}\int^{t}_{0}\Pi^{\prime}[\widehat{\Phi}(s)-\nu]ds\
\hbox{converges in distribution to }\ \mathcal{N}(0,S)\ \hbox{ as
}t\to\infty.$
That is, a normalized sequence defined on the left-hand side above converges
in distribution to a normal random vector with mean 0 and covariance $S$.
Another ramification is that in lieu of considering the second-moment bounds,
we can deal with $q$th-moment bounds. In fact, using the same techniques, we
can show that for any integer $p>0$,
${\mathbb{E}}\left[\left|{1\over\sqrt{t}}\int^{t}_{0}\Pi^{\prime}[\widehat{\Phi}(s)-\nu]ds\right|^{2p}\right]<\infty.$
Hence, as the solution (4.4) is given by
$\widehat{\Phi}(t)=\widehat{\Phi}(0)+\int^{t}_{0}\Pi^{\prime}\widehat{\Phi}(s)ds+N(t)$
which means that
$N(t)=\widehat{\Phi}(t)-\widehat{\Phi}(0)-\int^{t}_{0}\Pi^{\prime}\widehat{\Phi}(s)ds,$
we obtain that
${\mathbb{E}}\left[\left|{1\over\sqrt{t}}N(t)\right|^{2p}\right]<\infty.$
Next, for odd exponents and for any integer $p\geq 1$, it follows from
Hölder’s inequality that
$\left({\mathbb{E}}\left[\left|{1\over\sqrt{t}}N(t)\right|^{2p-1}\right]\right)^{2p}\leq\left({\mathbb{E}}\left[\left|{1\over\sqrt{t}}N(t)\right|^{2p}\right]\right)^{2p-1}<\infty.$
Finally, we conclude that for any positive integer $q$,
${\mathbb{E}}\left[\left|{1\over\sqrt{t}}N(t)\right|^{q}\right]<\infty.$
#### Almost Sure Bounds
For the almost sure stabilization, we need to show that
${1\over t}|N(t)|^{2}\leq K\hskip 28.45274pt\hbox{ a.s.}$
for some $K>0$ independent of $t$.
###### Proposition 4.5
. Consider (4.4) and suppose that the Markov chain $\alpha(t)$ is irreducible.
Then, the quadratic variation of the process $N(t)$ satisfies
$\big{\langle}N,N\big{\rangle}_{t}\leq Kt$ where $K$ is some positive constant
independent of $t$. Therefore,
$\lim_{t\rightarrow\infty}{1\over t}N(t)=0\hskip 28.45274pt\hbox{ a.s.}$ (4.7)
Proof. The proof is given in Appendix B. $\qquad\Box$
### 4.2 Stabilization
###### Lemma 4.6
. Consider the set $\Delta$ defined by
$\begin{array}[]{ll}\Delta&\\!\\!\\!\displaystyle=\Bigl{\\{}(x,\varphi)\in{\mathbb{R}}^{n}\times{\mathbb{R}}^{m},\varphi=(\varphi_{1},\ldots,\varphi_{m})\hbox{
satisfying }\varphi_{i}\geq 0\hbox{ and
}\sum^{m}_{i=1}\varphi_{i}=1\Bigr{\\}}.\end{array}$
Denote
$P(\varphi)=\sum^{m}_{i=1}P_{i}\varphi_{i}.$ (4.8)
For some $\theta>0$, let $V_{\theta}(x,\varphi):\Delta\mapsto{\mathbb{R}}$
with $V_{\theta}(x,\varphi)=\log(\theta+x^{\prime}P(\varphi)x)$. Then, we have
$\begin{array}[]{ll}{\cal
L}V_{\theta}(x,\varphi)&\\!\\!\\!\displaystyle={1\over\theta+x^{\prime}P(\varphi)x}\bigg{(}2x^{\prime}P(\varphi)C(x)\varphi+(x^{\prime}\widetilde{P}x)^{\prime}\Pi^{\prime}\varphi\bigg{)}\\\
&\displaystyle\ +{1\over\theta+x^{\prime}P(\varphi)x}\hbox{\rm
tr}\bigg{(}P(\varphi)+2C(x)D(\varphi)^{\prime}x^{\prime}\widetilde{P}\bigg{)}\\\
&\displaystyle\ -{1\over 2(\theta+x^{\prime}P(\varphi)x)^{2}}\hbox{\rm
tr}\bigg{(}\bigl{(}I_{n}\>,\>C(x)D(\varphi)^{\prime}\bigr{)}\Lambda(x,\varphi)\bigl{(}I_{n}\>,\>C(x)D(\varphi)^{\prime}\bigr{)}^{\prime}\Bigg{)},\end{array}$
(4.9)
where
$\Lambda(x,\varphi)=\pmatrix{2P(\varphi)x\cr
x^{\prime}\widetilde{P}x}\pmatrix{2P(\varphi)x\cr
x^{\prime}\widetilde{P}x}^{\prime},$ $P=(P_{1},\ldots,P_{m})^{\prime},\quad
x^{\prime}\widetilde{P}x=(x^{\prime}P_{1}x,\ldots,x^{\prime}P_{m}x)^{\prime}\in{\mathbb{R}}^{m},$
$Px=(P_{1}x,\ldots,P_{m}x)\in{\mathbb{R}}^{n\times m},\quad
x^{\prime}\widetilde{P}=(\widetilde{P}x)^{\prime}\in{\mathbb{R}}^{m\times n}.$
Proof. We have
$\nabla\log(\theta+x^{\prime}P(\varphi)x)={\pmatrix{2P(\varphi)x\cr
x^{\prime}\widetilde{P}x}\over\theta+x^{\prime}P(\varphi)x}$
and
$\nabla^{2}\log(\theta+x^{\prime}P(\varphi)x)=-{\pmatrix{2P(\varphi)x\cr
x^{\prime}\widetilde{P}x}\pmatrix{2P(\varphi)x\cr
x^{\prime}\widetilde{P}x}^{\prime}\over(\theta+x^{\prime}P(\varphi)x)^{2}}+{2\pmatrix{P(\varphi)&\widetilde{P}x\cr
x^{\prime}\widetilde{P}&0_{m}}\over\theta+x^{\prime}P(\varphi)x}\vspace{2ex}$
where $0_{m}$ stands for a square matrix of order $m$ with all entries equal
to zero. Consequently, it follows from (2.4) that
$\begin{array}[]{ll}{\cal
L}V_{\theta}(x,\varphi)&\\!\\!\\!\displaystyle={1\over\theta+x^{\prime}P(\varphi)x}\Bigl{(}2x^{\prime}P(\varphi)C(x)\varphi+(x^{\prime}\widetilde{P}x)^{\prime}\Pi^{\prime}\varphi\Bigr{)}\\\
&\displaystyle\quad+{1\over 2}\hbox{\rm
tr}\Bigl{(}\bigl{(}I_{n}\>,\>C(x)D(\varphi)^{\prime}\bigr{)}\
\nabla^{2}V_{\theta}(x,\varphi)\
\bigl{(}I_{n}\>,\>C(x)D(\varphi)^{\prime}\bigr{)}^{\prime}\Bigr{)},\end{array}$
(4.10)
which immediately implies (4.9). $\qquad\Box$
For the purpose of stabilization, we also need an estimate on ${\cal
L}V_{\theta}(X(t),\widehat{\Phi}(t))$.
###### Lemma 4.7
. Assume that equation (4.3) is satisfied and that
$Q-{1\over
2}\Bigl{[}P_{i}B_{i}-P_{j}B_{j}\Bigr{]}R^{-1}\Bigl{[}P_{i}B_{i}-P_{j}B_{j}\Bigr{]}^{\prime}$
are positive definite matrices for all $(i,j)\in{\mathcal{M}}^{2}$ where
$P_{i}$ for $i\in{\mathcal{M}}$ are the solutions of the algebraic Riccati
equations given by (4.2). Then, the infinitesimal generator of the process
$(X(t),\widehat{\Phi}(t))$ associated with the feedback control law
$U(t)=-R^{-1}\sum^{m}_{i=1}\widehat{\Phi}_{i}(t)B^{\prime}_{i}P_{i}X(t),$
(4.11)
satisfies for some constant $\gamma>0$
${\cal L}V_{\theta}(X(t),\widehat{\Phi}(t))\leq{\gamma\over\theta}.$ (4.12)
Proof. We can deduce from Lemma 4.6 that
$\displaystyle\begin{array}[]{ll}{\cal
L}V_{\theta}(X(t),\widehat{\Phi}(t))&\\!\\!\\!\displaystyle\leq{1\over\theta+X(t)^{\prime}P(\widehat{\Phi}(t))X(t)}(2X(t)^{\prime}P(\widehat{\Phi}(t))C(X(t))\widehat{\Phi}(t))\\\
&\displaystyle\
+{1\over\theta+X(t)^{\prime}P(\widehat{\Phi}(t))X(t)}((X(t)^{\prime}\widetilde{P}X(t))^{\prime}\Pi^{\prime}\widehat{\Phi}(t))\\\
&\displaystyle\
+{1\over\theta+X(t)^{\prime}P(\widehat{\Phi}(t))X(t)}\hbox{tr}\bigg{(}P(\widehat{\Phi}(t))+2C(X(t))D(\widehat{\Phi}(t))^{\prime}X(t)^{\prime}\widetilde{P}\bigg{)}.\end{array}$
Therefore, following exactly the same lines as in [5], we obtain that
$\displaystyle\begin{array}[]{ll}{\cal
L}V_{\theta}(X(t),\widehat{\Phi}(t))&\\!\\!\\!\displaystyle\leq-{1\over\theta+X(t)^{\prime}P(\widehat{\Phi}(t))X(t)}\Bigl{(}X(t)^{\prime}\Bigl{[}Q\\\
&\displaystyle\
-\sum^{m}_{i=1}\sum^{m}_{j=1}{\widehat{\Phi}_{i}(t)\widehat{\Phi}_{j}(t)\over
2}\bigg{[}P_{i}B_{i}-P_{j}B_{j}\bigg{]}R^{-1}\bigg{[}P_{i}B_{i}-P_{j}B_{j}\bigg{]}^{\prime}\\\
&\displaystyle\
+\Bigl{(}\sum^{m}_{j=1}\widehat{\Phi}_{j}(t)B^{\prime}_{j}P_{j}\Bigr{)}^{\prime}R^{-1}\Bigl{(}\sum^{m}_{i=1}\widehat{\Phi}_{i}(t)B^{\prime}_{i}P_{i}\Bigr{)}\Bigr{]}X(t)-\hbox{tr}(P(\widehat{\Phi}(t)))\Bigr{)}.\end{array}$
Finally,
${\cal
L}V_{\theta}(X(t),\widehat{\Phi}(t))\leq{1\over\theta}\sum_{i=1}^{m}\hbox{tr}(P_{i})$
which completes the proof of Lemma 4.7.$\qquad\Box$
###### Theorem 4.8
. Assume that the conditions of Lemma 4.7 are satisfied. Then, the feedback
control law defined in equation (4.11) stabilizes the system (2.3) almost
surely.
Proof. It follows from Ito’s rule that
$\begin{array}[]{ll}V_{\theta}(X(t),\widehat{\Phi}(t))&\\!\\!\\!\displaystyle=V_{\theta}(x,\varphi)+\int^{t}_{0}{\cal
L}V_{\theta}(X(s),\widehat{\Phi}(s))ds+M(t)\end{array}$ (4.15)
with the initial condition $X(0)=x$ and $\widehat{\Phi}(0)=\varphi$ and the
martingale term
$M(t)=\int^{t}_{0}\Sigma(s)dV(s)$
where
$\begin{array}[]{rl}\Sigma(s)&\\!\\!\\!\displaystyle={1\over\Theta+X(s)^{\prime}P(\widehat{\Phi}(s))X(s)}\pmatrix{2X(s)^{\prime}P(\widehat{\Phi}(s))&(X(s)^{\prime}\widetilde{P}X(s))^{\prime}}\pmatrix{I_{n}\cr
D(\widehat{\Phi}(s))C(X(s))^{\prime}},\\\
&\\!\\!\\!\displaystyle={1\over\theta+X(s)^{\prime}P(\widehat{\Phi}(s))X(s)}\bigl{(}2X(s)^{\prime}P(\widehat{\Phi}(s))+(X(s)^{\prime}\widetilde{P}X(s))^{\prime}D(\widehat{\Phi}(s))C(X(s))^{\prime}\bigr{)}.\end{array}$
We can split the martingale $M(t)$ into two terms, $M(t)=N_{1}(t)+N_{2}(t)$
with
$\displaystyle N_{1}(t)$ $\displaystyle=$
$\displaystyle\int^{t}_{0}{2X(s)^{\prime}P(\widehat{\Phi}(s))\over\theta+X(s)^{\prime}P(\widehat{\Phi}(s))X(s)}dV(s),$
$\displaystyle N_{2}(t)$ $\displaystyle=$
$\displaystyle\int^{t}_{0}{(X(s)^{\prime}\widetilde{P}X(s))^{\prime}\over\theta+X(s)^{\prime}P(\widehat{\Phi}(s))X(s)}D(\widehat{\Phi}(s))C(X(s))^{\prime}dV(s).$
It is easy to see that
${4X(t)^{\prime}P(\widehat{\Phi}(t))P(\widehat{\Phi}(t))X(t)\over(\theta+X(t)^{\prime}P(\widehat{\Phi}(t))X(t))^{2}}\leq
K_{1}\hskip 14.22636pt\textrm{where}\hskip
14.22636ptK_{1}={m\over\theta}\max_{i\in{\mathcal{M}}}(\lambda_{max}(P_{i})).$
Then, the quadratic variation of $N_{1}(t)$ satisfies
$\big{\langle}N_{1},N_{1}\big{\rangle}_{t}\leq K_{1}t$ a.s. Consequently, we
deduce from the strong law of large numbers for local martingales [12] that
$\lim_{t\rightarrow\infty}{1\over t}N_{1}(t)=0\hskip 28.45274pt\hbox{ a.s.}$
(4.16)
In view of Proposition 4.5, one can also find a positive constant $K_{2}$,
independent of $t$, such that
$\begin{array}[]{ll}\big{\langle}N_{2},N_{2}\big{\rangle}_{t}&\\!\\!\\!\displaystyle=\int^{t}_{0}{|X(s)^{\prime}\widetilde{P}X(s)|^{2}\over(\theta+X(s)^{\prime}P(\widehat{\Phi}(s))X(s))^{2}}|D(\widehat{\Phi}(s))C(X(s))^{\prime}|^{2}ds\\\
&\\!\\!\\!\displaystyle\leq K_{2}t\hskip 28.45274pt\hbox{ a.s.}\end{array}$
(4.17)
It also ensures that
$\lim_{t\rightarrow\infty}{1\over t}N_{2}(t)=0\hskip 28.45274pt\hbox{ a.s.}$
(4.18)
Therefore, (4.16) and (4.18) imply that
$\lim_{t\rightarrow\infty}{1\over t}M(t)=0\hskip 28.45274pt\hbox{ a.s.}$
(4.19)
Thus, we find from (4.15) that
${1\over t}V_{\theta}(X(t),\widehat{\Phi}(t))={1\over
t}V_{\theta}(x,\varphi)+{1\over t}\int^{t}_{0}{\cal
L}V_{\theta}(X(s),\widehat{\Phi}(s))ds+o(1)\hskip 28.45274pt\hbox{ a.s.}$
Moreover, $V_{\theta}(x,\varphi)/t=o(1)$ as $t\to\infty$ a.s. By virtue of
Lemma 4.7, it follows that for all $\theta>0$
$\limsup_{t\to\infty}{1\over
t}V_{\theta}(X(t),\widehat{\Phi}(t))=\limsup_{t\to\infty}{1\over
t}\int^{t}_{0}{\cal
L}V_{\theta}(X(s),\widehat{\Phi}(s))ds\leq{\gamma\over\theta}\hskip
28.45274pt\hbox{ a.s.}$ (4.20)
Furthermore, one can observe that
$x^{\prime}P(\varphi)x\geq\lambda_{\min}(P(\varphi))|x|^{2}$ and since
$P(\varphi)$ is positive definite, the minimal eigenvalue of $P(\varphi)$ is
positive. Consequently,
$\log(\lambda_{\min}(P(\varphi)))+2\log(|x|)\leq\log(\theta+\lambda_{\min}(P(\varphi))|x|^{2})\leq\log(\theta+x^{\prime}P(\varphi)x)$
which leads to
${1\over
t}\Bigl{(}\log(\lambda_{\min}(P(\widehat{\Phi}(t))))+2\log(|X(t)|)\Bigr{)}\leq{1\over
t}V_{\theta}(X(t),\widehat{\Phi}(t)).$ (4.21)
Finally, we conclude from (4.20) and (4.21) that for all $\theta>0$,
$\limsup{1\over t}\log|X(t)|\leq{\gamma\over 2\theta}\hskip 28.45274pt\hbox{
a.s.}$
We complete the proof of Theorem 4.8 by taking the limit as $\theta$ tends to
infinity. $\qquad\Box$
###### Remark 4.9
. Normally, dealing with stochastic differential equations, to obtain the
almost sure bounds of the solutions, one often relies on the use of
appropriate Lyapunov functions to have the diffusion term of the process be
bounded after a transformation. Here, we are dealing with a martingale term
with some what faster rate of growth in $x$. Nevertheless, thanks to the
second component of the diffusion (4.4), the probabilistic meaning of
$\widehat{\Phi}(t)$ enables us to work around the obstacle. To obtain the
desired bounds, an alternative is to obtain an almost sure central limit
theorem. Here however, we take a different approach. The main point is the use
of Proposition 4.5.
Recall the notion of recurrence for the diffusion process
$(X(t),\widehat{\Phi}(t))$ starting at $X(0)=x$ and
$\widehat{\Phi}(0)=\varphi$. Consider an open set $O$ with compact closure,
and let
$\sigma^{x,\varphi}_{O}=\inf\Bigl{\\{}t>0,(X(t),\widehat{\Phi}(t))\in
O\Bigr{\\}}$
be the first entrance time of the diffusion to the set $O$. If
$(X(t),\widehat{\Phi}(t))$ is regular, it is recurrent with respect to $O$ if
${\mathbb{P}}\\{\sigma^{x,\varphi}_{O}<\infty\\}=1$ for any $(x,\varphi)\in
O^{c}$, where $O^{c}$ is the complement of $O$. A recurrent process with
finite mean recurrence time for some set $O$, is said to be positive recurrent
with respect to $O$, otherwise, the process is null recurrent with respect to
$O$. It has been proven in [9] that recurrence and positive recurrence are
independent of the set $O$ chosen. Thus, if it is recurrent (resp. positive
recurrent) with respect to $D$, then it is recurrent (resp. positive
recurrent) with respect to any other open set $\Theta$ in the domain of
interest. Looking over the proof of the stabilization presented, we could show
that for the Lyapunov function
$V_{0}(x,\varphi)=\log(x^{\prime}P(\varphi)x),$
one can find $\gamma>0$ such that for all $(x,\varphi)\in O^{c}$,
${\cal L}V_{0}(x,\varphi)\leq-\gamma.$ (4.22)
In view of the known result of positive recurrence of diffusion processes [9],
(4.22) is precisely a necessary and sufficient condition for positive
recurrence. Thus, we obtain the following result as a by-product.
###### Corollary 4.10
Under the conditions of Theorem 4.8, with the control law (4.11) used, the
diffusion systems (2.3) is positive recurrent.
We would like to add that the positive recurrence of the process is an
important property. It has engineering implication for various applications.
Essentially, it ensures that starting from a point outside of a bounded set,
the control laws enables the system to return to a compact set almost surely.
This may be viewed as a practical stability condition. In fact, Wonham used
the term weak stability for such a property in his paper [18].
## 5 Further Remarks
This paper has been concerned with stabilization in the almost sure sense of
an adaptive control system with linear dynamics modulated by an unknown Markov
chain. Under the framework of Wonham filtering, the underlying system is
converted to a fully observable system. Using feedback control that is linear
in the continuous state variable, we establish pathwise stabilization of the
process. Along the way of our study, we have also obtained regularity of the
underlying process. In addition, as a corollary, we have shown that under the
stabilizing control law, the resulting system is positive recurrent. These
results pave a way for practical consideration of stabilization of adaptive
controls of LQ systems with a hidden Markov chain. Several directions may be
worthwhile for further study and investigation.
* •
In our study, irreducibility of the Markov chain is used. We note that the
irreducibility ensures the spectrum gap condition or exponential decay in
(A.6) and (A.9) of Proposition 4.3 to hold. It will be interesting to see if
it is possible to remove this condition. Our initial thoughts are: Under
certain conditions, this might be possible. For example, if the Markov chain
has several irreducible classes such that the states in each class vary
rapidly, and among different classes, they change slowly. One may be able to
use the different time scales to overcome the difficulty under the framework
of time-scale separation using a singular perturbation approach. However, the
details on this need to be thoroughly worked out; they are in fact out of the
scope of the current paper.
* •
It will be interesting to design admissible controls and find sufficient
conditions for stabilization of LQ systems with a hidden Markov in discrete-
time.
* •
In our setup, the process $X(t)$ represents the noisy observation–hidden
Markov chain observed in white noise. A class of controlled regime-switching
diffusion systems provides a somewhat more complex setup. In such a system,
the dynamics are represented by switching diffusions with a hidden Markov
chain. The Markov chain is not observable but can only be observed in another
Gaussian white noise. That is, let us consider the controlled system
$\begin{array}[]{ll}&\\!\\!\\!\displaystyle
dY(t)=[A_{\alpha(t)}Y(t)+B_{\alpha(t)}U(t)]dt+\sigma_{\alpha(t)}dV(t)\\\
&\\!\\!\\!\displaystyle dX(t)=g_{\alpha(t)}dt+\rho(t)dW(t),\end{array}$ (5.1)
where $Y(t)$ and $X(t)$ are vector-valued processes with compatible dimensions
representing the state and observations, respectively, $V(t)$ and $W(t)$ are
independent multi-dimensional Brownian motions, and $\alpha(t)$ is the hidden
Markov chain with a finite state space. As was alluded to in the introduction,
one of the motivations is Markowitz’s mean-variance portfolio selections [19].
One may then pose similar stabilization problems.
* •
Recently, using regime-switching jump diffusions, which are switching
diffusions with additional external jumps of a compound Poisson process, for
modeling surplus in insurance risk has drawn much attention. A related problem
in the adaptive setup is a regime-switching jump diffusion system in which the
hidden Markov chain is observed similar to the observation in (5.1). One may
then proceed with the study of stabilization problems.
* •
In the study of stabilization, positive definiteness of certain matrices is
used (see Lemma 4.2). A challenging problem is to investigate the
stabilization problem with the positive definiteness removed for the system
given by (5.1). Here, the crucial point seems to rely on recent developments
in LQ problems with indefinite control weights [4]. One needs to use the
backward stochastic differential equations from the toolbox of stochastic
analysis.
All of these problems deserve further study and investigation.
## Appendix A.
This appendix is devoted to the proof of Proposition 4.3. It is divided into
several steps.
Step 1. We already saw that the solution (4.4) is given by
$\widehat{\Phi}(t)=\widehat{\Phi}(0)+\int^{t}_{0}\Pi^{\prime}\widehat{\Phi}(s)ds+N(t).$
Consequently
$N(t)=\widehat{\Phi}(t)-\widehat{\Phi}(0)-\int^{t}_{0}\Pi^{\prime}\widehat{\Phi}(s)ds.$
(A.1)
In view of (A.1), the probabilistic interpretation of $\widehat{\Phi}(t)$
implies that $N(t)$ is a martingale bounded almost surely for each $t>0$. We
proceed to obtain the moment bounds of $N(t)$.
Step 2. As $\Pi$ is the generator of the irreducible Markov chain $\alpha(t)$,
its unique stationary distribution $\nu$ satisfies $\Pi^{\prime}\nu=0$. Hence,
it follows that
$\int^{t}_{0}\Pi^{\prime}\widehat{\Phi}(s)ds=\int^{t}_{0}\Pi^{\prime}(\widehat{\Phi}(s)-\nu)ds.$
On the one hand, we clearly have from (4.5)
${\mathbb{E}}[|N(t)|^{2}]={\mathbb{E}}\left[\int^{t}_{0}|D(\widehat{\Phi}(s))C(X(s))^{\prime}|^{2}ds\right].$
(A.2)
On the other hand, we deduce from (A.1) that
$\begin{array}[]{ll}\displaystyle{1\over
t}{\mathbb{E}}[|N(t)|^{2}]&\\!\\!\\!\displaystyle={1\over
t}{\mathbb{E}}\left[\Bigl{|}\widehat{\Phi}(t)-\widehat{\Phi}(0)-\int^{t}_{0}\Pi^{\prime}(\widehat{\Phi}(s)-\nu)ds\Bigr{|}^{2}\right],\\\
&\\!\\!\\!\displaystyle\leq{2\over
t}{\mathbb{E}}\Bigl{[}|\widehat{\Phi}(t)-\widehat{\Phi}(0)|^{2}\Bigr{]}+{2\over
t}{\mathbb{E}}\left[\Bigl{|}\int^{t}_{0}\Pi^{\prime}(\widehat{\Phi}(s)-\nu)ds\Bigr{|}^{2}\right],\\\
&\\!\\!\\!\displaystyle\leq{2\over t}+{2\over
t}E\int^{t}_{0}\int^{t}_{0}\hbox{tr}\\{\Pi^{\prime}\Pi(\widehat{\Phi}(r)-\nu)(\widehat{\Phi}^{\prime}(s)-\nu^{\prime})\\}drds.\end{array}$
(A.3)
Consider the symmetric matrix
$G(r,s)=(g_{ij}(r,s))={\mathbb{E}}[(\widehat{\Phi}(r)-\nu)(\widehat{\Phi}^{\prime}(s)-\nu^{\prime})].$
One can observe that
$\begin{array}[]{ll}g_{ij}(r,s)&\\!\\!\\!\displaystyle={\mathbb{E}}[(\widehat{\Phi}(r)-\nu)_{i}(\widehat{\Phi}^{\prime}(s)-\nu^{\prime})_{j}],\\\
&\\!\\!\\!\displaystyle={\mathbb{E}}[({\mathbb{E}}[\mbox{1}\kern-2.5pt\mbox{I}_{\\{\alpha(r)=i\\}}|{\mathcal{F}}^{X}_{r}]-\nu_{i})({\mathbb{E}}[\mbox{1}\kern-2.5pt\mbox{I}_{\\{\alpha(s)=j\\}}|{\mathcal{F}}^{X}_{s}]-\nu_{j})],\\\
&\\!\\!\\!\displaystyle={\mathbb{E}}[{\mathbb{E}}[\mbox{1}\kern-2.5pt\mbox{I}_{\\{\alpha(r)=i\\}}|{\mathcal{F}}^{X}_{r}]{\mathbb{E}}[\mbox{1}\kern-2.5pt\mbox{I}_{\\{\alpha(s)=j\\}}|{\mathcal{F}}^{X}_{s}]]-\nu_{j}{\mathbb{E}}[{\mathbb{E}}[\mbox{1}\kern-2.5pt\mbox{I}_{\\{\alpha(r)=i\\}}|{\mathcal{F}}^{X}_{r}]]\\\
&\displaystyle\qquad\ -\
\nu_{i}{\mathbb{E}}[{\mathbb{E}}[\mbox{1}\kern-2.5pt\mbox{I}_{\\{\alpha(s)=j\\}}|{\mathcal{F}}^{X}_{s}]]+\nu_{i}\nu_{j},\\\
&\\!\\!\\!\displaystyle={\mathbb{E}}[{\mathbb{E}}[\mbox{1}\kern-2.5pt\mbox{I}_{\\{\alpha(r)=i\\}}|{\mathcal{F}}^{X}_{r}]{\mathbb{E}}[\mbox{1}\kern-2.5pt\mbox{I}_{\\{\alpha(s)=j\\}}|{\mathcal{F}}^{X}_{s}]]-\nu_{j}{\mathbb{P}}(\alpha(r)=i)\\\
&\displaystyle\qquad\ -\
\nu_{i}{\mathbb{P}}(\alpha(s)=j)+\nu_{i}\nu_{j}.\end{array}$ (A.4)
Note also by the Fubini Theorem that
$\begin{array}[]{rl}\displaystyle{1\over
t}\int^{t}_{0}\int^{t}_{0}g_{ij}(r,s)drds&\\!\\!\\!\displaystyle={1\over
t}\left(\int^{t}_{0}\int^{t}_{r}g_{ij}(r,s)drds+\int^{t}_{0}\int^{r}_{0}g_{ij}(r,s)drds\right),\\\
&\\!\\!\\!\displaystyle={1\over
t}\int^{t}_{0}\left(\int_{r}^{t}g_{ij}(r,s)ds\right)dr+{1\over
t}\int^{t}_{0}\left(\int_{s}^{t}g_{ij}(r,s)dr\right)ds,\\\
&\\!\\!\\!\displaystyle=g_{1}(t)+g_{2}(t)=2g_{1}(t).\end{array}$
We have the decomposition
$g_{1}(t)=h_{1}(t)+\ell_{1}(t)$
where
$\begin{array}[]{rl}&\\!\\!\\!\displaystyle h_{1}(t)={1\over
t}\int^{t}_{0}\left(\int^{t}_{r}h(r,s)ds\right)dr,\\\
&\\!\\!\\!\displaystyle\ell_{1}(t)={1\over
t}\int^{t}_{0}\left(\int^{t}_{r}\nu_{i}(\nu_{j}-{\mathbb{P}}(\alpha(s)=j))ds\right)dr,\end{array}$
with
$\begin{array}[]{rl}&\\!\\!\\!\displaystyle
h(r,s)={\mathbb{P}}(\alpha(r)=i){\mathbb{P}}(\alpha(s)=j|\alpha(r)=i)-\nu_{j}P(\alpha(r)=i).\end{array}$
Before proceeding further, let us first note the following mixing properties
regarding the Markov chain $\alpha(t)$. For all $t\geq 0$ and $s\leq t$,
denote
$\begin{array}[]{rl}&\\!\\!\\!\displaystyle
p(t)=({\mathbb{P}}(\alpha(t)=1),\ldots,{\mathbb{P}}(\alpha(t)=m))^{\prime}\in{\mathbb{R}}^{m},\\\
&\\!\\!\\!\displaystyle P(t,s)=(({\mathbb{P}}(\alpha(t)=j|\alpha(s)=i),\
i,j\in{\mathcal{M}})\in{\mathbb{R}}^{m\times m},\end{array}$
which are the probability vector and transition matrix of the Markov chain
$\alpha(t)$, respectively. Since $\alpha(t)$ is irreducible, it is ergodic.
Consequently, as $t$ goes to infinity, for the solution of the system
$\left\\{\begin{array}[]{lcl}{\displaystyle{dp(t)\over
dt}}&=&\Pi^{\prime}p(t)\\\ p(0)&=&p_{0}\end{array}\right.$ (A.5)
satisfying
$p_{0,i}\geq 0\ \hbox{ and }\sum^{m}_{i=1}p_{0,i}=1,$
one can find two positive constants $\kappa$ and $K$ such that $p(t)\to\nu$
and
$|p(t)-\nu|\leq K\exp(-\kappa t)$ (A.6)
By virtue of (A.6), it is easily seen that
$\begin{array}[]{ll}|\ell_{1}(t)|&\\!\\!\\!\displaystyle=\left|{\nu_{i}\over
t}\int^{t}_{0}\left(\int^{t}_{r}(\nu_{j}-{\mathbb{P}}(\alpha(s)=j))ds\right)dr\right|,\\\
&\\!\\!\\!\displaystyle\leq{\nu_{i}\over
t}\int^{t}_{0}\left(\int^{t}_{u}|\nu_{j}-{\mathbb{P}}(\alpha(s)=j)|ds\right)dr,\\\
&\\!\\!\\!\displaystyle\leq{\nu_{i}K\over
t}\int^{t}_{0}\left(\int^{t}_{r}\exp(-\kappa s)ds\right)dr,\\\
&\\!\\!\\!\displaystyle\leq{\nu_{i}K\over\kappa t}\int^{t}_{0}\exp(-\kappa
r)dr,\\\ &\\!\\!\\!\displaystyle\leq{\nu_{i}K\over\kappa^{2}t}.\end{array}$
(A.7)
Consequently, $\ell_{1}(t)$ goes to zero as $t$ tends to infinity. Next, we
shall show that $h_{1}(t)$ is bounded. As before, the solution of the system
$\left\\{\begin{array}[]{lcl}{\displaystyle{\partial P(t,s)\over\partial
t}}&=&\Pi^{\prime}P(t,s)\\\ P(s,s)&=&I_{m}\end{array}\right.$ (A.8)
with $s\leq t$, also satisfies for two positive constants $\lambda$ and $K$,
$P(t,s)\to{\hbox{1{\kern-3.50006pt}1}}\nu^{\prime}$ and
$|P(t,s)-{\hbox{1{\kern-3.50006pt}1}}\nu^{\prime}|\leq K\exp(-\lambda(t-s)).$
(A.9)
It follows from (A.9) that
$\begin{array}[]{rl}|h_{1}(t)|&\\!\\!\\!\displaystyle=\left|{1\over
t}\int^{t}_{0}{\mathbb{P}}(\alpha(r)=i)\left(\int^{t}_{r}({\mathbb{P}}(\alpha(s)=j|\alpha(r)=i)-\nu_{j})ds\right)dr\right|,\\\
&\\!\\!\\!\displaystyle\leq{1\over
t}\int^{t}_{0}\left(\int^{t}_{r}|{\mathbb{P}}(\alpha(s)=j|\alpha(r)=i)-\nu_{j}|ds\right)dr,\\\
&\\!\\!\\!\displaystyle\leq{K\over
t}\int^{t}_{0}\left(\int^{t}_{r}\exp(-\lambda(s-r))ds\right)dr,\\\
&\\!\\!\\!\displaystyle\leq{K\over\lambda t}\int^{t}_{0}dr,\\\
&\\!\\!\\!\displaystyle\leq{K\over\lambda}.\end{array}$
Therefore, $h_{1}(t)$ as well as $g_{1}(t)$ are bounded sequences which
ensures that for some positive constant $K$ independent of $t$
$\left|{\mathbb{E}}\left[{1\over
t}\int^{t}_{0}\int^{t}_{0}\hbox{tr}\\{\Pi^{\prime}\Pi(\widehat{\Phi}(r)-\nu)(\widehat{\Phi}^{\prime}(s)-\nu^{\prime})\\}drds\right]\right|\leq
K.$ (A.10)
Finally, (A.2) together with (A.3) and (A.10) imply (4.6) which completes the
proof of Proposition 4.3. $\qquad\Box$
## Appendix B.
We shall now focus on the proof of Proposition 4.5. First of all, we know that
$\sup_{t\geq 0}|\Pi^{\prime}\widehat{\Phi}(t)|\leq 1\hskip
28.45274pt\hbox{a.s.}$
In addition, we also have $|\widehat{\Phi}(t)|\leq 1$ a.s. Consequently, it
follows from (A.1) that
$\begin{array}[]{ll}|N(t)|&\\!\\!\\!\displaystyle\leq|\widehat{\Phi}(t)-\widehat{\Phi}(0)|+\left|\int^{t}_{0}\Pi^{\prime}\widehat{\Phi}(s)ds\right|\leq
1+t\hskip 28.45274pt\hbox{a.s.}\end{array}$ (B.1)
For each $i\in{\mathcal{M}}$, denote
$\begin{array}[]{rl}&\\!\\!\\!\displaystyle
N_{i}(t)=\int^{t}_{0}\sum^{n}_{j=1}[D(\widehat{\Phi}(s)C(X(s))^{\prime}]_{ij}dV_{j}(s)\end{array}$
where $[D(\widehat{\Phi}(s)C(X(s))^{\prime}]_{ij}$ is the $ij$th entry of the
matrix $D(\widehat{\Phi}(s))C(X(s))^{\prime}$ and $V_{j}(s)$ stands the $j$th
component of $V(s)$. It follows from the well-known Doob’s martingale
inequality given for example in [13, Theorem 1.7.4, p. 44] that for each
$i\in{\mathcal{M}}$ and each positive integer $n$,
$\begin{array}[]{ll}&\\!\\!\\!\displaystyle{\mathbb{P}}\left(\sup_{0\leq t\leq
n}\left|\int^{t}_{0}N_{i}(t)-|(D(\widehat{\Phi}(s))C(X(s))^{\prime})_{i,.}|^{2}ds\right|\geq\log
n\right)\leq{1\over n^{2}},\end{array}$ (B.2)
where $(D(\widehat{\Phi}(s)C(X(s))^{\prime})_{i,.}$ denotes the row vector in
the $i$th row of the matrix $D(\widehat{\Phi}(s)C(X(s))^{\prime}$. Hence, we
deduce from the Borel-Cantelli Lemma that for almost all $\omega\in\Omega$,
there is a $K_{1}=K_{1}(\omega)>1$ such that for all $n\geq K_{1}$ and $t\leq
n$
$\begin{array}[]{ll}\displaystyle\int^{t}_{0}|(D(\widehat{\Phi}(s)C(X(s))^{\prime})_{i,.}|^{2}ds&\\!\\!\\!\displaystyle\leq\log
n+N_{i}(t)\hskip 28.45274pt\hbox{a.s.}\\\ &\\!\\!\\!\displaystyle\leq\log
n+1+t\hskip 28.45274pt\hbox{a.s.}\end{array}$ (B.3)
The last line above follows from (B.1). Dividing both sides of (B.3) by $t$,
we obtain that for $n\geq K_{2}$, $n-1\leq t\leq n$, so
$\begin{array}[]{ll}\displaystyle{1\over
t}\int^{t}_{0}|(D(\widehat{\Phi}(s)C(X(s))^{\prime})_{i,.}|^{2}ds&\\!\\!\\!\displaystyle\leq{1\over
n-1}(\log n+1+t)\hskip 14.22636pt\hbox{a.s.}\\\
&\\!\\!\\!\displaystyle\leq{1\over n-1}(\log n+1+n)\hskip
14.22636pt\hbox{a.s.}\\\ &\\!\\!\\!\displaystyle\leq K_{3}\hskip
28.45274pt\hbox{a.s.}\end{array}$ (B.4)
and the bound $K_{3}$ is independent of $t$. Consequently, for some positive
constant $K$ independent of $t$, the quadratic variation of the martingale is
bounded by $Kt$ almost surely. That is, (B.4) implies that
$\big{\langle}N,N\big{\rangle}_{t}\leq Kt$ a.s. Finally, we deduce from the
strong law of large numbers for local martingales [12] that
$\lim_{t\rightarrow\infty}{1\over t}N(t)=0\hskip 28.45274pt\hbox{ a.s.}$
which concludes the proof of Proposition 4.5. $\qquad\Box$
## References
* [1] W. P. Blair and D. D. Sworder, Feedback control of a class of linear discrete systems with jump parameters and quadratic cost criteria, Int. J. Control, 21 (1986), 833-841.
* [2] P. E. Caines and H. F. Chen, Optimal adaptive LQG control for systems with finite state process parameters, IEEE Trans. Automat. Control, 30 (1985), 185-189.
* [3] P.E. Caines and J.-F. Zhang, On the adaptive control of jump parameter systems via nonlinear filtering, SIAM J. Control Optim., 33 (1995), 1758-1777.
* [4] S. Chen, X. Li, and X.Y. Zhou, Stochastic linear quadratic regulators with indefinite control weight costs, SIAM J. Control Optim. 36 (1998), 1685-1702.
* [5] F. Dufour and P. Bertrand, Stabilizing control law for hybrid modes, IEEE Trans. Automatic. Control, 39 (1994), 2354-2357.
* [6] Y. Ji and H.J. Chizeck, Controllability, stabilizability, and continuous-time Markovian jump linear quadratic control, IEEE Trans. Automatic Control, 35 (1990), 777-788.
* [7] H. Abou-Kandil, G. Greling, and G. Jank, Solution and asymptotic behavior of coupled Riccati equations in jump linear systems, IEEE Trans. Automatic Control, 39 (1994), 1631-1636.
* [8] I.Ia. Kac and N.N. Krasovskii, On the stability of systems with random parameters, J. Appl. Math. Mech., 24 (1960), 1225-1246.
* [9] R.Z. Khasminskii, Stochastic Stability of Differential Equations, Sijthoff and Noordhoff, Alphen aan den Rijn, Netherlands, 1980.
* [10] R.Z. Khasminskii, C. Zhu, and G. Yin, Stability of regime-switcing diffusions, Stoachstic Proc. Appl., 117 (2007), 1037-1051.
* [11] H.J. Kushner, Stochastic Stability and Control, Academic Press, New York, NY, 1967.
* [12] R. Liptser, A strong law of large numbers for local martingales, Stochastics 3 (1980), 217–228.
* [13] X. Mao, Stochastic Differential Equations and Applications, 2nd Ed., Horwood, Chichester, UK, 2007.
* [14] X. Mao, Stability of stochastic differential equations with Markovian switching, Stochastic Process. Appl., 79 (1999), 45-67.
* [15] X. Mao, G. Yin, and C. Yuan, Stabilization and destabilization of hybrid systems of stochastic differential equations, Automatica, 43 (2007), 264-273.
* [16] M. Mariton and P. Bertrand, Robust jump linear quadratic control: A mode stabilizing solution, IEEE Trans. Automat. Control, AC-30 (1985), 1145-1147.
* [17] W.M. Wonham, Some applications of stochastic differential equations to optimal nonlinear filtering, SIAM J. Control, 2 (1965), 347–369.
* [18] W.M. Wonham, Liapunov criteria for weak stochastic stability, J. Differential Eqs., 2 (1966), 195–207.
* [19] X.Y. Zhou and G. Yin, Markowitz’s mean-variance portfolio selection with regime switching: A Continuous-time model, SIAM J. Control Optim, 42 (2003), 1466-1482.
* [20] C. Zhu and G. Yin, Asymptotic properties of hybrid diffusion systems, SIAM J. Control Optim., 46 (2007), 1155–1179.
| arxiv-papers | 2008-07-09T10:42:30 | 2024-09-04T02:48:56.638893 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Bernard Bercu, Francois Dufour, G. George Yin",
"submitter": "Bercu Bernard",
"url": "https://arxiv.org/abs/0807.1413"
} |
0807.1444 | # From $N$ M2’s to $N$ D2’s
Yi Pang111Electronic address: yipang@itp.ac.cn Interdisciplinary Center for
Theoretical Study,
University of Science and Technology of China,
Hefei, Anhui 230026, China
Institute of Theoretical Physics, Chinese Academy of Sciences,
P. O. Box 2735, Beijing 100080, China
Tower Wang222Electronic address: wangtao218@itp.ac.cn Interdisciplinary Center
for Theoretical Study,
University of Science and Technology of China,
Hefei, Anhui 230026, China
Institute of Theoretical Physics, Chinese Academy of Sciences,
P. O. Box 2735, Beijing 100080, China
###### Abstract
In this short note, we reduce the $\mathcal{N}=6$, $U(N)\times U(N)$ Chern-
Simons gauge theories to $\mathcal{N}=8$, $U(N)$ Yang-Mills gauge theories.
This process corresponds to recovering the world-volume theory of $N$
D2-branes from that of $N$ M2-branes in an intermediate energy range. The
supersymmetries are enhanced because in this limit the branes localize far
away from the orbifold singularity. Our main scheme is exactly in accordance
with Mukhi and Papageorgakis’s earlier work, although the Higgs mechanism
becomes trickier in the present case. We also speculate on applying the scheme
to a large class of new Bagger-Lambert models more generally.
††preprint: USTC-ICTS-08-12
## I Introduction
In the past few months, the study of M2-branes was revived333For a recent
review of membrane, please refer to Berman:2007bv . Following the pioneer work
Bagger:2006sk -Gustavsson:2008dy , substantial literature along this direction
appears in the past half year. As a partial list, see Bandres:2008vf -Krishnan
and references therein. We are sorry for missing a lot of relevant work in our
reference list., initiated by Bagger and Lambert Bagger:2006sk ; Bagger:2007jr
; Bagger:2007vi and Gustavsson Gustavsson:2007vu ; Gustavsson:2008dy , in
which the world-volume theory of two M2-branes was discovered. However,
because the old Bagger-Lambert-Gustavsson (BLG) theory Bagger:2006sk ;
Bagger:2007jr ; Bagger:2007vi ; Gustavsson:2007vu ; Gustavsson:2008dy relies
on a special form of 3-algebra, it is difficult to be generalized to arbitrary
number of M2-branes Gustavsson:2008dy ; Bandres:2008vf ; Ho:2008bn ;
Papadopoulos:2008sk ; Gauntlett:2008uf . Under certain conditions, the
$\mathcal{A}_{4}$ algebra in the old BLG theory is unique and there is a no-go
theorem Gauntlett:2008uf . One can get around such a no-go theorem by relaxing
the conditions Ho:2008bn ; Gran:2008vi ; Awata:1999dz . Switching from
Euclidean 3-algebra to Lorentzian 3-algebra, three groups Gomis:2008uv ;
Benvenuti:2008bt ; Ho:2008ei showed a possible way for generalization. Around
a month later, Aharony et al. Aharony:2008ug opened another possibility. Not
restricted to 3-algebra, they turn to consider Chern-Simons-matter theories
with a $U(N)\times U(N)$ or $SU(N)\times SU(N)$ gauge symmetry. In fact, the
old BLG theory can be reformulated into an $SU(2)\times SU(2)$ theory
Bandres:2008vf ; Mukhi:2008ux ; VanRaamsdonk:2008ft ; Berman:2008be ;
Distler:2008mk . What is more, it was shown long ago that the gauge group
$U(N)\times U(N)$ plays an important role in the infrared limit of $N$
coincident D3-branes at a conical singularity Klebanov:1998hh . Therefore, an
$SU(N)\times SU(N)$ Aharony-Bergman- Jafferis-Maldacena (ABJM) theory, as a
3-dimensional superconformal Chern-Simons-matter theory, is a very attractive
world-volume theory of $N$ coincident M2-branes. Given that the old BLG theory
is difficult for generalization, although ABJM’s $\mathcal{N}=6$ Chern-Simons-
matter theories do not rely on 3-algebra, it is still appealing to recast them
in terms of a certain 3-algebra, in hopes that it may told us how to better
overcome the no-go theorem Gauntlett:2008uf . This was elegantly accomplished
by Bagger and Lambert as reported in Bagger:2008se recently, which we will
call the new BL theory.
In this little exercise, we reduce the $\mathcal{N}=6$ $U(N)\times U(N)$
Chern-Simons gauge theories to $U(N)$ Yang-Mills gauge theories. We find the
resulting theories have $\mathcal{N}=8$ supersymmetry, or 16 real
supercharges. We will review ABJM theories in section II, and then in section
III show the details of Higgs mechanism for the present case. Finally, in
section IV, we will comment on how to apply our scheme to the new BL model
generally and conclude.
## II Review of $\mathcal{N}=6$ Chern-Simons theories
In this section, following the neat work Benna:2008zy , we review ABJM’s
$\mathcal{N}=6$, $U(N)\times U(N)$ Chern-Simons gauge theories Aharony:2008ug
. We are interested in the special case with an $SU(4)$ R-symmetry. In this
case, the coefficients of the Chern-Simons action and the superpotential in
Benna:2008zy are related by $K=1/L$. Comparing with the old BLG theory, one
immediately reads $L=8\pi/k$ at level444An example of Chern-Simons level
quantization was shown in Distler:2008mk . $k$. We mainly take the notations
and conventions of Benna:2008zy , but the generators of $U(N)$ algebra are
different in normalization,
$\displaystyle[T^{a},T^{b}]=if^{abc}T^{c},\leavevmode\nobreak\
\leavevmode\nobreak\ \leavevmode\nobreak\ \leavevmode\nobreak\
(a,b,c=1,2,...,N^{2}-1)$
$\displaystyle\mbox{Tr}(T^{a}T^{b})=\frac{1}{2}\delta^{ab},\leavevmode\nobreak\
\leavevmode\nobreak\ \leavevmode\nobreak\ \leavevmode\nobreak\
\mbox{Tr}\\{T^{a},T^{b}\\}=\delta^{ab},\leavevmode\nobreak\
\leavevmode\nobreak\ \leavevmode\nobreak\ \leavevmode\nobreak\
\mbox{Tr}(\\{T^{a},T^{b}\\}T^{c})=\frac{1}{2}d^{abc},$ $\displaystyle
T^{0}=\frac{1}{\sqrt{2N}}\mbox{diag}(1,1,...,1)$ (1)
in which we use the square bracket to denote commutators and the brace bracket
to denote anti-commutators. For convenience, we will raise and lower group
indices with a unit metric, so there is no distinction among upper and lower
indices, their position being dictated by notational convenience.
Nevertheless, as usual, repeated indices imply summation. The gauge potential
$A_{\mu}$ and $\tilde{A}_{\mu}$ in Benna:2008zy will be replaced by
$A_{\mu}^{(L)}$ and $A_{\mu}^{(R)}$ in our notations, while the covariant
derivative $\mathcal{D}_{\mu}$ will be replaced by
$\tilde{\mathcal{D}}_{\mu}$, since we will use these notations for other
purposes.
After integrating out auxiliary fields and combining the $SU(2)$ fields into
an $SU(4)$ representation,
$\displaystyle Y^{1}$ $\displaystyle=$ $\displaystyle X^{1}+iX^{5},$
$\displaystyle Y^{2}$ $\displaystyle=$ $\displaystyle X^{2}+iX^{6},$
$\displaystyle Y^{3}$ $\displaystyle=$ $\displaystyle X^{3}+iX^{7},$
$\displaystyle Y^{4}$ $\displaystyle=$ $\displaystyle X^{4}+iX^{8},$ (2)
one can write down the total action of ABJM’s $\mathcal{N}=6$, $U(N)\times
U(N)$ Chern-Simons gauge theories as Benna:2008zy ; Bagger:2008se
$\mathcal{S}=\int
d^{3}x\left[-\mbox{Tr}\left((\tilde{\mathcal{D}}^{\mu}Y_{A})^{{\dagger}}\tilde{\mathcal{D}}_{\mu}Y^{A}\right)+i\mbox{Tr}\left(\psi^{A{\dagger}}\gamma^{\mu}\tilde{\mathcal{D}}_{\mu}\psi_{A}\right)+\mathcal{L}_{\rm
CS}-V^{\rm ferm}-V^{\rm bos}\right].$ (3)
In the above action, the covariant derivative is defined by
$\tilde{\mathcal{D}_{\mu}}Y^{A}=\partial_{\mu}Y^{A}+iA_{\mu}^{(L)}Y^{A}-iY^{A}A_{\mu}^{(R)}.$
(4)
Note that the notation $\tilde{\mathcal{D}}_{\mu}$ with a tilde here is
different from $\mathcal{D}_{\mu}$ which will appear later. The Chern-Simons
term is
$\mathcal{L}_{\rm
CS}=\frac{2}{L}\epsilon^{\mu\nu\lambda}\mbox{Tr}\left(A_{\mu}^{(L)}\partial_{\nu}A_{\lambda}^{(L)}+\frac{2i}{3}A_{\mu}^{(L)}A_{\nu}^{(L)}A_{\lambda}^{(L)}-A_{\mu}^{(R)}\partial_{\nu}A_{\lambda}^{(R)}-\frac{2i}{3}A_{\mu}^{(R)}A_{\nu}^{(R)}A_{\lambda}^{(R)}\right).$
(5)
There is a potential like a Yukawa term, by which scalars and fermions are
coupled
$\displaystyle V^{\rm ferm}$ $\displaystyle=$
$\displaystyle\frac{iL}{4}\mbox{Tr}\left(Y_{A}^{\dagger}Y^{A}\psi^{B\dagger}\psi_{B}-Y^{A}Y_{A}^{\dagger}\psi_{B}\psi^{B\dagger}+2Y^{A}Y_{B}^{\dagger}\psi_{A}\psi^{B\dagger}-2Y_{A}^{\dagger}Y^{B}\psi^{A\dagger}\psi_{B}\right.$
(6)
$\displaystyle\left.+\epsilon^{ABCD}Y_{A}^{\dagger}\psi_{B}Y_{C}^{\dagger}\psi_{D}-\epsilon_{ABCD}Y^{A}\psi^{B\dagger}Y^{C}\psi^{D\dagger}\right).$
The sextic potential of scalars takes the form
$\displaystyle V^{\rm bos}$ $\displaystyle=$
$\displaystyle-\frac{L^{2}}{48}\mbox{Tr}\left(Y^{A}Y_{A}^{\dagger}Y^{B}Y_{B}^{\dagger}Y^{C}Y_{C}^{\dagger}+Y_{A}^{\dagger}Y^{A}Y_{B}^{\dagger}Y^{B}Y_{C}^{\dagger}Y^{C}\right.$
(7)
$\displaystyle\left.+4Y^{A}Y_{B}^{\dagger}Y^{C}Y_{A}^{\dagger}Y^{B}Y_{C}^{\dagger}-6Y^{A}Y_{B}^{\dagger}Y^{B}Y_{A}^{\dagger}Y^{C}Y_{C}^{\dagger}\right).$
## III Reduction to Yang-Mills Theories
In this section, we reduce the $U(N)\times U(N)$ Chern-Siomons theories
reviewed in the previous section into $U(N)$ Yang-Mills theories. This process
corresponds to recovering the world-volume theory of $N$ D2-branes from that
of $N$ M2-branes. Our main scheme is exactly in accordance with Mukhi:2008ux .
Firstly recast the gauge fields as555In the early days of preparing this work,
we found the choice (8) was also made by the reference Honma:2008jd , which
partly overlapped with our plan. What is more, actually the scaling limit
taken in Honma:2008jd is equivalent to the limit $\tilde{v}\rightarrow\infty$
in our scheme. Anyhow we decided to finish this little exercise to obtain more
details.
$A_{\mu}^{(L)a}=A_{\mu}^{a}+B_{\mu}^{a},\leavevmode\nobreak\
\leavevmode\nobreak\ \leavevmode\nobreak\ \leavevmode\nobreak\
A_{\mu}^{(R)a}=A_{\mu}^{a}-B_{\mu}^{a},$ (8)
and then integrate out the auxiliary field $B_{\mu}^{a}$ using its equation of
motion. At the same time, take the vacuum expectation value to be
$\langle Y^{A0}\rangle=i\tilde{v}\delta^{A}_{4}$ (9)
and other components vanished. $\tilde{v}$ denotes our choice of the vacuum
expectation value (VEV). In the following, $v$ denotes the choice in
Distler:2008mk . We should stress here $\tilde{v}$ is real in spite of the
fact that $Y^{A0}$ is a complex variable. The Yang-Mills coupling is defined
by666It was argued in Distler:2008mk that $g_{YM}=v/\sqrt{k}$. At first
glance, it contradicts with the relation (10). However, this disagreement is
understandable because the normalizations are different. By the replacement
$v\rightarrow 4\pi\tilde{v}/\sqrt{2kN}$, one can change from their
normalization to ours and recover (10). In the last step of (10), we have used
the quantization condition $L=8\pi/k$.
$g_{YM}=\frac{L\tilde{v}}{2\sqrt{2N}}=\frac{4\pi\tilde{v}}{k\sqrt{2N}}$ (10)
as explained in Distler:2008mk . As did in Mukhi:2008ux ; Distler:2008mk , we
have to rescale some of the components of $Y$ and $\psi$ as
$(X,\psi)\rightarrow(X/g_{YM},\psi/g_{YM})$ and keep only leading order terms
with respect to $\tilde{v}^{-1}$. In other words, we have to take the limit
$k\rightarrow\infty$, $\tilde{v}\rightarrow\infty$ with $g_{YM}$ fixed. One
possible interpretation of this limit is illustrated by Figure 1.
Figure 1: The left figure signifies the $C^{4}/Z_{k}$. In the rest of our
paper, we will see that $kg_{YM}^{2}$ is a typical energy scale, below which
our reduction is meaningful. According to the definition of $g_{YM}$,
$kg_{YM}^{2}\sim\tilde{v}^{2}/(kN)\sim v^{2}$. Since in reference
Distler:2008mk $v^{2}$ is also related to physical mass scale, we see again
that these two descriptions are equivalent. So taking the limit
$k\rightarrow\infty$ makes the cone become a fine cylinder. It is the usual
picture that by compactifying the $x^{11}$, M-theory is reduced to the type
IIA string theory.
Many of the techniques employed here can be found in Mukhi:2008ux ;
Honma:2008jd . In reference Honma:2008jd , a precise relation between ABJM
model and Lorentzian BLG model was studied. Since the Lorentzian BLG model is
intimately related to D2 branes, our will partly overlap with the analysis
there. In their analysis, they promoted the coupling constant to a spacetime
dependent vector and recovered the $SO(8)$ invariance. Since our aim is to get
the world-volume theory of D2-branes, we will take $\tilde{v}$ to be spacetime
independent and get the $SO(7)$ symmetry.
First of all, we decompose fields as
$Y^{A}=Y^{A0}T^{0}+iY^{Aa}T^{a},\leavevmode\nobreak\ \leavevmode\nobreak\
\leavevmode\nobreak\ \leavevmode\nobreak\
\psi_{A}=\psi_{A}^{0}T^{0}+i\psi_{A}^{a}T^{a},\leavevmode\nobreak\
\leavevmode\nobreak\ \leavevmode\nobreak\ \leavevmode\nobreak\
A_{\mu}=A_{\mu}^{a}T^{a}.$ (11)
Here $Y^{A0}$ and $Y^{Aa}$ are complex variables, $\psi_{A}^{0}$ and
$\psi_{A}^{a}$ are complex with two components, while $A_{\mu}^{a}$ is a real
variable. Let us also introduce
$\displaystyle\mathcal{D}_{\mu}Y^{A0}$ $\displaystyle=$
$\displaystyle\partial_{\mu}Y^{A0},$ $\displaystyle\mathcal{D}_{\mu}Y^{Aa}$
$\displaystyle=$ $\displaystyle\partial_{\mu}Y^{Aa}-f^{abc}A_{\mu}^{b}Y^{Ac},$
$\displaystyle(\mathcal{D}_{\mu}Y_{A}^{a})^{{\dagger}}$ $\displaystyle=$
$\displaystyle(\mathcal{D}_{\mu}Y_{A}^{a})^{*}=\partial_{\mu}\bar{Y}_{A}^{a}-f^{abc}A_{\mu}^{b}\bar{Y}_{A}^{c},$
$\displaystyle F_{\nu\lambda}^{a}$ $\displaystyle=$
$\displaystyle\partial_{\nu}A^{a}_{\lambda}-\partial_{\lambda}A^{a}_{\nu}-f^{abc}A_{\nu}^{b}A_{\lambda}^{c}.$
(12)
Making use of
$\tilde{\mathcal{D}}_{\mu}Y^{A}=\partial_{\mu}Y^{A}+i[A_{\mu},Y^{A}]+i\\{B_{\mu},Y^{A}\\}=\mathcal{D}_{\mu}Y^{A}+i\\{B_{\mu},Y^{A}\\},$
(13)
and conventions (II), one can rewrite the kinetic term of scalars in the form
$\displaystyle-\mbox{Tr}\left((\tilde{\mathcal{D}}^{\mu}Y_{A})^{{\dagger}}\tilde{\mathcal{D}}_{\mu}Y^{A}\right)$
(14) $\displaystyle=$
$\displaystyle-\frac{1}{2}\partial^{\mu}\bar{Y}_{A}^{0}\partial_{\mu}Y^{A0}-\frac{1}{2}(\mathcal{D}^{\mu}Y^{Aa})^{{\dagger}}\mathcal{D}_{\mu}Y^{Aa}-\frac{1}{N}B_{\mu}^{a}B^{\mu
a}\bar{Y}_{A}^{0}Y^{A0}$
$\displaystyle+\frac{i}{\sqrt{2N}}d^{abc}B_{\mu}^{a}B^{\mu
b}(Y^{A0}\bar{Y}_{A}^{c}-\bar{Y}_{A}^{0}Y^{Ac})-B_{\mu}^{a}B^{\mu
b}\bar{Y}_{A}^{c}Y^{Ad}\mbox{Tr}\left(\\{T^{a},T^{c}\\}\\{T^{b},T^{d}\\}\right)$
$\displaystyle-\frac{1}{\sqrt{2N}}B^{\mu
a}\left[(\mathcal{D}_{\mu}Y^{Aa})\bar{Y}_{A}^{0}+(\mathcal{D}_{\mu}Y_{A}^{a})^{{\dagger}}Y^{A0}\right]$
$\displaystyle+\frac{1}{\sqrt{2N}}B^{\mu
a}\left[(\partial_{\mu}Y^{A0})\bar{Y}_{A}^{a}+(\partial_{\mu}\bar{Y}_{A}^{0})Y^{Aa}\right]$
$\displaystyle+\frac{i}{2}d^{abc}B^{\mu
a}\left[(\mathcal{D}_{\mu}Y^{Ab})\bar{Y}_{A}^{c}-(\mathcal{D}_{\mu}Y_{A}^{b})^{{\dagger}}Y^{Ac}\right],$
while the kinetic term of fermions takes the form
$\displaystyle
i\mbox{Tr}\left(\psi^{A{\dagger}}\gamma^{\mu}\tilde{\mathcal{D}}_{\mu}\psi_{A}\right)$
(15) $\displaystyle=$
$\displaystyle\frac{i}{2}(\psi^{A0{\dagger}}\gamma^{\mu}\partial_{\mu}\psi_{A}^{0}+\psi^{Aa{\dagger}}\gamma^{\mu}\mathcal{D}_{\mu}\psi_{A}^{a})$
$\displaystyle+\frac{i}{2N}B_{\mu}^{a}(\psi^{Aa{\dagger}}\gamma^{\mu}\psi_{A}^{0}-\psi^{A0{\dagger}}\gamma^{\mu}\psi_{A}^{a})-\frac{1}{2}d^{abc}B_{\mu}^{a}\psi^{Ab{\dagger}}\gamma^{\mu}\psi_{A}^{c}.$
As for the Chern-Simons term, we obtain
$\displaystyle\mathcal{L}_{\rm CS}$ $\displaystyle=$
$\displaystyle\frac{2}{L}\epsilon^{\mu\nu\lambda}(B_{\mu}^{a}F_{\nu\lambda}^{a}-\frac{1}{3}f^{abc}B_{\mu}^{a}B_{\nu}^{b}B_{\lambda}^{c}).$
(16)
Although the gauge fields $A_{\mu}^{(L,R)}$ are purely topological, after the
recombination (8), one of the new fields $B_{\mu}^{a}$ becomes an auxiliary
field without derivatives, just as happened in Mukhi:2008ux . We will
eliminate this auxiliary field using its equation of motion, and obtain the
kinetic term of the other gauge field $A_{\mu}^{a}$. Namely, $A_{\mu}^{a}$
will get dynamical if one integrates out $B_{\mu}^{a}$.
The potential term mixing scalars and fermions becomes
$\displaystyle V^{\rm ferm}$ $\displaystyle=$
$\displaystyle\frac{iL}{4}\mbox{Tr}\biggl{[}\frac{1}{2\sqrt{2N}}f^{abc}(Y^{A0}\bar{Y}_{A}^{a}-\bar{Y}_{A}^{0}Y^{Aa})\bar{\psi}^{Bb}\psi_{B}^{c}$
(17)
$\displaystyle+\frac{1}{2\sqrt{2N}}f^{abc}\bar{Y}_{A}^{a}Y^{Ab}(\bar{\psi}^{Bc}\psi_{B}^{0}-\psi^{Bc}\bar{\psi}^{B0})$
$\displaystyle+\frac{1}{\sqrt{2N}}f^{abc}(Y^{A0}\bar{Y}_{B}^{a}\psi_{A}^{b}\bar{\psi}^{Bc}+\bar{Y}_{A}^{0}Y^{Ba}\bar{\psi}^{Ab}\psi_{B}^{c})$
$\displaystyle-\frac{1}{\sqrt{2N}}f^{abc}(Y^{Aa}\bar{Y}_{B}^{b}\psi_{A}^{c}\bar{\psi}^{B0}+\bar{Y}_{A}^{a}Y^{Bb}\bar{\psi}^{Ac}\psi_{B}^{0})$
$\displaystyle+(\bar{Y}_{A}^{a}Y^{Ab}\bar{\psi}^{Bc}\psi_{B}^{d}+2Y^{Aa}\bar{Y}_{B}^{b}\psi_{A}^{c}\bar{\psi}^{Bd})\mbox{Tr}\left(T^{a}T^{b}T^{c}T^{d}-T^{b}T^{a}T^{d}T^{c}\right)$
$\displaystyle-\frac{1}{2\sqrt{2N}}f^{abc}(\epsilon^{ABCD}\bar{Y}_{A}^{0}\psi_{B}^{a}\bar{Y}_{C}^{b}\psi_{D}^{c}+\epsilon_{ABCD}Y^{A0}\bar{\psi}^{Ba}Y^{Cb}\bar{\psi}^{Dc})$
$\displaystyle+\frac{1}{2\sqrt{2N}}f^{abc}(\epsilon^{ABCD}\bar{Y}_{A}^{a}\psi_{B}^{b}\bar{Y}_{C}^{c}\psi_{D}^{0}+\epsilon_{ABCD}Y^{Aa}\bar{\psi}^{Bb}Y^{Cc}\bar{\psi}^{D0})$
$\displaystyle+(\epsilon^{ABCD}\bar{Y}_{A}^{a}\psi_{B}^{b}\bar{Y}_{C}^{c}\psi_{D}^{d}-\epsilon_{ABCD}Y^{Aa}\bar{\psi}^{Bb}Y^{Cc}\bar{\psi}^{Dd})\mbox{Tr}\left(T^{a}T^{b}T^{c}T^{d}\right)\biggr{]}.$
The bosonic potential can be written as
$\displaystyle V^{\rm bos}$ $\displaystyle=$
$\displaystyle-\frac{L^{2}}{48}\biggl{[}\frac{3}{2N}Y^{A0}\bar{Y}_{A}^{0}Y^{Ba}\bar{Y}_{B}^{b}Y^{Cc}\bar{Y}_{C}^{d}\mbox{Tr}\Bigl{(}[T^{a},T^{d}][T^{b},T^{c}]+[T^{a},T^{c}][T^{b},T^{d}]\Bigr{)}$
(18)
$\displaystyle-\frac{3}{2N}(Y^{A0}Y^{B0}\bar{Y}_{A}^{a}\bar{Y}_{B}^{b}+\bar{Y}_{A}^{0}\bar{Y}_{B}^{0}Y^{Aa}Y^{Bb})Y^{Cc}\bar{Y}_{C}^{d}\mbox{Tr}\Bigl{(}[T^{a},T^{c}][T^{b},T^{d}]\Bigr{)}$
$\displaystyle-\frac{3}{2N}Y^{A0}\bar{Y}_{B}^{0}\bar{Y}_{A}^{a}Y^{Bb}Y^{Cc}\bar{Y}_{C}^{d}\mbox{Tr}\Bigl{(}[T^{a},T^{d}][T^{b},T^{c}]+[T^{a},T^{c}][T^{b},T^{d}]\Bigr{)}$
$\displaystyle+\frac{3i}{\sqrt{2N}}(\bar{Y}_{A}^{0}Y^{Aa}Y^{Bb}\bar{Y}_{B}^{c}Y^{Cd}\bar{Y}_{C}^{e}-Y^{A0}\bar{Y}_{A}^{a}\bar{Y}_{B}^{b}Y^{Bc}\bar{Y}_{C}^{d}Y^{Ce})$
$\displaystyle\times\mbox{Tr}\Bigl{(}2(T^{b}T^{e}T^{a}-T^{a}T^{e}T^{b})[T^{c},T^{d}]+(T^{d}T^{e}T^{a}-T^{a}T^{e}T^{d})[T^{b},T^{c}]\Bigr{)}$
$\displaystyle+Y^{Aa}\bar{Y}_{A}^{b}Y^{Bc}\bar{Y}_{B}^{d}Y^{Ce}\bar{Y}_{C}^{f}\mbox{Tr}\Bigl{(}T^{a}T^{b}T^{c}T^{d}T^{e}T^{f}+T^{b}T^{a}T^{d}T^{c}T^{f}T^{e}$
$\displaystyle+4T^{a}T^{d}T^{e}T^{b}T^{c}T^{f}-6T^{a}T^{d}T^{c}T^{b}T^{e}T^{f}\Bigr{)}\biggr{]}.$
Choosing the vacuum expectation value given in (9), we can expand the scalars
and fermions near the VEV as
$\displaystyle Y^{A}$ $\displaystyle=$ $\displaystyle
x^{A}_{0}T^{0}+i(\tilde{v}\delta^{A}_{4}+x^{A+4}_{0})T^{0}+ix^{Aa}T^{a}-x^{(A+4)a}T^{a},$
$\displaystyle\psi^{A}$ $\displaystyle=$
$\displaystyle\Psi^{A}_{0}T^{0}+i\Psi^{A+4}_{0}T^{0}+i\Psi^{Aa}T^{a}-\Psi^{(A+4)a}T^{a}.$
(19)
In the limit $\tilde{v}\rightarrow\infty$, the leading order terms in (3) can
be obtained using the above results. In particular, the leading order kinetic
terms in (14) and (15) give
$\displaystyle\mathcal{L}_{\rm kinetic}$ $\displaystyle=$
$\displaystyle-\sum_{I=1}^{8}\frac{1}{2}\left(\partial^{\mu}x^{I}_{0}\partial_{\mu}x^{I}_{0}+\mathcal{D}^{\mu}x^{Ia}\mathcal{D}_{\mu}x^{Ia}\right)-\frac{\tilde{v}^{2}}{N}B_{\mu}^{a}B^{\mu
a}-\frac{2\tilde{v}}{\sqrt{2N}}B^{\mu a}\mathcal{D}_{\mu}x^{8a}$ (20)
$\displaystyle+\sum_{A=1}^{4}\frac{i}{2}(\Psi^{A{\dagger}}_{0}-i\Psi^{(A+4){\dagger}}_{0})\gamma^{\mu}\partial_{\mu}(\Psi^{A}_{0}+i\Psi^{A+4}_{0})$
$\displaystyle+\sum_{A=1}^{4}\frac{i}{2}(\Psi^{Aa{\dagger}}-i\Psi^{(A+4)a{\dagger}})\gamma^{\mu}\mathcal{D}_{\mu}(\Psi^{Aa}+i\Psi^{(A+4)a}).$
As we will show later, the cubic term in (16) is negligible in the limit
$\tilde{v}\rightarrow\infty$, so only the first term survives
$\mathcal{L}_{\rm
CS}=\frac{2}{L}\epsilon^{\mu\nu\lambda}B_{\mu}^{a}F_{\nu\lambda}^{a}.$ (21)
The gauge potential $B_{\mu}^{a}$ appears only in $\mathcal{L}_{\rm kinetic}$
and $\mathcal{L}_{\rm CS}$ without derivatives. We can eliminate it using the
equation of motion
$B^{\mu
a}=\frac{N}{L\tilde{v}^{2}}\epsilon^{\mu\nu\lambda}F_{\nu\lambda}^{a}-\frac{\sqrt{2N}}{2\tilde{v}}\mathcal{D}^{\mu}x^{8a}.$
(22)
When deriving this equation we have neglected the quadratic term in $B^{\mu
a}$ coming from the cubic self-interaction as well as terms coming from higher
interaction with scalars. Because these terms will lead to higher order
contributions which are suppressed in the limit
$\tilde{v}\rightarrow\infty,L\rightarrow 0$.
Inserting the above equation into the lagrangian, up to a total derivative
term, the kinetic and Chern-Simons terms turn out to be
$\displaystyle\mathcal{L}_{\rm kinetic}+\mathcal{L}_{\rm CS}$ (23)
$\displaystyle=$ $\displaystyle-\frac{2N}{L^{2}\tilde{v}^{2}}F^{\nu\lambda
a}F_{\nu\lambda}^{a}-\sum_{j=1}^{7}\frac{1}{2}\left(\partial^{\mu}x^{j}_{0}\partial_{\mu}x^{j}_{0}+\mathcal{D}^{\mu}x^{ja}\mathcal{D}_{\mu}x^{ja}\right)-\frac{1}{2}\partial^{\mu}x^{8}_{0}\partial_{\mu}x^{8}_{0}$
$\displaystyle+\sum_{A=1}^{4}\frac{i}{2}(\Psi^{A{\dagger}}_{0}-i\Psi^{(A+4){\dagger}}_{0})\gamma^{\mu}\partial_{\mu}(\Psi^{A}_{0}+i\Psi^{A+4}_{0})$
$\displaystyle+\sum_{A=1}^{4}\frac{i}{2}(\Psi^{Aa{\dagger}}-i\Psi^{(A+4)a{\dagger}})\gamma^{\mu}\mathcal{D}_{\mu}(\Psi^{Aa}+i\Psi^{(A+4)a}),$
We find the kenetic terms of $x^{8a}$ exactly cancel out; hence they become
non-dynamical. But a question may arise: where are the degrees of freedom
newly turned on to compensate the disappeared $x^{8a}$. The doubt can be
resolved by observing that the gauge field $A_{\mu}^{a}$ becomes dynamical.
Since in three dimensions, each massless vector field $A_{\mu}^{a}$ has one
dynamical degree of freedom as the scalar $x^{8a}$, the total degrees of
freedom are unchanged. From the Yang-Mills term, by requiring its coefficient
to be $-1/(4g_{YM}^{2})$, one can quickly read the Yang-Mills coupling
constant as in (10). It is clear that $B^{\mu a}\propto L/g_{YM}^{2}$ right
now. For a finite $g_{YM}$, in the limit $L\rightarrow 0$, the first term in
(16) is finite while the second term vanishes, as we have assumed. Since this
limit is to keep the leading term in an expansion in powers of energy divided
by $\tilde{v}^{2}/k$, this also tells us that the resulting $U(N)$ gauge
theory is valid only at a energy below $g_{YM}^{2}/L$ i.e. $kg_{YM}^{2}$. On
the other hand, since our derivation is to depict the weakly coupled limit of
N D-branes, the theory is applicable only above the energy scale
$g_{YM}^{2}N$.
In the limit $\tilde{v}\rightarrow\infty,L\rightarrow 0$ while keeping
$L\tilde{v}$ finite, the potential terms for scalars and fermions reduce to
$\displaystyle V^{\rm ferm}$ $\displaystyle=$
$\displaystyle-\frac{L\tilde{v}}{4\sqrt{2N}}f^{abc}\mbox{Tr}\biggl{[}\sum_{A=1}^{3}2ix^{4a}\Psi^{Ab}\Psi^{(A+4)c}-2ix^{4a}\Psi^{4b}\Psi^{8c}$
$\displaystyle+\sum_{A=1}^{3}2ix^{Aa}(\Psi^{8b}\Psi^{Ac}-\Psi^{4b}\Psi^{(A+4)c})-\sum_{A=1}^{3}2ix^{(A+4)a}(\Psi^{4b}\Psi^{Ac}+\Psi^{8b}\Psi^{(A+4)c})$
$\displaystyle-\sum_{A,B,C=1}^{3}\frac{1}{2}\epsilon^{ABC4}(\Psi^{Aa}+i\Psi^{(A+4)a})(x^{Bb}-ix^{(B+4)b})(\Psi^{Cc}+i\Psi^{(C+4)c})$
$\displaystyle+\sum_{A,B,C=1}^{3}\frac{1}{2}\epsilon_{ABC4}(\Psi^{Aa}-i\Psi^{(A+4)a})(x^{Bb}+ix^{(B+4)b})(\Psi^{Cc}-i\Psi^{(C+4)c})\biggr{]}.$
$V^{\rm
bos}=-\sum_{i,j=1}^{7}\frac{L^{2}\tilde{v}^{2}}{16N}x^{ia}x^{ib}x^{jc}x^{jd}\mbox{Tr}\left([T^{a},T^{c}][T^{b},T^{d}]\right).$
(25)
In (17) and (III), we keep the notation Tr to remind the inner product between
two-component spinors. As one expected, the goldstones $x^{8a}$ disappear in
(23-25) and hence in the total action.
If we define the minimal spinor of $SO(2,1)\times SO(7)$ as,
$\Psi=\begin{pmatrix}\Psi^{1a},&\Psi^{2a},&\Psi^{3a},&\Psi^{4a},&\Psi^{5a},&\Psi^{6a},&\Psi^{7a},&\Psi^{8a}\\\
\end{pmatrix}^{T},$ (26)
where each $\Psi^{i}$ is also a two-component Majorana spinor. Then the
fermionic potential is actually the standard Yukawa coupling:
$V^{\rm
ferm}=-\frac{L\tilde{v}}{4\sqrt{2N}}f^{abc}x^{ia}\bar{\Psi}^{b}\Gamma^{i}\otimes\mathbf{1}_{2\times
2}\Psi^{c}.$ (27)
At this moment, it is easy to see that the fermionic potential term also has
an $SO(7)$ R-symmetry.
The last step is to rescale the scalars and spinor as
$(x^{j},\Psi)\rightarrow(x^{j}/g_{YM},\Psi/g_{YM})$. Gathering the above
results together, in the limit $\tilde{v}\rightarrow\infty$ with a fixed
$g_{YM}$, we have
$\mathcal{S}=\int d^{3}x\left(\mathcal{L}_{\rm
decoupled}+\frac{1}{g_{YM}^{2}}\mathcal{L}_{\rm coupled}\right),$ (28)
where
$\mathcal{L}_{\rm
decoupled}=-\sum_{j=1}^{7}\frac{1}{2}\partial^{\mu}x^{j}_{0}\partial_{\mu}x^{j}_{0}+\frac{i}{2}\Psi_{0}\Gamma^{0}\otimes\gamma^{\mu}\partial_{\mu}\Psi_{0}-\frac{1}{2}\partial^{\mu}x^{8}_{0}\partial_{\mu}x^{8}_{0},$
(29) $\displaystyle\mathcal{L}_{\rm coupled}$ $\displaystyle=$
$\displaystyle-\frac{1}{4}F^{\nu\lambda
a}F_{\nu\lambda}^{a}-\frac{1}{2}\mathcal{D}^{\mu}x^{ja}\mathcal{D}_{\mu}x^{ja}+\frac{i}{2}\Psi^{a}\Gamma^{0}\otimes\gamma^{\mu}\mathcal{D}_{\mu}\Psi_{a}$
(30)
$\displaystyle-\frac{1}{2}\mbox{Tr}\left([x^{i},x^{j}][x^{i},x^{j}]\right)-\frac{1}{2}f^{abc}x^{ja}\bar{\Psi}^{b}\Gamma^{j}\otimes\mathbf{1}\Psi^{c}.$
As one should have expected, at last we get a decoupled $U(1)$ sector and a
coupled $SU(N)$ sector, which is nothing else but an $\mathcal{N}=8$ super
Yang-Mills theory on the world-volume of $N$ coincident D2-branes. The action
(28) is valid in the energy range between $g_{YM}^{2}N$ and $kg_{YM}^{2}$, and
we have assumed $N\ll k$ in the above procedure.
## IV Discussion of New BL model and Conclusion
In new BL models Bagger:2008se , in general the structure constant of
3-algebra does not have three totally anti-symmetric indices, so one cannot
apply Mukhi and Papageorgakis’s method Mukhi:2008ux in a naive way. This
difficulty has been discussed in a newly appeared paper Cherkis:2008qr , where
Cherkis and Saemann also proposed a class of models similar to (but a little
different from) the new BL models.
One viable extrapolation of Mukhi and Papageorgakis’s method Mukhi:2008ux is
to decompose the 3-algebra into a couple of 2-algebras, say
$\mathcal{G}_{1}\oplus\mathcal{G}_{2}$, just as did in Bagger:2008se :
$f^{abcd}=\sum_{\lambda}\omega_{\lambda}\sum_{\alpha}(t^{\alpha}_{\lambda})^{ad}(t^{\alpha}_{\lambda})^{bc}.$
(31)
When $\mathcal{G}_{1}=\mathcal{G}_{2}=u(N)$, it exactly recovers the ABJM
theories, and the Higgs mechanism we discussed in the previous section exactly
applies.
More generally, if $\mathcal{G}_{1}=\mathcal{G}_{2}$ is another Lie algebra
instead of $u(N)$, the mechanism is similar. Some key points are:
1. 1.
taking the VEV as (9), which has also been mentioned in Aharony:2008ug ;
2. 2.
rearrange the gauge field as (8);
3. 3.
decompose four complex scalar (spinor) fields into eight real fields.
The procedure shown in the previous section still works if only
$\mathcal{G}_{1}=\mathcal{G}_{2}$, although that field theory would have less
supersymmetries. It is easy to see that in general the goldstones becomes non-
dynamical, at least they will disappear in the kinetic term. While the
$G_{1}\times G_{1}$ gauge symmetry is broken down to $G_{1}$. The degrees of
freedom of the goldstone bosons are transferred to the dynamical $G_{1}$ gauge
fields.
However, if $\mathcal{G}_{1}\neq\mathcal{G}_{2}$, the above scheme is helpless
and things would be much trickier.
As a conclusion, we can see that Mukhi and Papageorgakis’s method, after a
little adjustment, is still powerful in deriving the world-volume action of
$N$ D2-branes from that of $N$ M2-branes when $N$ is arbitrary.
_Note added_ : When this work was finished, a similar paper Li:2008ya
appeared with the emphasis on the pure Yang-Mils terms. The interested readers
are encouraged to compare our results with that paper, especially the Yang-
Mills coupling constant.
###### Acknowledgements.
We are grateful to Miao Li for many helpful comments. Also we would like to
thank Wei He, Yushu Song and Gang Yang for discussions. This work was
supported by grants of CNSF and grants of USTC.
## References
* (1) D. S. Berman, Phys. Rept. 456, 89 (2008) [arXiv:0710.1707 [hep-th]].
* (2) J. Bagger and N. Lambert, Phys. Rev. D 75, 045020 (2007) [arXiv:hep-th/0611108].
* (3) J. Bagger and N. Lambert, Phys. Rev. D 77, 065008 (2008) [arXiv:0711.0955 [hep-th]].
* (4) J. Bagger and N. Lambert, JHEP 0802, 105 (2008) [arXiv:0712.3738 [hep-th]].
* (5) A. Gustavsson, arXiv:0709.1260 [hep-th].
* (6) A. Gustavsson, JHEP 0804, 083 (2008) [arXiv:0802.3456 [hep-th]].
* (7) M. A. Bandres, A. E. Lipstein and J. H. Schwarz, JHEP 0805, 025 (2008) [arXiv:0803.3242 [hep-th]].
* (8) P. M. Ho, R. C. Hou and Y. Matsuo, JHEP 0806, 020 (2008) [arXiv:0804.2110 [hep-th]].
* (9) G. Papadopoulos, JHEP 0805, 054 (2008) [arXiv:0804.2662 [hep-th]].
* (10) J. P. Gauntlett and J. B. Gutowski, arXiv:0804.3078 [hep-th].
* (11) U. Gran, B. E. W. Nilsson and C. Petersson, arXiv:0804.1784 [hep-th].
* (12) H. Awata, M. Li, D. Minic and T. Yoneya, JHEP 0102, 013 (2001) [arXiv:hep-th/9906248].
* (13) J. Gomis, G. Milanesi and J. G. Russo, JHEP 0806, 075 (2008) [arXiv:0805.1012 [hep-th]].
* (14) S. Benvenuti, D. Rodriguez-Gomez, E. Tonni and H. Verlinde, arXiv:0805.1087 [hep-th].
* (15) P. M. Ho, Y. Imamura and Y. Matsuo, JHEP 0807, 003 (2008) [arXiv:0805.1202 [hep-th]].
* (16) O. Aharony, O. Bergman, D. L. Jafferis and J. Maldacena, arXiv:0806.1218 [hep-th].
* (17) S. Mukhi and C. Papageorgakis, JHEP 0805, 085 (2008) [arXiv:0803.3218 [hep-th]].
* (18) M. Van Raamsdonk, JHEP 0805, 105 (2008) [arXiv:0803.3803 [hep-th]].
* (19) D. S. Berman, L. C. Tadrowski and D. C. Thompson, arXiv:0803.3611 [hep-th].
* (20) J. Distler, S. Mukhi, C. Papageorgakis and M. Van Raamsdonk, JHEP 0805, 038 (2008) [arXiv:0804.1256 [hep-th]].
* (21) I. R. Klebanov and E. Witten, Nucl. Phys. B 536, 199 (1998) [arXiv:hep-th/9807080].
* (22) J. Bagger and N. Lambert, arXiv:0807.0163 [hep-th].
* (23) M. Benna, I. Klebanov, T. Klose and M. Smedback, arXiv:0806.1519 [hep-th].
* (24) Y. Honma, S. Iso, Y. Sumitomo and S. Zhang, arXiv:0806.3498 [hep-th].
* (25) J. Gomis, A. J. Salim and F. Passerini, arXiv:0804.2186 [hep-th].
* (26) J. Gomis, A. J. Salim and F. Passerini, arXiv:0804.2186 [hep-th].
* (27) E. A. Bergshoeff, M. de Roo and O. Hohm, Class. Quant. Grav. 25, 142001 (2008) [arXiv:0804.2201 [hep-th]].
* (28) Y. Honma, S. Iso, Y. Sumitomo and S. Zhang, arXiv:0805.1895 [hep-th].
* (29) H. Fuji, S. Terashima and M. Yamazaki, arXiv:0805.1997 [hep-th].
* (30) Y. Song, arXiv:0805.3193 [hep-th].
* (31) M. Li and T. Wang, arXiv:0805.3427 [hep-th].
* (32) K. Hosomichi, K. M. Lee, S. Lee, S. Lee and J. Park, arXiv:0805.3662 [hep-th].
* (33) S. Banerjee and A. Sen, arXiv:0805.3930 [hep-th].
* (34) H. Lin, arXiv:0805.4003 [hep-th].
* (35) J. Figueroa-O’Farrill, P. de Medeiros and E. Mendez-Escobar, arXiv:0805.4363 [hep-th].
* (36) M. A. Bandres, A. E. Lipstein and J. H. Schwarz, arXiv:0806.0054 [hep-th].
* (37) F. Passerini, arXiv:0806.0363 [hep-th].
* (38) C. Ahn, arXiv:0806.1420 [hep-th].
* (39) S. Cecotti and A. Sen, arXiv:0806.1990 [hep-th].
* (40) E. A. Bergshoeff, M. de Roo, O. Hohm and D. Roest, arXiv:0806.2584 [hep-th].
* (41) P. de Medeiros, J. Figueroa-O’Farrill and E. Mendez-Escobar, arXiv:0806.3242 [hep-th].
* (42) T. Nishioka and T. Takayanagi, arXiv:0806.3391 [hep-th].
* (43) C. Ahn, arXiv:0806.4807 [hep-th].
* (44) J. Bedford and D. Berman, arXiv:0806.4900 [hep-th].
* (45) G. Grignani, T. Harmark and M. Orselli, arXiv:0806.4959 [hep-th].
* (46) K. Hosomichi, K. M. Lee, S. Lee, S. Lee and J. Park, arXiv:0806.4977 [hep-th].
* (47) S. Terashima, arXiv:0807.0197 [hep-th].
* (48) B. Chen and J. B. Wu, arXiv:0807.0802 [hep-th].
* (49) C. S. Chu, P. M. Ho, Y. Matsuo and S. Shiba, arXiv:0807.0812 [hep-th].
* (50) S. Cherkis and C. Saemann, arXiv:0807.0808 [hep-th].
* (51) T. Li, Y. Liu and D. Xie, arXiv:0807.1183 [hep-th].
* (52) Chethan Krishnan, Carlo Maccaferri , arXiv:0805.3125 [hep-th].
| arxiv-papers | 2008-07-09T13:18:08 | 2024-09-04T02:48:56.644918 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Yi Pang, Tower Wang",
"submitter": "Pang Yi",
"url": "https://arxiv.org/abs/0807.1444"
} |
0807.1503 | # Theory for superfluidity in a Bose system
Zhidong Hao Department of Physics, University of Science and Technology of
China, Hefei, Anhui 230026, China
###### Abstract
We present a microscopic theory for superfluidity in an interacting many-
particle Bose system (such as liquid 4He). We show that, similar to
superconductivity in superconductors, superfluidity in a Bose system arises
from pairing of particles of opposite momenta. We show the existence of an
energy gap in single-particle excitation spectrum in the superfluid state and
the existence of a specific heat jump at the superfluid transition. We derive
an expression for superfluid particle density $n_{s}$ as a function of
temperature $T$ and superfluid velocity ${\bf v}_{s}$. We show that
superfluid-state free energy density $F$ is an increasing function of $v_{s}$
(i.e., $\partial F/\partial v_{s}>0$), which indicates that a superfluid has a
tendency to remain motionless (this result qualitatively explains the Hess-
Fairbank effect, which is analogous to the Meissner effect in
superconductors). We further speculate the existence of the equation ${\bf
j}=-\Lambda\nabla\times\text{\boldmath$\omega$}$, where ${\bf j}=n_{s}{\bf
v}_{s}$ is the superfluid current density,
$\text{\boldmath$\omega$}=\nabla\times{\bf v}_{s}$ the superfluid vorticity,
and $\Lambda$ a positive constant (with the help of this equation, the Hess-
Fairbank effect can be quantitatively described).
###### pacs:
67.25.D-
††preprint: Not ready for distribution
## I introduction
We present in this paper a microscopic theory for superfluidity in an
interacting many-particle Bose system such as liquid 4He. The theory is based
on an assumption that particles of opposite momenta are paired in the
superfluid state, and thus, is similar in many respects to the BCS theory of
superconductivity.bcs
It is well known that there is a marked similarity between liquid 4He II (the
superfluid phase of liquid 4He) and superconductors, both being chiefly
characterized by their ability to sustain flows of particles at a constant
velocity without a driving force.londonI ; londonII However, unlike
superconductors, for which there exists a successful microscopic theory, i.e.,
the BCS theory of superconductivity,bcs a satisfactory microscopic theory for
liquid 4He II is still lacking, despite many efforts (for example, Refs.
london1938, ; tisza, ; landau, ; bogo1947, ; feynman, ).
Fundamental to the BCS theory of superconductivity is an assumption that
electrons of opposite momenta and spins are paired in the superconducting
state.bcs This assumption allows microscopic derivation of all essential
properties of the superconducting state, such as the existence of an energy
gap in electronic excitation spectrum, a second-order phase transition
(manifested by a specific heat jump at the superconducting transition), the
Meissner effect, and the Josephson effect.
In this paper we show that it is also the pairing of particles of opposite
momenta that is responsible for superfluidity in a Bose system. Namely, the
cause for superconductivity in superconductors and superfluidity in liquid 4He
II is indeed essentially the same, irrespective of the nature of the particles
involved.
Some previous attempts to develop a microscopic theory for superfluidity in
liquid 4He II failed at the very start by assuming that the ground state of
liquid 4He II is a Bose-Einstein condensate (for example, Ref. bogo1947, ). As
we will see in this paper, the ground state of a superfluid is not a Bose-
Einstein condensate, but a state in which particles of opposite momenta are
paired, similar to that of superconductors.
Pairing of particles in a Bose system has been studied by a number of authors
(for example, Refs. vb, ; ei, ; ns, ; yinlan, ). However, the authors did not
treat properly self-consistency associated with pairing approximation, and
thus, failed to establish a connection between pairing and superfluidity.
In Sec. II, we present the theory for the case where a superfluid is at rest,
and show the existence of an energy gap in single-particle excitation spectrum
in the superfluid state, and the existence of a specific heat jump at the
superfluid transition. In Sec. III, we present the theory for the case where a
superfluid current is present. We derive an expression for the superfluid
particle density as a function of temperature and superfluid velocity. We show
that the superfluid-state free energy density is an increasing function of
superfluid velocity, which indicates that a superfluid has a tendency to
remain motionless. This result provides a qualitative explanation for the
Hess-Fairbank experimenthess-fairbank in which a reduction of moment of
inertia was observed when a rotating cylinder of liquid 4He was cooled through
the superfluid transition (this phenomenon, known in the literature as the
Hess-Fairbank effect, is analogous to the Meissner effect in superconductors).
We further consider how the Hess-Fairbank effect can be quantitatively
described. A brief summary is given in Sec. IV.
## II superfluid transition
We consider an interacting many-particle Bose system. We assume in this
section that superfluid velocity ${\bf v}_{s}=0$ (we will consider the case
where ${\bf v}_{s}\neq 0$ in the next section).
Similar to the pairing Hamiltonian in the BCS theory of superconductivity,bcs
we write the Hamiltonian of the interacting many-particle Bose system as
$\hat{H}=\sum_{\bf k}\left(\epsilon_{\bf k}-\mu\right)a^{\dagger}_{\bf
k}a_{\bf k}+\frac{1}{2}\sum_{\bf kk^{\prime}}V_{\bf
kk^{\prime}}a^{\dagger}_{\bf k}a^{\dagger}_{\bf-k}a_{-\bf k^{\prime}}a_{\bf
k^{\prime}},$ (1)
where $\epsilon_{\bf k}$ is the normal-state single-particle energy, $\mu$ the
chemical potential, $V_{\bf kk^{\prime}}$ the pairing interaction matrix
element, and $a^{\dagger}_{\bf k}$ and $a_{\bf k}$ are Bose operators for a
single-particle state of wave-vector k in the normal state and satisfy the
commutation rule $[a_{\bf k},a^{\dagger}_{\bf k^{\prime}}]=\delta_{\bf
k,k^{\prime}}$.
This Hamiltonian can be diagonalized in essentially the same manner as in the
BCS theory.bcs ; bogo ; valatin Namely, we assume
$\langle a_{-\bf k}a_{\bf k}\rangle\neq 0$ (2)
in the superfluid state for a pair of (k) and $(-{\bf k})$ particles (where
the angle brackets $\langle\cdots\rangle$ denote a thermal average); treat
$(a_{-\bf k}a_{\bf k}-\langle a_{-\bf k}a_{\bf k}\rangle)$ as a small quantity
so that terms bilinear in $(a_{-\bf k}a_{\bf k}-\langle a_{-\bf k}a_{\bf
k}\rangle)$ can be neglected; define an energy gap parameter
$\Delta_{\bf k}=-\sum_{\bf k^{\prime}}V_{\bf k,k^{\prime}}\langle a_{-\bf
k^{\prime}}a_{\bf k^{\prime}}\rangle$ (3)
(because of the similarity between the present theory and the BCS theory of
superconductivity, we will similarly refer to the quantity $\Delta_{\bf k}$ as
an “energy gap parameter” in this paper, although, as we will see below, it
does not directly relate to an “energy gap” in the present theory); and apply
a canonical transformationbogo1947 ; bogo ; valatin
$\left(\begin{array}[]{c}a_{\bf k}\\\ a^{\dagger}_{-\bf
k}\end{array}\right)=\left(\begin{array}[]{cc}u_{\bf k}&v_{\bf k}\\\
v^{\star}_{\bf k}&u^{\star}_{\bf
k}\end{array}\right)\left(\begin{array}[]{c}\alpha_{\bf k}\\\
\alpha^{\dagger}_{-\bf k}\end{array}\right),$ (4)
where $\alpha^{\dagger}_{\bf k}$ and $\alpha_{\bf k}$ are new Bose operators
for a single-particle excitation of wave-vector k in the superfluid state, and
coefficients $u_{\bf k}$ and $v_{\bf k}$ are so determined as to diagonalize
the Hamiltonian while maintaining the commutation rule $[\alpha_{\bf
k},\alpha^{\dagger}_{\bf k^{\prime}}]=\delta_{\bf k,k^{\prime}}$.
The diagonalized Hamiltonian is
${\hat{H}}=\sum_{\bf k}\left(U_{\bf k}+E_{\bf k}\alpha^{\dagger}_{\bf
k}\alpha_{\bf k}\right),$ (5)
where
$U_{\bf k}=-\frac{\xi_{\bf k}-E_{\bf k}}{2}+\frac{|\Delta_{\bf
k}|^{2}}{4E_{\bf k}}\left(1+2n_{\bf k}\right);$ (6) $\xi_{\bf k}=\epsilon_{\bf
k}-\mu$ (7)
is the single-particle energy in the normal state, measured relative to
chemical potential $\mu$;
$E_{\bf k}=\sqrt{\xi_{\bf k}^{2}-|\Delta_{\bf k}|^{2}}$ (8)
the single-particle excitation energy in the superfluid state; and
$n_{\bf k}=\left(e^{E_{\bf k}/k_{B}T}-1\right)^{-1}$ (9)
the Bose function (the number of single-particle excitations of wave-vector
k).
Coefficients $u_{\bf k}$ and $v_{\bf k}$ are found to satisfy the following
relations:
$|u_{\bf k}|^{2}=\frac{1}{2}\left(\frac{\xi_{\bf k}}{E_{\bf k}}+1\right),$
(10) $|v_{\bf k}|^{2}=\frac{1}{2}\left(\frac{\xi_{\bf k}}{E_{\bf
k}}-1\right),$ (11)
and
$\Delta_{\bf k}u_{\bf k}v^{\star}_{\bf k}=\frac{|\Delta_{\bf k}|^{2}}{2E_{\bf
k}}.$ (12)
After the diagonalization of Hamiltonian $\hat{H}$, Eq. (3) can be expressed
as
$\Delta_{\bf k}=\mbox{}-\sum_{\bf k^{\prime}}V_{\bf
k,k^{\prime}}\frac{1+2n_{\bf k^{\prime}}}{2E_{\bf k^{\prime}}}\Delta_{\bf
k^{\prime}}.$ (13)
This is a self-consistency equation that must be satisfied by $\Delta_{\bf k}$
as a function of wave-vector k and temperature $T$.
### II.1 Critical temperature $T_{c}$
Similar to that in the BCS theory of superconductivity, energy gap parameter
$\Delta_{\bf k}$ is an important quantity in the present theory. It is because
of the existence of $\Delta_{\bf k}$ that makes the superfluid state different
from the normal state. In this and the next subsections we consider
determination of $\Delta_{\bf k}$.
First, in the limit of $T\rightarrow T_{c}$ (because of the similarity between
the present theory and the BCS theory of superconductivity, we are similarly
using $T_{c}$, instead of $T_{\lambda}$, to denote the critical temperature of
the superfluid transition), we have $|\Delta_{\bf k}|\rightarrow 0$ so that
Eq. (13) can be linearized and we have an eigenvalue problem:
$\Delta_{\bf k}=-\sum_{{\bf k}^{\prime}}V_{{\bf k},{\bf
k}^{\prime}}\frac{\coth\left[(\epsilon_{\bf
k}-\mu_{0})/2k_{B}T_{c}\right]}{2(\epsilon_{\bf k}-\mu_{0})}\Delta_{{\bf
k}^{\prime}}\,,$ (14)
where $\mu_{0}$ is the value of chemical potential $\mu$ at critical
temperature $T_{c}$, and we have used $1+2n_{\bf k}=\coth(E_{\bf k}/2k_{B}T)$
and $E_{\bf k}=\epsilon_{\bf k}-\mu_{0}$ at $T=T_{c}$.
Critical temperature $T_{c}$ and phase $\theta_{\bf k}$ (as defined via
$\Delta_{\bf k}=|\Delta_{\bf k}|e^{i\theta_{\bf k}}$) are determined by
solving Eq. (14) for given interaction $V_{\bf k,k^{\prime}}$, single-particle
energy spectrum $\epsilon_{\bf k}$ and chemical potential $\mu_{0}$.
Note that it is not necessary to assume $V_{{\bf k},{\bf k^{\prime}}}<0$ in
order for Eq. (14) to have a $T_{c}>0$ solution. Therefore, the view that an
attractive interaction is responsible for particle pairing is incorrect. Here
we also emphasize that the pairing of particles of opposite momenta, as
expressed by Eq. (2), is a kind of ordering in momentum space (this point
agrees with London’s view that superconducting/superfluid state is an ordered
state in momentum spacelondonI ; londonII ); it does not mean that bound pairs
of particles (due to an attractive interaction) are formed.
### II.2 $|\Delta_{\bf k}|$ and $E_{\bf k}$
With respect to determination of $|\Delta_{\bf k}|$, we note that the self-
consistency equation, Eq. (13), can be converted into
$\sum_{\bf k}|\Delta_{\bf k}|^{2}\frac{\partial}{\partial
T}\left(\frac{1+2n_{\bf k}}{E_{\bf k}}\right)=0$ (15)
by first operating $\partial/\partial T$ on Eq. (13), and then, multiplying
the resulting equation by $\Delta^{\star}_{\bf k}(1+2n_{\bf k})/2E_{\bf k}$
and summing over k.
Interaction $V_{\bf kk^{\prime}}$ no longer appears in Eq. (15), because all
information about $V_{\bf k,k^{\prime}}$ is already contained in $T_{c}$, and
the latter is involved through the condition $|\Delta_{\bf k}(T_{c})|=0$.
From Eq. (15) we can see that the self-consistency equation alone does not
allow unique determination of $|\Delta_{\bf k}|$, because, as one can see, Eq.
(15) can have an infinite number of solutions. This property of the self-
consistency equation is true for arbitrary interaction $V_{\bf kk^{\prime}}$,
because Eq. (15) is derived for arbitrary $V_{\bf kk^{\prime}}$. Actually,
this property can also been seen directly from Eq. (13) by noticing that the
equation is linear with respect to $e^{i\theta_{\bf k}}$. [Even in the case of
$V_{\bf kk^{\prime}}=-V$, which leads to $\Delta_{\bf k}=V\sum_{\bf
k^{\prime}}[(1+2n_{\bf k^{\prime}})/2E_{\bf k^{\prime}}]\Delta_{\bf
k^{\prime}}$, $|\Delta_{\bf k}|$ still cannot be uniquely determined, because
phase $\theta_{\bf k^{\prime}}$ of $\Delta_{\bf k^{\prime}}$ in the summation
over ${\bf k^{\prime}}$ is measured relative to phase $\theta_{\bf k}$ of
$\Delta_{\bf k}$ on the left-hand side of the equation (it is relative phases
that matter). This is more clearly seen if we re-write the equation as
$|\Delta_{\bf k}|=V\sum_{\bf k^{\prime}}e^{i(\theta_{\bf
k^{\prime}}-\theta_{\bf k})}[(1+2n_{\bf k^{\prime}})/2E_{\bf
k^{\prime}}]|\Delta_{\bf k^{\prime}}|=V\sum_{\bf k^{\prime}}\cos(\theta_{\bf
k^{\prime}}-\theta_{\bf k})[(1+2n_{\bf k^{\prime}})/2E_{\bf
k^{\prime}}]|\Delta_{\bf k^{\prime}}|$, where $\cos(\theta_{\bf
k^{\prime}}-\theta_{\bf k})=1$ or $-1$ depending on $\theta_{\bf
k^{\prime}}-\theta_{\bf k}=0$ or $\pi$. We have $|\Delta_{\bf k}|=\Delta={\bf
k}\text{-independent}$ only when $\theta_{\bf k}=\theta={\bf
k}\text{-independent}$, but other solutions for $|\Delta_{\bf k}|$, with
$\theta_{\bf k}$ being k-dependent, are also possible. Similarly, for a
separable interaction of the form $V_{\bf kk^{\prime}}=-V\omega_{\bf
k}\omega_{\bf k^{\prime}}$, the solution $|\Delta_{\bf k}|=\Delta|\omega_{\bf
k}|$ corresponds to a solution with $e^{i\theta_{\bf
k}}=\text{sgn}(\omega_{\bf k})e^{i\theta}$ with $\theta$ being an arbitrary
constant, and is only one of an infinite number of possible solutions.]
On the other hand, we note that diagonalized Hamiltonian ${\hat{H}}$ is
$T$-dependent, i.e., both $U_{\bf k}$ and $E_{\bf k}$ in Eq. (5) are functions
of temperature $T$, because of their dependence upon $|\Delta_{\bf k}(T)|$.
Since diagonalized Hamiltonian ${\hat{H}}$ describes a set of independent
excitations, and there is no transition between different single-particle
states in thermodynamic equilibrium, we expect the thermal energy and entropy
associated with a single-particle state of wave-vector k to be
$\varepsilon_{\bf k}=U_{\bf k}+n_{\bf k}E_{\bf k}$ (16)
and
$S_{\bf k}=-k_{B}\left[n_{\bf k}\ln n_{\bf k}-(1+n_{\bf k})\ln(1+n_{\bf
k})\right],$ (17)
respectively.fetter ; huang ; feynmanBook However, because of the
$T$-dependence of ${\hat{H}}$, when we calculate, for a single-particle state
of wave-vector k, the partition function $Z_{\bf
k}=\text{Tr}(e^{-{\hat{H}}_{\bf k}/k_{B}T})$ (where ${\hat{H}}_{\bf k}=U_{\bf
k}+E_{\bf k}\alpha^{\dagger}_{\bf k}\alpha_{\bf k}$), free energy $F_{\bf
k}=-k_{B}T\ln Z_{\bf k}$, entropy $S_{\bf k}=-\partial F_{\bf k}/\partial T$
and thermal energy $\varepsilon_{\bf k}=F_{\bf k}+TS_{\bf k}$, we find that
there are additional terms involving $\partial U_{\bf k}/\partial T$ and
$\partial E_{\bf k}/\partial T$ in each of the expressions for
$\varepsilon_{\bf k}$ and $S_{\bf k}$, as compared to Eqs. (16) and (17). By
letting the sum of the additional terms in each of the expressions for
$\varepsilon_{\bf k}$ and $S_{\bf k}$ to be zero, we arrive at
$\frac{\partial U_{\bf k}}{\partial T}+\,n_{\bf k}\frac{\partial E_{\bf
k}}{\partial T}=0.$ (18)
This equation represents an additional self-consistency requirement of the
theory, and must be consistent with Eq. (13). To see that this is indeed true,
we substitute Eq. (6) into Eq. (18) to obtain
$|\Delta_{\bf k}|^{2}\frac{\partial}{\partial T}\left(\frac{1+2n_{\bf
k}}{E_{\bf k}}\right)=0.$ (19)
A solution of this equation is certainly also a solution of Eq. (15), and
therefore also a solution of Eq. (13), and thus, we see that Eq. (18) is
indeed consistent with Eq. (13).
Equation (19) shows that there are two possible solutions for $|\Delta_{\bf
k}|$ for each single-particle state of wave-vector k: one is a trivial
solution, $|\Delta_{\bf k}|=0$ (corresponding to the normal state), and the
other is a non-trivial solution ($|\Delta_{\bf k}|>0$, corresponding to the
superfluid state) satisfying
$\frac{\partial}{\partial T}\left(\frac{1+2n_{\bf k}}{E_{\bf k}}\right)=0,$
(20)
which is readily solved to give
$\frac{1+2n_{\bf k}}{E_{\bf k}}=\text{$T$-independent for $T\leq T_{c}$}.$
(21)
A consequence of Eq. (20), or (21), is that chemical potential $\mu$ in the
superfluid state is $T$-independent. This is shown in Appendix A (where we
discuss chemical potential $\mu$ and number-of-particle distribution $\langle
a^{\dagger}_{\bf k}a_{\bf k}\rangle$ in the superfluid state). Then, by using
$1+2n_{\bf k}=\coth(E_{\bf k}/2k_{B}T)$ and the condition that $|\Delta_{\bf
k}|=0$ at $T=T_{c}$, we can express Eq. (21) as
$\frac{\coth\left(E_{\bf k}/2k_{B}T\right)}{E_{\bf
k}}=\frac{\coth\left(\xi_{\bf k}/2k_{B}T_{c}\right)}{\xi_{\bf k}}$ (22)
with
$\xi_{\bf k}=\epsilon_{\bf k}-\mu_{0},$ (23) $E_{\bf k}=\sqrt{(\epsilon_{\bf
k}-\mu_{0})^{2}-|\Delta_{\bf k}|^{2}},$ (24)
and $\mu_{0}$ being the value of $\mu$ at $T=T_{c}$.
It is shown in Appendix B (where we discuss the ground state energy of the
superfluid state) that $\mu_{0}$ must be below a certain negative value in
order for the superfluid state to be energetically favorable as compared to
the normal state.
Equation (22) is an implicit solution for $|\Delta_{\bf k}|$ (or $E_{\bf k}$)
as a function of $\epsilon_{\bf k}$ and $T$ for given $T_{c}$ and $\mu_{0}$. A
complete solution for $\Delta_{\bf k}$ is therefore a combination of the
solution of Eq. (22) for $|\Delta_{\bf k}|/k_{B}T_{c}$ and the solutions of
Eq. (14) for $T_{c}$ and $\theta_{\bf k}$.
The present analysis with respect to the determination of $\Delta_{\bf k}$ is
similar to that of Ref. hao01, with respect to the determination of the
energy gap parameter in the BCS theory of superconductivity.
Figure 1: Energy dependence of energy gap parameter amplitude $|\Delta_{\bf
k}|$ for $-\mu_{0}/k_{B}T_{c}=0.5$ and different values of $T/T_{c}$ as
indicated on the curves. Figure 2: Temperature dependence of normalized
energy gap parameter amplitude $|\Delta_{\bf k}|/|\Delta_{\bf k}(0)|$ for
$-\mu_{0}/k_{B}T_{c}=0.5$ and different values of $\epsilon_{\bf
k}/k_{B}T_{c}$ as indicated on the curves. Figure 3: Superfluid-state
single-particle excitation energy $E_{\bf k}$ versus normal-state single-
particle energy $\epsilon_{\bf k}$ for $-\mu_{0}/k_{B}T_{c}=0.5$ and different
values of $T/T_{c}$ as indicated on the curves (solid curves). The two dotted
curves respectively show normal-state single-particle excitation energy
$E_{\bf k}^{(n)}=\epsilon_{\bf k}$ for $T$ below the Bose-Einstein
condensation temperature and $E_{\bf k}^{(n)}=\epsilon_{\bf k}-\mu_{0}$ at
$T=T_{c}$, as indicated on the curves. Figure 4: Temperature dependence of
minimum superfluid-state single-particle excitation energy $E_{min}$ for
$-\mu_{0}/k_{B}T_{c}=0.3$, 0.5 and 0.8, as indicated on the curves.
We solve Eq. (22) numerically by using an iterative methodconte80 to obtain
$|\Delta_{\bf k}|/k_{B}T_{c}$ as a function of $\epsilon_{\bf k}/k_{B}T_{c}$
and $T/T_{c}$ for given $\mu_{0}/k_{B}T_{c}$.
Figure 1 shows $\epsilon_{\bf k}$-dependence of $|\Delta_{\bf k}|$ for
$-\mu_{0}/k_{B}T_{c}=0.5$ and different values of $T$ as indicated on the
curves.
Figure 2 shows $T$-dependence of $|\Delta_{\bf k}|$ for
$-\mu_{0}/k_{B}T_{c}=0.5$ and different values of $\epsilon_{\bf k}$ as
indicated on the curves.
Figure 3 shows $E_{\bf k}$ versus $\epsilon_{\bf k}$ for
$-\mu_{0}/k_{B}T_{c}=0.5$ and different values of $T$ as indicated on the
curves. For comparison, normal-state single-particle excitation energy $E_{\bf
k}^{(n)}=\epsilon_{\bf k}$ for $T$ below the Bose-Einstein condensation
temperaturefetter ; huang ; feynmanBook and $E_{\bf k}^{(n)}=\epsilon_{\bf
k}-\mu_{0}$ at $T=T_{c}$ are shown as the dotted curves in the figure.
From Fig. 3 we can see the existence of an energy gap in the superfluid-state
single-particle excitation spectrum. Namely, minimum value $E_{min}$ of
$E_{\bf k}$, which is located at $\epsilon_{\bf k}=0$, is greater than zero.
According to Eq. (22),
$E_{min}=(-\mu_{0})\tanh\left[(-\mu_{0})/2k_{B}T_{c}\right]$ at $T=0$, and
increases monotonically to $E_{min}=-\mu_{0}$ at $T=T_{c}$. Since, as shown in
Appendix B, $\mu_{0}$ must be below a certain negative value in the superfluid
state, we see that $E_{min}>0$ in the superfluid state.
Figure 4 shows $T$-dependence of $E_{min}$ for different values
$-\mu_{0}/k_{B}T_{c}$ as indicated on the curves.
### II.3 Specific heat $C(T)$
Having obtained $|\Delta_{\bf k}(T)|$, we can calculated thermodynamic
quantities of the superfluid state. We consider the ground state energy of the
superfluid in Appendix B. We calculate in this subsection the specific heat of
the superfluid.
In the superfluid state, specific heat $C$ is given by
$\displaystyle\\!\\!\\!\\!\\!\\!C\\!$ $\displaystyle\\!\\!=\\!\\!$
$\displaystyle\frac{\partial\langle\hat{H}\rangle}{\partial T}=\sum_{\bf
k}E_{\bf k}\frac{\partial n_{\bf k}}{\partial T}$ (25)
$\displaystyle\\!\\!=\\!\\!$ $\displaystyle\\!k_{B}\\!\\!\sum_{\bf
k}\\!\\!\frac{E^{2}_{\bf k}/T^{2}}{e^{E_{\bf k}/T}(1\\!\\!-\\!\\!e^{-E_{\bf
k}/T})^{2}\\!\\!+\\!\\!2(\xi_{\bf k}/T)\tanh(\xi_{\bf k}/2)},$
where we have used Eq. (22), and have adopted a set of dimensionless units for
the last expression, in which energies are measured in units of $k_{B}T_{c}$
and temperature in units of $T_{c}$.
Figure 5: Temperature dependence of specific heat $C$ for
$-\mu_{0}/k_{B}T_{c}=0.3$, 0.5 and 0.8, as indicated on the curves. For
clarity, the $T/T_{c}>1$ part of the $-\mu_{0}/k_{B}T_{c}=0.5$ curve is not
shown. The horizontal dotted line shows the value of $C/Nk_{B}=3/2$ for
$T/T_{c}\gg 1$.
In the normal state ($T/T_{c}\geq 1$), the specific heat is given by
$\displaystyle C$ $\displaystyle=$ $\displaystyle\sum_{\bf k}\xi_{\bf
k}\frac{\partial n_{\bf k}}{\partial T}$ (26) $\displaystyle=$ $\displaystyle
k_{B}\sum_{\bf k}\frac{e^{\xi_{\bf k}/T}}{(e^{\xi_{\bf
k}/T}-1)^{2}}\left[\frac{\xi_{\bf k}^{2}}{T^{2}}-\frac{\xi_{\bf
k}}{T}\frac{\partial(-\mu)}{\partial T}\right],$
where we have used the above-mentioned dimensionless units for the last
expression.
Figure 5 shows $C/Nk_{B}$ versus $T/T_{c}$ for different values of
$-\mu_{0}/k_{B}T_{c}$ as indicated on the curves, where $N$ is the total
number of particles of the system, and is given by
$\displaystyle N$ $\displaystyle=$ $\displaystyle\sum_{\bf k}n_{\bf k}(T_{c})$
(27) $\displaystyle=$ $\displaystyle\sum_{\bf k}\frac{1}{e^{\,\xi_{\bf
k}}-1},$
where $n_{\bf k}(T_{c})$ is the Bose function at $T=T_{c}$, and we have used
the above-mentioned dimensionless units for the last expression.
In calculating $C/Nk_{B}$, we have assumed $\epsilon_{\bf
k}=\hbar^{2}k^{2}/2m$ (where $m$ is particle mass), and have made the
substitution $\sum_{\bf k}\rightarrow(2\pi)^{-3}\int d^{3}k$. The integrals
involved are calculated by using the Simpson method.conte80 The method for
calculating $\mu$ and $\partial\mu/\partial T$ for the specific heat in the
normal state is explained in Appendix A.
As shown in Fig. 5, there exists a finite jump in the specific heat at the
transition temperature, indicating a second-order phase transition. The
magnitude of the jump is larger for a larger value of $-\mu_{0}/k_{B}T_{c}$.
In the limit of $T\rightarrow 0$, we have $\partial C/\partial T\rightarrow
0$, because of the existence of an energy gap in the single-particle
excitation spectrum.
Experimentally, the $C$-versus-$T$ curve shows a $\lambda$-shaped peak at the
transition.expC
## III superfluidity
We next consider the case where the superfluid is in a state of uniform flow
with velocity ${\bf v}_{s}$.
We write the Hamiltonian of the system as
$\hat{H}=\sum_{\bf k}\left(\epsilon_{\bf k}-\mu\right)a^{\dagger}_{\bf
k}a_{\bf k}+\frac{1}{2}\sum_{\bf kk^{\prime}}V_{\bf
kk^{\prime}}a^{\dagger}_{\bf k}a^{\dagger}_{\bf-k}a_{-\bf k^{\prime}}a_{\bf
k^{\prime}},$ (28)
which is the same as the Hamiltonian of Eq. (1) for the case of ${\bf
v}_{s}=0$, except that wave-vector k in the above expression is now measured
in the coordinate frame moving with the superfluid.
We assume that pairing occurs between particles of opposite momenta measured
in the coordinate frame moving with the superfluid. I.e., we assume
$\langle a_{-\bf k}a_{\bf k}\rangle\neq 0$ (29)
in the superfluid state for a pair of (k) and $(-{\bf k})$ particles.
Note that, since wave-vector k is measured in the coordinate frame moving with
the superfluid, if we use a free Bose gas as an example, a single-particle
state of wave-vector k means, in the laboratory frame, a single-particle state
of wave function
$\phi_{\bf k}=e^{i({\bf k}+{\bf q})\cdot{\bf x}}$ (30)
and energy
$\epsilon_{\bf k}=\hbar^{2}({\bf k}+{\bf q})^{2}/2m,$ (31)
where
${\bf q}=m{\bf v}_{s}/\hbar.$ (32)
Therefore, a pair of (k) and $(-{\bf k})$ particles have zero net momentum in
the frame moving with the superfluid, but have a net momentum of $2\hbar{\bf
q}=2m{\bf v}_{s}$ in the laboratory frame.
Diagonalization of Hamiltonian $\hat{H}$ is the same as in the case of ${\bf
v}_{s}=0$, except that we now have $\epsilon_{-\bf k}\neq\epsilon_{\bf k}$ for
${\bf v}_{s}\neq 0$. The results of the diagonalization are as follows.
The diagonalized Hamiltonian is
${\hat{H}}=\sum_{\bf k}\left[U_{\bf k}+\frac{1}{2}\left(E_{\bf
k}\alpha^{\dagger}_{\bf k}\alpha_{\bf k}+E_{-\bf k}\alpha^{\dagger}_{-\bf
k}\alpha_{-\bf k}\right)\right],$ (33)
where
$U_{\bf k}=-\frac{\xi_{\bf k}+\xi_{-\bf k}}{4}+\frac{E^{(s)}_{\bf
k}}{2}+\frac{|\Delta_{\bf k}|^{2}}{4E^{(s)}_{\bf k}}\left(1+n_{\bf k}+n_{-\bf
k}\right);$ (34) $\xi_{\bf k}=\epsilon_{\bf k}-\mu;$ (35) $E_{\bf
k}=E^{(s)}_{\bf k}+\frac{\xi_{\bf k}-\xi_{-\bf k}}{2};$ (36) $E^{(s)}_{\bf
k}=\sqrt{\left(\frac{\xi_{\bf k}+\xi_{-\bf k}}{2}\right)^{2}-|\Delta_{\bf
k}|^{2}}$ (37)
is the symmetric part of $E_{\bf k}$; and
$n_{\bf k}=\left(e^{E_{\bf k}/k_{B}T}-1\right)^{-1}$ (38)
the Bose function.
Coefficients $u_{\bf k}$ and $v_{\bf k}$ are found to satisfy the following
relations:
$|u_{\bf k}|^{2}=\frac{1}{2}\left(\frac{\xi_{\bf k}+\xi_{-\bf
k}}{2E^{(s)}_{\bf k}}+1\right),$ (39) $|v_{\bf
k}|^{2}=\frac{1}{2}\left(\frac{\xi_{\bf k}+\xi_{-\bf k}}{2E^{(s)}_{\bf
k}}-1\right),$ (40)
and
$\Delta_{\bf k}u_{\bf k}v^{\star}_{\bf k}=\frac{|\Delta_{\bf
k}|^{2}}{2E^{(s)}_{\bf k}}.$ (41)
The self-consistency equation for the energy gap parameter is
$\Delta_{\bf k}=\mbox{}-\sum_{\bf k^{\prime}}V_{\bf
k,k^{\prime}}\frac{1+n_{\bf k^{\prime}}+n_{-\bf k^{\prime}}}{2E^{(s)}_{\bf
k^{\prime}}}\Delta_{\bf k^{\prime}}.$ (42)
Since $\epsilon_{-\bf k}\neq\epsilon_{\bf k}$ for ${\bf v}_{s}\neq 0$, we have
$\xi_{-\bf k}\neq\xi_{\bf k}$, $E_{-\bf k}\neq E_{\bf k}$ and $n_{-\bf k}\neq
n_{\bf k}$, but we have $U_{-\bf k}=U_{\bf k}$, $E^{(s)}_{-\bf k}=E^{(s)}_{\bf
k}$ and $\Delta_{-\bf k}=\Delta_{\bf k}$, as we can see from the expressions
shown above.
Chemical potential $\mu$, relative to which energies such as $\xi_{\bf k}$ and
$E_{\bf k}$ are measured, is q-dependent, because each pair of particles in
the superfluid state has a net energy increase of $2(\hbar^{2}q^{2}/2m)$ due
to the flow of paired particles. Namely, we have
$\mu=\mu_{0}+\hbar^{2}q^{2}/2m,$ (43)
where $\mu_{0}$ is the superfluid-state chemical potential for ${\bf q}=0$,
and the second term is the per-particle energy increase due to the flow of
paired particles [see Appendix A for a detailed derivation of Eq. (43)].
For simplicity in presenting the theory, we will use the normal-state single-
particle energy spectrum of a free Bose gas as given by Eq. (31) in the
following.
With the help of Eqs. (31) and (43), we have
$\frac{\xi_{\bf k}+\xi_{-\bf k}}{2}=\xi^{(0)}_{\bf k}$ (44)
and
$\frac{\xi_{\bf k}-\xi_{-\bf k}}{2}=\frac{\hbar^{2}}{m}{\bf k}\cdot{\bf q}\,,$
(45)
where
$\xi^{(0)}_{\bf k}=\epsilon^{(0)}_{\bf k}-\mu_{0}$ (46)
and $\epsilon^{(0)}_{\bf k}$ is the value of $\epsilon_{\bf k}$ for ${\bf
q}=0$.
### III.1 $|\Delta_{\bf k}(T,{\bf q})|$
The following equation is derived as an additional self-consistency
requirement of the theory:
$\frac{1+n_{\bf k}+n_{-\bf k}}{E^{(s)}_{\bf k}}=\text{independent of $T$ and
{\bf q}},$ (47)
which is a generalization of Eq. (21) to the case of ${\bf v}_{s}\neq 0$. We
present the details of the derivation of this equation in Appendix C.
With the help of Eqs. (44) and (45) and by using $1+2n_{\bf k}=\coth(E_{\bf
k}/2k_{B}T)$ and the condition that $|\Delta_{\bf k}|=0$ at $(T,{\bf
q})=(T_{c},0)$, we can express Eq. (47) as
$\frac{\coth\left(\frac{E^{(s)}_{\bf k}+2\sqrt{\epsilon^{(0)}_{\bf k}}qz_{\bf
k}}{2T}\right)+\coth\left(\frac{E^{(s)}_{\bf k}-2\sqrt{\epsilon^{(0)}_{\bf
k}}qz_{\bf k}}{2T}\right)}{2E^{(s)}_{\bf k}}$
$=\frac{\coth\left(\frac{\epsilon^{(0)}_{\bf
k}-\mu_{0}}{2}\right)}{\epsilon^{(0)}_{\bf k}-\mu_{0}}$ (48)
where
$E^{(s)}_{\bf k}=\sqrt{(\epsilon^{(0)}_{\bf k}-\mu_{0})^{2}-|\Delta_{\bf
k}|^{2}},$ (49)
$z_{\bf k}=\cos\alpha_{\bf k}$ with $\alpha_{\bf k}$ being the angle between k
and q, and we have used a set of dimensionless units in which energies such as
$E^{(s)}_{\bf k}$, $\epsilon^{(0)}_{\bf k}$ and $\mu_{0}$ are measured in
units of $k_{B}T_{c}$, temperature is measured in units of $T_{c}$, and $q$ in
units of $q_{0}$, which is defined via $\hbar^{2}q_{0}^{2}/2m=k_{B}T_{c}$. For
4He, $q_{0}$ is $\sim 1\text{ \AA}^{-1}$, and the corresponding superfluid
velocity is $v_{s0}=\hbar q_{0}/m\sim 10^{2}$ cm/s.
Figure 6: (a) $T_{c{\bf k}}$ versus $\epsilon^{(0)}_{\bf k}$ for different
values of $qz_{\bf k}/q_{0}$ as indicated on the curves; (b) $q_{c{\bf
k}}z_{\bf k}$ versus $\epsilon^{(0)}_{\bf k}$ for different values of
$T/T_{c}$ as indicated on the curves; $-\mu_{0}/k_{B}T_{c}=0.5$ in both (a)
and (b). Figure 7: $|\Delta_{\bf k}|$ versus $\epsilon^{(0)}_{\bf k}$ for
different values of $T$ as indicated on the curves; $qz_{\bf k}/q_{0}=0.2$ and
$0.4$ in (a) and (b), respectively; $-\mu_{0}/k_{B}T_{c}=0.5$ in both (a) and
(b). Figure 8: $|\Delta_{\bf k}(T,qz_{\bf k})|/|\Delta_{\bf k}(0,0)|$ versus
$T/T_{c}$ for different values of $\epsilon^{(0)}_{\bf k}/k_{B}T_{c}$ as
indicated on the curves; $qz_{\bf k}/q_{0}=0.2$ and $0.4$ in (a) and (b),
respectively; $-\mu_{0}/k_{B}T_{c}=0.5$ in both (a) and (b). Figure 9:
$|\Delta_{\bf k}(T,qz_{\bf k})|/|\Delta_{\bf k}(0,0)|$ versus $qz_{\bf
k}/q_{0}$ for different values of $T/T_{c}$ as indicated on the curves;
$\epsilon^{(0)}_{\bf k}/k_{B}T_{c}=0.5$ and $2.0$ in (a) and (b),
respectively; $-\mu_{0}/k_{B}T_{c}=0.5$ in both (a) and (b).
Equation (48) is an implicit solution for $|\Delta_{\bf k}(T,{\bf q})|$ [or
$E^{(s)}_{\bf k}(T,{\bf q})$]. The variables $({\bf k},T,{\bf q})$ for the
function $|\Delta_{\bf k}(T,{\bf q})|$ appear in Eq. (48) in the forms of
$(\epsilon^{(0)}_{\bf k},T,qz_{\bf k})$, where $qz_{\bf k}$ is the component
of q along k. We solve Eq. (48) by using an iterative methodconte80 to obtain
$|\Delta_{\bf k}|$ as a function of $\epsilon^{(0)}_{\bf k}$, $T$ and $qz_{\bf
k}$ for given $\mu_{0}$.
Note that temperature $T_{c\bf k}$, defined as such that Eq. (48) has no
$|\Delta_{\bf k}|>0$ solution for given $\epsilon^{(0)}_{\bf k}$, $T$ and
$qz_{\bf k}$ if $T\geq T_{c\bf k}$, is a function of $\epsilon^{(0)}_{\bf k}$
and $qz_{\bf k}$. Only for ${\bf q}=0$ is $T_{c\bf k}=T_{c}$ the same for all
single-particle states.
Similarly, superfluid wave-vector $q_{c\bf k}$, defined as such that Eq. (48)
has no $|\Delta_{\bf k}|>0$ solution for given $\epsilon^{(0)}_{\bf k}$, $T$
and $qz_{\bf k}$ if $q\geq q_{c\bf k}$, is a function of $\epsilon^{(0)}_{\bf
k}$, $T$ and $z_{\bf k}$ (or, $q_{c\bf k}z_{\bf k}$ is a function of
$\epsilon^{(0)}_{\bf k}$ and $T$).
Figure 6(a) shows $T_{c\bf k}$ versus $\epsilon^{(0)}_{\bf k}$ for different
values of $qz_{\bf k}$, and Fig. 6(b) shows $q_{c{\bf k}}z_{\bf k}$ versus
$\epsilon^{(0)}_{\bf k}$ for different values of $T$;
$-\mu_{0}/k_{B}T_{c}=0.5$ is assumed in both Fig. 6(a) and Fig. 6(b).
Note that there is an upper bound for $T_{c\bf k}$, i.e., $T_{c\bf k}\leq
T_{c}$, but there is no upper bound for $q_{c\bf k}$, i.e., $q_{c\bf
k}\rightarrow\infty$ for $\epsilon^{(0)}_{\bf k}\rightarrow\infty$, and
$q_{c\bf k}=\infty$ if ${\bf k}\perp{\bf q}$ (because $z_{\bf k}=0$ for this
case).
The minimum values of $T_{c\bf k}$ and $q_{c\bf k}$, $T_{{c\bf
k},\text{min}}(qz_{\bf k})$ and $q_{{c\bf k},\text{min}}(T)$, are of
particular importance. For $T<T_{{c\bf k},\text{min}}(qz_{\bf k})$ [or
$q<q_{{c\bf k},\text{min}}(T)$], the system is in an all-paired state, in
which $|\Delta_{\bf k}|>0$ for all particles. For $T_{{c\bf
k},\text{min}}(qz_{\bf k})<T<T_{c}$ [or $q_{{c\bf k},\text{min}}(T)<q$], the
system is in a partly-paired state, in which particles in states having
$T_{c\bf k}(qz_{\bf k})<T$ [or $q_{c\bf k}(T)<q$] become de-paired (having
$|\Delta_{\bf k}|=0$) while those in states having $T_{c\bf k}(qz_{\bf k})>T$
[or $q_{c\bf k}(T)>q$] remain paired (having $|\Delta_{\bf k}|>0$). For
$T>T_{c}$, the system is in the normal state, in which $|\Delta_{\bf k}|=0$
for all particles.
A finite viscosity should be observable in a partly-paired state, because of
the existence of de-paired particles, which are expected to behave as normal-
state particles. We therefore expect critical velocity $v_{sc}(T)$, defined as
the superfluid velocity at the onset of an observable viscosity, to be about
the same as $\hbar q_{{c\bf k},\text{min}}(T)/m$. From the numerical results
shown in Fig. 6(b), for example, which are obtained for the case of
$-\mu_{0}/k_{B}T_{c}=0.5$, we can see that $q_{{c\bf k},\text{min}}(T)$ is a
few tenth of $q_{0}$, corresponding to a superfluid velocity of a few tenth of
$v_{s0}=\hbar q_{0}/m$. Since $v_{s0}\sim 10^{2}$ cm/s for 4He, we see that
the critical velocity for 4He in this case is about a few tens of centimeters
per second (which is several orders of magnitude smaller than the value
previously predicted by Landaulandau ).
Numerical results for $|\Delta_{\bf k}|$ versus $\epsilon^{(0)}_{\bf k}$ for
different values of $T$ and $qz_{\bf k}$ are shown in Fig. 7 [Figs. 7(a) and
7(b)]. Figure 7(a) shows an example of the case of $0<q<q_{c{\bf
k},\text{min}}(0)$. In this case, the $|\Delta_{\bf
k}|$-versus-$\epsilon^{(0)}_{\bf k}$ curve for $T=0$ is the same as in the
case of ${\bf q}=0$ (which is shown in Fig. 1). As $T$ increases,
$|\Delta_{\bf k}|$ for smaller $\epsilon^{(0)}_{\bf k}$ is more strongly
suppressed, and decreases faster. As $T$ increases further so that $T>T_{c{\bf
k},\text{min}}$, the $|\Delta_{\bf k}|$-versus-$\epsilon^{(0)}_{\bf k}$ curve
has a $|\Delta_{\bf k}|=0$ part for low energies (except for
$\epsilon^{(0)}_{\bf k}\rightarrow 0$), for which $T_{c{\bf k}}(qz_{\bf k})<T$
[or $q_{c{\bf k}}(T)<q$]. Figure 7(b) shows an example of the case of
$q>q_{c{\bf k},\text{min}}(0)$. In this case, the $|\Delta_{\bf
k}|$-versus-$\epsilon^{(0)}_{\bf k}$ curve has a $|\Delta_{\bf k}|=0$ part
even at $T=0$. Namely, at $T=0$, $|\Delta_{\bf k}|=0$ for those single-
particle states with $q_{c{\bf k}}(0)<q$. The vertical rises (or drops) in the
$|\Delta_{\bf k}|$-versus-$\epsilon^{(0)}_{\bf k}$ curves in both Fig. 7(a)
and Fig. 7(b) indicate discontinuities.
Figures 8 and 9 show, respectively, the numerical results for the
$T$-dependence and $qz_{\bf k}$-dependence of $|\Delta_{\bf k}|$. As shown in
the figures, $|\Delta_{\bf k}|$ is a monotonic decreasing function of $T$ and
$qz_{\bf k}$, except that, at $T=0$, $|\Delta_{\bf k}|>0$ is a constant for
$q<q_{c{\bf k}}(0)$ (Fig. 9). The vertical drops in some of the curves shown
in Figs. 8 and 9 indicate discontinuities.
### III.2 Superfluid particle density $n_{s}$
Particle current density j is the expectation value of particle current
density operator $\hat{\bf J}$.fetter I.e.,
${\bf j}=\langle\hat{\bf J}\rangle,$ (50)
where
$\hat{\bf J}=\frac{1}{2}\left[\hat{\Psi}^{\dagger}\left(\hat{\bf
v}\hat{\Psi}\right)+\left(\hat{\bf
v}\hat{\Psi}\right)^{\dagger}\hat{\Psi}\right],$ (51)
the particle field operator
$\hat{\Psi}=\sum_{\bf k}\phi_{\bf k}({\bf x})a_{\bf k}$ (52)
with $\phi_{\bf k}({\bf x})$ being the single-particle wave-function [given by
Eq. (30)], and the velocity operator
$\hat{\bf v}=-i\frac{\hbar}{m}\nabla.$ (53)
A straightforward calculation gives
${\bf j}=n\frac{\hbar}{m}\,{\bf q}-\frac{\hbar}{2m}\sum_{\bf k}\left(n_{-\bf
k}-n_{\bf k}\right){\bf k},$ (54)
where $n$ is the particle density and can be expressed as (see Appendix A)
$n=\sum_{\bf k}n_{\bf k}(T_{c}),$ (55) $n_{\pm\bf
k}=\left[e^{\left(\sqrt{(\epsilon^{(0)}_{\bf k}-\mu_{0})^{2}-|\Delta_{\bf
k}|^{2}}\pm\frac{\hbar^{2}}{m}{\bf k}\cdot{\bf
q}\right)/k_{B}T}-1\right]^{-1},$ (56)
and $n_{\bf k}(T_{c})$ is the value of $n_{\bf k}$ at $(T,{\bf q})=(T_{c},0)$.
The first term on the right-hand side of Eq. (54) represents a uniform flow of
all particles. The second term represents contribution from single-particle
excitations and de-paired particles, and tends to cancel the first term. When
all particles are in the superfluid ground state (at $T=0$ and for $q$ below a
threshold), the second term is zero. On the other hand, when $|\Delta_{\bf
k}|=0$ for all single-particle states, the two terms cancel each other so that
${\bf j}=0$.
Since superfluid current density ${\bf j}=0$ without pairing, as shown above,
according to standard quantum theory of many-particle systems, it is clear
that superfluidity arises from pairing of particles, not from Bose-Einstein
condensation (there is no microscopic theoretical justification for the view
that Bose-Einstein condensation leads to superfluidity). Therefore, we believe
that the superfluid properties of liquid 4He,londonII ; hess-fairbank as well
as the recently observed superfluid properties of ultra-cold atomic gases
(such as the persistent flow of atoms in a toroidal trapphillips and the
vortices in rotating atomic gasesketterle ; cornel ), are associated with
pairing of the atoms involved, not Bose-Einstein condensation.
By using $\epsilon^{(0)}_{\bf k}=\hbar^{2}k^{2}/2m$ and making the
substitution $\sum_{\bf k}\\!\\!\rightarrow(2\pi)^{-3}\\!\int d^{3}k$, we can
rewrite Eq. (54) as
${\bf j}=n_{s}\frac{\hbar}{m}\,{\bf q}=n_{s}{\bf v}_{s},$ (57)
where $n_{s}$ is the effective superfluid particle density, and, by using the
above-introduced dimensionless units (in which energies are measured in units
of $k_{B}T_{c}$, temperature $T$ in units of $T_{c}$ and superfluid wave-
vector $q$ in units of $q_{0}$), can be expressed as
$\frac{n_{s}}{n}=1-\frac{1}{2q\tilde{n}}\int^{\infty}_{0}d\epsilon^{(0)}_{\bf
k}\epsilon^{(0)}_{\bf k}\int^{1}_{0}dz_{\bf k}z_{\bf k}\left(n_{-\bf k}-n_{\bf
k}\right),$ (58)
where
$\tilde{n}=\int^{\infty}_{0}\frac{d\epsilon^{(0)}_{\bf k}(\epsilon^{(0)}_{\bf
k})^{1/2}}{e^{(\epsilon^{(0)}_{\bf k}-\mu_{0})}-1}$ (59)
is a function of $\mu_{0}$ and relates to $n$ (i.e., $n\propto\tilde{n}$), and
$n_{\pm\bf k}=\left[e^{\left(\sqrt{(\epsilon^{(0)}_{\bf
k}-\mu_{0})^{2}-|\Delta_{\bf k}|^{2}}\pm 2\sqrt{\epsilon^{(0)}_{\bf k}}qz_{\bf
k}\right)/T}\\!\\!\\!-1\right]^{-1}\\!\\!\\!\\!\\!\\!.$ (60)
#### III.2.1 $n_{s}$ for $q\rightarrow 0$
In the limit of $q\rightarrow 0$, Eq. (58) becomes
$\frac{n_{s}}{n}=1-\frac{2}{3\tilde{n}T}\int^{\infty}_{0}\frac{d\epsilon^{(0)}_{\bf
k}(\epsilon^{(0)}_{\bf k})^{3/2}e^{E^{(s)}_{\bf k}/T}}{\left(e^{E^{(s)}_{\bf
k}/T}-1\right)^{2}},$ (61)
where $E^{(s)}_{\bf k}$, as a function of $\epsilon^{(0)}_{\bf k}$ and $T$ for
given $\mu_{0}$, is determined by Eq. (48) for $q=0$.
Figure 10: Temperature dependence of superfluid particle density $n_{s}$ for
the case of $q\rightarrow 0$ for $-\mu_{0}/k_{B}T_{c}=0.3$, 0.5 and 0.8, as
indicated on the curves. (The dotted curve shows a linear $1-T/T_{c}$
dependence.)
Numerical results for $n_{s}/n$ for the case of $q\rightarrow 0$ as a function
of $T$ are shown in Fig. 10 for different values of $\mu_{0}$. Note that
$dn_{s}/dT\rightarrow 0$ in the limit of $T\rightarrow 0$, because of the
existence of an energy gap in the superfluid single-particle excitation
spectrum.
#### III.2.2 $n_{s}$ for finite $q$
In Appendix A, chemical potential $\mu$ in the superfluid state is determined
based on the assumption that all particles are paired in the superfluid state.
For the case of finite $q$, de-paired particles may exist. The result for
$\mu$ obtained in Appendix A is no longer valid when de-paired particles
exist. The question how to determine the chemical potential in a partly-paired
state is not addressed in this paper. To proceed, we make the following
approximation with respect to $n_{\pm\bf k}$ in calculating $n_{s}$ when de-
paired particles exist. For paired particles (for which $|\Delta_{\bf k}|>0$),
we use the same result for $n_{\pm\bf k}$ as in the case of an all-paired
state; and for de-paired particles (for which $|\Delta_{\bf k}|=0$), we use
the following approximation:
$\displaystyle n_{-\bf k}-n_{\bf k}$ $\displaystyle=$ $\displaystyle\langle
a^{\dagger}_{-\bf k}a_{-\bf k}\rangle-\langle a^{\dagger}_{\bf k}a_{\bf
k}\rangle$ (62) $\displaystyle\simeq$ $\displaystyle n^{(n)}_{-\bf
k}(T_{c})-n^{(n)}_{\bf k}(T_{c})\,,$ (63)
where
$n^{(n)}_{\pm\bf k}(T_{c})=\left[e^{\left(\epsilon_{\pm\bf
k}-\mu_{0}\right)}-1\right]^{-1}$ (64)
(in dimensionless units) with $\epsilon_{\pm\bf k}=\hbar^{2}(\pm{\bf k}+{\bf
q})^{2}/2m$.
Figure 11: (a) $n_{s}/n$ versus $T/T_{c}$ for different values of $q/q_{0}$ as
indicated on the curves; (b) $n_{s}/n$ versus $q/q_{0}$ for different values
of $T/T_{c}$; $-\mu_{0}/k_{B}T_{c}=0.5$ in both (a) and (b).
Numerical results for $n_{s}/n$ obtained based on the above-described
approximation are shown in Fig. 11(a) as $n_{s}/n$ versus $T/T_{c}$ for
difference values of $q/q_{0}$ and in Fig. 11(b) as $n_{s}/n$ versus $q/q_{0}$
for difference values of $T/T_{c}$. In both Fig. 11(a) and Fig. 11(b),
$-\mu_{0}/k_{B}T_{c}=0.5$ is assumed.
### III.3 Free energy density $F$
From diagonalized Hamiltonian $\hat{H}$ [Eq. (33)], we derive the following
expression for the free energy density in the superfluid state:
$\displaystyle F$ $\displaystyle\\!\\!\\!=$ $\displaystyle\\!\\!\\!\sum_{\bf
k}\\!\left[U_{\bf
k}\\!+\\!\frac{k_{B}T}{2}\ln\\!\left(1\\!-\\!e^{-\frac{E_{\bf
k}}{k_{B}T}}\\!\right)\\!\\!\left(1\\!-\\!e^{-\frac{E_{-\bf
k}}{k_{B}T}}\\!\right)\\!\right]$ (65)
$\displaystyle\mbox{}+\,\,\,n\frac{\hbar^{2}q^{2}}{2m},$
where the first term comes from
$\,\,-k_{B}T\ln\left[\text{Tr}\left(e^{-\hat{H}/k_{B}T}\right)\right]$, which
is the usual statistical free energy density,fetter ; huang ; feynmanBook and
the second term comes from $\,n(\mu\\!-\\!\mu_{0})$, which is the energy
increase due to the flow of the superfluid, and which is added to $F$ because
single-particle energies in the expression for Hamiltonian $\hat{H}$ are
measured relative to $\mu$.
For an isotropic system (such as liquid 4He), $F$ is a function of $T$ and
$q=|{\bf q}|$, i.e., $F=F(T,q)$. As can be shown, superfluid current density j
and effective superfluid particle density $n_{s}$ are related to $F$ via the
relations
${\bf j}=\frac{1}{\hbar}\frac{\partial F}{\partial{\bf q}}$ (66)
and
$n_{s}=\frac{m}{\hbar^{2}q}\frac{\partial F}{\partial q}\,,$ (67)
respectively [where $F$ is as given by Eq. (65)]. When $|\Delta_{\bf k}|=0$
for all single-particle states, as in the normal state, $F$ becomes
q-independent, and we have ${\bf j}=0$ and $n_{s}=0$.
### III.4 Spatially varying ${\bf v}_{s}({\bf x})$
The theory presented so far is based on the assumption that superfluid
velocity ${\bf v}_{s}$ is spatially constant. We can extend the theory to the
case where superfluid velocity ${\bf v}_{s}$ is spatially varying, by making
an assumption that there exists a length $l$ such that the following is true:
Length $l$ is large compared to inter-particle distance so that the properties
of particles in volume $l^{3}$ are essentially those of an infinite system,
but small by macroscopic standards so that the volume can be regarded as a
“point” macroscopically and all thermodynamic functions of the system vary
negligibly over the distance $l$.
Based on this assumption, quantities such as $|\Delta_{\bf k}|$, $n_{s}$, j
and $F$ can all be considered as local quantities, obtained with respect to
particles in a volume $l^{3}$ around a local point x for ${\bf q}={\bf q}({\bf
x})$, i.e., $|\Delta_{\bf k}|=|\Delta_{\bf k}(T,{\bf q}({\bf x}))|$,
$n_{s}=n_{s}(T,q({\bf x}))$, ${\bf j}={\bf j}(T,{\bf q}({\bf x}))$ and
$F=F(T,q({\bf x}))$, where ${\bf q}({\bf x})=m{\bf v}_{s}({\bf x})/\hbar$ is
assumed to vary spatially with a length much larger than $l$.
The theory presented so far for the case of ${\bf v}_{s}\neq 0$ is similar to
the theory presented in Ref. hao11, for superconductivity in the presence of
a magnetic field.
### III.5 Hess-Fairbank effect
Since $n_{s}>0$ in the superfluid state, we see from Eq. (67) that we have
$\frac{\partial F}{\partial q}>0,$ (68)
which shows that a larger value of $q$ is energetically less favorable in the
superfluid state. This implies that a superfluid tends to expel superfluid
wave-vector ${\bf q}({\bf x})$, or, equivalently, superfluid velocity ${\bf
v}_{s}({\bf x})=\hbar{\bf q}({\bf x})/m$, from its interior so as to minimize
the overall free energy of the system. This result qualitatively explains the
Hess-Fairbank effect,hess-fairbank the reduction of moment of inertia of a
rotating cylinder of liquid 4He when it is cooled through the superfluid
transition. Namely, when the liquid is in the superfluid state, because a
motionless state is energetically more favorable, it stops rotating with the
container, except in the immediate vicinity of the wall of the container where
the liquid rotates with the container due to interaction between the liquid
and the wall at the interface.
Although the theory presented so far provides a qualitative explanation for
the Hess-Fairbank effect, it does not allow quantitative description of the
Hess-Fairbank effect. Namely, the theory says that a superfluid tends to expel
superfluid velocity ${\bf v}_{s}({\bf x})$, but it does not tell us how ${\bf
v}_{s}({\bf x})$ can be determined for given boundary condition and
temperature (in the case of a rotating cylinder of liquid 4He, for example,
the boundary condition is determined by the angular speed and geometry of the
container).
We note that the Hess-Fairbank effect is analogous to the Meissner effect in
superconductors, and the latter is quantitatively describable by combination
of the London equationlondonI ${\bf j}=-n_{s}{\bf a}$ and the Ampere’s law
${\bf j}=\nabla\times{\bf b}$ (here j, $n_{s}$, a and ${\bf
b}=\nabla\times{\bf a}$ are, respectively, the electrical current density,
superconducting electron density, vector potential and magnetic flux density,
and a set of dimensionless units is used for the present discussion). Our Eq.
(57) is analogous to the London equation, with ${\bf v}_{s}$ playing the role
of a. What is missing for a superfluid is an equation analogous to the
Ampere’s law.
We therefore speculate the existence of the following equation:
${\bf j}=-\Lambda\nabla\times{\bf\text{\boldmath$\omega$}},$ (69)
where
$\text{\boldmath$\omega$}=\nabla\times{\bf v}_{s}$ (70)
is superfluid vorticity, and $\Lambda$ a positive constant.
Note that Eq. (69) applies only to particles (or atoms) in the superfluid
state; i.e., here j and $\omega$ are associated with the superfluid component
of a fluid, and correspond, respectively, to ${\bf j}_{s}$ and $\omega$s in a
two-fluid modellondonII ; london1938 ; tisza ; landau .
Equation (69) is clearly only a speculation based on the similarity between
superfluidity and superconductivity, and thus, must derive its validity from
experimental confirmation of the consequences that it implies.
Contrary to the common view that superfluid is irrotational (vorticity-free)
(for which there is no microscopic theoretical justification), Eq. (69) shows
that, analogous to that electrical current creates magnetic field, superfluid
current creates vorticity.
As we will see below, an important consequence of Eq. (69) is the existence of
a penetration depth that characterizes the typical distance to which
superfluid velocity and superfluid vorticity penetrate into a superfluid. This
penetration depth is analogous to the London penetration depthlondonI that
characterizes the typical distance to which magnetic vector potential and
magnetic field penetrate into a superconductor.
We further speculate the existence of an additional term in the expression for
free energy density $F$, i.e.,
$\displaystyle F$ $\displaystyle\\!\\!\\!=$ $\displaystyle\\!\\!\\!\sum_{\bf
k}\\!\left[U_{\bf
k}\\!+\\!\frac{k_{B}T}{2}\ln\\!\left(1\\!-\\!e^{-\frac{E_{\bf
k}}{k_{B}T}}\\!\right)\\!\\!\left(1\\!-\\!e^{-\frac{E_{-\bf
k}}{k_{B}T}}\\!\right)\\!\right]$ (71)
$\displaystyle\mbox{}+\,\,\,n\frac{\hbar^{2}q^{2}}{2m}+\frac{\Lambda
m}{2}\omega^{2},$
where the first two terms are the same as in Eq. (65), and the third term,
which is analogous to the magnetic filed energy density in the case of
superconductors, is the additional term whose existence is speculated. The
reason for this speculation is as follows. We note that Eq. (57) is an
equilibrium property of a superfluid, and thus, must also be derivable as a
result of the variational problem that, in the thermodynamic equilibrium, the
overall free energy of the superfluid, given by the volume integral of free
energy density $F$, is stationary with respect to arbitrary variation of ${\bf
v}_{s}({\bf x})$. This is true when $F$ is as given by Eq. (71), as can be
shown with the help of Eq. (69).
Combination of Eqs. (57) and (69) allows quantitative description of the Hess-
Fairbank effect. In the following we present two simple examples.
#### III.5.1 Superfluid in a rotating cylinder
We consider a superfluid in a rotating cylinder. Assuming the length of the
cylinder is much larger than its radius $R$, and neglecting the bottom portion
of the cylinder, in terms of cylindrical coordinates $(r,\phi,z)$ and unit
vectors ($\hat{\bf r}$, $\hat{\phi}$, $\hat{\bf z}$), we can write superfluid
current density ${\bf j}=j(r)$$\hat{\phi}$, superfluid velocity ${\bf
v}_{s}=v_{s}(r)$$\hat{\phi}$ and superfluid vorticity
$\omega$$=\omega(r)\hat{\bf z}$, and we have
$\Lambda\omega^{\prime}(r)=n_{s}(r)v_{s}(r)$ (72)
and
$v_{s}^{\prime}(r)=\omega(r)-\frac{v_{s}(r)}{r},$ (73)
where a “prime” indicates a derivative with respect to $r$,
$n_{s}(r)=n_{s}(v_{s}(r))$ is given by Eq. (58), the first equation comes from
combining Eqs. (57) and (69), and the second equation comes from Eq. (70).
These equations describe only the behavior of the superfluid component of the
fluid. We will not consider in this paper the behavior of the normal-fluid
component of the fluid.
This is a second-order boundary value problem (which is expressed here as a
system of two first-order differential equations) with the boundary conditions
$v_{s}(0)=0$ (74)
and
$v_{s}(R)=R\Omega_{0}\,,$ (75)
where $R$ is the inner radius of the cylindrical container and $\Omega_{0}$
the angular speed of the container. Here we have assumed that, in equilibrium,
$v_{s}(R)$ is the same as the linear speed of the inner wall of the container,
as otherwise there would be momentum transfer between the superfluid and the
container.
In this paper, we will not attempt to solve this boundary value problem for
arbitrary $T$ and $v_{s}(R)$. Instead, for simplicity in presenting the main
features of the theory, we will consider only the case where $n_{s}$ is
spatially constant. As we can see from Fig. 11(b), that $n_{s}$ is spatially
constant is true only at $T=0$ for $v_{s}<\hbar q_{c{\bf k},\text{min}}(0)/m$
[i.e., $n_{s}$ is independent of $v_{s}$ at $T=0$ for $v_{s}<\hbar q_{c{\bf
k},\text{min}}(0)/m$], and is approximately true at higher temperatures for
sufficiently low values of $v_{s}$.
For a spatially constant $n_{s}$, the above-described boundary value problem
can be solved analytically, and the solutions are:
$v_{s}(r)=R\Omega_{0}\frac{I_{1}(r/\lambda)}{I_{1}(R/\lambda)}$ (76)
and
$\omega(r)=\frac{R\Omega_{0}}{\lambda}\frac{I_{0}(r/\lambda)}{I_{1}(R/\lambda)},$
(77)
where $I_{n}(x)$ is the modified Bessel function of the first kind of order
$n$,spiegel and
$\lambda=\sqrt{\Lambda/n_{s}}\,\,\,.$ (78)
For $R/\lambda\gg 1$, by using the asymptotic expansionspiegel $I_{n}(x)\sim
e^{x}/\sqrt{2\pi x}$, where $x\gg 1$, we have, near the inner wall of the
container,
$v_{s}(r)\simeq R\Omega_{0}\,e^{-(R-r)/\lambda}$ (79)
and
$\omega(r)\simeq\frac{R\Omega_{0}}{\lambda}\,e^{-(R-r)/\lambda},$ (80)
from which we see that superfluid velocity and superfluid vorticity
“penetrate” only a distance of the order of $\lambda$ into the superfluid; at
a depth of little more than $\lambda$, superfluid velocity and superfluid
vorticity are practically zero; and thus, $\lambda$ has the meaning of
“penetration depth” that characterizes the distance to which superfluid
velocity and superfluid vorticity penetrate into a superfluid, analogous to
the London penetration depthlondonI that characterizes the distance to which
magnetic field penetrates into a superconductor.
For $R/\lambda\ll 1$, by using the approximationspiegel $I_{0}(x)\sim
1+O(x^{2})$ and $I_{1}(x)\sim x/2+O(x^{3})$, where $x\ll 1$, we have
$v_{s}(r)\simeq\Omega_{0}r$ (81)
and
$\omega(r)\simeq 2\Omega_{0},$ (82)
which are the same as the results for a rotating rigid body.
#### III.5.2 Flow of superfluid in a pipe
We next consider the case where a superfluid flows through an infinitely long
pipe of a constant circular cross section. Let the axis of the pipe be the
$z$-axis, the inner radius of the pipe be $R$, and the total superfluid
current be $I$. In terms of cylindrical coordinates $(r,\phi,z)$ and unit
vectors ($\hat{\bf r}$, $\hat{\phi}$, $\hat{\bf z}$), we can write superfluid
current density ${\bf j}=j(r)\hat{\bf z}$, superfluid velocity ${\bf
v}_{s}=v_{s}(r)\hat{\bf z}$ and superfluid vorticity
$\omega$$=\omega(r)$$\hat{\phi}$, and we have
$\omega^{\prime}(r)=-\frac{\omega(r)}{r}-\frac{n_{s}(r)}{\Lambda}v_{s}(r)$
(83)
and
$v_{s}^{\prime}(r)=-\omega(r),$ (84)
where a “prime” indicates a derivative with respect to $r$,
$n_{s}(r)=n_{s}(v_{s}(r))$ is given by Eq. (58), the first equation comes from
combining Eqs. (57) and (69), and the second equation comes from Eq. (70).
This is a second-order boundary value problem (which is expressed here as a
system of two first-order differential equations) with the boundary conditions
$\omega(0)=0$ (85)
and
$\omega(R)=-\frac{I}{2\pi R\Lambda},$ (86)
where the last condition comes from $I=\int_{S}{\bf j}\cdot\\!d{\bf s}$.
Similar to that for the case of a superfluid in a rotating cylinder, discussed
above, we consider only the case where $n_{s}$ is spatially constant. As
mentioned above, that $n_{s}$ is spatially constant is true only at $T=0$ for
$v_{s}<\hbar q_{c{\bf k},\text{min}}(0)/m$, and is approximately true at
higher temperatures for sufficiently low values of $v_{s}$. In this case, the
above-described boundary value problem can be solved analytically, and the
solutions are:
$v_{s}(r)=\frac{I\lambda}{2\pi
R\Lambda}\frac{I_{0}(r/\lambda)}{I_{1}(R/\lambda)}$ (87)
and
$\omega(r)=-\frac{I}{2\pi R\Lambda}\frac{I_{1}(r/\lambda)}{I_{1}(R/\lambda)},$
(88)
where $I_{n}(x)$ is the modified Bessel function of the first kind of order
$n$,spiegel and $\lambda$ is as defined by Eq. (78).
For $R/\lambda\gg 1$, by using the asymptotic expansionspiegel $I_{n}(x)\sim
e^{x}/\sqrt{2\pi x}$, where $x\gg 1$, we have, near the inner wall of the
pipe,
$v_{s}(r)\simeq\frac{I\lambda}{2\pi R\Lambda}\,e^{-(R-r)/\lambda}$ (89)
and
$\omega(r)\simeq-\frac{I}{2\pi R\Lambda}\,e^{-(R-r)/\lambda}\,,$ (90)
from which we see that superfluid flows mainly in the region near the wall of
the pipe; at a distance of little more than $\lambda$ away from the wall, both
superfluid velocity and superfluid vorticity are practically zero; and
$\lambda$ has the meaning of “penetration depth” that characterizes the
distance to which superfluid velocity and superfluid vorticity penetrate into
a superfluid.
For $R/\lambda\ll 1$, by using the approximationspiegel $I_{0}(x)\sim
1+O(x^{2})$ and $I_{1}(x)\sim x/2+O(x^{3})$, where $x\ll 1$, we have
$v_{s}(r)\simeq\frac{I}{\pi R^{2}n_{s}}$ (91)
and
$\omega(r)\simeq-\frac{I}{2\pi R^{2}\Lambda}r\,,$ (92)
which show that, in this case, the superfluid flow is nearly uniform; and
superfluid vorticity is nearly linear in $r$.
## IV summary
We have presented a microscopic theory for superfluidity in an interacting
many-particle Bose system (such as liquid 4He). The theory shows that, similar
to superconductivity in superconductors, superfluidity in a Bose system arises
from pairing of particles of opposite momenta.
In Sec. II, we presented the theory for the case where superfluid velocity
${\bf v}_{s}=0$. The theory shows the existence of an energy gap in single-
particle excitation spectrum, and the existence of a specific heat jump at the
transition.
In Sec. III, we presented the theory for the case where superfluid velocity
${\bf v}_{s}\neq 0$. We derived an equation that gives a relation between
superfluid current density j and superfluid velocity ${\bf v}_{s}$ (this
equation is analogous to the London equation for the superconducting state
that gives a relation between current density of superconducting electrons and
magnetic vector potential), and an expression for superfluid particle density
$n_{s}$ as a function of temperature $T$ and superfluid velocity ${\bf
v}_{s}$. We showed that superfluid-state free energy density $F$ is an
increasing function of $v_{s}$ (i.e., $\partial F/\partial v_{s}>0$), which
indicates that a superfluid tends to expel superfluid velocity (i.e., a
superfluid has a tendency to remain motionless); this result provides a
qualitative explanation for the Hess-Fairbank effect (which is analogous to
the Meissner effect in superconductors). We further speculated, based on the
similarity between superconductivity and superfluidity, the existence of an
equation [i.e., Eq. (69)] that specifies a relation between superfluid current
density j and superfluid vorticity $\omega$ (this equation is analogous to the
Ampere’s law). With the help of this equation, the Hess-Fairbank effect can be
quantitatively described.
## Appendix A Chemical potential $\mu$ and number-of-particle distribution
$\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle$
### A.1 $\mu$ for ${\bf v}_{s}=0$
We consider in this Appendix chemical potential $\mu$ in the superfluid state.
We first consider the case where superfluid velocity ${\bf v}_{s}=0$ in this
subsection.
Chemical potential $\mu$ as a function of temperature $T$ is so determined
such that the number of particles of a Bose system is conserved.
Number of particles $N$ is given by
$N=\sum_{\bf k}\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle,$ (93)
which, in the normal state, becomes
$N=\sum_{\bf k}\left[e^{(\epsilon_{\bf k}-\mu)/k_{B}T}-1\right]^{-1}.$ (94)
This equation determines chemical potential $\mu$ as a function of $T$ for
given $N$ in the normal state. Here, we have assumed volume $V=1$ (in
arbitrary unit) so that $N$ can also be considered as the particle density of
the system.
In the superfluid state, Eq. (93) can be expressed as
$N=\sum_{\bf k}\frac{1}{2}\left[\left(\epsilon_{\bf
k}-\mu\right)\left(\frac{1+2n_{\bf k}}{E_{\bf k}}\right)-1\right]$ (95)
by using the results of the canonical transformation of Eq. (4) and
$\langle\alpha^{\dagger}_{\bf k}\alpha_{\bf k}\rangle=n_{\bf k}$.
Since the quantity $(1+2n_{\bf k})/E_{\bf k}$ in the above expression is
$T$-independent, according to Eq. (21), we see that particle conservation
condition $\partial N/\partial T=0$ implies that $\partial\mu/\partial T=0$ in
the superfluid state, which can also be expressed as
$\mu=\mu_{0}\text{ for $T\leq T_{c}$},$ (96)
where $\mu_{0}$ is the value of the chemical potential at $T=T_{c}$. Namely,
chemical potential $\mu$ is $T$-independent in the superfluid state.
Equation (95) can then be expressed as
$N=\sum_{\bf k}n_{\bf k}(T_{c}),$ (97)
where
$n_{\bf k}(T_{c})=\left[e^{(\epsilon_{\bf
k}-\mu_{0})/k_{B}T_{c}}-1\right]^{-1}.$ (98)
Figure 12: Temperature dependence of chemical potential $\mu$, calculated for
$-\mu_{0}/k_{B}T_{c}=0.5$, assuming $\epsilon_{\bf k}=\hbar^{2}k^{2}/2m$. For
$T\leq T_{c}$, $\mu=\mu_{0}$ is a constant. The dotted curve shows normal
state chemical potential $\mu^{(n)}$ for $T<T_{c}$; $\mu^{(n)}=0$ for $T\leq
T_{c}^{(\text{BEC})}$, where $T_{c}^{(\text{BEC})}$ is the critical
temperature of Bose-Einstein condensation; and
$T_{c}^{(\text{BEC})}/T_{c}\simeq 0.49$ for $-\mu_{0}/k_{B}T_{c}=0.5$.
Figure 12 shows the temperature dependence of chemical potential $\mu$,
calculated for $-\mu_{0}/k_{B}T_{c}=0.5$. In calculating $\mu$, we have used
the Runge-Kutta method.conte80 Namely, from Eq. (94) and the condition
$\partial N/\partial T=0$, we drive an expression for $\partial\mu/\partial
T$, which, together with a given value of $\mu=\mu_{0}$ at $T/T_{c}=1$, allows
us to numerically compute $\mu$ (and $\partial\mu/\partial T$, which is
required in calculating the normal-state specific heat in Sec. II.3 ) for
arbitrary $T$ by using the Runge-Kutta method.conte80 We have also assumed
$\epsilon_{\bf k}=\hbar^{2}k^{2}/2m$ and made the substitution $\sum_{\bf
k}\propto\int_{0}^{\infty}d\epsilon\,\epsilon^{1/2}$ in calculating $\mu(T)$.
The integrals involved in the expression for $\partial\mu/\partial T$ are
calculated by using the Simpson method.conte80
As shown in Fig. 12, for $T>T_{c}$, $\mu$ is a decreasing function of $T$. For
$T\leq T_{c}$, $\mu=\mu_{0}$ is a constant. The dotted curve in Fig. 12 shows
normal state chemical potential $\mu^{(n)}$ for $T<T_{c}$; $\mu^{(n)}=0$ for
$T\leq T_{c}^{(\text{BEC})}$, where $T_{c}^{(\text{BEC})}$ is the critical
temperature of Bose-Einstein condensation;fetter ; huang ; feynmanBook and
$T_{c}^{(\text{BEC})}/T_{c}\simeq 0.49$ for $-\mu_{0}/k_{B}T_{c}=0.5$. As
shown in Appendix B, $\mu_{0}$ must be below a certain negative number in a
superfluid state, which means $T_{c}>T_{c}^{\text{BEC}}$, as one can see from
Fig. 12.
Superfluid-state chemical potential $\mu_{0}$ is an important parameter in the
present theory with respect to the properties of the superfluid state. For
example, zero-temperature minimum single-particle excitation energy (or energy
gap) $E_{min}$ directly relates to $\mu_{0}$ via the relation
$E_{min}=(-\mu_{0})\tanh[(-\mu_{0})/2k_{B}T_{c}]$ according to Eq. (22).
Generally, $\mu_{0}$ relates to the particle density of the Bose system, and,
as one can see from Eq. (97), a lower $\mu_{0}$ (larger $|\mu_{0}|$)
corresponds to a lower particle density.
### A.2 $\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle$ for ${\bf v}_{s}=0$
The number-of-particle distribution, $\langle a^{\dagger}_{\bf k}a_{\bf
k}\rangle$, is temperature $T$-dependent in the normal state. This is no
longer the case in the superfluid state. The fact that chemical potential
$\mu$ is $T$-independent in the superfluid state implies that $\langle
a^{\dagger}_{\bf k}a_{\bf k}\rangle$ is also $T$-independent in the superfluid
state. Namely, from Eqs. (21) and (96) we can derive
$\partial\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle/\partial T=0\text{ for
$T\leq T_{c}$},$ (99)
which can also be expressed as
$\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle=n_{\bf k}(T_{c})\text{ for $T\leq
T_{c}$}.$ (100)
In other words, the number-of-particle distribution, $\langle a^{\dagger}_{\bf
k}a_{\bf k}\rangle$, becomes frozen at $T=T_{c}$ when the Bose system is
cooled through the superfluid transition.
Note the difference between the particles described by operators
$a^{\dagger}_{\bf k}$ and $a_{\bf k}$ and the particles described operators
$\alpha^{\dagger}_{\bf k}$ and $\alpha_{\bf k}$: $\langle a^{\dagger}_{\bf
k}a_{\bf k}\rangle$ is $T$-independent (in the superfluid state), whereas
$\langle\alpha^{\dagger}_{\bf k}\alpha_{\bf k}\rangle$ is $T$-dependent (for
example, $\langle\alpha^{\dagger}_{\bf k}\alpha_{\bf k}\rangle\rightarrow 0$
as $T\rightarrow 0$ and $\langle\alpha^{\dagger}_{\bf k}\alpha_{\bf
k}\rangle\rightarrow n_{\bf k}(T_{c})$ as $T\rightarrow T_{c}$).
At $T=T_{c}$, we have
$\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle=\langle\alpha^{\dagger}_{\bf
k}\alpha_{\bf k}\rangle_{T=T_{c}}=n_{\bf k}(T_{c}),$ (101)
which shows that all particles in a single-particle state of wave-vector k
exist as single-particle excitations.
For $0<T<T_{c}$, we have
$\displaystyle\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle$ $\displaystyle=$
$\displaystyle|u_{\bf k}|^{2}\langle\alpha^{\dagger}_{\bf k}\alpha_{\bf
k}\rangle+|v_{\bf k}|^{2}\left(1+\langle\alpha^{\dagger}_{-\bf k}\alpha_{-\bf
k}\rangle\right)$ (102) $\displaystyle=$ $\displaystyle n_{\bf k}(T_{c}),$
which shows that particles in a single-particle state of wave-vector k partly
exist as single-particle excitations and partly are in the superfluid
condensate, but the sum of the particles remains the same as $n_{\bf
k}(T_{c})$.
At $T=0$, we have
$\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle=|v_{\bf k}(0)|^{2}=n_{\bf
k}(T_{c}),$ (103)
which shows that all particles in a single-particle state of wave-vector k are
in the superfluid condensate.
### A.3 $\mu$ for ${\bf v}_{s}\neq 0$
We consider in this subsection chemical potential $\mu$ in the superfluid
state for the case where superfluid velocity ${\bf v}_{s}\neq 0$.
In this case, we have
$\displaystyle\\!\\!\\!N\\!\\!$ $\displaystyle=$ $\displaystyle\\!\\!\sum_{\bf
k}\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle$ (104) $\displaystyle=$
$\displaystyle\\!\\!\sum_{\bf k}\frac{1}{2}\\!\left[\frac{\xi_{\bf
k}+\xi_{-\bf k}}{2}\\!\left(\\!\frac{1+n_{\bf k}+n_{-\bf k}}{E^{(s)}_{\bf
k}}\\!\right)\\!-1\right]\\!,$
where $\xi_{\bf k}$ and $E^{(s)}_{\bf k}$ are given by Eqs. (35) and (37),
respectively.
Let $X$ denote any one of $T$, $q_{1}$, $q_{2}$ and $q_{3}$, where $q_{i}$
($i=1,2,3$) are components of ${\bf q}=m{\bf v}_{s}/\hbar$. Since in the above
expression for $N$ the quantity $(1+n_{\bf k}+n_{-\bf k})/E^{(s)}_{\bf k}$ is
$X$-independent according to Eq. (47), and $(\xi_{\bf k}+\xi_{-\bf
k})/2=\epsilon^{(0)}_{\bf k}+\hbar^{2}q^{2}/2m-\mu$ by using Eq. (31), the
particle conservation condition $\partial N/\partial X=0$ leads to
$\partial(\hbar^{2}q^{2}/2m-\mu)/\partial X=0,$ (105)
which is readily solved to give
$\mu=\mu_{0}+\hbar^{2}q^{2}/2m,$ (106)
where $\mu_{0}$ is the value of $\mu$ at $(T,{\bf q})=(T_{c},0)$, and which is
the same as Eq. (43).
Equation (104) can then be expressed as
$N=\sum_{\bf k}n^{(0)}_{\bf k}(T_{c}),$ (107)
where
$n^{(0)}_{\bf k}(T_{c})=\left[e^{(\epsilon^{(0)}_{\bf
k}-\mu_{0})/k_{B}T_{c}}-1\right]^{-1},$ (108)
and the superscript ${(0)}$ indicates ${\bf q}=0$. [Equation (107) is the same
as Eq. (97), as it should.]
### A.4 $\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle$ for ${\bf v}_{s}\neq
0$
When superfluid velocity ${\bf v}_{s}\neq 0$, as can be shown, number-of-
particle distribution $\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle$ in the
superfluid state becomes
$\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle=n^{(0)}_{\bf
k}(T_{c})+\frac{1}{2}\left(n_{\bf k}-n_{-\bf k}\right)$ (109)
for $T<T_{c\bf k}(qz_{\bf k})$ (i.e., for $|\Delta_{\bf k}|>0$), where
$n^{(0)}_{\bf k}(T_{c})$ is given by Eq. (108), $n_{\bf k}$ is given by Eq.
(38), and $qz_{\bf k}$ is the component of ${\bf q}=m{\bf v}_{s}/\hbar$ along
wave-vector k (as discussed in Sec. III.1).
We have
$\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle+\langle a^{\dagger}_{-\bf
k}a_{-\bf k}\rangle=2\,n^{(0)}_{\bf k}(T_{c})$ (110)
and
$\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle-\langle a^{\dagger}_{-\bf
k}a_{-\bf k}\rangle=\,n_{\bf k}-n_{-\bf k}.$ (111)
Namely, $\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle+\langle a^{\dagger}_{-\bf
k}a_{-\bf k}\rangle$ is $T$-independent, but $\langle a^{\dagger}_{\bf
k}a_{\bf k}\rangle-\langle a^{\dagger}_{-\bf k}a_{-\bf k}\rangle$ is
$T$-dependent.
We have $\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle=\langle a^{\dagger}_{-\bf
k}a_{-\bf k}\rangle$ at $T=0$, where all particles are paired [for
$q<q_{c,\text{min}}(0)$], and where $n_{\bf k}=0$ , i.e.,
$\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle=\langle a^{\dagger}_{-\bf
k}a_{-\bf k}\rangle=|v_{\bf k}(0)|^{2}=n^{(0)}_{\bf k}(T_{c})$ (112)
for $T=0$.
We have $\langle a^{\dagger}_{\bf k}a_{\bf k}\rangle\neq\langle
a^{\dagger}_{-\bf k}a_{-\bf k}\rangle$ for $T>0$, because $n_{\bf k}\neq
n_{-\bf k}$. The difference between $\langle a^{\dagger}_{\bf k}a_{\bf
k}\rangle$ and $\langle a^{\dagger}_{-\bf k}a_{-\bf k}\rangle$ contributes to
the reduction in the effective superfluid particle density $n_{s}$ as one can
see from the expression for $n_{s}$, Eq. (58).
## Appendix B Ground state energy $U(0)$
### B.1 $U(0)$ for ${\bf v}_{s}=0$
The ground state energy of the superfluid state in the case where superfluid
velocity ${\bf v}_{s}=0$ is
$\displaystyle U(0)$ $\displaystyle=$
$\displaystyle\langle\hat{H}\rangle_{T=0}\,=\,\sum_{\bf k}U_{\bf k}(0)$ (113)
$\displaystyle=$ $\displaystyle\sum_{\bf k}\frac{\xi_{\bf k}}{e^{2\xi_{\bf
k}/k_{B}T_{c}}-1},$
where $\xi_{\bf k}=\epsilon_{\bf k}-\mu_{0}$ and we have used $E_{\bf
k}(0)=\xi_{\bf k}\tanh(\xi_{\bf k}/2k_{B}T_{c})$, according to Eq. (22).
It is interesting to note that the right-hand-side of Eq. (113) can be
interpreted as the thermal energy of a Bose system of paired particles at
$T=T_{c}$ having a pair excitation spectrum of $2\xi_{\bf k}$.
Since the single-particle energy in the expression for Hamiltonian ${\hat{H}}$
is measured relative to chemical potential $\mu=\mu_{0}$, we must add the term
$\mu_{0}N$ to the above expression for $U(0)$ when we compare $U(0)$ with
$U^{(\text{BEC})}=0$, the energy of a Bose-Einstein condensate, which is the
ground state of the normal state.fetter ; huang ; feynmanBook Namely, in
order to have a superfluid state, we must have
$\displaystyle U(0)+\mu_{0}N$ $\displaystyle=$ $\displaystyle\sum_{\bf
k}\left[\frac{\xi_{\bf k}}{e^{2\xi_{\bf
k}/k_{B}T_{c}}-1}+\frac{\mu_{0}}{e^{\xi_{\bf k}/k_{B}T_{c}}-1}\right]$ (114)
$\displaystyle<$ $\displaystyle 0\,,$
which shows that $\mu_{0}$ must be below a certain negative value
$\mu^{\star}$ in a superfluid state. Assuming $\epsilon_{\bf
k}=\hbar^{2}k^{2}/2m$ and making the substitution $\sum_{\bf
k}\rightarrow(1/2\pi)^{3}\int d^{3}k$, we find [from the condition that
$U(0)+\mu_{0}N=0$ at $\mu_{0}=\mu^{\star}$] that
$\mu^{\star}/k_{B}T_{c}\simeq-0.21$.
### B.2 $U(0)$ for ${\bf v}_{s}\neq 0$
When superfluid velocity ${\bf v}_{s}\neq 0$, the ground state energy is
$U(0)=U^{(0)}(0)+(\hbar^{2}q^{2}/2m)N,$ (115)
where $U^{(0)}(0)$ is the ground state energy for ${\bf v}_{s}=0$, as given by
Eq. (113), and the second term comes from $(\mu-\mu_{0})N$, which is the
energy increase due to the flow of the superfluid (here $\mu$ is the chemical
potential in the superfluid state when ${\bf v}_{s}\neq 0$, $\mu_{0}$ is the
chemical potential in the superfluid state when ${\bf v}_{s}=0$, and
$\hbar^{2}q^{2}/2m=mv_{s}^{2}/2$ is the per-particle energy increase due to
the flow of the superfluid).
## Appendix C Derivation of Eq. (47)
We present in this Appendix the details of the derivation of Eq. (47).
For convenience, we define
$C_{\bf k}=\frac{1+n_{\bf k}+n_{-\bf k}}{2E_{\bf k}^{(s)}}$ (116)
Then, the self-consistency equation, Eq. (42), can be rewritten as
$\Delta_{\bf k}=-\sum_{\bf k^{\prime}}V_{\bf kk^{\prime}}C_{\bf
k^{\prime}}\Delta_{\bf k^{\prime}}.$ (117)
When superfluid velocity ${\bf v}_{s}\neq 0$, we expect $\Delta_{\bf k}$ to be
a function of temperature $T$ and superfluid wave-vector ${\bf q}=m{\bf
v}_{s}/\hbar$. Let $X$ denote any one of $T$, $q_{1}$, $q_{2}$ and $q_{3}$,
where $q_{i}$ ($i=1,2,3$) are components of q. We operate $\partial/\partial
X$ on both sides of Eq. (117) to obtain
$\frac{\partial\Delta_{\bf k}}{\partial X}=-\sum_{\bf k^{\prime}}V_{{\bf
k},{\bf k^{\prime}}}\left(\frac{\partial C_{\bf k^{\prime}}}{\partial
X}\Delta_{\bf k^{\prime}}+C_{\bf k^{\prime}}\frac{\partial\Delta_{\bf
k^{\prime}}}{\partial X}\right)$ (118)
(note that $V_{{\bf k},{\bf k^{\prime}}}$ is assumed to be $X$-independent).
We next multiply both sides of the above equation by $C_{\bf k}\Delta_{\bf
k}^{\star}$, and then take summation over k, i.e.,
$\displaystyle\sum_{\bf k}C_{\bf k}\Delta_{\bf
k}^{\star}\frac{\partial\Delta_{\bf k}}{\partial X}=\sum_{\bf
k^{\prime}}\left(-\sum_{\bf k}V_{{\bf k},{\bf k^{\prime}}}C_{\bf k}\Delta_{\bf
k}^{\star}\right)$ $\displaystyle\times\left(\frac{\partial C_{\bf
k^{\prime}}}{\partial X}\Delta_{\bf k^{\prime}}+C_{\bf
k^{\prime}}\frac{\partial\Delta_{\bf k^{\prime}}}{\partial X}\right).$ (119)
The quantity inside the first pair of parentheses on the right-hand side of
the above equation equals to $\Delta_{\bf k^{\prime}}^{\star}$, according to
Eq. (117). Therefore, the second of the two terms on the right-hand side is
the same as the term on the left-hand side. Thus, we have
$\sum_{\bf k}|\Delta_{\bf k}|^{2}\frac{\partial C_{\bf k}}{\partial X}=0\;.$
(120)
We want a $|\Delta_{\bf k}|>0$ solution. Clearly,
$C_{\bf k}=\mbox{$X$-independent},$ (121)
which is Eq. (47) and satisfies
$\frac{\partial C_{\bf k}}{\partial X}=0,$ (122)
is a solution of Eq. (120).
However, Eq. (121) is not the only possible solution of Eq. (120) [as one can
see, Eq. (120) actually can have an infinite number of solutions]. We
therefore need to justify that Eq. (121) is the only physical solution, which
is done as follows.
Diagonalized Hamiltonian $\hat{H}$ of Eq. (33) describes a system of
independent quasi-particle excitations. In thermodynamic equilibrium, there is
no transition between different quasi-particle states, except pairing
correlation. Therefore, we can calculate, for each pair of $({\bf k},{\bf-k})$
excitations, the partition function $Z_{\bf k}=\text{Tr}(e^{-{\hat{H}}_{\bf
k}/k_{B}T})$, where ${\hat{H}}_{\bf k}=2U_{\bf k}+E_{\bf
k}\alpha^{\dagger}_{\bf k}\alpha_{\bf k}+E_{-\bf k}\alpha^{\dagger}_{-\bf
k}\alpha_{-\bf k}$, free energy $F_{\bf k}=-k_{B}T\ln Z_{\bf k}$, entropy
$S_{\bf k}=-\partial F_{\bf k}/\partial T$ and thermal energy
$\varepsilon_{\bf k}=F_{\bf k}+TS_{\bf k}$.
We expect entropy $S_{\bf k}$ and thermal energy $\varepsilon_{\bf k}$ to be
$\displaystyle\\!\\!\\!S_{\bf k}$ $\displaystyle\\!\\!=\\!\\!$ $\displaystyle-
k_{B}\left[n_{\bf k}\ln n_{\bf k}-(1+n_{\bf k})\ln(1+n_{\bf k})\right.$ (123)
$\displaystyle\\!\\!\\!+\left.n_{-\bf k}\ln n_{-\bf k}-(1+n_{-\bf
k})\ln(1+n_{-\bf k})\right]$
and
$\varepsilon_{\bf k}=2U_{\bf k}+n_{\bf k}E_{\bf k}+n_{-\bf k}E_{-\bf k},$
(124)
respectively.
However, because $U_{\bf k}$ and $E^{(s)}_{\bf k}$ are $T$-dependent, there
are additional terms involving $\partial U_{\bf k}/\partial T$ and $\partial
E^{(s)}_{\bf k}/\partial T$ in the expressions for $S_{\bf k}$ and
$\varepsilon_{\bf k}$ obtained from $F_{\bf k}$ , as compared to Eqs. (123)
and (124). By letting the sum of the additional terms in each of the
expressions for $S_{\bf k}$ and $\varepsilon_{\bf k}$ to be zero, we arrive at
$2\frac{\partial U_{\bf k}}{\partial T}+\,\left(n_{\bf k}+\,n_{-\bf
k}\right)\frac{\partial E^{(s)}_{\bf k}}{\partial T}=0.$ (125)
Similarly, we expect the contribution from each pair of $({\bf k},{\bf-k})$
excitations to the particle current density of the superfluid to be
${\bf j}_{\bf k}=\left(n_{\bf k}-n_{-\bf k}\right)\frac{\hbar}{m}{\bf k}\,$
(126)
(where $\hbar{\bf k}/m$ is the velocity of an excitation in the state of wave-
vector k).
We further expect ${\bf j}_{\bf k}=\hbar^{-1}\partial F_{\bf k}/\partial{\bf
q}$. However, as compared to Eq. (126), the expression for ${\bf j}_{\bf k}$
obtained from $F_{\bf k}$ contains additional terms involving $\partial U_{\bf
k}/\partial{\bf q}$ and $\partial E^{(s)}_{\bf k}/\partial{\bf q}$. By letting
the sum of the additional terms in the expression for ${\bf j}_{\bf k}$ to be
zero, we arrive at
$2\frac{\partial U_{\bf k}}{\partial{\bf q}}+\,\left(n_{\bf k}+\,n_{-\bf
k}\right)\frac{\partial E^{(s)}_{\bf k}}{\partial{\bf q}}=0.$ (127)
Substituting the expressions for $U_{\bf k}$ and $E^{(s)}_{\bf k}$ [Eqs. (34)
and (37)] into Eqs. (125) and (127), we obtain
$|\Delta_{\bf k}|^{2}\frac{\partial C_{\bf k}}{\partial X}=0,$ (128)
and therefore, for $|\Delta_{\bf k}|>0$, Eq. (121) [which is Eq. (47)].
The derivation of Eq. (47) presented in this Appendix is similar to that
presented in Ref. hao11, for a similar equation in the theory for the
superconductivity in the presence of a magnetic field.
## References
* (1) J. Bardeen, L. N. Cooper, and J. R. Schrieffer, Phys. Rev. 108, 1175 (1957).
* (2) F. London, Superfluids, vol. 1, John Wiley & Sons, New York, 1950.
* (3) F. London, Superfluids, vol. 2, John Wiley & Sons, New York, 1954.
* (4) F. London, Nature 141, 643 (1938); Phys. Rev. 54, 947 (1938);
* (5) L. Tisza, Nature 141, 913 (1938).
* (6) L. D. Landau, J. Phys. (Moscow) 5, 71 (1941) (reprinted in I. M. Khalatnikov, An Introduction to the Theory of Superfluidity, translated by P. C. Hohenberg, Perseus Publishing, Cambridge, Massachusetts, 2000).
* (7) N. N. Bogoliubov, J. Phys. (Moscow) 11, 23 (1947) (reprinted in D. Pines, The Many-Body Problem, W. A. Benjamin, New York, 1961).
* (8) R. P. Feynman, Phys. Rev. 94, 262 (1954); in Progress in Low Temperature Physics (C. J. Gorter, ed.), Vol. I, Chap. II, North-Holland, Amsterdam, (1955).
* (9) J. G. Valatin and D. Butler, Nuovo Cimento 10, 37 (1958).
* (10) W. A. B. Evans and Y. Imry, Nuovo Cimento B 63, 155 (1969).
* (11) P. Nozi$\grave{\text{e}}$res and D. Saint James, J. Physique 43, 1133 (1982).
* (12) G. S. Jeon, L. Yin, S. W. Rhee and D. J. Thouless, Phys. Rev. A 66, 011603(R) (2002).
* (13) G. B. Hess and W. M. Fairbank, Phys. Rev. Lett. 19, 216 (1967).
* (14) N. N. Bogoliubov, Nuovo Cimento 7, 794 (1958); Zh. Eksp. Teor. Fiz. 34, 58 (1958) [Sov. Phys. JETP 7, 41 (1958)].
* (15) J. G. Valatin, Nuovo Cimento 7, 843 (1958).
* (16) A. L. Fetter and J. D. Walecka, Quantum Theory of Many-Partcle Systems, McGraw-Hill, New York, 1971\.
* (17) K. Huang, Statistical Mechanics, Wiley, New York, 1963.
* (18) R. P. Feynman, Statistical Mechanics, Benjamin, Reading, MA (1972).
* (19) Zhidong Hao, “New interpretation for energy gap $\Delta$ of the cut-off approximation in the BCS theory of superconductivity,” arXiv:0706.2392.
* (20) S. D. Conte and C. de Boor, Elementary Numerical Analysis: An Algorithmic Approach, 3rd edition, McGraw-Hill, New York (1980).
* (21) M. J. Buckingham and W. M. Fairbank, “The nature of the lambda transition,” in Progress in Low Temperature Physics (C. J. Gorter, ed.), Vol. III, North-Holland, Amsterdam, (1961).
* (22) C. Ryu, M. F. Andersen, P. Clade, V. Natarajan, K. Helmerson and W. D. Phillips, Phys. Rev. Lett. 99, 260401 (2007).
* (23) J. R. Abo-Shaeer, C. Raman, J. M. Vogels and W. Ketterle, Science 292(5516), 476 (2001).
* (24) P. Engels, I. Coddington, P. C. Haljan, V. Schweikhard and E. A. Cornel, Phys. Rev. Lett. 90(17), 170405 (2003).
* (25) Zhidong Hao, “Theory for superconductivity in a magnetic field: A local approximation approach,” arXiv:0706.2394.
* (26) M. R. Piegel, Mathematical Handbook of Formulas and Tables, McGraw-Hill, New York, 1968.
| arxiv-papers | 2008-07-09T17:12:13 | 2024-09-04T02:48:56.650453 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Zhidong Hao",
"submitter": "Zhidong Hao",
"url": "https://arxiv.org/abs/0807.1503"
} |
0807.1590 | # Nonequilibrium carriers in an intrinsic graphene under interband
photoexcitation
A. Satou1,2 F. T. Vasko1,3 ftvasko@yahoo.com V. Ryzhii1,2 1 University of
Aizu, Ikki-machi, Aizu-Wakamatsu 965-8580, Japan
2 Japan Science and Technology Agency, CREST, Tokyo 107-0075, Japan
3 Institute of Semiconductor Physics, NAS of Ukraine, Pr. Nauki 41, Kiev,
03028, Ukraine
###### Abstract
We study nonequilibrium carriers (electrons and holes) in an intrinsic
graphene at low temperatures under far- and mid-infrared (IR) radiation in a
wide range of its intensities. The energy distributions of carriers are
calculated using a quasiclassic kinetic equation which accounts for the energy
relaxation due to acoustic phonons and the radiative generation-recombination
processes associated with thermal radiation and the carrier photoexcitation by
incident radiation. It is found that the nonequilibrium distributions are
determined by an interplay between weak energy relaxation on acoustic phonons
and generation-recombination processes as well as by the effect of pumping
saturation. Due to the effect of saturation, the carrier distribution
functions can exhibit plateaus around the pumping region at elevated
intensities. As shown, at sufficiently strong mid-IR pumping, the population
inversion can occur below the pumping energy. The graphene dc conductivity as
a function of the pumping intensity exhibits a pronounced nonlinearity with a
sub-linear region at fairly low intensities and a saturation at a strong
pumping. However, an increase in the pumping intensity in very wide range
leads only to a modest increase in the carrier concentration and,
particularly, the dc conductivity. The graphene conductivity at mid-IR
irradiation exhibit strong sensitivity to mechanisms of carrier momentum
relaxation.
###### pacs:
73.50.Pz, 73.63.-b, 81.05.Uw
## I Introduction
The features of the dynamics of carriers (electrons and holes) in graphene and
the mechanisms of their relaxation 1 result in the exceptional properties of
graphene and a new device prospects (see 2 for review). The studies of
optical phenomena, including the Raman scattering (see3 and references
therein), ultrafast spectroscopy, 4 ; 5 and magnetooptics, 6 can be used to
reveal both the energy spectrum parameters and the mechanisms of carrier
scattering. The gapless energy spectrum of graphene, with the characteristic
velocity $v_{W}\simeq 10^{8}$ cm/s which corresponds to the neutrinolike
bandstructure (Weyl-Wallace model) 7 , provides its nontrivial optical
properties due to effective interband transitions in the far- and mid-infrared
(IR) spectral regions. The linear response of epitaxial graphene 8 and
graphite 9 was measured and the pertinent calculations were performed in
Ref.10 . Recently, 11 a fairly low threashold of the nonlinear response under
far- or mid-IR excitation have been found and the photoconductivity of an
intrinsic graphene was calculated for a low-pumping region. But an essentially
nonlinear regime of response was not calculated and no experimental data
concerning nonlinear properties of graphene in this spectral region are
available. Thus, an investigation of the nonlinear response of graphene is
timely now.
In this paper, we consider the nonequilibrium energy distributions of the
carriers under interband photoexcitation by far- or mid-IR pumping in a wide
range of the intensities. Using the obtained distributions, we analyze the
nonlinear dependencies of the carrier concentration and the dc conductivity on
the pumping intensity and frequency at different temperatures. The dynamic
conductivity of graphene photoexcited by far- or mid-IR radiation is
calculated as well.
As shown below, the energy distribution of carriers are determined by an
interplay between quasielastic energy relaxation, which is ineffective at low
energies, and generation-recombination processes. There is a marked increase
in the carrier population in the range of energies where the phonon and
radiative mechanisms of relaxation are virtually compensated. The interband
absorption saturation also affects the energy distribution of carriers. The
graphene dc conductivity as a function of the pumping intensity exhibits a
pronounced nonlinearity with a sub-linear region at fairly low intensities and
a saturation at a strong pumping. It is also found that in a certain energy
range of the pumping power the real part of the interband contribution to the
dynamic conductivity becomes negative, i.e. the negative absorption takes
place.
The paper is organized in the following way. The model under consideration is
described in Sec. II. In Sec. III, we present the nonequilibrium carrier
distribution as function of pumping intensity for a few pumping frequencies.
The results of calculations of the photoconductivity and the dynamic
conductivity, as a response on a weak dc electric field or a probe high-
frequency field, are discussed in Sec. IV. The brief discussion of the
assumptions used in calculations and concluding remarks are given in the last
section.
## II Model
In order to obtain the energy distributions of the carriers in intrinsic
graphene we use the quasiclassic kinetic equation derived and analytically
analyzed previously 11 for the case of low intensity pumping. The kinetic
equation under consideration accounts for the energy relaxation due to
scattering on acoustic phonons, the radiative generation-recombination
processes associated with thermal radiation, and the far- or mid-IR pumping.
We disregard the inter-carrier scattering since at low temperatures the
carrier concentration can be small even at relatively strong interband
pumping. Since the scattering mechanisms in $c$\- and $v$-bands are symmetric,
the electron and hole distributions in the intrinsic material are identical
and we consider below the carrier distribution function $f_{p}$. Taking into
account the abovementioned mechanisms, the kinetic equation under
consideration, which governs the distribution function $f_{p}$, is presented
in the following form:
$J_{LA}\\{f_{p}\\}+J_{R}\\{f_{p}\\}+G\\{f_{p}\\}=0.$ (1)
Here the collision integrals $J_{LA}\\{f_{p}\\}$ and $J_{R}\\{f_{p}\\}$ are
associated with the relaxation of carriers caused by the acoustic phonons and
the equilibrium thermal radiation, respectively, the term $G\\{f_{p}\\}$
describes the interband carrier excitation. Since the interband transition due
to the acoustic phonon scattering are forbidden (the sound velocity is weak in
comparison to $v_{W}$), the concentration balance equation takes form
$\int_{0}^{\infty}dpp[J_{R}\\{f_{p}\\}+G\\{f_{p}\\}]=0$ (2)
and it can be considered as the normalization condition for $f_{p}$. Another
condition for $f_{p}$ streams from the requirement $f_{p\rightarrow\infty}$=0,
so that any term in Eq. (1) is equal to zero at $p\rightarrow\infty$, i.e. the
zero acoustic flow at high energies takes place,
$J_{LA}\\{f_{p}\\}|_{p\rightarrow\infty}$=0.
Considering the quasielastic scattering of carriers on acoustic phonons, one
can use the Fokker-Planck form of $J_{LA}\\{f_{p}\\}$ 12 and Eq. (1) can be
presented as 11
$\displaystyle\frac{\gamma}{p}\frac{d}{dp}\left\\{p^{4}\left[\frac{df_{p}}{dp}+\frac{f_{p}(1-f_{p})}{p_{T}}\right]\right\\}$
$\displaystyle+\frac{p}{p_{T}}\left[N_{2p/p_{T}}(1-2f_{p})-f_{p}^{2}\right]$
(3) $\displaystyle+G(1-2f_{p})\Delta\left(\frac{p-p_{\Omega}}{\delta
p_{\Omega}}\right)=0.$
Here $p_{T}=T/v_{W}$ is the characteristic thermal momentum ($T$ is the
temperature in the energy units), $N_{2p/p_{T}}=[\exp(2p/p_{T})-1]^{-1}$ is
the Planck distribution function, and the form-factor $\Delta(\varepsilon)$
describes the broadening of interband transitions. So that
$p_{\Omega}=\hbar\Omega/2v_{W}$ is the momentum of just photogenerated
carriers corresponding to the pumping frequency $\Omega$ and the broadening is
described by $\delta p_{\Omega}$ which is proportional to $p_{\Omega}$ if the
scattering rates are proportional to the density of states. It was assumed
below that $\Delta(\varepsilon)=\pi^{-1/2}\exp(-\varepsilon^{2})$ and $\delta
p_{\Omega}/p_{\Omega}\simeq$0.1 13 . The dimensionless parameter $\gamma$ and
$G$ are the relative strength of the energy relaxation with respect to the
generation-recombination efficiency and the characteristic value of the
pumping intensity, respectively. It is important below, that a following
dependencies on $T$, $\Omega$ and the pumping intensity, $S$, take place:
$\gamma\propto T~{},~{}~{}~{}~{}~{}G\propto S/(\Omega^{3}T),$ (4)
so that the acoustic contribution increases with $T$ while the excitation
efficiency decreases with $\Omega$ and $T$. The explicit expressions of $g$
and $G$ can be found in Ref. 11 . Using the typical parameters of graphene 14
, at $T$=77 K one obtains $\gamma\simeq$0.32. At the same temperature at
$S$=1W/cm2 and $\hbar\Omega$=120 meV one can get $G\simeq$2.8.
Numerical procedure for the problem described can be simplified if one takes
into account that the equilibrium distribution remains valid at $p\rightarrow
0$. The condition $f_{p\rightarrow 0}$=1/2 (it is the Fermi function at zero
energy) can be explicitly derived considering that $G\\{f_{p}\\}$ vanishes at
$p\rightarrow 0$ (if $p_{\Omega}\gg\delta p_{\Omega}$) and equations
$J_{LA}\\{f_{p}\\}=0$ and $J_{R}\\{f_{p}\\}=0$ are satisfied by the
equilibrium distribution. Taking this into account, one can write the boundary
condition at $p\rightarrow\infty$, as the requirement that the factor
$\\{\ldots\\}$ under derivative in the first term in the left-hand side of Eq.
(3) turns to zero. As a consequence, for numerical solution of Eq. (2), one
can use the following requirements as the boundary conditions for the
distribution function:
$f_{p}\biggr{|}_{p\to 0}=\frac{1}{2},\qquad
p^{4}\biggl{[}\frac{df_{p}}{dp}+\frac{f_{p}(1-f_{p})}{p_{T}^{2}}\biggr{]}\biggr{|}_{p\to\infty}=0,$
(5)
while the balance condition (2) should be used for checking of a numerical
results obtained. The numerical solution of Eq. (3) is performed below using a
finite difference method and the pertinent iteration procedure.
Figure 1: (Color online) Distribution functions $f_{p}$ vs energy $v_{W}p$ (a)
for the pumping intensities $S$=10-5, 10${}^{-4},\ldots$, 103 W/cm2 (from left
to right) at $T=4.2$ K and (b) for $S$=10-2, 10${}^{-1},\ldots$, 103 W/cm2 at
77 K. Dotted curves correspond to the equilibrium distributions. Vertical
arrow corresponds to the energy equal to $\hbar\Omega/2$ at $\hbar\Omega$=12
meV.
## III Nonequilibrium distribution
We present here the results of numerical solution of Eq. (3) with conditions
(5) and discuss the obtained distribution functions at different excitation
conditions (frequency and intensity of pumping) and temperature for the
typical parameters of graphene 14 . The variation of the sheet carrier
concentration with varying excitation conditions is also considered.
Figure 2: (Color online) The same as in Fig. 1 under mid-IR pumping with
$\hbar\Omega=$60 meV.
Figure 3: (Color online) The same as in Fig. 1 under the CO2 laser pumping,
$\hbar\Omega=$120 meV, with intensities (a) $S$=10-4, 10${}^{-3},\ldots$, 105
W/cm2 (from left to right) and (b) for $S$=10-2, 10${}^{-1},\ldots$, 105
W/cm2.
The obtained distribution functions under the far-IR pumping with the photon
energy $\hbar\Omega=$12 meV are shown in Fig. 1 with the one-order step in
pumping intensities from $10^{-5}$ W/cm2 to $10^{3}$ W/cm2. First of all, one
can see a visible modification of the carrier distributions at low intensities
(up to $S\sim 10^{-3}$ W/cm2 at $T$=4.2 K and for $S\leq 10^{-1}$ W/cm2 at
$T$=77 K). Next, the peak of the distribution function shifts towards
$\hbar\Omega/2$. Such a behavior is in agreement with the previous analytical
consideration 11 . Further, at sufficiently strong pumping, $f_{p\sim
p_{\Omega}}$ tends also to the values close to $1/2$, so that plateau-like
energy distibutions are formed. The range of carrier energies, where
$f_{p}\simeq 1/2$, widens with increasing pumping intensity starting $S\sim
10^{-2}$ W/cm2 at $T$=4.2 K and $S\sim 0.1$ W/cm2 at $T$=77 K. Since $f_{p}$
remains equilibrium at $p\to 0$ (all terms of Eq.(1) are equal zero
separately), one can see a deepening of $f_{p}$ in the region $v_{W}p\sim T$.
Under mid-IR excitation, $\hbar\Omega=$60 meV, a similar character of the low-
pumping regime of response takes place, see Fig. 2. With increased pumping
intensity, a peak of distribution appears, moreover, the peak distribution
function can markedly exceed $1/2$, i.e. the population inversion occurs if
$S>10^{-2}$ W/cm2 at $T$=4.2 K and $S>10^{-1}$ W/cm2 at $T$=77 K. Further, in
the pumping region when $S>1$ W/cm2 at $T$=4.2 K (or $S>10$ W/cm2 at $T$=77 K)
$f_{p\sim p_{\Omega}}$ tends to $1/2$, so that a plateau of $f_{p}$ is formed
around the energy $\hbar\Omega/2$. As $S$ increases and the plateau region
widens, the peak amplitude below $\hbar\Omega/2$ somewhat decreases but a
maximal value of $f_{p}$ exceeds 1/2. Since the equilibrium distribution is
predetermined by Eqs. (3) and (5) at $p\to 0$, a non-monotonic distribution
with a deepening at $v_{W}p\leq T$, a peak under the energy $\hbar\Omega/2$,
and a plateau around $\hbar\Omega/2$ is realized.
Similar character of distribution takes place as $\hbar\Omega$ increases: the
same peculiarities are shifted to higher intensities according to Eq. (5). In
Figs. 3a and 3b we plot $f_{p}$ under the CO2 laser pumping, $\hbar\Omega=$120
meV. One can see that the population inversion regime begins for $S>10^{-2}$
W/cm2 at $T$=4.2 K (or for $S>10^{-1}$ W/cm2 at $T$=77 K) and the saturation
region around $\hbar\Omega/2$ takes place if $S>10^{3}$ W/cm2 at $T$=4.2 K and
77 K. These peculiarities are retained for higher intensities, up to $S\sim
10^{5}$ W/cm2.
Figure 4: (Color online) Carrier concentration $n_{2D}$ vs pumping intensity
$S$ for the excitation energies $\hbar\Omega=$12, 60, and 120 meV (solid,
dashed, and dotted curves, respectively) at different temperatures.
Figure 5: (Color online) Normalized dc conductivity $\sigma/\sigma_{o}$ vs
pumping intensity $S$: (a) for low pumping region and $l_{c}$=20 nm at
different temperatures and $\hbar\Omega=$12, 60, and 120 meV (solid, dashed,
and dotted curves, respectively), (b) for mid-IR pumping at temperatures
$T=4.2$ K and 77 K (solid and dotted curves), and (c) for different $l_{c}$
and $\hbar\Omega=$60 meV at $T=4.2$ K and 77 K (solid and dotted curves).
The obtained distribution functions allows to calculate the sheet carrier
concentration at different pumping conditions and temperatures according to
the standard formula:
$n_{2D}=\frac{2}{\pi\hbar^{2}}\int_{0}^{\infty}dppf_{p}.$ (6)
Figure 4 demonstrates the dependences of the sheet concentration in graphene
as a function of the pumping intensity corresponding to the energy
distributions shown in Figs. 1-3. As follows from the numerical calculation of
the integral (6) at $T$=4.2 K, $n_{2D}$ increases with $S$ in an intermediate
pumping region and tends to saturation under higher pumping intensities. The
power dependence $n_{2D}\propto S^{r}$ with $r\simeq$0.57 takes place at
$T=4.2K$ between $S\sim 10^{-6}$ W/cm2 and 0.5 W/cm2 for the mid-IR pumping or
between $S\sim 10^{-7}$ W/cm2 and $10^{-5}$ W/cm2 for the far-IR pumping (see
Figs. 4a and 4b, respectively). Under mid-IR pumping at $T$=77 K, the
concentration increases according to the same power law and saturates at the
same pumping intensities. However, the nonlinear regime begins starting $S\sim
5$ mW/cm2. This is because the equilibrium concentration is proportional to
$T^{2}$. Note that in the case of pumping by CO2 laser, the saturation occurs
at $n_{2D}>10^{11}$ cm-2. In this situation, the inter-carrier scattering
might be important.
## IV DC and dynamic conductivities
Here we turn to the consideration of the response of the nonequilibrium
carriers with the distribution functions obtained above to a weak dc electric
field or a probe radiation. Taking into account that the momentum relaxation
of carriers is caused by elastic scattering mechanisms, one can use the
following formula for the dc conductivity $\sigma$ 15 :
$\sigma=\sigma_{0}\biggl{[}2f_{p=0}-\frac{l_{c}}{\hbar}\int_{0}^{\infty}dpf_{p}\frac{\Psi^{\prime}(pl_{c}/\hbar)}{\Psi(pl_{c}/\hbar)^{2}}\biggr{]}.$
(7)
Here $l_{c}$ is the correlation length of static disorder scattering,
$\Psi(z)=e^{-z^{2}}I_{1}(z^{2})/z^{2}$, where $I_{1}(z^{2})$ the first order
Bessel function of imaginary argument, and $\sigma_{0}$ is the conductivity in
the case of short-range disorder scattering, when $l_{c}=0$ 15 . According to
Eq.(5) $f_{p=0}=1/2$, for the short-range scattering case, when
$\overline{p}l_{c}/\hbar\ll 1$ ($\overline{p}$ is the average momentum), the
dc conductivity $\sigma$ is determined by the low energy carriers so that
$\sigma$ appears to be independent of optical pumping intensity despite a
significant concentration of the photogenerated carriers,
$\sigma\simeq\sigma_{0}$. For the definiteness, it was assumed that $l_{c}$=
10, 20, and 30 nm.
Figure 5 shows the dependences of the dc conductivity $\sigma$ normalized by
its characteristic value $\sigma_{0}$ on the pumping power $S$ calculated
using Eq. (7) with the distribution functions obtained in Sec. III for $T=4.2$
K and $77$ K and different pumping energies $\hbar\Omega$. The dependences are
markedly nonlinear beginning from rather low pumping intensities (see the
pertinent curves in Fig. 5a). The nonlinearity of the dc conductivity at low
$S$ ($\leq$1 mW/cm2) is particularly strong in the case $T=4.2$ K and far-IR
pumping, $\hbar\Omega=12$ meV, but the variation of conductivity does not
exceed $5\%$ and becomes saturated at $S>$5 mW/cm2. In the range of
intermediate intensities of mid-IR pumping, see Fig. 5b, there is a power
increase in the dc conductivity, which is followed by the saturation region.
It is instructing that an increase in the pumping intensity by several orders
of magnitude leads to a modest increase in $\sigma/\sigma_{0}$.
In Fig. 5c, the normalized dc conductivity as a function of the pumping
intensity calculated for different values of the correlation length of the
disorder scattering, $l_{c}$. One can see that the conductivity becomes more
sensitive to the pumping when parameter $l_{c}$ increases. If $l_{c}$ tends to
zero, a quenching of photoconductivity takes place. Since the dc conductivity
$\sigma$ depends on $\overline{p}l_{c}/\hbar$, it, as a function of the
pumping intensity, saturatates at higher intensities that the carrier
concentration $n_{2D}$ (compare Figs. 5 and 4).
Figure 6: (Color online) (a) Spectral dependences of the real part of the
dynamic interband conductivity $Re\sigma_{\omega}$ at $T=4.2$ K for
$\hbar\Omega$=60 meV and pumping intensities: $S=10^{-3}$, 10-1, 101, and 103
W/cm2 (solid, dashed, dotted and dash-dotted curves, respectively). (b)
Regions of parameters (shaded by vertical and horizontal lines for $T$=4.2 K
and 77 K, respectively) for which the dynamic interband conductivity is
negative if $\hbar\Omega$=60 meV.
As seen from Figs. 2 and 3, the distribution function $f_{p}$ can be larger
than $1/2$ in a certain energy range if the pumping power is sufficiently
high. This corresponds to the population inversion of carriers and might be
used for far infrared lasing. The real part of the interband contribution to
the dynamic conductivity which is given by 9
${\rm
Re}\sigma_{\omega}=\frac{e^{2}v_{W}^{2}}{\hbar^{2}\omega}\int_{0}^{\infty}dp\,p(1-2f_{p})\,\Delta\biggl{(}\frac{p-p_{\omega}}{\delta\,p_{\omega}}\biggr{)},$
(8)
can be negative. Here $p_{\omega}\equiv\hbar\omega/2v_{W}$ and $\delta
p_{\omega}\propto p_{\omega}$ are introduced in a similar way to $p_{\Omega}$
and $\delta p_{\Omega}$ in Eq. (3).13 . At $\delta p_{\omega}\to$0 one obtains
${\rm Re}\sigma_{\omega}<0$, if $f_{p_{\omega}}>1/2$. Figure 6a shows the
dependences of the real part of the interband dynamic conductivity ${\rm
Re}\sigma_{\omega}$ (normalized by factor $e^{2}/\pi\hbar$) on the probe
frequency $\omega$ at different pumping intensities $S$. In contrast to the
pumping scheme when the pumping energy $\hbar\Omega$ exceeds the energy of
optical photons, so that a cascade of optical phonons is emitted and the
photogenerated electrons (holes) are accumulated near the conduction band
bottom (valence band top), the negative dynamic interband conductivity in the
case under consideration corresponds to the interband transitions with
relatively high energies $\hbar\omega$ (compare with 16 ). Figure 6 shows the
regions (shaded) on the pumping intensity - probe frequency plane
corresponding to the negative dynamic interband conductivity at different
temperatures. The dependence of the real part of net dynamic conductivity on
the probe frequency $\omega$, can be markedly modified by the contribution of
the intraband transitions (which correspond to the Drude conductivity).
However, the latter contribution can be effective only in the range of rather
small frequencies $\omega$.16
## V Discussion and Conclusions
The main restriction the model used above is the neglect of the Coulomb
interaction of carriers, so that the obtained results are valid at their
sufficiently low concentrations. As can be found from Fig. 4, in the case of
an intrinsic graphene, this assumption limits the validity of our model by the
temperatures $T\lesssim 100-200$ K, whereas the limitation imposed on the
pumping intensity is fairly liberal. The latter is owing to the effect of
saturation of the interband absorption associated with the Pauli principle. In
our consideration we also disregarded more complex processes, namely, the
carrier interactions with the substrate vibrations and the effect of disorder
on the generation-recombination processes. The mechanism included in the model
describe the general features of the relaxation processes considered, so that
their refinement should not lead to a qualitative change of the pattern of the
phenomena studied above. As for the photogeneration of carriers by incident
far- or mid-infrared radiation, only the single-photon interband processes
were taken into account. Hence, possible nonlinear frequency multiplication
and renormalization of the energy spectra were disregarded. Our estimates show
that the latter correspond to the limitations of the pumping intensity by the
values $S<0.5$ MW/cm2 (at $\hbar\Omega=12$ meV) and $5$ MW/cm2 (at
$\hbar\Omega=120$ meV). The other assumptions of our model (the isotropic
energy spectra of carriers, the valley degeneration, and exclusion of the
interaction with optical phonons in the range of energy under consideration)
appears to be rather natural.
In summary, we calculated the energy distributions of carriers in an intrinsic
graphene at low temperatures under far- and mid-IR ($\hbar\Omega=12-120$ meV)
pumping in a wide range of its intensities (from zero to $10^{2}$ kW/cm2). It
was shown that these distributions are determined by interplay between weak
energy relaxation on acoustic phonons and radiative generation-recombination
processes as well as by the effect of pumping saturation due to the Pauli
principle. The obtained energy distributions at elevated pumping intensities
can exhibit the plateau regions and the regions corresponding to the
population inversion. The graphene dc conductivity as a function of the
pumping intensity demonstrates a pronounced nonlinearity at fairly weak
pumping (particularly at helium temperature), a steep increase in a certain
range of intermediate intensities, and a saturation at sufficiently strong
pumping. Due to this, the graphene dc conductivity varies through 1-2 orders
of magnitude when the pumping intensity is varied many orders. We showed that
the alteration of the graphene dc conductivity with far- and mid-IR
irradiation, i.e., the effect of the photoconductivity is sensitive to the
correlation length of disorder scattering.
To conclude, the obtained results, can be useful for the extraction of the
parameters determined the relaxation mechanisms in graphene and for the
estimation of a potential of novel optoelectronic devices 16 ; 17 ; 18 , in
particular, in terahertz and far infrared lasers 16 ; 17 .
## Acknowledgments
The authors (A. S. and V. R.) are grateful to Professor T. Otsuji for
stimulating discussions. This work was partially supported by the Japan
Science and Technology Agency, CREST, Japan.
## References
* (1) C. Berger et al., J. Phys. Chem. 108, 19912 (2004); K.S. Novoselov, et al., Nature 438, 197 (2005).
* (2) F. Guinea and N.M. Peres, Physics World 19 33 (2006); A.K. Geim and A.H. MacDonald, Physics Today 60, 35 (2007); A.H. Castro Neto et al., arXiv:0709.1163.
* (3) J.M. Dawlaty, et al., arXiv:0712.0119; T. Stauber et al., arXiv:0803.1802.
* (4) J.-C. Charlier et al., Topics Appl. Physics 111, 673 (2008).
* (5) J.M. Dawlaty et al., arXiv:0801.3303; D. Sun et al., arXiv:0803.2883, P.George et al., arXiv:0805.4647; S. Butscher et al., Appl. Phys. Lett. 91, 203103 (2007).
* (6) Z. Jiang et al., Phys. Rev. Lett. 98, 197403 (2007); R.S. Deacon et al., Phys. Rev. B 76, 081406(R) (2007).
* (7) E.M. Lifshitz, L.P. Pitaevskii, and V.B. Berestetskii, Quantum Electrodynamics, (Butterworth-Heinemann, Oxford 1982); P.R. Wallace, Phys. Rev. 71, 622 (1947).
* (8) P. Plochocka, et al., arXiv:0709.1324.
* (9) A. B. Kuzmenko et al., arXiv:0712.0835; T.G. Pedersen, Phys. Rev. B 67, 113106 (2003).
* (10) L. A. Falkovsky and A. A. Varlamov, Eur. Phys. J. B 56, 281 (2007); L. A. Falkovsky, arXiv:0806.3663; V. P. Gusynin et al., Int. Jour. of Mod. Phys. B 21, 4611 (2007).
* (11) F.T. Vasko and V. Ryzhii, Phys. Rev. B 77, 195433 (2008).
* (12) E.M. Lifshitz and L.P. Pitaevskii, Physical Kinetics, (Pergamon, Oxford 1981); F.T. Vasko and O.E. Raichev, Quantum Kinetic Theory and Applications (Springer, N.Y. 2005).
* (13) One can estimate the broadening energy as $v_{d}p_{\Omega}$, so that $\delta p/p_{\Omega}\simeq v_{d}/v_{\scriptscriptstyle W}$. Here the characteristic velocity $v_{d}$ was introduced in Ref. 15 through the momentum relaxation time for the short-range scattering mechanism, $\hbar/v_{d}p$.
* (14) The characteristic values of the energy relaxation and pumping, $\gamma$ and $G$, are introduced in Eq. (3) with the use of the energy relaxation, the generation-recombination, and the photoexcitation rates given by Eqs. (13), (10), and (7) of Ref. 11. Under numerical calculations we used the following parameyers of graphene: the deformation potential 12 eV, the sheet density 7$\cdot 10^{-8}$ g/cm2, the sound velocity 7.3$\times 10^{5}$ cm/s and the dielectric permittivity $\sim 3.7$ (SiO2 \- graphene structure).
* (15) F.T. Vasko and V. Ryzhii, Phys. Rev. B 76, 233404 (2007).
* (16) V. Ryzhii, M. Ryzhii, and T. Otsuji, J. Appl. Phys. 101, 083114 (2007); Phys. Stat. Sol.(c) 5, 261 (2008).
* (17) F. Rana, IEEE Trans. Nanotechnology 7, 91 (2008).
* (18) V. Ryzhii et al., Appl. Phys. Express 1, 063002 (2008).
| arxiv-papers | 2008-07-10T08:08:53 | 2024-09-04T02:48:56.658395 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "A. Satou, F.T. Vasko, and V. Ryzhii",
"submitter": "Fedir Vasko T",
"url": "https://arxiv.org/abs/0807.1590"
} |
0807.1620 |
Finding Short Cycles in an Embedded Graph in Polynomial Time 111Supported by
NNSF of China under granted number 10671073 and partially supported by Science
and Technology Commission of Shanghai Municipality (07XD14011)
Han Ren 222Supported by Shanghai Leading Academic Discipline Project, Project
Number B407 and Ni Cao
Department of Mathematics,East China Normal University,
Shanghai 200062,P.R.China
ABSTRACT
Let ${\cal{C}}_{1}$ be the set of fundamental cycles of breadth-first-search
trees in a graph $G$ and ${\cal{C}}_{2}$ the set of the sums of two cycles in
${\cal{C}}_{1}$. Then we show that
$(1)\,{\cal{C}}={\cal{C}}_{1}\bigcup{\cal{C}}_{2}$ contains a shortest
$\Pi$-twosided cycle in a $\Pi$-embedded graph $G$;$(2)$ $\cal{C}$ contains
all the possible shortest even cycles in a graph $G$;$(3)$ If a shortest cycle
in a graph $G$ is an odd cycle, then $\cal{C}$ contains all the shortest odd
cycles in $G$. This implies the existence of a polynomially bounded algorithm
to find a shortest $\Pi-$twosided cycle in an embedded graph and thus solves
an open problem of B.Mohar and C.Thomassen[2,pp112]
Key Words $\Pi-$twosided cycle, breadth-first-search tree, embedded graph.
AMS Classification: (2000)05C10,05C30,05C45.
1\. Introduction
C.Thomassen showed that if cycles in a set of cycles satisfy the $3$-path-
condition, then there exists a polynomial time algorithm that finds a shortest
cycle in this set[3]. As applications, he showed that the following types of
shortest cycles may be found in polynomial time.
(1) A shortest $\Pi-$noncontractible cycle in a $\Pi-$embedded graph;
(2) A shortest $\Pi-$nonseparating cycle in a $\Pi-$embedded graph;
(3) A shortest $\Pi-$onesided cycle in a $\Pi-$ embedded graph.
But what about a family of cycles which do not satisfy the $3$-path-condition?
In their monograph[2, pp112], B.Mohar and C.Thomassen raised the following
open problems:
(a) Is there a polynomially bounded algorithm that finds a shortest
$\Pi-$contractible cycle in a $\Pi-$embedded graph?
(b) Is there a polynomially bounded algorithm that finds a shortest
$\Pi-$surface-separating cycle in a $\Pi-$embedded graph?
(c) Is there a polynomially bounded algorithm that finds a shortest
$\Pi-$twosided cycle in a $\Pi-$embedded graph?
Here in this paper we consider connected graphs and all the concepts used are
standard following from [1,2].
Let $G$ be a $\Pi-$embedded graph and $T_{x}$ a breadth-first-search tree
rooted at a vertex $x$ of $G$. Let
${\cal{C}}_{1}=\\{C(T_{x})|\forall
x\in{V(G)},C(T_{x})\,\,is\,\,a\,\,fundamental\,\,cycle\,\,of\,\,T_{x}\\};$
${\cal{C}}_{2}=\\{C|\exists x,y\in{V(G)},C=C(T_{x})\oplus C(T_{y})\\};\\\
{\cal C}={\cal C}_{1}\bigcup{\cal C}_{2},$
where ”$\oplus$” is the operation defined as ”$A\oplus B=(A-B)\cup(B-A)$” for
any subsets $A,B$ of $E(G)$.
Theorem A. The collection $\cal C$ of cycles defined above satisfies the
following conditions:
(a). There exists a shortest $\Pi-$twosided cycle (in a $\Pi-$embedded graph )
in $\cal C$;
(b). Each shortest even cycle of a graph is contained in $\cal C$;
(c). If a shortest cycle of a graph is an odd cycle, then every shortest odd
cycle is contained in ${\cal C}_{1}$.
Therefore, Theorem A implies the existence of a polynomially bounded algorithm
to find short cycles defined above.
Theorem B. There exists a polynomially bounded algorithm to find a shortest
$\Pi-$twosided cycle in a $\Pi-$embedded graph and all the shortest even
cycles in a graph.
This solves problem (c). Since every possible two-sided cycle in an embedded
graph in the projective plane is contractible, we have the following
Corollary. There is a polynomially bounded algorithm to find a shortest
contractible cycle in an embedded graph in the projective plane.
This answers the problems (a)-(c) in the case of projective plane graphs.
2\. Proof of Main Result
A generalized embedding scheme of a graph $G$ in a surface $\Sigma$ is a set
of a rotation systems $\pi=\\{{\pi}_{v}|v\in{V(G)}\\}$ together with a mapping
$\lambda:E(G)\rightarrow\\{-1,+1\\}$,called a signature, where ${\pi}_{v}$ is
a clockwise ordering of edges incident with $v$. We define $\Pi=(\pi,\lambda)$
as an embedding scheme for the graph $G$. A cycle $C$ of an embedded graph $G$
ia called $\Pi-$twosided if $C$ contains even number of edges with negative
signature;otherwise, it is called $\Pi-$onesided.
For a vertex $v\in{V(G)}$, we may change the clockwise ordering to
anticlockwise, i.e., ${\pi}_{v}$ is replaced by its inverse ${\pi}^{-1}_{v}$,
and $\lambda(e)$ is replaced by $-\lambda(e)$ for each edge $e$ incident with
$v$. Therefore, we obtain another embedding scheme
${\Pi}^{\prime}=({\pi}^{\prime},{\lambda}^{\prime})$. It is clear that a cycle
is $\Pi-$twosided if and only if it is ${\Pi}^{\prime}$-twosided. Two of such
embedding schemes are equivalent if and only if one can be changed into
another by a sequence of such local changes. Thus, for any spanning tree $T$
of an embedded graph $G$, we may always assume that each edge $e$ of $T$ has
signature $\lambda(e)=+1$ for convenience.
The key of the proof of Theorem A is the following classification of short
cycles according to the distance between two vertices on such cycles. Let $C$
be a shortest $\Pi-$twosided cycle in an embedded graph $G$. Then $C$ must
satisfy one of the following conditions:
(1).$\forall x,y\in{V(C)}\Rightarrow d_{C}(x,y)=d_{G}(x,y);$
(2).$\exists x,y\in{V(C)}\Rightarrow d_{C}(x,y)>d_{G}(x,y$.
Remark: In the following, we will see that a cycle satisfying (2) may be
written as a sum of two shorter cycles. Therefore, a shortest cycle in a
collection of cycles satisfying the $3$-path-condition can’t satisfy (2).
Lemma 1. Let $G$ and $\cal C$ be as defined in Theorem A and $C$ a shortest
$\Pi$-twosided cycle in $G$ satisfying (2). Then $\cal C$ contains a shortest
$\Pi-$twosided cycle of $G$.
Proof of Lemma 1. We assume that $C$ has a clockwise ( anticlockwise
)orientation $\overrightarrow{C}$ ($\overleftarrow{C}$)and for any two
vertices $u$ and $v$ of $C$, $u\overrightarrow{C}v$ ($u\overleftarrow{C}v$)
denotes the closed interval from $u$ to $v$ along $\overrightarrow{C}$
($\overleftarrow{C}$). Similarly, we may define a segment $uPv$ as the closed
interval from $u$ to $v$ in a path $P$. If $u$ is a vertex of $C$, then
$u^{-}(u^{+})$ is the predecessor ( successor) of $u$ along
$\overrightarrow{C}$.
We consider two vertices $x,y$ of $C$ such that $d_{G}(x,y)$ is minimum
subject to (2). Then for any shortest $(x-y)$ path $P$ in $G$, $P$ and $C$ has
no inner vertex on $C$( since otherwise, there would be another two vertices
$x_{1},y_{1}$ od $C$ satisfying (2) and $D_{G}(x_{1},y_{1})<d_{G}(x,y)$). Let
$T_{x}$ be a breadth-first-search tree rooted at $x$. Then it contains a
$(x-y)$ path P and each of the two cycles $P\cup x\overrightarrow{C}y$ and
$P\cup x\overleftarrow{C}y$ is a $\Pi-$onesided cycle shorter than that of $C$
and satisfies (1) (since otherwise one of them may be written as a sum of two
shorter cycles,among then, one is a $\Pi-$twosided ). Under these
structure,each of $P\cup x\overrightarrow{C}y$ and $P\cup x\overleftarrow{C}y$
is a fundamental cycle of a breadth-first-search tree rooted at a vertex on
itself. This completes the proof of Lemma 1.
Now we turn to the cycles satisfying (1). Firstly, the following result is
easy to be verified and we omit the proof of it.
Lemma 2. Let $C$ be a shortest $\Pi-$twosided cycle in a $\Pi-$embedded graph
$G$. Then for any three vertices $x,u,v$ of $C$ with
$d_{c}(x,u)=d_{c}(x,v)=[\frac{|C|-1}{2}],$
any shortest $(x-u)$path $P$ and $x\overleftarrow{C}v$ has no inner vertex in
common.
Thus, for a shortest $\Pi-$twosided cycle $C$ satisfying the conditions in
Lemma 2 and the last common vertex $\alpha$ ( called branched vertex)
contained in $(x-u)$path $P_{1}$ and $(x-v)$path $P_{2}$ of $T_{x}$,
$x\overrightarrow{P_{1}}\alpha(=x\overrightarrow{P_{2}}\alpha)$ has no inner
vertex on $x\overleftarrow{C}v(x\overrightarrow{C}u)$.
Lemma 3. Let $G$ and $\cal C$ be as defined in Theorem A and $C$ a shortest
$\Pi-$twosided cycle in $G$ satisfying (1). Then $\cal C$ contains a shortest
$\Pi-twosided$ cycle of $G$.
Proof of Lemma 3.
Case 1 $|C|=2n,\,n\in N$.
Let $x$ and $y$ be two vertices of $C$ with $d_{G}(x,y)=n$ and $T_{x}$ a
breadth-first-search tree rooted at $x$ whose edges are all assigned signature
$\lambda=+1$. Then $T_{x}$ contains a $(x-y)$path $P$, a $(x-y^{-})$path
$P_{1}$, and a $(x-y^{+})$path $P_{2}$.
Subcase 1.1 Either $P_{1}\subset P$ or $P_{2}\subset P$.
We may suppose that $P_{1}\subset P$ and $\lambda(y,y^{+})=-1$ without
generality. If $P$ is not contained in $C$,then $P\cap
C=\\{x=x_{1},x_{2},...,x_{m}=y\\}$. By Lemma 2, we may choose an index $i$
such that
$C_{i}=x_{i}\overrightarrow{P}x_{i+1}\cup{x_{i}\overrightarrow{C}x_{i+1}}$ is
an even $\Pi-$onesided cycle such that $P$ has no inner vertex on
$x_{i}\overrightarrow{C}x_{i+1}$ other than $x_{i}$ and $x_{i+1}$. Since
$|C_{i}|<|C|$, $C_{i}$ can’t be a sum of two shorter cycles. Therefore,
$C_{i}$ satisfies (1) and further, $C_{i}$ is a fundamental cycle of a
breadth-first-tree rooted at some vertex of $C_{i}$. Now $C_{i}\cup(P_{2}\cup
P\cup\\{(y,y^{+})\\})$ has a shortest $\Pi-$twosided cycle in $G$.
If $P\subset C$, then $P_{2}$ can’t be contained in $C$ (since $C$ is
$\Pi-$twosided). As we have reasoned above, there exists a segment
$x_{j}\overleftarrow{C}x_{j+1}$ of $C$ such that
$C_{j}=x_{j}\overleftarrow{C}x_{j+1}\cup{x_{j}\overleftarrow{P_{2}}x_{j+1}}$
is a fundamental cycle and so, $C_{j}\cup(P_{2}\cup P\cup\\{(y,y^{+})\\})$
contains a shortest $\Pi-$twosided cycle in $G$.
Subcase 1.2 $P_{1}\not\subset P$ and $P_{2}\not\subset P.$
If $\lambda(y,y^{-})=+1$ or $\lambda(y,y^{+})=+1$, then
$P_{1}\cup{P\cup{(y,y^{-})}}$ or $P_{2}\cup{P\cup{(y,y^{+})}}$ will contain a
shortest $\Pi-$twosided cycle. Assume further that
$\lambda(y,y^{-})=\lambda(y,y^{+})=-1$. Then $P_{1}\cup
P_{2}\cup\\{(y,y^{-}),(y,y^{+})\\}$ also contains a shortest $\Pi-$twosided
cycle in $G$.
Case 2. $|C|=2n+1,\,n\in N$
Let $x,y\in C$ with $d_{C}(x,y)=d_{C}(x,y^{+})=n$ and $T_{x}$ be a breadth-
first-search tree rooted at $x$ with all its edges been assigned $\lambda=+1$.
Then $T_{x}$ contains a $(x-y)$path $P_{1}$ and a $(x,y^{+})$path $P_{2}$ and
a branched vertex $\alpha\in{P_{1}\cap P_{2}}$. Suppose that $P_{1}$ is not a
part of $C$ and $\lambda(y,y^{+})=-1$. Then as we have shown in Case 1 there
exists an index $i$ such that
$C_{i}=x_{i}\overrightarrow{C}x_{i+1}\cup{x_{i}\overrightarrow{P_{1}}x_{i+1}}$
is a fundamental cycle and thus $C_{i}\cup P_{1}\cup P_{2}\cup\\{(y,y^{+})\\}$
contains a shortest $\Pi-$twosided cycle in $\cal C$. This completes the proof
of Lemma 3 and so, finishes the proof of part (a) of Theorem A.
As for the proof of parts (b) and (c) of Theorem A, it follows from the
proving procedure of (a) of Theorem A.
Based on the above statement, Theorem A is proved.
References
[1] J.A.Bondy and U.S.R.Murty, Graph theory with applications, Macmillan,
London, (1978).
[2] B.Mohar and C.Thomassen, Graphs on Surfaces The Johns Hopkins University
Press, Baltimore and London, 2001.
[3] C.Thomassen, Embeddings of graphs with no short noncontractible cycles, J.
of Combin. Theory., Ser B 48 (1990), pp155-177
| arxiv-papers | 2008-07-10T10:33:18 | 2024-09-04T02:48:56.663903 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Han Ren, Ni Cao",
"submitter": "Ren Han",
"url": "https://arxiv.org/abs/0807.1620"
} |
0807.1629 |
From Grad-Shafranov Equations set to a pseudo-general form of
Non-Linear Schrödinger Equation
About the validity conditions of some general transformation
functionals of the free fields in the Grad-Shafranov Equations
M.Romeo (
( permanent e-mail address: michele.romeo.mr@gmail.com ) )
Theoretical Sector, Department of Physics, University of Salento,
Lecce (Italy)
In the year 2003 a paper by G. Lapenta demonstrated that there is a
“new class of soliton-like solutions for the Grad-Shafranov
Equations (GSE)”. The author determined an appropriate
pair of transformations of the free fields $\mathit{p}$ (fluid field of
hydrodynamical pressure) and $\mathit{B}_{z}$ (z-component of magnetic
induction field) that leads from the Helmholtz Equation to the Non-Linear
Schrödinger Equation (NLSE) with cubic non-linearity. In the following
year (2004), the work of Lapenta was opposed by G.N.
Throumoulopoulos et al., who criticized his idea of the field
transformations as a mathematically incoherent choice; contextually the
authors suggested a new point of view for this one. In his response, in the
same year, G. Lapenta carried out numerical simulations that showed
the existence of solitonic structures in a Magnetohydrodynamical (MHD)
plasma-context. In the present work I want to demonstrate a critical
condition for a 'complex' poloidal flux function $\mathit{\Psi }_{p}$ in a
plane framework $\mathit{(x,y)}$ that leads to a class of pseudo-general
NLSEs which establishes the validity of both G.Lapenta and
G.N. Throumoulopoulos et al. choices for the field transformation
§ INTRODUCTION
In theoretical models in which the accretion dynamics in
magnetohydrodynamical plasmas are explored ,i.e. plasmas that obey at the
various manetohydrodynamical models ,with good approximation, in Plasma
Physics, like resistive model, Hall-model and others mixed and dominant
models which include the General-Relativistic Theory, we can see that the
astrophysical plasmas develop themselves around compact objects, like
neutron stars and black holes usually; in such theoretical models in which
the formations of anti-parallel plasma-jets are often examined in the
context of their singular morphology, there are many references at well
defined characteristic mathematical structures that could explain the shape
of some sub-formation inside the accretion and ejection plasma mechanics.
These mathematical models result very interesting for both their
internal coherence and physical plausibility, because they can impose an
indiscutible mathematical link between the fundamental theory of the above
astrophysical phenomena and the physical charateristics of the singular
plasma formations in the above star structures, without affecting the
essential set of solutions in the base phisical models that explain very
well yet the most considerable star dynamics in the astrophysical plasma
For example, there are many mathematical singular structures often
investigated because they match at several characteristic cluster formations
inside the accretion disc and the plasma jets; these cluster formations are
well localized physical structures, with magnetic induction field $\mathbf{B%
}$ and electric field $\mathbf{E}$ topologies that take
place in the surrounding space and that donate, with the necessary
hydrodynamical force fields, a precise shape at these formations, which are
therefore confined in limited space areas in the curved space around the
concerned compact object (this situation exist because the gravitational
field densities near these astrophysical objects are relevant and a full
relativistic treatment in any possible theoretical model must be considered)
and that show in many cases a peculiar space periodicity in a well defined
global space configuration when the magnetohydrodynamical mechanics are in
steady state condition, or in an equilibrium state.
These localized plasma formation are indentifiable, for example, in
the space development of a plasma jet which derives from a compact star and
it stretch itself from the central object to the far space areas by a series
of quasi-spherical plasma structures named 'plasma bubbles' [1],[2]
(in this case the jet acquired a singular morphology known by the name of 'knotty structure') .
In other cases, similar structures are localized in the accretion
discs around same compact stars, leading to the formation of related
structures morphologically, in which the characteristic 'toroidal' sha- pe
is linked to the spiral motions in the disc, which response to a specifical
magnetohydrodynamical equations set of equlibrium state that persist on a
long time-scale, meaning in this way the same steady-state configuration. In
this regard, localized structures in the dynamics of accretion discs in an
astrophysical context have been identified analytically and
'experimentally', by several numerical simulations, from
Petviashvili et al. [3],[4] in the 80s as a
solution to the problem of the Grad-Shafranov Equation (GSE), a fundamental
non-linear equilibrium equation in Plasma Physics, by the context of a
general problem of two-dimensional dynamical plasma equilibrium.
Indeed the Grad-Shafranov Equation [
( H. Grad and H. Rubin (1958), V.D. Shafranov (1966) )] is a 2nd-order
elliptic PDE, from the mathematical point of view, which responds to a
plasma equlibrium problem in a two-dimensional framework $(x,y)$
[14] and it's derived by the Euler Equation (EE) in the ideal limit
(ideal Magnetohydrodynamics) and the Ampere's Law (i.e. 'source equation' in
the Maxwell's set equations) with neglect the 'displacement current',
because the non-relativistic limit $(v<<c)$ or the quasi-classic
relativistic limit (it takes into account by a 'pseudo-newtonian'
gravitational potential, like Paczynski-Wiita potential) of the gravity
potential term in the Euler Equation, which derived for the concerned
hydrodynamical problem (i.e. accretion disc or plasma jet in the
astrophysical objects framework).
This equation establish therefore a 'non-linear problem'
for a certain scalar 'flux function' $\mathit{\Psi }$
which defines the induction field $\boldsymbol{B}$ in the ordinary
space $(x,y,z)$ around the compact star and which have been yet
analized from Petviashvili et al. [3] in an
analytical work that demonstrated the existence of solitary toroidal
structures in the validity framework of the GSE;
furthermore, the GSE for this flux function $\mathit{\Psi }$
establish the necessary conditions (i.e. a 'critic conditions') to
which the 'parameters space' and the free fields in the fluid equations must
obey so that the equilibrium scenario is respected in the steady-state
At this point, it's indispensable to watch that in many non-linear
problems of Mathematical-Physics that usually develop some sets of
expressions containing PDEs, like non linear wave equations, we are led to
consider, by an appropriate choice of a function transformations set and a
'reduction of potentials', an equivalent and more general non-linear
differential problem that represents the 'root differential problem' for a
more general category of non-linear problems. From the point of view of this
one, the more considerable non-linear equation that is present in many
problems of Mathematical-Physics regarding non-linear wave phenomena that
involve PDEs is obviously the well known 'Non-Linear Schrö
dinger Equation' (NLSE), an equation that is often present, for
example, in non-linear optical problems [11].
You know the best known solution scenario of this non-linear
equation is a 'solitonic solution' or 'soliton solution',
where a 'soliton' is the well known special solitary wave
solution of several non-linear dispersive wave equations; the main
characteristics of this wave solution are its singular mathematical
structure and its physical behavior, which synthesize them primarily in
their peculiarity of localized wave-structure.
It seems therefore that the GSE, a non-linear (elliptic) equation,
can be lead to the more general form of the NLSE for the stationary case
(steady-state for a plasma in an equlibrium state) by an appropriate choice
of 'transformation functionals' for the free fields
inside it; this consideration is furthermore reasoned by the fact that the
peculiar plasma knots in the accretion discs and plasma jets cited
above could be justified by an appropriate 're-interpretation' of these
formations as periodic solitonic structures with justified contour
At this point, if such appropriate choice of a transformations set
exist and it is coherent from the mathematical point of view, there are no
reasons for to refuse such a possibility. Such a choice for turning GSE in a
NLSE with a specific cubic non-linearity, i.e. the NLSE for the Kerr
effect in a non-linear optical mean, was made recently by
G.Lapenta [8], who was demonstrated how it's possible to
obtain a pair of appropriate transformations in a two-dimensional framework
$(x,y)$, like that in the equatorial plane of an accretion disc,
which lead from a general GSE to a cubic NLSE by a two-dimensional
Helmholtz Equation (2dHE) [7], which is important in the non-linear
optical problems and it is related to many problems for steady-state
oscillations (mechanical, acoustical, thermal, electromagnetic); but in the
following year, by a purely mathematical work, Lapenta's idea was opposed by Throumoulopoulos G.N., Hizanidis K.
and Tasso H. [9], whom criticized one of his
pair of free fields transformations because its supposed mathematical
inconsistency. The authors proposed furthermore a new pair of correct
transformations for the same problem.
In this work both Lapenta G. and
Throumoulopoulos G.N. et al. positions will be examined and a new
and more general choice will be proposed by me, who will show that a
'critical condition' exist for the flux function $\mathit{\Psi }$,
so that this general choice be a coherent choice, and both
choices are not only mathematically coherent but the more general choice
cited above include any possible set of reasonable transformations, leading
in this way to a more desiderable pseudo-general form of the NLSE [6].
All the analytical considerations will be made for the case of
knotty structures in the accretion discs as in Lapenta's
work, therefore in the equatorial plane or in the $(x,y)$
framework; in this case $z$ is a negligible variable.
A brief mention to the 'moment method' in general NLSE
examined in the paper of García-Ripoll, J.J. and Pérez-Garc
ía, V.M. [6] will be done in the next paragraph.
§ FROM GSES TO A PSEUDO-GENERAL FORM OF THE NLSE -
STATIONARY CASE
We derive in the next sections a general choice for the
transformation functionals of the free fields in the GSEs which leads to a
named pseudo-general form of the NLSE in the stationary case (steady-state
case), showing that this choice include both Lapenta G.and Throumoulopoulos G.N. et al. choices and putting in
evidence that the same choices are compatible with one another.
§.§ Grad-Shafranov Equations set
The GSEs set, in the more general form for a fully ionized plasma $%
(n_{e}=Zn_{i})$, can be derive by both the Euler Equations in the
ideal limit for the ions and the electrons in a two-fluid treatment and the
Ampere's Law; in this respect, we take the two-fluid equations cited in
Lighthill, M.J. [13]$\ \ \ \ \ \ \ \ \ $
\begin{equation}
{\Large n}_{e}{\Large m}_{e}\left( \frac{{\LARGE \partial }}{{\LARGE %
\partial t}}\mathbf{v}_{e}+\mathbf{v}_{e}\cdot {\Large \nabla }\mathbf{v}%
_{e}\right) =-{\Large \nabla }{\large p}_{e}-\mathbf{M}-{\Large n}_{e}%
{\Large e(}\mathbf{E+v}_{e}\times \mathbf{B}{\Large )} \tag{2.1}
\label{2.1}
\end{equation}
$\ \ \ \ \ \ \ $
\begin{equation}
\ {\Large n}_{i}{\Large m}_{i}\left( \frac{{\LARGE \partial }}{{\LARGE %
\partial t}}\mathbf{v}_{i}+\mathbf{v}_{i}\cdot {\Large \nabla }\mathbf{v}%
_{i}\right) =-{\Large \nabla }{\large p}_{i}+\mathbf{M}+{\Large n}_{i}%
{\Large Ze(}\mathbf{E+v}_{i}\times \mathbf{B}{\Large )} \tag{2.2}
\label{2.2}
\end{equation}
where $\mathbf{M}$ is the rate of loss of electron
momentum, per unit volume, by collisions with ions; neglecting electrons and
ions inertia (non-inertial approximation) and adding both
corresponding sides of (2.1) and (2.2), we obtain
\begin{eqnarray*}
\mathbf{0} &=&-{\Large \nabla }{\large p}_{e}-{\Large \nabla }{\large p}_{i}-%
{\Large n}_{e}{\Large e(}\mathbf{E}+\mathbf{v}_{e}\times \mathbf{B}{\Large )}%
+{\Large n}_{i}{\Large Ze(}\mathbf{E}+\mathbf{v}_{i}\times \mathbf{B}{\Large %
)}= \\
&=&-{\Large \nabla }{\large p}+{\large n}_{e}{\large e(}\mathbf{v}_{i}-%
\mathbf{v}_{e}{\large )}\times \mathbf{B=}-{\Large \nabla }{\large p}+%
\mathbf{J}\times \mathbf{B}
\end{eqnarray*}
which lead to
\begin{equation}
{\Large \nabla }{\large p\ }=\mathbf{J}\times \mathbf{B} \tag{2.3}
\label{2.3}
\end{equation}
being the current density
\begin{equation*}
\mathbf{J}={\large n}_{e}{\large e(}\mathbf{v}_{i}-\mathbf{v}_{e}{\large )}
\end{equation*}
At this point the calculus is the same of Lapenta
's work. Assuming a plane framework $(x,y)$ because the
analysis is made on the equatorial plane of an accretion
disc,we can take as a particular solution of (2.3) [12]
\begin{equation}
\mathbf{B}=\widehat{\mathbf{z}}\times {\Large \nabla }\Psi +B_{z}\widehat{%
\mathbf{z}} \tag{2.4} \label{2.4}
\end{equation}
The Ampere's Law lead by the (2.4) to
\begin{eqnarray*}
\mathbf{J} &=&{\Large \nabla }\times \mathbf{B}=\widehat{\mathbf{z}}\left(
{\Large \nabla }\cdot {\Large \nabla }\Psi \right) -{\Large \nabla }\Psi
\left( {\Large \nabla }\cdot \widehat{\mathbf{z}}\right) +\left( {\Large %
\nabla }\Psi \cdot {\Large \nabla }\right) \widehat{\mathbf{z}}+ \\
&&-\left( \widehat{\mathbf{z}}\cdot {\Large \nabla }\right) {\Large \nabla }%
\Psi +B_{z}\left( {\Large \nabla }\times \widehat{\mathbf{z}}\right) -%
\widehat{\mathbf{z}}\times {\Large \nabla }B_{z}
\end{eqnarray*}
which delivers, being $z$ a negligible variable,
\begin{equation}
\mathbf{J}=\widehat{\mathbf{z}}{\Large \nabla }^{2}\Psi -\widehat{\mathbf{z}}%
\times {\Large \nabla }B_{z} \tag{2.5} \label{2.5}
\end{equation}
If we use the current density expression (2.5) and the solution
(2.4) in (2.3), we obtain the equation
\begin{eqnarray*}
{\Large \nabla }{\large p\ } &=&\left( \widehat{\mathbf{z}}{\Large \nabla }%
^{2}\Psi -\widehat{\mathbf{z}}\times {\Large \nabla }B_{z}\right) \times
\left( \widehat{\mathbf{z}}\times {\Large \nabla }\Psi +B_{z}\widehat{%
\mathbf{z}}\right) = \\
&=&\widehat{\mathbf{z}}{\Large \nabla }^{2}\Psi \times \left( \widehat{%
\mathbf{z}}\times {\Large \nabla }\Psi \right) -\left( \widehat{\mathbf{z}}%
\times {\Large \nabla }B_{z}\right) \times \left( \widehat{\mathbf{z}}\times
{\Large \nabla }\Psi \right) -\left( \widehat{\mathbf{z}}\times {\Large %
\nabla }B_{z}\right) \times B_{z}\widehat{\mathbf{z}}= \\
&=&{\Large \nabla }^{2}\Psi \left( \widehat{\mathbf{z}}\widehat{\mathbf{z}}%
\cdot {\Large \nabla }\Psi -{\Large \nabla }\Psi \widehat{\mathbf{z}}\cdot
\widehat{\mathbf{z}}\right) -\left[ \left( \widehat{\mathbf{z}}\times
{\Large \nabla }B_{z}\cdot {\Large \nabla }\Psi \right) \widehat{\mathbf{z}}%
-\left( \widehat{\mathbf{z}}\times {\Large \nabla }B_{z}\cdot \widehat{%
\mathbf{z}}\right) {\Large \nabla }\Psi \right] + \\
&&+B_{z}\left( \widehat{\mathbf{z}}\widehat{\mathbf{z}}\cdot {\Large \nabla }%
B_{z}-{\Large \nabla }B_{z}\widehat{\mathbf{z}}\cdot \widehat{\mathbf{z}}%
\right)
\end{eqnarray*}
which delivers
\begin{equation}
{\Large \nabla }{\large p\ }+{\Large \nabla }^{2}\Psi {\Large \nabla }\Psi
+\left( \widehat{\mathbf{z}}\times {\Large \nabla }B_{z}\cdot {\Large \nabla
}\Psi \right) \widehat{\mathbf{z}}+B_{z}{\Large \nabla }B_{z}=\mathbf{0}
\tag{2.6} \label{2.6}
\end{equation}
Observing that equation (2.6) is made of two parts linearly
independent among them, from this one it follows the equations system
\begin{equation*}
\left( \widehat{\mathbf{z}}\times {\Large \nabla }B_{z}\cdot {\Large \nabla }%
\Psi \right) \widehat{\mathbf{z}}=\mathbf{0}
\end{equation*}
\begin{equation*}
{\Large \nabla }{\large p\ }+{\Large \nabla }^{2}\Psi {\Large \nabla }\Psi
+B_{z}{\Large \nabla }B_{z}=\mathbf{0}
\end{equation*}
which represents the 'equivalent GSEs set', or
\begin{equation}
{\Large \nabla }\Psi \times {\Large \nabla }B_{z}\cdot \widehat{\mathbf{z}}=0
\tag{2.7} \label{2.7}
\end{equation}
\begin{equation}
{\Large \nabla }{\large p\ }+{\Large \nabla }^{2}\Psi {\Large \nabla }\Psi
+B_{z}{\Large \nabla }B_{z}=\mathbf{0} \tag{2.8} \label{2.8}
\end{equation}
where the equation (2.7) derives from
\begin{equation}
\widehat{\mathbf{z}}\times {\Large \nabla }B_{z}\cdot {\Large \nabla }\Psi =0
\tag{2.9} \label{2.9}
\end{equation}
The equations (2.7) and (2.8) are the vectorial-scalar form system
of the GSE for the ideal magnetohydrodynamical equilibria in a
two-dimensional framework and they lead to a fundamental 2nd-order PDEs
system for searching the steady-state solution of the stationary problem for
a fully ionized plasma (astrophysical plasma).
§.§.§ Canonical form of the GSE
We can observe that the canonical form of the GSE
derives from the system (2.7),(2.8) as it follows: by the equation (2.7) we
have necessarily
\begin{equation}
{\Large \nabla }\Psi \times {\Large \nabla }B_{z}=\mathbf{0}\text{ \ \ \ }%
\forall (x,y)\in Dom(\Psi )\cap Dom(B_{z}) \tag{2.10} \label{2.10}
\end{equation}
because ${\Large \nabla }\Psi \times {\Large \nabla }B_{z}$
is on the $z$ axis; this one delivers a linear dependence between $%
{\Large \nabla }\Psi $ and$\ {\Large \nabla }B_{z}$ as it
\begin{equation}
{\Large \nabla }\Psi =\alpha (\Psi ,B_{z}){\Large \nabla }B_{z} \tag{2.11}
\label{2.11}
\end{equation}
where we suppose that $\alpha $ is a rational real
functional of $\Psi $ and ${\large B}_{z}$ defined in $%
Dom(\Psi )\cap Dom(B_{z})$; therefore we obtain
by the (2.8)
\begin{equation*}
{\Large \nabla }{\large p\ }+{\Large \nabla }^{2}\Psi \alpha (\Psi ,B_{z})%
{\Large \nabla }B_{z}+B_{z}{\Large \nabla }B_{z}=\mathbf{0}
\end{equation*}
which delivers
\begin{equation}
{\Large \nabla }{\large p+}\left( {\Large \nabla }^{2}\Psi \alpha (\Psi
,B_{z})+B_{z}\right) {\Large \nabla }B_{z}=\mathbf{0} \tag{2.12}
\label{2.12}
\end{equation}
this one means that ${\Large \nabla }B_{z}$ and ${\Large %
\nabla }{\large p}$ are linearly dipendent among them, or they are
co-axial vectors; furthermore, by applying ${\Large \nabla \Psi \times }$
to this one, we obtain
\begin{equation}
{\large \nabla \Psi \times \nabla p}=\mathbf{0}\text{ \ \ \ \ }\forall
(x,y)\in \left( Dom(\Psi )\cap Dom({\large p})\right) \tag{2.13}
\label{2.13}
\end{equation}
which means that ${\Large \nabla \Psi }$ and ${\Large %
\nabla }{\large p}$ are necessarily linearly dependent as in
(2.10); thus determines
\begin{equation}
{\large \nabla \Psi }=\beta (\Psi ,{\large p}){\Large \nabla }{\large p}
\tag{2.14} \label{2.14}
\end{equation}
where we suppose that $\beta $ is a rational real
functional of $\Psi $ and ${\large p}$ defined in $%
Dom(\Psi )\cap Dom({\large p})$; now, we can observe that (2.14)
lead to the differential equation
\begin{equation*}
\mathbf{0}={\Large \nabla }\Psi -\beta (\Psi ,{\large p}){\Large \nabla }%
{\large p}=\widehat{D}(\Psi ,{\large p})
\end{equation*}
which is
\begin{equation}
\mathbf{0}=\widehat{D}(\Psi ,{\large p})\text{\ \ \ }\forall (x,y)\in \left(
Dom(\Psi )\cap Dom({\large p})\right) \tag{2.15} \label{2.15}
\end{equation}
where we suppose that $\widehat{D}$ is a real differential
functional dependent on $\beta $ or
\begin{equation*}
\widehat{D}(\Psi ,{\large p})(x,y)\in
%TCIMACRO{\U{211d} }%
\mathbb{R}
\text{ \ \ \ \ }\forall (x,y)\in Dom(\Psi )\cap Dom({\large p})
\end{equation*}
a general integral of (2.15) can be written as
\begin{equation}
{\large p}=F_{\beta }(\Psi )\ \ \ \ \forall (x,y)\in \Omega _{\beta
}\subseteq \left( Dom(\Psi )\cap Dom({\large p})\right) \tag{2.16}
\label{2.16}
\end{equation}
where $F_{\beta }$ is a real functional dependent on $%
\beta $ what is defined in $\Omega _{\beta }$;
If we take the equation (2.11) instead, we have a differential
equation equal to
\begin{equation*}
\mathbf{0}={\Large \nabla }\Psi -\alpha (\Psi ,B_{z}){\Large \nabla }B_{z}=%
\widetilde{D}(\Psi ,B_{z})
\end{equation*}
which is
\begin{equation}
\mathbf{0}=\widetilde{D}(\Psi ,B_{z})\text{\ \ \ }\forall (x,y)\in \left(
Dom(\Psi )\cap Dom(B_{z})\right) \tag{2.17} \label{2.17}
\end{equation}
where we suppose, as above, that $\widetilde{{\large D}}$
is a differential functional dependent on $\alpha $; a general
integral of (2.17) therefore can be written as
\begin{equation}
B_{z}=F_{\alpha }(\Psi )\text{ \ \ \ \ }\forall (x,y)\in \Omega _{\alpha
}\subseteq \left( Dom(\Psi )\cap Dom(B_{z})\right) \tag{2.18} \label{2.18}
\end{equation}
where $F_{\alpha }$ is a real functional dependent on $%
\alpha $ and it's defined in $\Omega _{\alpha }$;
now, if $\widehat{b}\neq \overrightarrow{\mathbf{0}}$ is
the unit vector for both ${\Large \nabla }B_{z}$ and ${\Large %
\nabla }{\large p}$, from (2.12) it follows
\begin{eqnarray*}
\left\Vert {\Large \nabla }{\large p}\right\Vert \widehat{b}+\left( {\Large %
\nabla }^{2}\Psi \alpha (\Psi ,B_{z})+B_{z}\right) \left\Vert {\Large \nabla
}B_{z}\right\Vert \widehat{b} &=&\mathbf{0} \\
\left[ \left\Vert {\Large \nabla }{\large p}\right\Vert +\left( {\Large %
\nabla }^{2}\Psi \alpha (\Psi ,B_{z})+B_{z}\right) \left\Vert {\Large \nabla
}B_{z}\right\Vert \right] \widehat{b} &=&\mathbf{0} \\
\left\Vert {\Large \nabla }{\large p}\right\Vert +\left( {\Large \nabla }%
^{2}\Psi \alpha (\Psi ,B_{z})+B_{z}\right) \left\Vert {\Large \nabla }%
B_{z}\right\Vert &=&0
\end{eqnarray*}
which lead to
\begin{equation}
{\Large \nabla }^{2}\Psi =\left( -\left\Vert {\Large \nabla }{\large p}%
\right\Vert \left\Vert {\Large \nabla }B_{z}\right\Vert ^{-1}-B_{z}\right)
\alpha (\Psi ,B_{z})^{-1} \tag{2.19} \label{2.19}
\end{equation}
and taking in account (2.16) and (2.18) in this one, we obtain
finally the GSE in canonical form
\begin{equation}
{\Large \nabla }^{2}\Psi =\left( -\left\Vert {\Large \nabla }F_{\beta }(\Psi
)\right\Vert \left\Vert {\Large \nabla }F_{\alpha }(\Psi )\right\Vert
^{-1}-F_{\alpha }(\Psi )\right) \alpha (\Psi ,F_{\alpha }(\Psi ))^{-1}
\tag{2.20} \label{2.20}
\end{equation}
in which we can recognize an Helmholtz Equation. In this one we
suppose that
\begin{equation*}
\underset{(x,y)\rightarrow (\overline{x},\overline{y})}{\lim }\frac{%
\left\Vert {\Large \nabla }F_{\beta }(\Psi )\right\Vert }{\left\Vert {\Large %
\nabla }F_{\alpha }(\Psi )\right\Vert }\in
%TCIMACRO{\U{211d} }%
\mathbb{R}
\text{,\ \ \ }\forall (\overline{x},\overline{y})\in \Omega _{\alpha }\cap
\Omega _{\beta }:\left\Vert {\Large \nabla }F_{\alpha }(\Psi )\right\Vert =0
\end{equation*}
because the flux function $\Psi $ has not any pole insidethe equilibria domain $\Omega _{\alpha }\cap \Omega _{\beta }$.
§.§ A general choice in the complex plane for the free fields
A pseudo-general form of the NLSE show at first view a more great
complexity than an ordinary cubic NLSE cited in Lapenta's paper but on the other hand its mathematical morphology is
affected by a more general set of solutions, which can represent a
wide-spectrum of possibilities for the research of explanations in the
physical framework for the above equilibrium structures. Furthermore, the
general form of the NLSE cited below includes both Lapenta
choice [8] and Throumoulopoulos et al. choice <cit.> (this last leads to a pseudo-cubic NLSE); in this way, several
specialized choices set for a same physical problem can be put in a more
general mathematical solving context from the point of view of the related
non-linear problem. In this case, such a strategy therefore permits us to
define a valid vay for to find a more general possible solving method for a
GSEs set in object, working in the framework of the solitonic solutions as
in the interesting idea of Lapenta.
A pseudo-general form of the NLSE, in the time-dependent case, can
be represented as it follows [6]
\begin{equation}
{\large i}\frac{\partial \Psi }{\partial t}{\large =\Delta }_{3}{\large \Psi
+g}\left( \left\vert \Psi \right\vert ^{2},t\right) {\large \Psi +i\sigma }%
\left( \left\vert \Psi \right\vert ^{2},t\right) {\large \Psi } \tag{2.21}
\label{2.21}
\end{equation}
where ${\large \Psi }$ is a complex function, ${\Large %
\Delta }_{3}$ is the Laplace operator in the $(x,y,z)$ space
framework and both $g$ and$\ \sigma $ are
supposed to be time-dependent potential complex functionals of $\left\vert
\Psi \right\vert ^{2}$. It's remarkable the fact that this equation
is named 'pseudo-general' because the functions $g$ and$\ \sigma
$ are complex-valued functionals instead that real-valued
functionals as is usually. However, it's clear that
(2.21) can be put 'always' in a form in which $g$ and$\ \sigma $
lead to a pair of real-valued functionals. Indeed
\begin{eqnarray*}
{\large g\Psi +i\sigma \Psi } &{\large =}&\left( {\large g}_{R}+i{\large g}%
_{i}\right) {\large \Psi +i}\left( {\large \sigma }_{R}+i{\large \sigma }%
_{i}\right) {\large \Psi =} \\
&{\large =}&\left( {\large g}_{R}-{\large \sigma }_{i}\right) {\large \Psi +i%
}\left( {\large g}_{i}+{\large \sigma }_{R}\right) {\large \Psi =}\text{ }%
\widetilde{{\large g}}{\large \Psi +i}\widetilde{{\large \sigma }}{\large %
\Psi }
\end{eqnarray*}
where $\widetilde{{\large g}}$ and $\widetilde{{\large %
\sigma }}$ are obviously real functionals. In the stationary case,
this equation leads to the steady-state form
\begin{equation}
{\large \Delta }_{3}{\large \Psi +}\left( {\large g}\left( \left\vert \Psi
\right\vert ^{2}\right) \text{ }{\large +}\text{ }{\large i\sigma }\left(
\left\vert \Psi \right\vert ^{2}\right) \right) {\large \Psi =0} \tag{2.22}
\label{2.22}
\end{equation}
which is, as we see, a more general form of the Helmholtz Equation;
it can be viewed as [7]
\begin{equation}
\left( {\large \Delta }_{3}{\large +\varpi }^{2}\left( \left\vert \Psi
\right\vert ^{2}\right) \right) {\large \Psi =0} \tag{2.23} \label{2.23}
\end{equation}
\begin{equation}
{\large \varpi }^{2}\left( \left\vert \Psi \right\vert ^{2}\right) ={\large g%
}\left( \left\vert \Psi \right\vert ^{2}\right) {\large +i\sigma }\left(
\left\vert \Psi \right\vert ^{2}\right) \tag{2.24} \label{2.24}
\end{equation}
is a complex-valued function. This non-linear PDE therefore
represents the fundamental link between the GSE and the pseudo-general NLSE
in the stationary case for a three-dimensional equilibrium problem. In the
two-dimensional framework $(x,y)$, that is the framework in which
we analize the localized accretion plasma structures, we have
therefore the differential equation
\begin{equation}
\left( {\large \Delta }_{2}{\large +\varpi }^{2}\left( \left\vert \Psi
\right\vert ^{2}\right) \right) {\large \Psi =0}\text{ \ \ {\large that is \
\ }}{\Large \nabla }_{{\small 2}}^{2}{\large \Psi =-\varpi }^{2}\left(
\left\vert \Psi \right\vert ^{2}\right) {\large \Psi } \tag{2.25}
\label{2.25}
\end{equation}
At this point, we observe that an appropriate choice for the free
fields in the GSEs can permit the translating of such equilibrium equations
in a certain form of the NLSE, as the general stationary form (2.22), which
could be an excellent solution for re-interpretate localized structures in
the accretion discs as 'solitonic structures', going as viewed by the
Helmholtz Equation. Indeed we'll see that a general transformations set for
the free fields ${\Large p}$ and ${\Large B}_{z}$ can
lead from the equations (2.7) and (2.8) to the equation (2.25) by an
appropriate expression for the ${\large \varpi }^{2}$ function on
the same ${\Large p}$ and ${\Large B}_{z}$. It's important
to observe that the 'analiticity' of the complex function $\Psi $
is not required for solving the Helmholtz Equation in general [7].
Furthermore, I cite that the 'moment method' [6] can
be used as a valid approximation analytical way for to solve a wide family
of non-linear wave equations of NLSE type in its pseudo-general form; this
method have been developed for n-dimensional cases in general by
García-Ripoll, J.J. and Pérez-García, V.M.
. So, it's remarkable that by this mathematical approximation
strategy it could be possible to identify a good set of physically coherent
solutions of the GSE for the magnetohydrodynamical equilibrium problem, i.e.
solitonic solutions as we would like.
§.§.§ Connection between GSEs and a pseudo-general form of the NLSE
Now, we consider the expressions (2.16) and (2.18) in the
intersection of the $\Psi ,{\large B}_{z}$ and ${\large p}$
\begin{equation*}
{\large p}=F_{\beta }(\Psi )\text{ \ \ \ \ }B_{z}=F_{\alpha }(\Psi )\text{ \
\ \ \ \ }\forall (x,y)\in \Omega _{\alpha }\cap \Omega _{\beta }
\end{equation*}
and replacing the free fields ${\large p}$ and ${\large B}%
_{z}$ in the equation (2.8) with this relations we obtain
\begin{equation*}
{\Large \nabla }F_{\beta }(\Psi ){\large \ }+{\Large \nabla }^{2}\Psi
{\Large \nabla }\Psi +F_{\alpha }(\Psi ){\Large \nabla }F_{\alpha }(\Psi )=%
\mathbf{0}
\end{equation*}
obviously for the linear dependence of ${\Large \nabla }{\large p\ }%
+B_{z}{\Large \nabla }B_{z}$ by ${\Large \nabla }^{2}\Psi {\Large %
\nabla }\Psi $ this one delivers
\begin{equation}
{\Large \nabla }F_{\beta }(\Psi ){\large \ }+F_{\alpha }(\Psi ){\Large %
\nabla }F_{\alpha }(\Psi )={\Large k}\left( \Psi \right) {\Large \nabla }%
\Psi =-{\Large \nabla }^{2}\Psi {\Large \nabla }\Psi \tag{2.26}
\label{2.26}
\end{equation}
\begin{eqnarray}
\left( {\Large \nabla }^{2}\Psi +{\Large k}\left( \Psi \right) \right)
{\Large \nabla }\Psi &=&\mathbf{0}\text{ \ \ \ \ } \TCItag{2.27}
\label{2.27} \\
\text{{\large which is} \ \ \ \ }{\Large \nabla }^{2}\Psi +{\Large k}\left(
\Psi \right) &=&0\text{ \ \ \ }\forall (x,y)\in \Omega _{\alpha }\cap \Omega
_{\beta } \notag
\end{eqnarray}
that is exactly the Helmholtz Equation, in which we suppose that $%
{\large k}$ is a generic functional of $\Psi $; if we
compare the second equation in (2.27) with the Helmholtz Equation (2.25) and
we consider the expression (2.24), then we obtain necessarily
\begin{equation}
{\Large k}\left( \Psi \right) ={\large \varpi }^{2}\left( \left\vert \Psi
\right\vert ^{2}\right) {\large \Psi }=\left( {\large g}\left( \left\vert
\Psi \right\vert ^{2}\right) {\large +i\sigma }\left( \left\vert \Psi
\right\vert ^{2}\right) \right) {\large \Psi }\text{ \ \ \ }\forall (x,y)\in
\Omega _{\alpha }\cap \Omega _{\beta } \tag{2.28} \label{2.28}
\end{equation}
which indicates that ${\Large k}$ is a complex functional
and $\Psi $ is clearly a complex function by the equation (2.27),
because by this one it derives
\begin{eqnarray}
{\Large \nabla }^{2}\Psi &=&-{\Large k}\left( \Psi \right) =-\left( {\large g%
}\left( \left\vert \Psi \right\vert ^{2}\right) {\large +i\sigma }\left(
\left\vert \Psi \right\vert ^{2}\right) \right) \Psi \notag \\
\text{{\large or} \ \ \ \ \ }\Psi ^{-1}{\Large \nabla }^{2}\Psi &=&-\left(
{\large g}\left( \left\vert \Psi \right\vert ^{2}\right) {\large +i\sigma }%
\left( \left\vert \Psi \right\vert ^{2}\right) \right) \text{ \ \ \ {\large %
with\ \ }}{\large g},{\large \sigma \neq }\text{ }0 \TCItag{2.29}
\label{2.29}
\end{eqnarray}
At this point, taking into account the Lapentaand Throumoulopoulos et al.'s works [8], [9] and
starting from their fields choices, I found a well posed pair of
transformation functionals like this
\begin{equation}
{\large B}_{z}{\large \nabla B}_{z}=\frac{1}{2}{\large \nabla B}_{z}^{2}=%
\underset{i,j}{\sum }a_{ij}\overline{\Psi }^{i}\Psi ^{j+1}{\Large \nabla }%
\Psi \tag{2.30} \label{2.30}
\end{equation}
\begin{equation}
{\large \nabla p}=\underset{i,j}{\sum }b_{ij}\overline{\Psi }^{i}\Psi ^{j+1}%
{\Large \nabla }\Psi \tag{2.31} \label{2.31}
\end{equation}
where $\overline{\Psi }$ is the complex conjugate function
related to the complex flux function $\Psi $ and $a_{ij},b_{ij}$
are complex constants; we observe that (2.30) and (2.31)
are respectively a term proportional to the magnetic force density and the
pure hydrodynamical pressure force density, which are two 'real' physical
variables; this is for me a fundamental physical condition and it's
important that it doesn't affect the complexity of the flux function $\Psi $
; we will explore this 'reality condition' in the next
section as a 'validity condition' for (2.30) and (2.31). From these
transformations and the first equation in (2.26) it derives
\begin{equation}
{\Large \nabla }{\large p\ }+B_{z}{\Large \nabla }B_{z}=\underset{i,j}{\sum }%
\left( a_{ij}+b_{ij}\right) \overline{\Psi }^{i}\Psi ^{j+1}{\Large \nabla }%
\Psi ={\Large k}\left( \Psi \right) {\Large \nabla }\Psi \tag{2.32}
\label{2.32}
\end{equation}
which together with (2.28) delivers
\begin{eqnarray}
{\Large k}\left( \Psi \right) &=&\underset{i,j}{\sum }\left(
a_{ij}+b_{ij}\right) \overline{\Psi }^{i}\Psi ^{j+1}= \TCItag{2.33}
\label{2.33} \\
&=&\left( {\large g}\left( \left\vert \Psi \right\vert ^{2}\right) {\large %
+i\sigma }\left( \left\vert \Psi \right\vert ^{2}\right) \right) \Psi \text{
\ \ \ }\forall (x,y)\in \Omega _{\alpha }\cap \Omega _{\beta } \notag
\end{eqnarray}
Now, it's interesting to observe that the fields
transformations (2.30), (2.31) lead from equations (2.7),(2.8) to the
pseudo-general form (2.25) of the NLSE in a space framework $(x,y)$
by the equation
\begin{eqnarray}
\underset{i,j}{\sum }\gamma _{ij}\overline{\Psi }^{i}\Psi ^{j} &=&{\large g}%
\left( \left\vert \Psi \right\vert ^{2}\right) {\large +i\sigma }\left(
\left\vert \Psi \right\vert ^{2}\right) \text{ \ \ \ }\forall (x,y)\in
\Omega _{\alpha }\cap \Omega _{\beta } \TCItag{2.34} \label{2.34} \\
\text{{\large where} \ \ \ \ \ }\gamma _{ij} &=&a_{ij}+b_{ij}\in
%TCIMACRO{\U{2102} }%
\mathbb{C}
\text{ \ \ {\large and}\ \ }\func{Im}\left( {\large \Psi }\right) \subset
%TCIMACRO{\U{2102} }%
\mathbb{C}
\notag
\end{eqnarray}
Because of the second equation in (2.29), we will proof below that,
if $\Psi $ is a complex function, $\underset{i,j}{\sum }\gamma
_{ij}\overline{\Psi }^{i}\Psi ^{j}$ can be put in the form
\begin{equation*}
{\large g}\left( \left\vert \Psi \right\vert ^{2}\right) {\large +i\sigma }%
\left( \left\vert \Psi \right\vert ^{2}\right)
\end{equation*}
where both $g$ and$\ \sigma $ are supposed to be
complex functionals as above.
In this regard, we consider the expression at left hand
side of (2.34) in $\Omega _{\alpha }\cap \Omega _{\beta }$ domain
\begin{equation}
\underset{i,j}{\sum }\gamma _{ij}\overline{\Psi }^{i}\Psi ^{j} \tag{2.35}
\label{2.35}
\end{equation}
and if we develop the sum, we obtain the next identities chain
\begin{eqnarray*}
\underset{i,j}{\sum }\gamma _{ij}\overline{\Psi }^{i}\Psi ^{j} &=&\underset{j%
}{\sum }\gamma _{0j}\Psi ^{j}+\underset{j}{\sum }\gamma _{1j}\overline{\Psi }%
\Psi ^{j}+\underset{j}{\sum }\gamma _{2j}\overline{\Psi }^{2}\Psi ^{j}+...=
\\
&=&\underset{j}{\sum }\gamma _{0j}\Psi ^{j}+\underset{j=1}{\overset{...}{%
\sum }}\gamma _{1j}\left\vert \Psi \right\vert ^{2}\Psi ^{j-1}+\underset{j=2}%
{\overset{...}{\sum }}\gamma _{2j}\left( \left\vert \Psi \right\vert
^{2}\right) ^{2}\Psi ^{j-2}+...=
\end{eqnarray*}
\begin{equation*}
=\left( \gamma _{00}+\gamma _{10}\left\vert \Psi \right\vert ^{2}+\gamma
_{20}\left( \left\vert \Psi \right\vert ^{2}\right) ^{2}+...\right) +
\end{equation*}
\begin{equation*}
+\left( \gamma _{01}+\gamma _{11}\left\vert \Psi \right\vert ^{2}+\gamma
_{21}\left( \left\vert \Psi \right\vert ^{2}\right) ^{2}+...\right) \Psi
\end{equation*}
\begin{eqnarray*}
&=&\underset{i}{\sum }\gamma _{i0}\left( \left\vert \Psi \right\vert
^{2i}\right) \Psi ^{0}+\underset{i}{\sum }\gamma _{i1}\left( \left\vert \Psi
\right\vert ^{2i}\right) \Psi ^{1}+\underset{i}{\sum }\gamma _{i2}\left(
\left\vert \Psi \right\vert ^{2i}\right) \Psi ^{2}+...= \\
&=&\underset{j}{\sum }\left( \underset{i}{\sum }\gamma _{ij}\left(
\left\vert \Psi \right\vert ^{2i}\right) \right) \Psi ^{j}=\underset{j}{\sum
}f_{j}\left( \left\vert \Psi \right\vert ^{2}\right) \Psi ^{j}
\end{eqnarray*}
which deliver
\begin{equation}
\underset{i,j}{\sum }\gamma _{ij}\overline{\Psi }^{i}\Psi ^{j}=\underset{j}{%
\sum }f_{j}\left( \left\vert \Psi \right\vert ^{2}\right) \Psi ^{j}
\tag{2.36} \label{2.36}
\end{equation}
\begin{equation}
f_{j}\left( \left\vert \Psi \right\vert ^{2}\right) =\underset{i}{\sum }%
\gamma _{ij}\left( \left\vert \Psi \right\vert ^{2i}\right) \tag{2.37}
\label{2.37}
\end{equation}
is a complex polynomial functional on $\left\vert \Psi \right\vert
^{2i}$. Now, we 'impose' the complexity of the flux function $\Psi $
($\func{Im}\left( {\large \Psi }\right) \subset
%TCIMACRO{\U{2102} }%
\mathbb{C}
$) and we write therefore
\begin{equation}
\Psi (x,y)=u(x,y)+iv(x,y)\text{ \ \ \ }\forall (x,y)\in \Omega _{\alpha
}\cap \Omega _{\beta } \tag{2.38} \label{2.38}
\end{equation}
where $u$ and $v$ are generic real-valued
functions in the two-dimensional framework $(x,y)$; from this one
and (2.36) we obtain
\begin{eqnarray}
\underset{i,j}{\sum }\gamma _{ij}\overline{\Psi }^{i}\Psi ^{j}\underset{j}{%
=\sum }f_{j}\left( \left\vert \Psi \right\vert ^{2}\right) \left(
u+iv\right) ^{j} &=& \TCItag{2.39} \label{2.39} \\
\underset{j}{=\sum }f_{j}\left( \left\vert \Psi \right\vert ^{2}\right)
\underset{k=0}{\overset{j}{\sum }}\binom{j}{k}u^{j-k}(iv)^{k} &=&\underset{j}%
{\text{ }\sum }f_{j}\left( \left\vert \Psi \right\vert ^{2}\right) \underset{%
k=0}{\overset{j}{\sum }}\zeta _{jk}u^{j-k}(iv)^{k} \notag
\end{eqnarray}
in which we used the 'Newton binomial formula' and $\zeta _{jk}=%
\binom{j}{k}$ are the binomial coefficients.
At this point, we fix the maximum level $N$ of the
polynomial sum in (2.39) and we observe that
\begin{eqnarray}
\underset{j=0}{\text{ }\overset{N}{\sum }}f_{j}\left( \left\vert \Psi
\right\vert ^{2}\right) \underset{k=0}{\overset{j}{\sum }}\zeta
_{jk}u^{j-k}(iv)^{k} &=&\underset{j=0}{\text{ }\overset{N}{\sum }}\underset{%
k=0}{\overset{j}{\sum }}f_{j}\left( \left\vert \Psi \right\vert ^{2}\right)
\zeta _{jk}u^{j-k}(iv)^{k}= \notag \\
&=&\underset{j=0}{\text{ }\overset{N}{\sum }}\underset{k=0}{\overset{j}{\sum
}}\eta _{jk}\left( \left\vert \Psi \right\vert ^{2}\right) u^{j-k}(iv)^{k}
\TCItag{2.40} \label{2.40}
\end{eqnarray}
\begin{equation}
\text{{\large where \ \ \ \ \ }}\eta _{jk}\left( \left\vert \Psi \right\vert
^{2}\right) =f_{j}\left( \left\vert \Psi \right\vert ^{2}\right) \zeta _{jk}=%
\binom{j}{k}\underset{i}{\sum }\gamma _{ij}\left( \left\vert \Psi
\right\vert ^{2i}\right) \tag{2.41} \label{2.41}
\end{equation}
is a new complex polynomial functional; from (2.40) therefore it
\begin{equation*}
\underset{j=0}{\text{ }\overset{N}{\sum }}\underset{k=0}{\overset{j}{\sum }}%
\eta _{jk}\left( \left\vert \Psi \right\vert ^{2}\right) u^{j-k}(iv)^{k}=
\end{equation*}
\begin{equation*}
=\text{ }\underset{l=0}{\overset{N}{\sum }}\eta _{l0}\left( \left\vert \Psi
\right\vert ^{2}\right) u^{l}(iv)^{0}+\underset{l=0}{\overset{N-1}{\sum }}%
\eta _{\left( l+1\right) 1}\left( \left\vert \Psi \right\vert ^{2}\right)
u^{l}(iv)^{1}+\underset{l=0}{\overset{N-2}{\sum }}\eta _{\left( l+2\right)
2}\left( \left\vert \Psi \right\vert ^{2}\right) u^{l}(iv)^{2}+...
\end{equation*}
that is
\begin{equation*}
\underset{h=0}{\text{ }\overset{N/2}{\sum }}\left( \underset{l=0}{\overset{%
N-2h}{\sum }}\eta _{l\left( 2h\right) }\left( \left\vert \Psi \right\vert
^{2}\right) u^{l}\right) (iv)^{2h}+\underset{h=0}{\text{ }\overset{N/2-1}{%
\sum }}\left( \underset{l=0}{\overset{N-\left( 2h+1\right) }{\sum }}\eta
_{l\left( 2h+1\right) }\left( \left\vert \Psi \right\vert ^{2}\right)
u^{l}\right) (iv)^{2h+1}
\end{equation*}
if $N$ is 'even'; in clear complex form this one deliver,
\begin{equation}
\underset{i,j}{\sum }\gamma _{ij}\overline{\Psi }^{i}\Psi ^{j}= \tag{2.42}
\label{2.42}
\end{equation}
\begin{eqnarray*}
&=&\underset{h=0}{\text{ }\overset{\frac{N}{2}}{\sum }}\left( \underset{l=0}{%
\overset{N-2h}{\sum }}\eta _{l\left( 2h\right) }\left( \left\vert \Psi
\right\vert ^{2}\right) u^{l}\right) \left( -1\right) ^{h}v^{2h}+ \\
&&+\text{ }{\Large i}\underset{h=0}{\overset{\frac{N}{2}-1}{\sum }}\left(
\underset{l=0}{\overset{N-\left( 2h+1\right) }{\sum }}\eta _{l\left(
2h+1\right) }\left( \left\vert \Psi \right\vert ^{2}\right) u^{l}\right)
\left( -1\right) ^{h}v^{2h+1}
\end{eqnarray*}
and if it's compared with equation (2.34), taking into account
(2.41), it delivers finally the next expressions for the $g$ and$\
\sigma $ functionals
$\forall (x,y)\in \Omega _{\alpha }\cap \Omega _{\beta }$
\begin{equation*}
{\large g}\left( \left\vert \Psi \right\vert ^{2},u,v\right) =\underset{h=0}{%
\text{ }\overset{\frac{N}{2}}{\sum }}\left( \underset{l=0}{\overset{N-2h}{%
\sum }}\eta _{l\left( 2h\right) }\left( \left\vert \Psi \right\vert
^{2}\right) u^{l}\right) \left( -1\right) ^{h}v^{2h}=
\end{equation*}
\begin{equation}
=\underset{i}{\sum }\underset{h=0}{\overset{\frac{N}{2}}{\sum }}\underset{l=0%
}{\overset{N-2h}{\sum }}\binom{l}{2h}\gamma _{il}\left( \left\vert \Psi
\right\vert ^{2i}\right) u^{l}\left( -1\right) ^{h}v^{2h} \tag{2.43a}
\label{2.43a}
\end{equation}
\begin{equation*}
{\large \sigma }\left( \left\vert \Psi \right\vert ^{2},u,v\right) =\underset%
{h=0}{\overset{\frac{N}{2}-1}{\sum }}\left( \underset{l=0}{\overset{N-\left(
2h+1\right) }{\sum }}\eta _{l\left( 2h+1\right) }\left( \left\vert \Psi
\right\vert ^{2}\right) u^{l}\right) \left( -1\right) ^{h}v^{2h+1}=
\end{equation*}
\begin{equation}
=\underset{i}{\sum }\underset{h=0}{\overset{\frac{N}{2}}{\sum }}\underset{l=0%
}{\overset{N-\left( 2h+1\right) }{\sum }}\binom{l}{2h+1}\gamma _{il}\left(
\left\vert \Psi \right\vert ^{2i}\right) u^{l}\left( -1\right) ^{h}v^{2h+1}
\tag{2.44a} \label{2.44a}
\end{equation}
which are clearly two complex functionals on the square function $%
\left\vert \Psi \right\vert ^{2}=\overline{\Psi }\Psi $ and on the $%
u$ and $v$ functions; they are complex because the presence
of the complex coefficients $\gamma _{il}$ in the above
If the level of the polynomial sum $N$ is odd, we have
similarly the complex functionals
\begin{equation}
{\large g}\left( \left\vert \Psi \right\vert ^{2},u,v\right) =\underset{h=0}{%
\text{ }\overset{\frac{N-1}{2}}{\sum }}\left( \underset{l=0}{\overset{N-2h}{%
\sum }}\eta _{l\left( 2h\right) }\left( \left\vert \Psi \right\vert
^{2}\right) u^{l}\right) \left( -1\right) ^{h}v^{2h}= \notag
\end{equation}
\begin{equation}
=\underset{i}{\sum }\underset{h=0}{\text{ }\overset{\frac{N-1}{2}}{\sum }}%
\underset{l=0}{\overset{N-2h}{\sum }}\binom{l}{2h}\gamma _{il}\left(
\left\vert \Psi \right\vert ^{2i}\right) u^{l}\left( -1\right) ^{h}v^{2h}
\tag{2.43b} \label{2.43b}
\end{equation}
\begin{equation}
{\large \sigma }\left( \left\vert \Psi \right\vert ^{2},u,v\right) =\underset%
{h=0}{\overset{\frac{N-1}{2}}{\sum }}\left( \underset{l=0}{\overset{N-\left(
2h+1\right) }{\sum }}\eta _{l\left( 2h+1\right) }\left( \left\vert \Psi
\right\vert ^{2}\right) u^{l}\right) \left( -1\right) ^{h}v^{2h+1}= \notag
\end{equation}
\begin{equation}
=\underset{i}{\sum }\underset{h=0}{\text{ }\overset{\frac{N-1}{2}}{\sum }}%
\underset{l=0}{\overset{N-\left( 2h+1\right) }{\sum }}\binom{l}{2h+1}\gamma
_{il}\left( \left\vert \Psi \right\vert ^{2i}\right) u^{l}\left( -1\right)
^{h}v^{2h+1} \tag{2.44b} \label{2.44b}
\end{equation}
We observe that in these expressions the maximum level of polynomial
sum on $i$-index is not fixed, leaving in this way the
choice of dependence of the above functionals on the even powers of the
function $\left\vert \Psi \right\vert ^{2}$ completely free. Such a
dependence can be imposed on the base of mathematical or physical criteria
which are well defined if they are related to a specific physical problem.
it's clear that the general choice (2.30),(2.31) includes both
Lapenta [8] and Throumoulopoulos et al.
[9] choices; indeed for the first choice we have
\begin{equation}
a_{00}=\alpha _{0}^{2},\text{ \ \ \ }a_{ij}=0\text{ \ \ \ }\forall i,j>0
\tag{2.45a} \label{2.45a}
\end{equation}
\begin{equation}
b_{11}=\alpha _{0}^{2},\text{ \ \ \ }b_{ij}=0\text{ \ \ \ }\forall \left(
i,j\right) \neq \left( 1,1\right) \tag{2.45b} \label{2.45b}
\end{equation}
while for the second choice we have
\begin{equation}
a_{00}=\alpha _{0}^{2},\text{ \ \ \ }a_{ij}=0\text{ \ \ \ }\forall i,j>0
\tag{2.46a} \label{2.46a}
\end{equation}
\begin{equation}
b_{03}=\alpha _{0}^{2},\text{ \ \ \ }b_{ij}=0\text{ \ \ }\forall \left(
i,j\right) \neq \left( 0,3\right) \tag{2.46b} \label{2.46b}
\end{equation}
§ VALIDITY CONDITIONS IN THE MATHEMATICAL AND PHYSICAL FRAMEWORKS
In this section we explore the validity conditions for the
general choice (2.30),(2.31) from the physical and mathematical points of
view. We observe that these conditions are named 'critical conditions
' in the present paper because they are 'necessary' conditions for the
acceptability, or internal coherence, of the proposed mathematical models,
which are analized in respect of a related reasonable link between the
mathematical framework and the physical one. Furthermore, the analysis of
these theoretical positions will take into account necessary and sufficient
conditions too for the validity of unique solutions for the
magnetohydrodynamical equilibrium problem proposed.
§.§ On a critical condition for the general choice: mathematical
We consider the general choice for the free fields ${\large p}$
and ${\large B}_{z}$ in the GSEs set (2.7),(2.8)
\begin{equation}
{\large B}_{z}{\large \nabla B}_{z}=\frac{1}{2}{\large \nabla B}_{z}^{2}=%
\underset{i,j}{\sum }a_{ij}\overline{\Psi }^{i}\Psi ^{j+1}{\Large \nabla }%
\Psi \tag{3.1} \label{3.1}
\end{equation}
\begin{equation}
{\large \nabla p}=\underset{i,j}{\sum }b_{ij}\overline{\Psi }^{i}\Psi ^{j+1}%
{\Large \nabla }\Psi \tag{3.2} \label{3.2}
\end{equation}
\begin{equation*}
\text{{\large with \ \ \ }}a_{ij},b_{ij}\in
\mathbb{C}%
\end{equation*}
and for to establish a validity condition for the internal
mathematical coherence of these transformation functionals, we apply the
operator ${\large \nabla \times }$ at the both sides of
(3.1),(3.2), as in the Throumoulopoulos et al. work <cit.>, taking into account that they are 'similar' except for the coefficients
sets $\left\{ {\large a}_{ij}\right\} _{ij\in \text{ }%
%TCIMACRO{\U{2115} }%
\mathbb{N}
}$ and $\left\{ {\large b}_{ij}\right\} _{ij\in \text{ }%
%TCIMACRO{\U{2115} }%
\mathbb{N}
}$; for this reason, it will be sufficient therefore to consider for
both relations the term
\begin{equation}
\underset{i,j}{\sum }c_{ij}\overline{\Psi }^{i}\Psi ^{j+1}{\Large \nabla }%
\Psi \tag{3.3} \label{3.3}
\end{equation}
with $c_{ij}=a_{ij}$ or $b_{ij}$ as appropriate,
and to reduce in normal form the expression
\begin{equation}
{\large \nabla \times }\underset{i,j}{\sum }c_{ij}\overline{\Psi }^{i}\Psi
^{j+1}{\Large \nabla }\Psi \tag{3.4} \label{3.4}
\end{equation}
for to identify a validity condition which is unique for both field
transformations. At this point the calculus is quite simple: starting from
the expression (3.4) we obtain for (3.1) and (3.2)
\begin{equation}
{\large \nabla \times }\underset{i,j}{\sum }c_{ij}\overline{\Psi }^{i}\Psi
^{j+1}{\Large \nabla }\Psi =\mathbf{0}\text{ \ \ }\forall (x,y)\in \Omega
_{\alpha }\cap \Omega _{\beta } \tag{3.5} \label{3.5}
\end{equation}
\begin{equation*}
{\large \nabla \times \nabla p=\nabla \times }\left( \frac{1}{2}{\large %
\nabla B}_{z}^{2}\right) \text{ }{\large =}\text{ }\mathbf{0}
\end{equation*}
for the 'scalar' functions ${\large p}$ and ${\large B}%
_{z}^{2}$. If we proceed in the calculus we obtain
\begin{equation*}
\mathbf{0=}\text{ }{\large \nabla \times }\underset{i,j}{\sum }c_{ij}%
\overline{\Psi }^{i}\Psi ^{j+1}{\Large \nabla }\Psi =\underset{i,j}{\sum }%
c_{ij}{\large \nabla \times }\left( \overline{\Psi }^{i}\Psi ^{j+1}{\large %
\nabla }\Psi \right)
\end{equation*}
for the linearity of vectorial product; taking in account the
vectorial identity
\begin{equation*}
{\large \nabla \times }\left( \alpha \overrightarrow{a}\right) \text{ }%
{\large =\alpha \nabla \times }\overrightarrow{a}\text{ }{\large -}\text{ }%
\overrightarrow{a}{\large \times \nabla \alpha }\text{ \ \ \ with{\large \ }}%
\alpha \text{ as a scalar variable,}
\end{equation*}
we have therefore (we remember that $\Psi $ is a scalar
\begin{eqnarray*}
\mathbf{0} &\mathbf{=}&\text{ }\underset{i,j}{\sum }c_{ij}{\large \nabla
\times }\left( \overline{\Psi }^{i}\Psi ^{j+1}{\large \nabla }\Psi \right) =%
\underset{i,j}{\sum }c_{ij}\left( \overline{\Psi }^{i}\Psi ^{j+1}{\large %
\nabla \times \nabla }\Psi -{\large \nabla }\Psi \times {\large \nabla }%
\left( \overline{\Psi }^{i}\Psi ^{j+1}\right) \right) = \\
&=&-\underset{i,j}{\text{ }\sum }c_{ij}{\large \nabla }\Psi \times {\large %
\nabla }\left( \overline{\Psi }^{i}\Psi ^{j+1}\right) =-\underset{i,j}{\text{
}\sum }c_{ij}{\large \nabla }\Psi \times \left( \Psi ^{j+1}{\large \nabla }%
\overline{\Psi }^{i}+\overline{\Psi }^{i}{\large \nabla }\Psi ^{j+1}\right) =
\\
&=&-\underset{i,j}{\text{ }\sum }c_{ij}{\large \nabla }\Psi \times \left(
\Psi ^{j+1}i\overline{\Psi }^{i-1}{\large \nabla }\overline{\Psi }+\overline{%
\Psi }^{i}(j+1)\Psi ^{j}{\large \nabla }\Psi \right) = \\
&=&-\underset{i,j}{\text{ }\sum }c_{ij}\left( \Psi ^{j+1}i\overline{\Psi }%
^{i-1}{\large \nabla }\Psi \times {\large \nabla }\overline{\Psi }+\overline{%
\Psi }^{i}(j+1)\Psi ^{j}{\large \nabla }\Psi \times {\large \nabla }\Psi
\right) = \\
&=&-\underset{i,j}{\text{ }\sum }c_{ij}\Psi ^{j+1}i\overline{\Psi }^{i-1}%
{\large \nabla }\Psi \times {\large \nabla }\overline{\Psi }=-\text{ }%
J\left( \Psi ,\left\vert \Psi \right\vert ^{2}\right) {\large \nabla }\Psi
\times {\large \nabla }\overline{\Psi }
\end{eqnarray*}
\begin{equation}
J\left( \Psi ,\left\vert \Psi \right\vert ^{2}\right) {\large \nabla }\Psi
\times {\large \nabla }\overline{\Psi }=\mathbf{0}\text{ \ \ }\forall
(x,y)\in \Omega _{\alpha }\cap \Omega _{\beta } \tag{3.6} \label{3.6}
\end{equation}
\begin{equation}
J\left( \Psi ,\left\vert \Psi \right\vert ^{2}\right) =\underset{i,j}{\text{
}\sum }c_{ij}\Psi ^{j+1}i\overline{\Psi }^{i-1} \tag{3.7} \label{3.7}
\end{equation}
is a complex polynomial functional of finite degree on $\Psi $
and$\ \left\vert \Psi \right\vert ^{2}$. As we can see,
the equation (3.6) has the solutions
\begin{equation}
J\left( \Psi ,\left\vert \Psi \right\vert ^{2}\right) =0 \tag{3.8a}
\label{3.8a}
\end{equation}
\begin{equation}
{\large \nabla }\Psi \times {\large \nabla }\overline{\Psi }=\mathbf{0}
\tag{3.8b} \label{3.8b}
\end{equation}
The (3.8a) solution functional equation has no solutions for an
'arbitrary' flux function $\Psi $, because the terms $\overline{%
\Psi }^{i}\Psi ^{j}$ inside it are linearly independent, or
\begin{equation*}
\forall i\neq \widetilde{i},\forall j\neq \widetilde{j}\text{ \ \ }\nexists
\text{ }\omega \in
%TCIMACRO{\U{2102} }%
\mathbb{C}
\setminus \left\{ 0\right\} :\overline{\Psi }^{i}\Psi ^{j}=\omega \overline{%
\Psi }^{\widetilde{i}}\Psi ^{\widetilde{j}}
\end{equation*}
indeed, for this reason it derives, taking in account (3.7),
\begin{equation*}
\forall \Psi ,\text{ \ }\nexists \left\{ c_{ij}\right\} _{i,j\in \text{ }%
%TCIMACRO{\U{2115} }%
\mathbb{N}
},\text{ }c_{ij}\in
%TCIMACRO{\U{2102} }%
\mathbb{C}
:J\left( \Psi ,\left\vert \Psi \right\vert ^{2}\right) =0
\end{equation*}
this means that the functional equation (3.8a) cannot represent a
necessary condition so the equation (3.5) is verified; at this point it's
clear that the unique 'critical condition' for the mathematical coherence of
the general fields choice for an arbitrary flux function $\Psi $ is
\begin{equation*}
{\large \nabla }\Psi \times {\large \nabla }\overline{\Psi }=\mathbf{0}\text{
\ }\forall (x,y)\in \Omega _{\alpha }\cap \Omega _{\beta }\text{\ }
\end{equation*}
Now, if we impose the complexity of this function ($\func{Im}\left(
{\large \Psi }\right) \subset
%TCIMACRO{\U{2102} }%
\mathbb{C}
$) and we write therefore (see position (2.38))
\begin{equation*}
\Psi (x,y)=u(x,y)+iv(x,y)\text{ \ \ \ }\forall (x,y)\in \Omega _{\alpha
}\cap \Omega _{\beta }
\end{equation*}
we obtain, starting from the equation (3.8b),
\begin{eqnarray*}
\mathbf{0}_{%
%TCIMACRO{\U{2102} }%
\mathbb{C}
} &\mathbf{=}&\text{ }{\large \nabla }\Psi \times {\large \nabla }\overline{%
\Psi }={\large \nabla }\left( u+iv\right) \times {\large \nabla }\left(
u-iv\right) =\left( {\large \nabla }u+i{\large \nabla }v\right) \times
\left( {\large \nabla }u-i{\large \nabla }v\right) = \\
&=&-i{\large \nabla }u\times {\large \nabla }v+i{\large \nabla }v\times
{\large \nabla }u=-2i{\large \nabla }u\times {\large \nabla }v
\end{eqnarray*}
(where $\mathbf{0}_{%
%TCIMACRO{\U{2102} }%
\mathbb{C}
}=\mathbf{0+}i\mathbf{0}$); in this way, the condition (3.8b) is
equivalent to the condition ${\large \nabla }u\times {\large \nabla }v=%
\mathbf{0}$, or
\begin{equation*}
{\large \nabla }\Psi \times {\large \nabla }\overline{\Psi }=\mathbf{0}_{%
%TCIMACRO{\U{2102} }%
\mathbb{C}
}\mathbf{\Leftrightarrow }\text{ }{\large \nabla }u\times {\large \nabla }v=%
\mathbf{0}\text{ \ \ }\forall (x,y)\in \Omega _{\alpha }\cap \Omega _{\beta }
\end{equation*}
We have therefore the 'critical condition' (which is a 'necessary
condition') for both real and imaginary parts of $\Psi $
\begin{equation}
{\large \nabla }u\times {\large \nabla }v=\mathbf{0}\text{ \ \ }\forall
(x,y)\in \Omega _{\alpha }\cap \Omega _{\beta } \tag{3.9} \label{3.9}
\end{equation}
so the equation (3.6) is verified; an expansion in a two-dimensional
framework $(x,y)$ of the ${\large \nabla }$ operator
finally leads to the critical PDE for the real functions ${\large u}$
and ${\large v}$
\begin{eqnarray*}
\mathbf{0} &=&{\large \nabla }u\times {\large \nabla }v=\left( \frac{%
\partial u}{\partial x}\widehat{\mathbf{x}}+\frac{\partial u}{\partial y}%
\widehat{\mathbf{y}}\right) \times \left( \frac{\partial v}{\partial x}%
\widehat{\mathbf{x}}+\frac{\partial v}{\partial y}\widehat{\mathbf{y}}%
\right) = \\
&=&\frac{\partial u}{\partial x}\frac{\partial v}{\partial y}\left( \widehat{%
\mathbf{x}}\times \widehat{\mathbf{y}}\right) +\frac{\partial u}{\partial y}%
\frac{\partial v}{\partial x}\left( \widehat{\mathbf{y}}\times \widehat{%
\mathbf{x}}\right) =\left( \frac{\partial u}{\partial x}\frac{\partial v}{%
\partial y}-\frac{\partial u}{\partial y}\frac{\partial v}{\partial x}%
\right) \left( \widehat{\mathbf{x}}\times \widehat{\mathbf{y}}\right)
\end{eqnarray*}
or the equation
\begin{equation}
\left\{ \frac{\partial u}{\partial x}\frac{\partial v}{\partial y}-\frac{%
\partial u}{\partial y}\frac{\partial v}{\partial x}\right\} \left(
x,y\right) =0\text{ \ \ }\forall (x,y)\in \Omega _{\alpha }\cap \Omega
_{\beta } \tag{3.10} \label{3.10}
\end{equation}
At this point, it's clear that the differential equation (3.10) is
the unique necessary condition for an arbitrary complex flux function $\Psi
$ so the general free fields transformations (3.1),(3.2) are a
mathematical coherent choice for the equilibrium problem in object;
furthermore, it's remarkable that this condition doesn't imply the
analiticity of the function $\Psi $; indeed, if the analiticity is
requested in $\Omega _{\alpha }\cap \Omega _{\beta }$ for $\Psi $,
the functions ${\large u}$ and ${\large v}$
must be $%
%TCIMACRO{\U{2102} }%
\mathbb{C}
$ - differentiable in the domain $\Lambda =\overline{\left( \Omega
_{\alpha }\cap \Omega _{\beta }\right) }\setminus \partial \left( \Omega
_{\alpha }\cap \Omega _{\beta }\right) $, or the
Cauchy-Riemann conditions must be valid in $\Lambda $; these
deliver for ${\large u}$
\begin{equation}
\left\{ \left( \frac{\partial u}{\partial x}\right) ^{2}+\left( \frac{%
\partial u}{\partial y}\right) ^{2}\right\} \left( x,y\right) =0\text{ \ \ }%
\forall (x,y)\in \Lambda \tag{3.11} \label{3.11}
\end{equation}
while for ${\large v}$ we have similarly
\begin{equation}
\left\{ \left( \frac{\partial v}{\partial x}\right) ^{2}+\left( \frac{%
\partial v}{\partial y}\right) ^{2}\right\} \left( x,y\right) =0\text{ \ \ }%
\forall (x,y)\in \Lambda \tag{3.12} \label{3.12}
\end{equation}
and it's clear that the differential equations (3.11) and (3.12) are
specific cases which are included in the differential equation (3.10).
Usefully, we note that if the complex flux function $\Psi $
is in the Gauss form
\begin{equation}
\Psi (x,y)=\rho (x,y)e^{i\phi (x,y)}\text{ \ \ \ }\forall (x,y)\in \Omega
_{\alpha }\cap \Omega _{\beta } \tag{3.13} \label{3.13}
\end{equation}
the critical condition (3.9) must return
\begin{equation}
{\large \nabla }\rho \times {\large \nabla }\phi =\mathbf{0}\text{ \ \ }%
\forall (x,y)\in \Omega _{\alpha }\cap \Omega _{\beta } \tag{3.14}
\label{3.14}
\end{equation}
Let us observe that if we impose the analiticity of the flux
function in the equilibria domain $\left( \Omega _{\alpha }\cap \Omega
_{\beta }\right) ^{\circ }$, the Helmholtz problem (2.27)
\begin{equation}
\text{\ }{\Large \nabla }^{2}\Psi +{\Large k}\left( \Psi \right) =0\text{ \
\ \ }\forall (x,y)\in \Omega _{\alpha }\cap \Omega _{\beta } \notag
\end{equation}
deliver the functional equations system
\begin{eqnarray}
{\Large \nabla }^{2}\Psi &=&0\text{ \ \ \ } \TCItag{3.14.1} \label{3.14.1}
\\
{\Large k}\left( \Psi \right) &=&0\text{ \ \ }\forall (x,y)\in \Omega
_{\alpha }\cap \Omega _{\beta } \TCItag{3.14.2} \label{3.14.2}
\end{eqnarray}
these equations destroy obviously the related NLSE-problem but they
represents also a not-banal question from the closely functional point of
view, because the equation (3.14.2) must be satisfied only in a 'sub-domain'
$\Omega _{\alpha }\cap \Omega _{\beta }$ of the entire domain $%
Dom\left( \Psi \right) $; indeed these domains are
subject to the condition
\begin{equation*}
\left( \Omega _{\alpha }\cap \Omega _{\beta }\right) \subseteq Dom\left(
\Psi \right)
\end{equation*}
leaving free in this way the choice of the not-bound portion of $%
\Psi $ inside $Dom\left( \Psi \right) \setminus \left( \Omega
_{\alpha }\cap \Omega _{\beta }\right) $. This means that a certain
family of $\Psi $ functions which satisfy the equation (3.14.2)
can be identified as
\begin{equation*}
\widetilde{\Psi }=h\left( \Omega _{\alpha }\cap \Omega _{\beta }\right)
=\Psi \text{ \ \ }\forall (x,y)\in \left( \Omega _{\alpha }\cap \Omega
_{\beta }\right) \cap Dom\left( \Psi \right)
\end{equation*}
and it means also that this family must depend only on the
possible forms of the flux functions inside the complementary domain $%
Dom\left( \Psi \right) \setminus \left( \Omega _{\alpha }\cap \Omega _{\beta
}\right) $. Now, because the (3.14.1), we can note that if $\left(
\Omega _{\alpha }\cap \Omega _{\beta }\right) \neq $ $Dom\left(
\Psi \right) $, it can be possible to define a family of harmonic
functions $\widetilde{\Psi }$ inside $\Omega _{\alpha }\cap \Omega
_{\beta }$ which satisfy the functional equation (3.14.2); this
functions family is related obviously to the solutions set of the equation
\begin{equation}
{\large g}\left( \left\vert \Psi \right\vert ^{2}\right) {\large +i\sigma }%
\left( \left\vert \Psi \right\vert ^{2}\right) =0\text{ \ \ }\forall
(x,y)\in \Omega _{\alpha }\cap \Omega _{\beta } \tag{3.14.3} \label{3.14.3}
\end{equation}
At this point, it's remarkable that this equation
necessarily delivers the above not-trivial solution for $\Psi $,because there is the close condition
\begin{equation}
\left( \Omega _{\alpha }\cap \Omega _{\beta }\right) \subset Dom\left( \Psi
\right) \tag{3.14.4} \label{3.14.4}
\end{equation}
which can be justified by observing that the equilibria
domain does not concern in general the entire framework of the equatorial
plane in the accretion disc unlike $\Psi $.
Let us note that in our specific case the solutions set of
the (3.14.3) is not trivial because the general equation scenario (3.14.3)
assume the particular form
\begin{equation*}
{\large g}\left( \left\vert \Psi \right\vert ^{2},u,v\right) {\large %
+i\sigma }\left( \left\vert \Psi \right\vert ^{2},u,v\right) =0\text{ \ \ }%
\forall (x,y)\in \Omega _{\alpha }\cap \Omega _{\beta }
\end{equation*}
related to the expressions (2.43a)-(2.44b). It's clear that this
case is generally independent of the condition (3.14.4).
It's important too, in conclusion, to observe that the 'knotty
structures' can be present only inside a sub-domain $\Omega _{\alpha }\cap
\Omega _{\beta }{\large \ }$for which the condition (3.14.4) is
valid, giving sense in this way to the plasma equilibria in the
related space regions; hence, it's necessary that the contour conditions for
the plasma equilibrium-dependent differential Helmholtz problem
\begin{equation*}
{\Large \nabla }^{2}\Psi +\left( {\large g}\left( \left\vert \Psi
\right\vert ^{2}\right) {\large +i\sigma }\left( \left\vert \Psi \right\vert
^{2}\right) \right) \Psi =0
\end{equation*}
must be specified on the contours of an appropriate domain $\Omega $
, taking for this $\Omega \subseteq \Omega _{\alpha }\cap \Omega
_{\beta }$.
The critical condition (3.9) is identically satisfied for the
Throumoulopoulos et al. transformations [9], while
for the Lapenta transformations [8] we have that the
above condition remains valid in its general form, leading to a free choice
for the real flux functions ${\large u}$ and ${\large v}$,
which in this case are related only to the bond equation (3.10).
This is because in the second choice is present a term equal to $\overline{%
\Psi }$, while this one there is not in the first choice. Such
situation means that both choices represent coherent transformation models
but the Throumoulopoulos et al. choice is
mathematically more strong than Lapenta choice; from this
consideration clearly doesn't derive that the choice (2.45a),(2.45b) is less
valid than the choice (2.46a),(2.46b). It's important instead that this
observation reconciles in this way the results of the above authors.
One of the solutions for the equation (3.6) is
\begin{equation*}
J\left( \Psi ,\left\vert \Psi \right\vert ^{2}\right) =0\text{ \ \ \ }%
\forall (x,y)\in \Omega _{\alpha }\cap \Omega _{\beta }
\end{equation*}
Let us note that this functional equation has not a trivial solution
if we consider an appropriate family of flux functions $\widehat{\Psi }$
as we did above in a similar problem; in this way, we could obtain
another valid critical condition on the $\Psi $ function for the
general fields transformations. An idea is to study the character of the
functional power series
\begin{equation*}
J_{\infty }\left( \Psi ,\left\vert \Psi \right\vert ^{2}\right) =\underset{%
i,j}{\text{ }\overset{\infty }{\sum }}c_{ij}\Psi ^{j+1}i\overline{\Psi }%
\end{equation*}
for an appropriate choice of the constants set $\left\{
c_{ij}\right\} _{i,j\in \text{ }%
%TCIMACRO{\U{2115} }%
\mathbb{N}
},$ $c_{ij}\in
%TCIMACRO{\U{2102} }%
\mathbb{C}
$, taking into account that it must be
\begin{equation*}
J_{\infty }\left( \Psi ,\left\vert \Psi \right\vert ^{2}\right) =0\text{ \ \
}\forall (x,y)\in \Omega _{\alpha }\cap \Omega _{\beta }
\end{equation*}
§.§ On a reality condition for the free fields: physical aspects
In previous section we talked about a plausible physical
validity condition for the general choice (3.1),(3.2), saying that the
'reality' of the terms proportional to the magnetic force density and the
pure hydrodynamical pressure force density inside it can be a fundamental
condition from the physical point of view and it's remarkable that such
condition doesn't affect the complexity of the flux function $\Psi $;
indeed these terms are the two 'real' vectorial physical variables,
by the free fields transformations,
\begin{equation}
{\large \nabla B}_{z}^{2}=2\underset{i,j}{\sum }a_{ij}\overline{\Psi }%
^{i}\Psi ^{j+1}{\Large \nabla }\Psi \tag{3.15} \label{3.15}
\end{equation}
\begin{equation}
{\large \nabla p}=\underset{i,j}{\sum }b_{ij}\overline{\Psi }^{i}\Psi ^{j+1}%
{\Large \nabla }\Psi \tag{3.16} \label{3.16}
\end{equation}
which must be therefore
\begin{equation}
{\large \nabla B}_{z}^{2},{\large \nabla p\notin
%TCIMACRO{\U{211d} }%
\mathbb{R}
%TCIMACRO{\U{2102} }%
\mathbb{C}
}^{3},\text{ \ where \ }{\large
%TCIMACRO{\U{211d} }%
\mathbb{R}
%TCIMACRO{\U{2102} }%
\mathbb{C}
}^{3}=\left\{ \mathbf{x}_{%
%TCIMACRO{\U{2102} }%
\mathbb{C}
%TCIMACRO{\U{2102} }%
\mathbb{C}
}=\mathbf{x}+i\mathbf{y},\forall \mathbf{x,y\in }\text{ }{\large
%TCIMACRO{\U{211d} }%
\mathbb{R}
}^{3}\right\} \tag{3.17} \label{3.17}
\end{equation}
Now, for to determine this reality condition for the above force
densities, we impose
\begin{equation}
\func{Im}\left( {\large \nabla B}_{z}^{2}\right) =\func{Im}\left( {\large %
\nabla p}\right) =\mathbf{0} \tag{3.18} \label{3.18}
\end{equation}
and taking the polynomial functional (2.34)
\begin{equation}
\underset{i,j}{\sum }\gamma _{ij}\overline{\Psi }^{i}\Psi ^{j}={\large g}%
\left( \left\vert \Psi \right\vert ^{2}\right) {\large +i\sigma }\left(
\left\vert \Psi \right\vert ^{2}\right) ,\text{\ \ }\gamma _{ij}\in
%TCIMACRO{\U{2102} }%
\mathbb{C}
,\text{ }\forall (x,y)\in \Omega _{\alpha }\cap \Omega \text{ } \notag
\end{equation}
for (3.15) and (3.16) we obtain
\begin{equation}
{\large \nabla B}_{z}^{2}=2\left[ {\large g}\left( \left\vert \Psi
\right\vert ^{2}\right) {\large +i\sigma }\left( \left\vert \Psi \right\vert
^{2}\right) \right] \Psi {\Large \nabla }\Psi \tag{3.19} \label{3.19}
\end{equation}
\begin{equation}
{\large \nabla p}=\left[ {\large g}\left( \left\vert \Psi \right\vert
^{2}\right) {\large +i\sigma }\left( \left\vert \Psi \right\vert ^{2}\right) %
\right] \Psi {\Large \nabla }\Psi \tag{3.20} \label{3.20}
\end{equation}
which are different only for a factor $2$. At this point,
for the validity of both (3.18) relations, it's sufficient obviously to
analyze the condition
\begin{equation}
\func{Im}\left\{ \left[ {\large g}\left( \left\vert \Psi \right\vert
^{2}\right) {\large +i\sigma }\left( \left\vert \Psi \right\vert ^{2}\right) %
\right] \Psi {\Large \nabla }\Psi \right\} =\mathbf{0} \tag{3.21}
\label{3.21}
\end{equation}
For this one, taking in account that
\begin{equation*}
\Psi (x,y)=u(x,y)+iv(x,y)\text{ \ \ \ }\forall (x,y)\in \Omega _{\alpha
}\cap \Omega _{\beta }
\end{equation*}
let us calculate in clear complex form the expression
\begin{equation*}
\left[ {\large g}\left( \left\vert \Psi \right\vert ^{2}\right) {\large %
+i\sigma }\left( \left\vert \Psi \right\vert ^{2}\right) \right] \Psi
{\Large \nabla }\Psi
\end{equation*}
We have therefore
\begin{equation*}
\left( {\large g+i\sigma }\right) \Psi {\Large \nabla }\Psi =\left( {\large %
g+i\sigma }\right) \left( {\large u+iv}\right) {\Large \nabla }\left(
u+iv\right) =
\end{equation*}
\begin{eqnarray*}
&=&\left[ \left( {\large gu-\sigma v}\right) {\large +i}\left( {\large %
gv+\sigma u}\right) \right] \left( {\large \nabla }u+i{\large \nabla }%
v\right) = \\
&=&\left[ \left( {\large gu-\sigma v}\right) {\large \nabla }u-\left(
{\large gv+\sigma u}\right) {\large \nabla }v\right] +i\left[ \left( {\large %
gu-\sigma v}\right) {\large \nabla }v+\left( {\large gv+\sigma u}\right)
{\large \nabla }u\right]
\end{eqnarray*}
which delivers for the condition (3.21)
\begin{equation*}
\left( {\large gu-\sigma v}\right) {\large \nabla }v+\left( {\large %
gv+\sigma u}\right) {\large \nabla }u=\mathbf{0}
\end{equation*}
\begin{equation}
{\large \nabla }u=\frac{\left( {\large \sigma v-gu}\right) }{\left( {\large %
gv+\sigma u}\right) }{\large \nabla }v\text{ \ \ }\forall (x,y)\in \Omega
_{\alpha }\cap \Omega _{\beta } \tag{3.22} \label{3.22}
\end{equation}
where the form of the functions ${\large \sigma }$ and $%
{\large g}$ is derived by the relations (2.43a)-(2.44b); it's clear
that (3.22) represents a necessary and sufficient condition for (3.21) and
furthermore it leads to the 'new' critical condition
\begin{equation}
{\large \nabla }u\times {\large \nabla }v=\mathbf{0}\text{ \ \ }\forall
(x,y)\in \Omega _{\alpha }\cap \Omega _{\beta } \tag{3.23} \label{3.23}
\end{equation}
It's important to observe that this one is the same of the critical
condition from the mathematical point of view, which can be found in (3.9);
this means that the critical condition (3.23) ia a 'common necessary
condition' for both the mathematical and physical aspects of the coherence
for the transformation functionals (3.15),(3.16).
Now, if we note that the generic solution choosed in a plane
framework $(x,y)$ for the magnetic induction field is (by the
(2.4)) $\forall (x,y)\in \Omega _{\alpha }\cap \Omega _{\beta }$
\begin{equation*}
\mathbf{B}=\widehat{\mathbf{z}}\times {\Large \nabla }\Psi +B_{z}\widehat{%
\mathbf{z}}
\end{equation*}
is reasonable to think that it must be for a complex $\Psi $,
as in the conditions (3.18),
\begin{equation}
\func{Im}\left( \mathbf{B}\right) =\mathbf{0}\text{ \ \ }\forall (x,y)\in
\Omega _{\alpha }\cap \Omega _{\beta } \tag{3.24} \label{3.24}
\end{equation}
which delivers necessarily
\begin{eqnarray*}
\mathbf{0} &\mathbf{=}&\func{Im}\left( \widehat{\mathbf{z}}\times {\Large %
\nabla }\Psi +B_{z}\widehat{\mathbf{z}}\right) =\func{Im}\left( \widehat{%
\mathbf{z}}\times {\Large \nabla }\Psi \right) =\func{Im}\left( \widehat{%
\mathbf{z}}\times \left( {\large \nabla }u+i{\large \nabla }v\right) \right)
= \\
&=&\func{Im}\left( \widehat{\mathbf{z}}\times {\large \nabla }u+i\widehat{%
\mathbf{z}}\times {\large \nabla }v\right) =\widehat{\mathbf{z}}\times
{\large \nabla }v
\end{eqnarray*}
\begin{equation}
\widehat{\mathbf{z}}\times {\large \nabla }v=\mathbf{0} \tag{3.25}
\label{3.25}
\end{equation}
this condition leads to below vectorial differential equation by the
expansion of the operator ${\large \nabla }$
\begin{equation*}
\widehat{\mathbf{z}}\times \left( \frac{\partial v}{\partial x}\widehat{%
\mathbf{x}}+\frac{\partial v}{\partial y}\widehat{\mathbf{y}}\right) =\frac{%
\partial v}{\partial x}\widehat{\mathbf{y}}-\frac{\partial v}{\partial y}%
\widehat{\mathbf{x}}=\mathbf{0}
\end{equation*}
which means
\begin{equation}
\frac{\partial v}{\partial x}=\frac{\partial v}{\partial y}=0\text{ \ \ or \
\ }{\large \nabla }v=\mathbf{0}\text{ \ \ }\forall (x,y)\in \Omega _{\alpha
}\cap \Omega _{\beta } \tag{3.26} \label{3.26}
\end{equation}
At this point it's indispensable to observe that if the imaginary
component of the flux function $\Psi $ has a zero gradient in $%
\Omega _{\alpha }\cap \Omega _{\beta }$, the critical common
condition (3.23) is identically verified; so, the only position which can
respect a reality condition for all fields in the equilibrium problem, as
viewed, is therefore a unique flux function condition for which it must be
\begin{equation}
\func{Im}\left( \mathbf{\Psi }(x,y)\right) =v(x,y)=\text{cost \ \ }\forall
(x,y)\in \Omega _{\alpha }\cap \Omega _{\beta } \tag{3.27} \label{3.27}
\end{equation}
which delivers finally
\begin{equation}
\Psi (x,y)=u(x,y)+iA,\text{ \ }A\in
%TCIMACRO{\U{211d} }%
\mathbb{R}
,\text{\ \ }\forall (x,y)\in \Omega _{\alpha }\cap \Omega _{\beta }
\tag{3.28} \label{3.28}
\end{equation}
§.§ On a global critical condition for the GS equilibrium problem:critical equations system
Here we briefly explore a further validity condition for the
acceptability of the general choice (3.1),(3.2). This condition regards the
equation (2.8), which leads to a final 'validity equation' if it's related
to the common critical condition (3.9) or to the flux function condition
(3.27), taking into account the relations (2.11) and (2.14).
If we start from the equation (2.8), we obtain by applying the
operator ${\large \nabla \times }$ to both sides of this one
\begin{eqnarray*}
\mathbf{0}_{%
%TCIMACRO{\U{2102} }%
\mathbb{C}
} &\mathbf{=}&{\large \nabla \times }\left( {\Large \nabla }{\large p\ }+%
{\Large \nabla }^{2}\Psi {\Large \nabla }\Psi +B_{z}{\Large \nabla }%
B_{z}\right) ={\Large \nabla }{\large \times }{\Large \nabla }{\large p\ }+%
{\Large \nabla }{\large \times }{\Large \nabla }^{2}\Psi {\Large \nabla }%
\Psi +{\Large \nabla }{\large \times }B_{z}{\Large \nabla }B_{z}= \\
&=&{\Large \nabla }^{2}\Psi {\Large \nabla }{\large \times }{\Large \nabla }%
\Psi -{\Large \nabla }\Psi \times {\Large \nabla \nabla }^{2}\Psi +B_{z}%
{\Large \nabla }{\large \times }{\Large \nabla }B_{z}-{\Large \nabla }%
B_{z}\times {\Large \nabla }B_{z}= \\
&=&-{\Large \nabla }\Psi \times {\Large \nabla \nabla }^{2}\Psi ={\Large %
\nabla \nabla }^{2}\Psi \times {\Large \nabla }\Psi
\end{eqnarray*}
\begin{equation}
{\Large \nabla \nabla }^{2}\Psi \times {\Large \nabla }\Psi =\mathbf{0}_{%
%TCIMACRO{\U{2102} }%
\mathbb{C}
} \tag{3.29} \label{3.29}
\end{equation}
This further condition is obviously a necessary condition for the
Grad-Shafranov equilibrium problem and it can be translate into below
differential equations set: if $\Psi $ is the complex function $%
\Psi (x,y)=u(x,y)+iv(x,y)$ $\forall (x,y)\in \Omega _{\alpha }\cap
\Omega _{\beta }$, we have
\begin{eqnarray*}
\mathbf{0}_{%
%TCIMACRO{\U{2102} }%
\mathbb{C}
} &=&{\large \nabla \nabla }^{2}\Psi {\large \times \nabla \Psi =\nabla
\nabla }^{2}\left( u+iv\right) {\large \times \nabla }\left( u+iv\right) = \\
&=&\left( {\large \nabla \nabla }^{2}u+i{\large \nabla \nabla }^{2}v\right)
\times \left( {\large \nabla }u+i{\large \nabla }v\right) = \\
&=&{\large \nabla \nabla }^{2}u\times {\large \nabla }u-{\large \nabla
\nabla }^{2}v\times {\large \nabla }v+i\left( {\large \nabla \nabla }%
^{2}u\times {\large \nabla }v+{\large \nabla \nabla }^{2}v\times {\large %
\nabla }u\right)
\end{eqnarray*}
which delivers the 'critical equations system'
\begin{eqnarray}
{\large \nabla \nabla }^{2}u\times {\large \nabla }u-{\large \nabla \nabla }%
^{2}v\times {\large \nabla }v &=&\mathbf{0} \TCItag{3.30a} \label{3.30a} \\
{\large \nabla \nabla }^{2}u\times {\large \nabla }v+{\large \nabla \nabla }%
^{2}v\times {\large \nabla }u &=&\mathbf{0} \TCItag{3.30b} \label{3.30b}
\end{eqnarray}
Now, if we consider the flux function condition (3.27), we have
that the second equation is identically verified, while for the
first equation we obtain
\begin{equation}
{\large \nabla \nabla }^{2}u\times {\large \nabla }u=\mathbf{0}\text{ \ \ }%
\forall (x,y)\in \Omega _{\alpha }\cap \Omega _{\beta } \tag{3.31}
\label{3.31}
\end{equation}
\begin{eqnarray*}
\mathbf{0} &=&{\large \nabla \nabla }^{2}u\times {\large \nabla }u=\left(
\frac{\partial }{\partial x}\widehat{\mathbf{x}}+\frac{\partial }{\partial y}%
\widehat{\mathbf{y}}\right) {\large \nabla }^{2}u\times \left( \frac{%
\partial u}{\partial x}\widehat{\mathbf{x}}+\frac{\partial u}{\partial y}%
\widehat{\mathbf{y}}\right) = \\
&=&\frac{\partial }{\partial x}{\large \nabla }^{2}u\frac{\partial u}{%
\partial y}\left( \widehat{\mathbf{x}}\times \widehat{\mathbf{y}}\right) +%
\frac{\partial }{\partial y}{\large \nabla }^{2}u\frac{\partial u}{\partial x%
}\left( \widehat{\mathbf{y}}\times \widehat{\mathbf{x}}\right) = \\
&=&\left( \frac{\partial }{\partial x}{\large \nabla }^{2}u\frac{\partial u}{%
\partial y}-\frac{\partial }{\partial y}{\large \nabla }^{2}u\frac{\partial u%
}{\partial x}\right) \left( \widehat{\mathbf{x}}\times \widehat{\mathbf{y}}%
\right)
\end{eqnarray*}
which delivers the unique differential equation for all critical
\begin{equation}
\left( \frac{\partial }{\partial x}{\large \nabla }^{2}u\frac{\partial u}{%
\partial y}-\frac{\partial }{\partial y}{\large \nabla }^{2}u\frac{\partial u%
}{\partial x}\right) =0\text{ \ \ }\forall (x,y)\in \Omega _{\alpha }\cap
\Omega _{\beta } \tag{3.32} \label{3.32}
\end{equation}
If we consider instead the common critical condition (3.9) only,
\begin{equation*}
{\large \nabla }u\times {\large \nabla }v=\mathbf{0}\text{ \ \ }\forall
(x,y)\in \Omega _{\alpha }\cap \Omega _{\beta }
\end{equation*}
we obtain by the system (3.30a),(3.30b)
\begin{equation}
{\large \nabla \nabla }^{2}u\times {\large \nabla }u=\mathbf{0}\text{ \ \ }%
\forall (x,y)\in \Omega _{\alpha }\cap \Omega _{\beta } \tag{3.33}
\label{3.33}
\end{equation}
which delivers, as in (3.32), the differential equation
\begin{equation}
\left( \frac{\partial }{\partial x}{\large \nabla }^{2}v\frac{\partial v}{%
\partial y}-\frac{\partial }{\partial y}{\large \nabla }^{2}v\frac{\partial v%
}{\partial x}\right) =0\text{ \ \ }\forall (x,y)\in \Omega _{\alpha }\cap
\Omega _{\beta } \tag{3.34} \label{3.34}
\end{equation}
taking into account that the condition (3.9) means
\begin{equation*}
{\large \nabla }u=f\left( u,v\right) {\large \nabla }v
\end{equation*}
where $f$ is a real functional on $u$ and $v$, $%
\forall (x,y)\in \Omega _{\alpha }\cap \Omega _{\beta }$.
At this point, it's clear that the conditioning equations (3.32) and
(3.34) represent two global critical conditions for the
magnetohydrodynamical equilibrium problem in the pseudo-general NLSE
framework, for which obviously it's true that
\begin{equation}
\left( \frac{\partial }{\partial x}{\large \nabla }^{2}u\frac{\partial u}{%
\partial y}-\frac{\partial }{\partial y}{\large \nabla }^{2}u\frac{\partial u%
}{\partial x}\right) =0\Longrightarrow \tag{3.35} \label{3.35}
\end{equation}
\begin{equation*}
\Longrightarrow \left( \frac{\partial }{\partial x}{\large \nabla }^{2}v%
\frac{\partial v}{\partial y}-\frac{\partial }{\partial y}{\large \nabla }%
^{2}v\frac{\partial v}{\partial x}\right) =0\text{\ \ \ }\forall (x,y)\in
\Omega _{\alpha }\cap \Omega _{\beta }
\end{equation*}
for the bond equations system (3.30a),(3.30b).
§ COMMENTS
In conclusion: we have seen in previous sections that for to obtain
a pseudo-general form of the NLSE in the stationary case (see (2.22) in the
plane framework $(x,y)$) by the GSEs set (see the system
(2.7),(2.8)) for a general plasma equilibrium problem in the equatorial
plane of an accretion disc, or
\begin{equation*}
{\Large \nabla }^{2}\Psi +\left( {\large g}\left( \left\vert \Psi
\right\vert ^{2}\right) {\large +i\sigma }\left( \left\vert \Psi \right\vert
^{2}\right) \right) \Psi =0
\end{equation*}
a set of transformation relations (i.e. the general choice) of the
free fields in the GSE is imposable in a general form and this is
\begin{equation*}
{\large B}_{z}{\large \nabla B}_{z}=\frac{1}{2}{\large \nabla B}_{z}^{2}=%
\underset{i,j}{\sum }a_{ij}\overline{\Psi }^{i}\Psi ^{j+1}{\Large \nabla }%
\Psi
\end{equation*}
\begin{equation*}
{\large \nabla p}=\underset{i,j}{\sum }b_{ij}\overline{\Psi }^{i}\Psi ^{j+1}%
{\Large \nabla }\Psi \text{ \ \ \ \ {\large with \ \ \ }}a_{ij},b_{ij}\in
\mathbb{C}%
\end{equation*}
where $\Psi $ is the flux function in the solution (2.4) of
the general equilibrium problem (2.3) and it's a complex function on the
real variables $x$ and $y$
\begin{equation*}
\Psi (x,y)=u(x,y)+iv(x,y)\text{ \ \ \ }\forall (x,y)\in \Omega _{\alpha
}\cap \Omega _{\beta }
\end{equation*}
which is a solution for the Helmholtz problem (2.25) and it's
defined in a specific local not-banal domain (in the accretion disc
framework) as $\Omega _{\alpha }\cap \Omega _{\beta }$; it's clear
that the topology of this domain is a very fundamental characteristic of the
equilibrium problem, because the critical condition equation (3.10) must be
verified for the mathematical coherence of the above fields transformations
inside it, or
\begin{equation*}
\left\{ \frac{\partial u}{\partial x}\frac{\partial v}{\partial y}-\frac{%
\partial u}{\partial y}\frac{\partial v}{\partial x}\right\} \left(
x,y\right) =0\text{ \ \ }\forall (x,y)\in \Omega _{\alpha }\cap \Omega
_{\beta }
\end{equation*}
which, in the vectorial form, is
\begin{equation*}
{\large \nabla }u\times {\large \nabla }v=\mathbf{0}\text{ \ \ }\forall
(x,y)\in \Omega _{\alpha }\cap \Omega _{\beta }
\end{equation*}
and therefore it's remarkable that the analiticity of the flux
function is not necessary for the validity of the general fields choice.
Furthermore, this equation represents a global critical condition for the
equilibrium problem from the solitonic point of view, as showed in (3.35).
We remember that the critical condition (3.10) is equal to the condition on
$\Psi $
\begin{equation*}
{\large \nabla }\Psi \times {\large \nabla }\overline{\Psi }=\mathbf{0}\text{
\ \ }\forall (x,y)\in \Omega _{\alpha }\cap \Omega _{\beta }
\end{equation*}
It's clear at this point that the Throumoulopoulos et al.
and Lapenta theoretical positions (see
(2.45a-b) and (2.46a-b)) about a specific transformations set for the free
fields are cannot opposite, because both these choices satisfy the critical
condition (3.8b), which is identically verified for the first choice, while
it is in its general form in the second case because however it remains
verified for an appropriate choice of the $u$ and $v$ real
functions inside $\Omega _{\alpha }\cap \Omega _{\beta }$.
[1] Bridle, A.H., Hough, D.H., Lonsdale, C.J., Burns, J.O., Laing,
R.A., Astron. J.,108, 766 (1994)
[2] Bogoyavlenskij, O.I., Phys. Rev. Lett.,
84, 1914 (2000)
[3] Petviashvili, V.I. et al., Sov. Phys. JETP 55,
1056 (1982)
[4] Petviashvili, V.I. et al., Sov. J. Plasma Phys. JETP
12, 545 (1986)
[5] Shukla, P.K., Stenflo, L., Pokhotelov, O.A., Physica
Scripta T116, 135 (2005)
[6] García-Ripoll, J.J., Pérez-García, V.M.,
arXiv:patt-so1/9904006v1, 14 Apr 1999
[7] Boyer, C.P., Kalnins, E.G., Miller, W.Jr., Nagoya Math.
J. 60, 35 (1976)
[8] Lapenta, G., Phys. Rev. Lett. 90, 135005 (2003)
[9] Throumoulopoulos, G.N. et al., Phys. Rev. Lett.
92, 2495001 (2004)
[10] Lapenta, G., Phys. Rev. Lett. 92, 2495002
[11] Okamoto, K., Fundamentals of Optical Waveguides
(Academic Press, San Diego, 2000)
[12] Finn, J.M., Manheimer, W.M., Ott, E., Phys. Fluids,
24, 1336 (1981)
[13] Lighthill, M.J., Phil. Trans. Roy. Soc. London,
252, 397 (1960)
[14] Shafranov, V.D., Rev. of Plasma Phys., 2, 103
| arxiv-papers | 2008-07-10T12:51:42 | 2024-09-04T02:48:56.669062 | {
"license": "Public Domain",
"authors": "Michele Romeo",
"submitter": "Michele Romeo",
"url": "https://arxiv.org/abs/0807.1629"
} |
0807.1652 |
Fundamental Cycles and Graph Embeddings 111Supported by Natural Science
Foundation of China ( Under the Granted Number 10271048, 10671073)
Ren Han222This work is partially supported by Science and Technology
Commission of Shanghai Municipality (07XD14011) and Shanghai Leading Academic
Discipline Project, Project Number B407, Zhao Hongtao and Li Haoling
Dept. of Mathematics, East China Normal University
Shanghai 200062, P.R.China
E-mail: hren@math.ecnu.edu.cn
Abstract: In this paper, we investigate fundamental cycles in a graph $G$ and
their relations with graph embeddings. We show that a graph $G$ may be
embedded in an orientable surface with genus at least $g$ if and only if for
any spanning tree $T$, there exists a sequence of fundamental cycles
$C_{1},~{}C_{2},~{}\cdots,C_{2g}$ with $C_{2i-1}\cap C_{2i}\neq\phi$ for
$1\leq~{}i\leq g$. In particular, among $\beta(G)$ fundamental cycles of any
spanning tree $T$ of a graph $G$, there are exactly $2\gamma_{M}(G)$ cycles
$C_{1},~{}C_{2},~{}\cdots,C_{2\gamma_{M}(G)}$ such that $C_{2i-1}\cap
C_{2i}\neq\phi$ for $1\leq~{}i\leq\gamma_{M}(G)$, where $\beta(G)$ and
$\gamma_{M}(G)$ are, respectively, the Betti number and the maximum genus of
$G$. This implies that it is possible to construct an orientable embedding
with large genus of a graph $G$ from an arbitrary spanning tree $T$( which may
have very large number of odd components in $G\backslash E(T)$). This is
different from the earlier work of Xuong and Liu[9,6], where spanning trees
with small odd components are needed. In fact, this makes a common
generalization of Xuong[9],Liu[6] and Fu et al[2]. Further more, we show that
(1).This result is useful in locating the maximum genus of a graph having a
specific edge-cut. Some known results for embedded graphs are also
concluded;(2).The maximum genus problem may be reduced to the maximum matching
problem. Based on this result and the algorithm of Micali-Vazirani[8], we
present a new efficient algorithm to determine the maximum genus of a graph in
$O({(\beta(G))}^{\frac{5}{2}})$ steps. Our method is straight and quite
deferent from the algorithm of Furst,Gross and McGeoch[3] which depends on a
result of Giles[4]where matroid parity method is needed.
Keyword : Fundamental cycles, Maximum genus,upper-embedded .
AMS 2000: Primary 05C10, secondary 05C70
## 1 Definitions and Notations
The graph considered here is finite and undirected and, furthermore, is
connected unless it is stated otherwise. In general, multiple edges and loops
are allowed. Terminology and notation without explicit explanation follows as
from [1,6,7].
By a surface, denoted by $S$, we mean a compact and connected 2-manifold
without boundary. It is well known from elementary topology that surfaces can
be divided into two classes: orientable and nonorientable ones. An orientable
surface can be viewed as a sphere attached $h$ handles, while a nonorientable
surface as a sphere attached $k$ crosscaps. The number $h$ or $k$ is called
the genus of the surface. A cellular embedding of a graph $G$ into a surface
$S$ is a continuous one-to-one mapping $\phi$: $G\rightarrow S$ such that each
component of $G\backslash\phi(G)$ is homeomorphic to an open disc, called a
face of $G$ ( with respect to this embedding $\phi$) and $\phi$ is called a
cellular embedding( or embedding as some scholars called ). A cycle ( curve )
$C$ in an embedded graph in a surface $\sum$ is called surface separating if
$\sum-C$ is disconnected. In particular, if $\sum-C$ has an open disc, denoted
by int$(C)$, then $C$ is called contractible (otherwise, $C$ is
noncontractible), and int$(C)+C=$Int$(C)$ is the inner part of $C$ . The other
part of $\sum-C$ is called exterior of $C$ and is denoted by Ext$(C)$.
Recall that the maximum genus $\gamma_{M}(G)$ of a graph $G$ is the largest
integer $k$ such that $G$ has an embedding in an orientable surface with genus
$k$. Since any graph $G$ embedded in a surface has at least one face, Euler’s
formula shows that $\gamma_{M}(G)\leq\lfloor\frac{\beta(G)}{2}\rfloor$, where
$\beta(G)=|E(G)|-|V(G)|+1$ is known as Betti number of $G$ ( which is equal to
the cyclic number of $G$). A graph is upper-embeddable if
$\gamma_{M}(G)=\lfloor\frac{\beta(G)}{2}\rfloor$.
Let $G$ be a graph and $T$ be a spanning tree of $G$. It is clear that for any
edge $e\in{E(G)\backslash E(T)}$, $T+e$ contains a unique cycle of $G$,
denoted by $C_{T}(e)$, which is called a fundamental cycle of $G$ ( with
respect to the spanning tree $T$ of $G$). If a pair of edges $e_{1}$ and
$e_{2}$ have a common end vertex in a graph $G$, then we say that the pair
$\langle e_{1},e_{2}\rangle$ is an adjacent-edge pair in $G$.Let $G_{1}$ and
$G_{2}$ be a pair of disjoint subgraphs of $G$. Then $E[G_{1},G_{2}]$ is the
set of edges with their ends in $G_{1}$ and $G_{2}$,respectively.
Denote by $\xi(G,T)$ the number of components of $G\backslash E(T)$ with an
odd number of edges. Then the Betti deficiency of $G$ denoted by $\xi(G)$ is
defined as the value $min_{T}\xi(G,T)$, where the minimum is taken over all
spanning trees $T$ of $G$. A spanning tree $T$ of $G$ is said to be an optimal
spanning tree if $\xi(G,T)=\xi(G)$.
In the following, the paper is organized as follows: in §2 we give a good
characterization (i.e., Theorems 1 and 2) of maximum genus; §3 will
concentrate on the applications of Theorems 1 and 2 and their refined form; §4
will show that finding the maximum genus of a graph $G$ is, in some extend,
equivalent to the problem of finding a maximum matching in a specific graph
$G_{M}$ called the fundamental intersecting graph of $G$ and presents an
efficient algorithm in finding the maximum genus of a graph.
## 2 A Good Characterization
Lemma 1[6,9] Let $G$ be a graph, then
(1) $\gamma_{M}(G)=\frac{1}{2}(\beta(G)-\xi(G))$;
(2) $G$ is upper embeddable if and only if $\xi(G)\leq 1$.
Theorem 1.If a graph $G$ contains a spanning tree $T$ such that there exist
$2g$ fundamental cycles $C_{1},C_{2},\cdots,C_{2g}$ with $C_{2i-1}\cap
C_{2i-1}\neq\phi$, for $i=1,2\cdots g$, then $G$ may be embedded in an
orientable surface with genus at least $g$.
Proof Let $G$ and $T$ be as assumed and $e_{1},e_{2},\cdots,e_{2g}$ be edges
in $E(G)\setminus E(T)$ such that $C_{i}$ is the unique cycle in
$T+e_{i}(1\leq i\leq 2g).$ We may suppose further that
$G=T+\\{e_{1},e_{2},\cdots,e_{2g}\\}$ by Xuong’s constructive proof of maximum
genus formula[9]. Let $G_{0}=T$, and $G_{1}=G_{0}+\\{e_{1},e_{2}\\}$. Then we
have the following.
Claim 1. $\xi(G_{1})\leq\xi(G_{0})$
$(\Leftrightarrow\gamma_{M}(G_{1})\geq\gamma_{M}(G_{0})+1).$ To see this, we
observe that $\beta(G_{0})=\beta(G_{1})-2$, and so,
$\xi(G_{1})\equiv\xi(G_{0})(~{}mod~{}2~{})$.
If $\xi(G_{1})\geq\xi(G_{0})~{}+~{}2$, then we have one of the following
situations:
(1). Both $e_{1}$ and $e_{2}$ have, respectively, their ends in distinct even
components in $E(G_{0})\backslash E(T)$ ( As shown in left hand side of
Fig.1).
(2). Both $e_{1}$ and $e_{2}$ have, respectively, their ends in the same even
components in $E(G_{0})\backslash E(T)$ ( As shown in center of Fig.1).
(3). Exactly, one of $e_{1}$ and $e_{2}$, say $e_{1}$, joins two even
components of $E(G_{0})\backslash E(T)$, while $e_{2}$ has two ends in the
same even components in $E(G_{0})\backslash E(T)$ ( As shown in right hand
side of Fig.1).
$e_{1}$$\sigma_{1}$$\sigma_{2}$$e_{2}$$\sigma_{3}$$\sigma_{4}$$e_{1}$$e_{2}$$\sigma_{1}$$\sigma_{2}$$Fig.~{}1$$\sigma_{1}$$\sigma_{2}$$\sigma_{3}$$e_{2}$$e_{1}$
Without loss of generality, we may suppose that $e_{1}\cap e_{2}=\phi$, and
consider case (1).
Let $e_{1}\in~{}E[\sigma_{1},\sigma_{2}],~{}e_{2}\in
E[\sigma_{3},\sigma_{4}]$, and $C_{i}$ be the fundamental cycle in
$T+e_{i}~{}(~{}1\leq i\leq 2~{})$.
Subcase A. $C_{1}\cap C_{2}$ is a path.
Let $P=C_{1}\cap C_{2}$ be a path with an end vertex $x$ in $C_{1}\cap C_{2}$.
Let $e^{\prime}_{1}$ and $e^{\prime}_{2}$ be two edges such that
$e^{\prime}_{1},~{}e^{\prime}_{2}\in~{}E(T)$ ,and
$x\in~{}e^{\prime}_{1}\cap~{}e^{\prime}_{2}$. Now
$e^{\prime}_{1},~{}e^{\prime}_{2}\not\in~{}E(P)$.( As shown in left hand side
of Fig.2).Consider a new spanning tree
$T^{\prime}=T+\\{e_{1},~{}e_{2}\\}-\\{e^{\prime}_{1},~{}e^{\prime}_{2}\\}.$
$e_{2}$$e_{1}$$P$$C_{1}$$C_{2}$$x$$e^{\prime}_{1}$$e^{\prime}_{2}$$Fig.2$$e_{1}$$x$$P$$e_{2}$$e^{\prime}_{1}$$e^{\prime}_{2}$$C_{1}$$C_{2}$
Subcase B. $x\in~{}e_{1}$ or $x\in~{}e_{2}$, say $x\in~{}e_{2}$. ( As shown in
right hand side of Fig.2).
If $|E(P)|\geq 1$ ,then we take edges $e^{\prime}_{1}~{}\in C_{1}\backslash
E(T),~{}x\in e^{\prime}_{1},~{}e^{\prime}_{2}\in E(P),~{}x\in
e^{\prime}_{2}~{}$. We construct a new spanning tree
$T^{\prime}=T+\\{e_{1},~{}e_{2}\\}-\\{e^{\prime}_{1},~{}e^{\prime}_{2}\\}$. If
$|E(P)|=0$, then this may be a special case of A.
Let $T^{{}^{\prime}}$ be the spanning tree as defined in either subcase A or
B. It is easy to see that $E(G_{1})\backslash E(T^{\prime})$ has at most
$\xi(G_{0})$ odd components . It is contradictory to our suppose. Therefore
$\xi(G_{1})\leq\xi(G_{0})$.
Similarly, We may prove the claims in the cases of (2) and (3).
Repeat this procedure for
$G_{2}=G_{1}+\\{e_{3},e_{4}\\},\cdots,G_{g}=G_{g-1}+\\{e_{2g-1},e_{2g}\\}$
until we get $\xi(G_{g})\leq\xi(G_{g-1})\leq\cdots\leq\xi(G_{0})$, so
$\gamma_{M}(G_{g})\geq\gamma_{M}(G_{0})+g=g$. Theorem 2.Let $G$ be a
connected graph embedded in an orientable surface $S_{g}$ and $T$ be a
spanning tree of $G$. Then there are at least $2g$ noncontractible
foundamental cycles $C_{1},~{}C_{2},\cdots,C_{2g}$, such that $C_{2i-1}\cap
C_{2i}\neq\phi$ for $1\leq i\leq g$. In particular, if $G$ is a one-face-
embedded graph in $S_{g}$, then for any spanning tree $T$ of $G$, there are
$2g$ edges in $G\backslash E(T)$ such that the corresponding $2g$ fundamental
cycles $C_{1},~{}C_{2},\cdots,C_{2g}$ satisfy $C_{2i-1}\cap C_{2i}\neq\phi$
for $1\leq i\leq g$.
Proof . We contract $T$ into a single vertex $v_{T}$ and delete all the
possible edges on distinct faces. Then we get a vertex-graph $G_{T}$ with
exactly one vertex $v_{T}$ and one face in $S_{g}$. There are two crossed
loops, say $e_{\alpha},~{}e_{\beta}$, such that the local rotation of semi-
edges incident to $v_{T}$ is $e_{\alpha}\cdots e_{\beta}\cdots
e_{\alpha}\cdots e_{\beta}$ ( as shown in Fig.3). Furthermore, $e_{\beta}$ is
the only possible edge crossing $e_{\alpha}$ ( since otherwise $G_{T}$ would
have at least two faces! ) . Hence , all ( loop ) edges of $G_{T}$ may be
listed as follows: $e_{1},e_{2},\cdots,e_{2g-1},e_{2g}$ such that $e_{2i-1}$
crossing $e_{2i}$ for $i=1,2,\cdots,g$. It is easy to see that $e_{2i-1}$ and
$e_{2i}$ determine two fundamental cycles $C_{2i-1}$ and $C_{2i}$ with a
vertex in common.
$e_{\beta}$$e_{\alpha}$$e_{\alpha}$$e_{\beta}$$v_{T}$$Fig.3$
Remark: Theorems 1 and 2 give a good characterization of maximum genus of a
graph(i.e., they implies the existence of a polynomially bounded algorithm to
find the maximum genus of a graph).
Let $T$ be a spanning tree in $G$ with a group of fundamental cycles
$C_{1},C_{2},\cdots,C_{2g}$. If $C_{2i-1}\cap C_{2i}\neq\phi$ for $1\leq i\leq
g$, then we say $\langle C_{2i-1},~{}C_{2i}\rangle$ is an adjacent fundamental
cycle pairs $(1\leq i\leq g)$. If $g$ is chosen as the largest number
satisfying above condition, then we call $g$ the maximum number of adjacent
fundamental cycle pairs of $T$. Hence Theorem $2$ implies the following:
Theorem 3 Any two spanning trees $T_{1}$ and $T_{2}$ in a graph $G$ have the
same maximum number of adjacent fundamental cycle pairs. ( In fact , this
unique number is $\gamma_{M}(G)$, the maximum genus of $G$ ).
This generalizes a result of Fu et al[2] where they introduced the concept
intersecting graph which is determined by bases of cycle space of a graph to
describe the maximum genus of a graph. In fact, our result stands for any
spanning tree’s fundamental cycles.
###### Corollary 1
If a connected graph $G$ has a spanning tree $T$ such that any two fundamental
cycles have a vertex in common. Then $G$ is upper-embeddable.
Sometimes however, we need a refined form of Theorems 1 and 2 in practice. The
following result gives us a recursive relation between the maximum genera of a
graph and its subgraph(s).
Theorem 4 Let $G$ be a connected graph and $T$ be an arbitrary spanning tree
in $G$. If $e_{1},e_{2}$ are two edges not in $G$ and the two cycles
$C_{T}(e_{1})$ and $C_{T}(e_{2})$ have a vertex in common. Then
$\gamma_{M}(G)=\gamma_{M}(G+e_{1}+e_{2})-1$. In particular, $G$ is upper-
embeddable if and only if $G+e_{1}+e_{2}$ is upper-embeddable.
One may easily see that this generalizes a recursive relation for maximum
genus of Xuong[9] and ( we will see in the next section )is much more
practical in use.
## 3 Applications
Now in this section, we begin to apply Theorems $1-2$ to determine the maximum
genus of some type of graphs.
Let us recall that the essence of Xuong’s method[9] consists of two parts: one
is to find an optimal tree $T$ in a graph $G$ having the smallest number of
odd components; the other is to organize edges of $E(G)\backslash E(T)$ into
adjacent pairs such as
$E(G)\backslash
E(T)=\\{e_{1},e_{2},\cdots,~{}e_{2s}\\}\cup\\{~{}f_{1},~{}f_{2},~{}\cdots~{}f_{m}\\},$
where $e_{2i-1}\cap~{}e_{2i}\neq\phi$($1\leq~{}i\leq~{}s$ ) and
$C_{T}(f_{i})\cap{C_{T}(f_{j})}=\phi$, for $1\leq~{}i<j\leq~{}m$ and
$s=\gamma_{M}(G),~{}m=\xi(G)$. Compared with the above procedure, Theorems 1
and 2 consider adjacent foundamental cycle pairs(rather than adjacent pairs of
edges). We may construct large genus embedding from any spanning tree $T$,
although it may have very large number of odd components in $G\backslash
E(T)$. This greatly releases the conditions of Xuong. Of course, an optimal
tree is also valid in our constructions. Hence, Theorems 1 and 2 generalize
Xuong’s characterization of maximum genus. Based on this idea, we may
construct a large orientable genus as follows: Take a specific spanning tree
$T$ in graph $G$ and first organize some non-tree edges into adjacent pairs (
as Xuong did ) and then match other possible non-tree edges into pairs such
that their fundamental cycles also become adjacent fundamental cycle pairs. It
is easy for one to see that the second part of non-tree edges may be chosen as
an edge-cut of $G$. Therefor, Theorems $1-2$ may be useful in determination of
a maximum genus of a graph $G$ with a specific edge-cut. Now, the following
result is easy to be verified.
Theorem 5 Let $A=\\{e_{1},~{}e_{2},\cdots,~{}e_{k}~{}\\}$ be an edge-cut of
$G$ such that $G-A$ has exactly two components $G_{1}$ and $G_{2}$. If both
$G_{1}$ and $G_{2}$ are upper-embeddable, then
$~{}\gamma_{M}(G)\geq\lfloor\frac{\beta(G)}{2}\rfloor-1$ . Furthermore,if $G$
satisfies one of the following conditions , then $G$ is upper-embeddable:
(1). $\beta(G_{1})\equiv\beta(G_{2})\equiv~{}0(~{}mod~{}2~{})$
(2). $|A|\equiv~{}1(~{}mod~{}2~{})$ and
$\beta(G_{1})+\beta(G_{2})\equiv~{}1(~{}mod~{}2~{})$.
The next result is due to Huang. As a consequence of the above results, we
will give another proof.
Theorem 6(Huang[5]) Let $G$ be a strongly embedded graph in an orientable
surface $S_{g}$ ( i.e., all facial walks are cycles ). If the dual graph
$G^{*}$ of $G$ has a surface separating Hamiltonian cycle, then $G$ is upper-
embeddable.
Proof We will show the existence of a spanning tree $T$ of $G$ satisfying the
conditions in Theorems 1 and 2\. Let
$\mathcal{F}=\\{f_{1},~{}f_{2},\cdots~{}f_{\varphi}\\}$ be the face-set of $G$
and $C^{*}$ be a surface separating Hamiltonian cycle in $G^{*}$. Let
$E(C^{*})=\\{e^{*}_{1},~{}e^{*}_{2},\cdots~{}e^{*}_{\varphi}~{}\\}$ and
$e^{*}_{i}~{}=(f_{i},f_{i+1})$ for $1\leq i\leq\varphi$. Let $e_{i}$ be the
edge in $\partial f_{i}\cap\partial f_{i+1}$ corresponding to $e^{*}_{i}$ for
$1\leq i\leq\varphi$ (where $\partial f_{i}$ denotes the boundary of $f_{i}$).
Claim 2. $G-\\{e_{1},~{}e_{2},\cdots~{}e_{\varphi-1}\\}$ is a one-face
embedded subgraph of $G$ in $S_{g}$, Furthermore,
$G-\\{e_{1},~{}e_{2},\cdots~{}e_{\varphi}\\}$ has exactly two components
$G_{1}$ and $G_{2}$.
Now $E[G_{1},~{}G_{2}]=\\{e_{1},~{}e_{2},\cdots~{}e_{\varphi}\\}$. Let
$G_{1}\subset~{}Int(C^{*})$ and $G_{2}\subset~{}Ext(C^{*})$ and $\partial
f_{i}$ denotes the boundary cycle of $f_{i}$ for $1\leq i\leq\varphi$. Then we
may construct a graph as follows. $H_{0}=(\partial f_{1}\cup\partial
f_{2}\cup\cdots\cup\partial
f_{\varphi-1})\setminus\\{e_{1},~{}e_{2},\cdots~{}e_{\varphi-1}\\}$. It is
easy to see that $H_{0}$ is a connected spanning subgraph of
$G-\\{e_{1},~{}e_{2},\cdots~{}e_{\varphi-1}\\}$. ( Hence, a spanning subgraph
of $G$). Let $e_{\varphi}=(\alpha,\beta)$ with
$\alpha\in~{}V(G_{1}),~{}\beta\in~{}V(G_{2})$. Then $H_{0}-e_{\varphi}$ has
exactly two components $H^{\prime},~{}H_{1}$ with $H^{\prime}=G_{1}$.
Claim 3. If $H_{1}$ has a cycle $C$, then $C$ must be a noncontractible cycle.
This follows from the fact that $S_{g}-H_{0}$ has only one component. If
$H_{1}$ has a cycle $C_{1}$, then delete an edge $e^{\prime}_{1}\in~{}C_{1}$
and get a subgraph $H_{2}$ of $H_{1}$ with $V(H_{2})=V(H_{1})$. Repeat this
procedure until we arrive at a connected subgraph $H_{k}$ of $H_{1}$ with
$V(H_{k})=V(H_{1})$ and $H_{k}$ has no cycle.
Claim 4: $T=H^{\prime}\cup~{}H_{k}\cup~{}\\{e_{\varphi}\\}$ is a spanning tree
of $G$, such that each fundamental cycle $C_{T}(e_{i})$ in $T+e_{i}$ has an
edge $e_{\varphi}$ in common for $i=1,2,\cdots,\varphi-1~{}$.
To see this, we consider an edge
$e_{i}=(x_{i},y_{i})\in[H^{\prime}~{},~{}H_{k}]\subseteq[G_{1},~{}G_{2}]$,
such that $x_{i}\in H^{\prime},~{}y_{i}\in H_{k}$. Since $H^{\prime}(H_{k})$
is connected, there is a path$P_{i}(Q_{i})$ in $H^{\prime}(H_{k})$ joining
$\alpha(\beta)$ and $x_{i}(y_{i})$. Hence,
$C_{T}(e_{i})=\\{e_{\varphi}\\}\cup~{}P_{i}\cup~{}Q_{i}\cup~{}\\{e_{i}\\}$ is
a cycle containing $e_{\varphi}$ for $1\leq i\leq\varphi$.
Now we find a spanning tree $T$ of $G$ such that:
$(1)$. All the fundamental cycles
$C_{T}(e_{1}),~{}C_{T}(e_{2}),\cdots,~{}C_{T}(e_{\varphi-1})$ has an edge in
common;
$(2)$. By Theorems 1 and 2, and the fact that $T$ is also a spanning tree in
$G-\\{e_{1},~{}e_{2},\cdots~{}e_{\varphi-1}\\}$, there are another group of
fundamental ( noncontractible ) cycles $C_{1},~{}C_{2},\cdots,~{}C_{2g}$ such
that $C_{2i-1}\cap~{}C_{2i}\neq{\o}$ for $1\leq i\leq g$. By theorem 1, $G$ is
upper-embeddable.
One may readily see that a surface separating cycle may not be Hamiltonian and
the hosting surface on which a graph is embedded may not be orientable. Thus,
Theorem 6 can be extended to a much more generalized form.
Theorem 7 Let $G$ be an embedded graph in a surface $\sum$ such that the dual
graph $G^{*}$ of $G$ contains a surface separating cycle $C^{*}$ such that
both of the left subgraph $G_{L}(C^{*})$ and right subgraph $G_{R}(C^{*})$ of
$G$ are upper embeddable . Then
$~{}\gamma_{M}(G)\geq\lfloor\frac{\beta(G)}{2}\rfloor-1$. In particular, if
${\beta(G_{L}(C^{*}))}\equiv{\beta(G_{R}(C^{*}))}\equiv 0\,(\,mod\,2)$, then
$G$ is upper-embeddable.
Remark: The term “ left( right ) subgraph ” follows from [7]
###### Corollary 2
If $G$ is an embedded graph on the Klein bottle such that the dual graph
$G^{*}$ has a surface separating Hamiltonian cycle. Hence
$\gamma_{M}(G)\geq\lfloor\frac{\beta(G)}{2}\rfloor-1$.
In practical use, our attentions need not to be restricted to graphs with an
edge-cut. Theorems 1-4 provide us a tool to evaluating large genus embeddings
in more extended range of graphs. The following results show us how to do so (
we omit the proof of them ).
Theorem 8. The following graphs are upper-embeddable:
(1). The cartisian product $G\times P_{n}$ of a simple connected graph $G$ and
a path $P_{n}$ with $n(\geq 1)$ egdes;
(2). The composition of two disjoint Halin graphs $H_{1}$ and $H_{2}$ with
some edges $e_{1},e_{2},\cdots,e_{k}(k\geq 2)$ connecting them;
(3). The $n$-cube $Q_{n}$ which is composed of two $(n-1)$-cube $Q_{n-1}$
together with some edges joining the two copies of vertice in $Q_{n-1}$.
(4). The generalized Petersen graphs $P(n,k)$ which is determined by $n-$cycle
$(u_{1},u_{2},\cdots,u_{n})$ and vertices $v_{1},v_{2},\cdots,v_{n}$ such that
(i) each $(u_{i},v_{i})\in E,1\leq i\leq n$;(ii) $(u_{i},v_{i+k})\in E,1\leq
i\leq n$.
Note: A graph $G=(V,~{}E)$ is a Halin graph if $G$ is obtained by joining the
leaves(1-valent vertices) of a plane tree $T$ with a cycle in this orientation
and the definition of cartisian product of two graphs may be fund in and
textbook of graph theory.
## 4 A polynomially bounded algorithm
In this section we shall present a polynomially bounded algorithm to find the
maximum genus of a given graph. A basic fact is that Theorems 1 and 2 present
a good characterization of maximum genus problem, i.e., we have the following
Theorem 9 To determine the maximum genus of a graph $G$ is equivalent to
determine a maximum matching of the graph $G_{M}=(V_{M},E_{M})$, called
fundamental intersecting graph of $G$, where $V_{M}$ is the set of fundamental
cycles of a spanning tree $T$ of $G$ and any two cycles in $V_{M}$ are
adjacent if and only if they have at least a vertex in common.
We observe that the fastest algorithm to find a maximum matching in a graph
$G$ is due to Micali-Vazirani[8] which will end in $O(m\sqrt{n})$ steps, where
$m$ and $n$ are, respectively, the number of edges and vertices of $G$. Based
on this fact and Theorems 1 and 2 we may construct a new algorithm to
determine the maximum genus of a graph $G$.
Fundamental cycle algorithm
Step 1.Input the date of the graph $G$ and then searching for a spanning tree
$T$ and the set $V_{M}$ of fundamental cycles in $G$;
Step 2. For cycles in $V_{M}$ we build the graph $G_{M}$;
Step 3 Perform Micali-Vazirani algorithm to find a maximum matching in $G_{M}$
and then terminate.
Remark: Since the number of fundamental cycles in a graph $G$ of order $n$ is
${(\beta(G))}$, this algorithm will end in at most
$O({(\beta(G))}^{\frac{5}{2}})$ steps. Although Furst, Gross and McGeoch had
already construct the first polynomially bounded algorithm[3], this result is
a new approach to do so.
## References
* [1] J.A. Bondy, U.S.R Murty, Graph Theory with Application, MacMillan, London, 1976.
* [2] Hung-Lin Fu, Hsinchu, M. Skoviera, Bratislava, and M.Tsai, The maximum genus, matchings and the cycle space of a graph, Csechoslovak Math. J., 48(123)(1998),329-339
* [3] M.L.Furst, J.L.Gross,L.A.McGeoch, Finding a maximum-genus graph imbedding, J.Assoc.Comput.Mach.35(1988), 523 - 534.
* [4] R.Giles, Optimum matching forest I:Special weights, Math.Programming 22(1982), 1 - 11.
* [5] Y. Huang, Maximum genus of a graph in term of its embedding properties, Discreate Math. 262 (2003) 171 - 180.
* [6] Y.P.Liu The maximum orientable genus of a graph(Chinese with English abstract). _Scientia Sinica, Special Issue on Math.II,_ , 41–55 (1979)
* [7] B.Mohar, C.Thomassen. Graphs on Surface, Johns Hopkins University Press ,2001.
* [8] S.Micali, V.V.Vazirani, An $O(\sqrt{|V|}|E|)$ algorithm for finding maximum matching in general graphs. In Proc.21th IEEE Symp.Found.Comp.Sci. ACM(1980), 17 - 27
* [9] N.H. Xoung, How to determine the maximum genus of a graph, J.Combin. Theory Ser.B 26 (1979) 217 - 225.
| arxiv-papers | 2008-07-10T13:51:17 | 2024-09-04T02:48:56.676539 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Han Ren, Hongtao Zhao, Haoling Li",
"submitter": "Ren Han",
"url": "https://arxiv.org/abs/0807.1652"
} |
0807.1697 | 2.5cm2.5cm2.5cm2.5cm
# Investigation of solutions of boundary value problems for a composite type
equation with non-local boundary conditions
Aliev N.A., Aliev A.M
###### Abstract.
Since the order of elliptic type model equation (Laplace equation) is two [1],
[2], then it is natural the order of composite type model equation must be [3]
[4] [5] three. At each point of the domain under consideration these equations
have both real and complex characteristics.
Notice that a boundary value problem for a composite type equation of second
order first appeared in the paper [6].
The method for investigating the Fredholm property of boundary value problems
is distinctive and belongs to one of the authors of the present paper.
Key words: Composite type equations, non local boundary conditions for partial
differential equations, fundamental solution, necessary condition,
regularization, Fredholm property.
_Faculty of Applied Mathematics and Cybernetics,Institute of Applied
Mathematics,
Baku State University,Z.Khalilov str.23,AZ1148, Azerbaijan,
e-mail:nihan_aliev@rambler.ru, aahmad07@rambler.ru_
## Introduction.
The paper is devoted to the investigation of boundary value problems for a
composite type equation of second order.
This was possible as earlier we investigated an elliptic type boundary value
problem of first order ( Cauchy-Riemann equation) for which the boundary is a
carrier of boundary conditions [7]. Notice that in this case it is impossible
to determine local boundary conditions (since undeterminacy is obtained).
Therefore non-local boundary conditions were considered.
Necessary conditions that contain singular integrals are obtained proceeding
from fundamental solution of Cauchy-Riemann equation [8]. Considering that we
are on a spectrum, regularization of these singularities is also conducted in
distinctive way [6]. Joining regularized necessary conditions with the given
boundary conditions we get a sufficient condition for Fredholm property of the
stated boundary value problems.
Notice that in [8] the cited investigation of the process in a nuclear reactor
leads to a boundary value problem for first order integro-differential
equation in three-dimensional space where not all the space is a carrier of
the given local boundary condition.
## Problem statement
Let’s consider the following boundary value problem:
$\ell\,u\equiv\frac{\partial^{2}u(x)}{\partial
x_{2}^{2}}+i\frac{\partial^{2}u(x)}{\partial x_{1}\,\partial
x_{2}}=0,\,\,\,\,\,x\in D,$ (1)
$\ell_{k}u\equiv\left.\frac{\partial
u(x)}{\partial\,x_{2}}\right|_{x_{2}=\gamma_{k}(x_{1})}+\alpha_{k}u(x_{1},\gamma_{k}(x_{1}))=\varphi_{k}(x_{1}),\,\,\,k=1,\,2;\,\,x_{1}\in\left[a_{1},b_{1}\right]$
(2)
where $i=\sqrt{-1}$, $D\subset R^{2}$\- is a bounded domain convex in the
direction $x_{2}$, the boundary $\Gamma=\bar{D}\backslash D$ \- is a Liapunov
line, $\gamma_{k}(x_{1})$, $k=1,\,2$ are the equations of open lines
$\Gamma_{k}$ ($\Gamma_{1}\bigcup\Gamma_{2}=\Gamma$), obtained from the
boundary $\Gamma$ of the domain $D$ by means of orthogonal projection of this
domain on the axis $x_{1}$ parallel to the axis $x_{2}$ and
$\left[a_{1},\,b_{1}\right]=np_{x_{1}}\Gamma_{1}=np_{x_{1}}\Gamma_{2}$. In the
given boundary conditions (17) $\alpha_{k}$ ($k=1,\,2$) are constants,
$\varphi_{k}(x_{1})$, $k=1,\,2$; $x_{1}\in\left[a_{1},\,b_{1}\right]$ are
sufficiently smooth functions. Boundary conditions (17) are assumed to be
linear independent .
## Fundamental solution.
Applying the Fourier transform to equation (16) we get a fundamental solution
in the form
$U(x-\xi)=\frac{-1}{4\pi^{2}}\int_{R^{2}}\frac{e^{i(\alpha,\,x-\xi)}}{\alpha_{2}(\alpha_{2}+i\,\alpha_{1})}d\alpha,$
(3)
where
$(\alpha_{1}x-\xi)=\sum_{j=1}^{2}\alpha_{j}(x_{j}-\xi_{j}).$
Further, since
$\frac{1}{2\pi
i}\int_{R}\frac{e^{i\alpha_{2}(x_{2}-\xi_{2})}}{\alpha_{2}}d\alpha_{2}=e(x_{2}-\xi_{2}),$
where $e(x_{2}-\xi_{2})$ is a unique symmetric Heaviside function, from (3) we
get:
$\frac{\partial U(x-\xi)}{\partial x_{2}}+i\frac{\partial U(x-\xi)}{\partial
x_{2}}=e(x_{2}-\xi_{2})\,\delta(x_{1}-\xi_{1}).$ (4)
Finally considering that (3) is a fundamental solution of the composite type
equation (16), we get that $\frac{\partial U(x-\xi)}{\partial x_{2}}$ is a
fundamental solution of the Cauchy-Riemann equation. Making negligible changes
in the fundamental solution of the Cauchy-Riemann equation [8] we get a
fundamental solution in the direction $x_{2}$
$\frac{\partial U(x-\xi)}{\partial
x_{2}}=\frac{1}{2\pi}\frac{\theta(x_{2}-\xi_{2})+\theta(\xi_{2}-x_{2})}{x_{2}-\xi_{2}+i(x_{1}-\xi_{1})}.$
(5)
here $\theta(x_{2}-\xi_{2})+\theta(x_{2}-\xi_{2})=1$, if none differentiation
operation is produced on it, since each addend has a break and contribution of
this break appears in differentiation. Thus, for fundamental solution (3) of
composite type equation (16) we get:
$U(x-\xi)=\frac{1}{2\pi}\int_{0}^{x_{2}}\frac{\theta(t-\xi_{2})+\theta(\xi_{2}-t)}{t-\xi_{2}+i(x_{1}-\xi_{1})}dt,$
(6)
i.e. it holds the following statement:
Theorem 1. For a composite type equation of second order (16) a fundamental
solution in the direction $x_{2}$ is of the form (6).
This means that if we differentiate $U(x-\xi)$ twice with respect to $x_{2}$
and twice with respect to the mixed derivatives $x_{1}$ and $x_{2}$, the Dirac
delta function (two-dimensional) appears only in the derivative of second
order with respect to $x_{2}$.
## Necessary conditions.
Multiplying equation (16) by fundamental solution (6) and integrating if in
domain $D$, applying Ostrogradskii-Gauss formula [8], we get formula similar
to Green’s second formula that after application of fundamentality properties
of function (6) get the form:
$\int_{a_{1}}^{b_{1}}\left.\left[u(x)\frac{\partial U(x-\xi)}{\partial
x_{2}}-\frac{\partial u(x)}{\partial
x_{2}}U(x-\xi)\right]\right|_{x_{2}=\gamma_{1}(x_{1})}^{\gamma_{2}(x_{1})}dx_{1}+i\int_{a_{1}}^{b_{1}}\left.u(x)\frac{\partial
U(x-\xi)}{\partial
x_{1}}\right|_{x_{2}=\gamma_{1}(x_{1})}^{\gamma_{2}(x_{1})}dx_{1}+$
$+i\int_{a_{1}}^{b_{1}}\left.\frac{\partial u(x)}{\partial
x_{2}}U(x-\xi)\right|_{x_{2}=\gamma_{2}(x_{1})}\gamma^{\prime}_{2}(x_{1})dx_{1}-i\int_{a_{1}}^{b_{1}}\left.\frac{\partial
U(x)}{\partial
x_{2}}U(x-\xi)\right|_{x_{2}=\gamma_{1}(x_{1})}\gamma^{\prime}_{2}(x_{1})dx_{1}=$
$=\left\\{\begin{array}[]{l}{u(\xi),\,\,\,\,\,\,\,\xi\in D,}\\\
{\frac{1}{2}u(\xi),\,\,\xi\in\Im}\end{array}\right.$ (7)
The second expression in formula (7) is one of the necessary conditions. This
condition has the form:
$u(\xi_{1},\gamma_{1}(\xi_{1}))=u(\xi_{1},\gamma_{2}(\xi_{1}))-2\int_{a_{1}}^{b_{1}}\left.\frac{\partial
u(x)}{\partial
x_{2}}\right|_{x_{2}=\gamma_{2}(x_{1})}U(x_{1}-\xi_{1},\gamma_{2}(x_{1})-\gamma_{1}(\xi_{1}))\left[1-i\gamma^{\prime}_{2}(x_{1})\right]dx_{1}+$
$+2\int_{a_{1}}^{b_{1}}\left.\frac{\partial u(x)}{\partial
x_{2}}\right|_{x_{2}=\gamma_{1}(x_{1})}U(x_{1}-\xi_{1},\gamma_{1}(x_{1})-\gamma_{1}(\xi_{1}))\left[1-i\gamma^{\prime}_{1}(x_{1})\right]dx_{1}$
(8)
In exactly the same way to [9] and [10], we get the following necessary
conditions:
$\left.\frac{\partial
u(\xi)}{\partial\xi_{1}}\right|_{\xi_{2}=\gamma_{1}(\xi_{1})}=\left.\frac{\partial
u(\xi)}{\partial\xi_{1}}\right|_{\xi_{2}=\gamma_{2}(\xi_{1})}-i\left.\frac{\partial
u(\xi)}{\partial\xi_{2}}\right|_{\xi_{2}=\gamma_{2}(\xi_{1})}-$
$-2i\int_{a_{1}}^{b_{1}}\left.\frac{\partial u(x)}{\partial
x_{2}}\right|_{x_{2}=\gamma_{1}(x_{1})}\left.\frac{\partial U(x-\xi)}{\partial
x_{2}}\right|_{\begin{array}[]{l}{x_{2}=\gamma_{1}(x_{1})}\\\
{\xi_{2}=\gamma_{1}(\xi_{1})}\end{array}}\left[1-i\gamma^{\prime}_{1}(x_{1})\right]dx_{1}+$
$+2i\int_{a_{1}}^{b_{1}}\left.\frac{\partial u(x)}{\partial
x_{2}}\right|_{x_{2}=\gamma_{2}(x_{1})}\left.\frac{\partial U(x-\xi)}{\partial
x_{2}}\right|_{\begin{array}[]{l}{x_{2}=\gamma_{2}(x_{1})}\\\
{\xi_{2}=\gamma_{1}(\xi_{1})}\end{array}}\left[1-i\gamma^{\prime}_{2}(x_{1})\right]dx_{1},$
(9)
$\left.\frac{\partial
u(\xi)}{\partial\xi_{2}}\right|_{\xi_{2}=\gamma_{1}(\xi_{1})}=2\int_{a_{1}}^{b_{1}}\left.\frac{\partial
u(x)}{\partial x_{2}}\right|_{x_{2}=\gamma_{2}(x_{1})}\left.\frac{\partial
U(x-\xi)}{\partial
x_{2}}\right|_{\begin{array}[]{l}{x_{2}=\gamma_{2}(x_{1})}\\\
{\xi_{2}=\gamma_{1}(\xi_{1})}\end{array}}\left[1-i\gamma^{\prime}_{2}(x_{1})\right]dx_{1}-$
$-2\int_{a_{1}}^{b_{1}}\left.\frac{\partial u(x)}{\partial
x_{2}}\right|_{x_{2}=\gamma_{1}(x_{1})}\left.\frac{\partial U(x-\xi)}{\partial
x_{2}}\right|_{\begin{array}[]{l}{x_{2}=\gamma_{1}(x_{1})}\\\
{\xi_{2}=\gamma_{1}(\xi_{1})}\end{array}}\left[1-i\gamma^{\prime}_{1}(x_{1})\right]dx_{1},$
(10)
$\left.\frac{\partial
u(\xi)}{\partial\xi_{1}}\right|_{\xi_{2}=\gamma_{2}(\xi_{1})}=\left.\frac{\partial
u(\xi)}{\partial\xi_{1}}\right|_{\xi_{2}=\gamma_{1}(\xi_{1})}-i\left.\frac{\partial
u(\xi)}{\partial\xi_{2}}\right|_{\xi_{2}=\gamma_{1}(\xi_{1})}-$
$-2i\int_{a_{1}}^{b_{1}}\left.\frac{\partial u(x)}{\partial
x_{2}}\right|_{x_{2}=\gamma_{1}(x_{1})}\left.\frac{\partial U(x-\xi)}{\partial
x_{2}}\right|_{\begin{array}[]{l}{x_{2}=\gamma_{1}(x_{1})}\\\
{\xi_{2}=\gamma_{2}(\xi_{1})}\end{array}}\left[1-i\gamma^{\prime}_{1}(x_{1})\right]dx_{1}+$
$+2i\int_{a_{1}}^{b_{1}}\left.\frac{\partial u(x)}{\partial
x_{2}}\right|_{x_{2}=\gamma_{2}(x_{1})}\left.\frac{\partial U(x-\xi)}{\partial
x_{2}}\right|_{\begin{array}[]{l}{x_{2}=\gamma_{2}(x_{1})}\\\
{\xi_{2}=\gamma_{2}(\xi_{1})}\end{array}}\left[1-i\gamma^{\prime}_{2}(x_{1})\right]dx_{1},$
(11)
$\left.\frac{\partial
u(\xi)}{\partial\xi_{2}}\right|_{\xi_{2}=\gamma_{2}(\xi_{1})}=2\int_{a_{1}}^{b_{1}}\left.\frac{\partial
u(x)}{\partial x_{2}}\right|_{x_{2}=\gamma_{2}(x_{1})}\left.\frac{\partial
U(x-\xi)}{\partial
x_{2}}\right|_{\begin{array}[]{l}{x_{2}=\gamma_{2}(x_{1})}\\\
{\xi_{2}=\gamma_{2}(\xi_{1})}\end{array}}\left[1-i\gamma^{\prime}_{2}(x_{1})\right]dx_{1}-$
$-2\int_{a_{1}}^{b_{1}}\left.\frac{\partial u(x)}{\partial
x_{2}}\right|_{x_{2}=\gamma_{1}(x_{1})}\left.\frac{\partial U(x-\xi)}{\partial
x_{2}}\right|_{\begin{array}[]{l}{x_{2}=\gamma_{1}(x_{1})}\\\
{\xi_{2}=\gamma_{2}(\xi_{1})}\end{array}}\left[1-i\gamma^{\prime}_{1}(x_{1})\right]dx_{1}.$
(12)
Thus, we established the following statement:
Theorem 2. Let $D$ be a plane domain convex in the direction $x_{2}$, the
boundary $\Gamma$ be Liapunov line, then each solution of equation (16)
determined in the domain $D$ satisfies the necessary conditions (8)–(12),
containing singular integrals besides (8).
## Regularization.
As it was said above, necessary conditions (9)–(12) contain singular addends.
Considering (5), we have:
$\left.\frac{\partial U(x-\xi)}{\partial
x_{2}}\right|_{\begin{array}[]{l}{x_{2}=\gamma_{k}(x_{1})}\\\
{\xi_{2}=\gamma_{k}(\xi_{1})}\end{array}}=\frac{1}{2\pi}\cdot\frac{1}{\gamma_{k}(x_{1})-\gamma_{k}(\xi_{1})+i(x_{1}-\xi_{1})}=$
$=\frac{1}{2\pi}\frac{1}{x_{1}-\xi_{1}}\cdot\frac{1}{\gamma^{\prime}_{k}(\sigma_{k}(x_{1},\,\xi_{1}))+i},k=1,\,2,$
where $\sigma_{k}(x_{1},\,\xi_{1})$ is located between $x_{1}$ and $\xi_{1}$.
Then from (9) – (12) we find:
$\left.\frac{\partial
u(\xi)}{\partial\xi_{1}}\right|_{\xi_{2}=\gamma_{1}(\xi_{1})}-\left.\frac{\partial
u(\xi)}{\partial\xi_{1}}\right|_{\xi_{2}=\gamma_{2}(\xi_{1})}+i\left.\frac{\partial
u(\xi)}{\partial\xi_{2}}\right|_{\xi_{2}=\gamma_{2}(\xi_{1})}=$
$=-\frac{1}{\pi}\int_{a_{1}}^{b_{1}}\left.\frac{\partial u(x)}{\partial
x_{2}}\right|_{x_{2}=\gamma_{1}(x_{1})}\frac{dx_{1}}{x_{1}-\xi_{1}}+......,$
$\left.\frac{\partial
u(\xi)}{\partial\xi_{2}}\right|_{\xi_{2}=\gamma_{1}(\xi_{1})}=\frac{i}{\pi}\int_{a_{1}}^{b_{1}}\left.\frac{\partial
u(x)}{\partial
x_{2}}\right|_{x_{2}=\gamma_{1}(x_{1})}\frac{dx_{1}}{x_{1}-\xi_{1}}+......,$
(13)
$\left.\frac{\partial
u(\xi)}{\partial\xi_{1}}\right|_{\xi_{2}=\gamma_{2}(\xi_{1})}-\left.\frac{\partial
u(\xi)}{\partial\xi_{1}}\right|_{\xi_{2}=\gamma_{1}(\xi_{1})}+i\left.\frac{\partial
u(\xi)}{\partial\xi_{2}}\right|_{\xi_{2}=\gamma_{1}(\xi_{1})}=$
$=\frac{1}{\pi}\int_{a_{1}}^{b_{1}}\left.\frac{\partial u(x)}{\partial
x_{2}}\right|_{x_{2}=\gamma_{2}(x_{1})}\frac{dx_{1}}{x_{1}-\xi_{1}}+......,$
$\left.\frac{\partial
u(\xi)}{\partial\xi_{2}}\right|_{\xi_{2}=\gamma_{2}(\xi_{1})}=-\frac{i}{\pi}\int_{a_{1}}^{b_{1}}\left.\frac{\partial
u(x)}{\partial
x_{2}}\right|_{x_{2}=\gamma_{2}(x_{1})}\frac{dx_{1}}{x_{1}-\xi_{1}}+......$
where the sum of non-singular addends are denoted by dots.
Considering boundary conditions (17), from necessary conditions (8) for
boundary values of the unknown function we get the following regular
relations:
$\begin{array}[]{l}{u(\xi_{1},\gamma_{1}(\xi_{1})=u(\xi_{1},\,\gamma_{2}(\xi_{1}))-}\\\
{-2\int_{a_{1}}^{b_{1}}\left[\varphi_{2}(x_{1})-\alpha_{2}u(x_{1},\,\gamma_{2}(x_{1})\right]U(x_{1}-\xi_{1},\,\gamma_{2}(x_{1})-\gamma_{1}(\xi_{1}))\,\left[1-i\gamma^{\prime}_{2}(x_{1})\right]dx_{1}+}\end{array}$
$+2\int_{a_{1}}^{b_{1}}\left[\varphi_{1}(x_{1})-\alpha_{1}u(x_{1},\,\gamma_{1}(x_{1})\right]U(x_{1}-\xi_{1},\,\gamma_{1}(x_{1})-\gamma_{1}(\xi_{1}))\,\left[1-i\gamma^{\prime}_{1}(x_{1})\right]dx_{1},$
(14)
In exactly the same way, from (13) we get:
$\varphi_{1}(\xi_{1})-\alpha_{1}u(\xi_{1},\,\gamma_{1}(\xi_{1}))=\frac{i}{\pi}\int_{a_{1}}^{b_{1}}\left[\varphi_{1}(x_{1})-\alpha_{1}u(x_{1},\gamma_{1}(x_{1}))\right]\frac{dx_{1}}{x_{1}-\xi_{1}}+......,$
$\varphi_{2}(\xi_{1})-\alpha_{2}u(\xi_{1},\,\gamma_{2}(\xi_{1}))=-\frac{i}{\pi}\int_{a_{1}}^{b_{1}}\left[\varphi_{2}(x_{1})-\alpha_{2}u(x_{1},\gamma_{2}(x_{1}))\right]\frac{dx_{1}}{x_{1}-\xi_{1}}+.......$
Finally, proceeding from (14) for boundary values of the unknown function we
get the following regular relation [9],[10]
$\begin{array}[]{l}{\frac{\varphi_{1}(\xi_{1})}{\alpha_{1}}+\frac{\varphi_{2}(\xi_{1})}{\alpha_{2}}-\left[u(\xi_{1},\gamma_{1}(\xi_{1}))+u(x_{1},\gamma_{2}(\xi_{1})\right]=\frac{i}{\pi}\int_{a_{1}}^{b_{1}}\left[\frac{\varphi_{1}(x_{1})}{\alpha_{1}}-\frac{\varphi_{2}(x_{1})}{\alpha_{2}}\right]\frac{dx_{1}}{x_{1}-\xi_{1}}-}\\\
{-\frac{i}{\pi}\int_{a_{1}}^{b_{1}}\left\\{-2\int_{a_{1}}^{b_{1}}\left[\varphi_{2}(\eta_{1})-\alpha_{2}u(\eta_{1},\gamma_{2}(\eta_{1}))\right]\,U(\eta_{1}-x_{1},\,\gamma_{2}(\eta_{1})-\gamma_{1}(x_{1}))\left[1-i\gamma^{\prime}_{2}(\eta_{1})\right]\,d\eta_{1}+\right.}\end{array}$
$\left.+2\int_{a_{1}}^{b_{1}}\left[\varphi_{1}(\eta_{1})-\alpha_{1}u(\eta_{1},\gamma_{1}(\eta_{1})\right]\,U(\eta_{1}-x_{1},\,\gamma_{1}(\eta_{1})-\gamma_{1}(x_{1}))\left[1-i\gamma^{\prime}_{1}(\eta_{1})\right]\,d\eta_{1}\right\\}\frac{dx_{1}}{x_{1}-\xi_{1}}+...$
(15)
that is regular if we interchange the integrals contained in the right hand
side of (15) and consider the singular integrals of unknown functions
calculated in [11]. Thus we proved the following statement:
Theorem 3. When fulfilling the conditions of Theorem 2 if
$\varphi_{k}(x_{1})$, $k=1,\,2$ are continuously differentiable functions
vanishing at the end of the interval $(a_{1},\,b_{1})$, then (15) are regular
relations.
## Fredholm property.
Considering boundary conditions (17), the first necessary condition (8) not
containing singular integrals leads to regular relation (14).
Further, proceeding from boundary conditions, after regularizing two necessary
conditions given in (13), that contain singular integrals, we get a relation
that has no singularity in the form (15).
It holds :
Theorem 4. When fulfilling conditions of Theorem 3 boundary value problem
(16)–(17) is Fredholm.
Really, it is easy to get from (14) and (15) a system of Fredholm integral
equations of second kind with respect to the unknown functions
$u(x_{1},\gamma_{k}(x_{1}))$, $k=1,\,2$, in which a kernel may have only weak
singularity.
## Unsolved problems.
## 1\. The inverse problem in Tikhonov-Lavrent’ev sense.
Let’s consider the problem
$\frac{\partial^{2}u(x)}{\partial x_{2}^{2}}+i\frac{\partial^{2}u(x)}{\partial
x_{1}\partial x_{2}}=0,\,\,\,\,x\in D,$ (16)
$\left.\frac{\partial u(x)}{\partial
x_{2}}\right|_{x_{2}=\gamma_{k}(x_{1})}+\alpha_{k}u(x_{1},\gamma_{k}(x_{1}))=\varphi_{k}(x_{1}),\,\,\,\,k=1,\,2;\,\,x_{1}\in\left[a_{1},\,b_{1}\right],$
(17)
with the following complementary restriction
$\begin{array}[]{l}{\left.\alpha_{1}(x_{1})\frac{\partial u(x)}{\partial
x_{2}}\right|_{x_{2}=\gamma_{1}(x_{1})}+\left.\alpha_{2}(x_{1})\frac{\partial
u(x)}{\partial x_{2}}\right|_{x_{2}=\gamma_{2}(x_{1})}+}\\\
{+\alpha_{3}u(x_{1},\gamma_{1}(x_{1}))+\alpha_{4}(x_{1})\,u(x_{1},\,\gamma_{2}(x_{1}))=\varphi_{3}(x_{1}),\,\,\,\,\,\,x_{1}\in\left[a_{1},\,b_{1}\right]}\end{array}$
where , , $\varphi_{1}(x_{1})$, , $k=1,\,3,\,4$ and are the known, $u(x)$,
$x\in D$ $\varphi_{2}(x_{1})=\alpha_{2}(x_{1})$ \- are the unknown functions.
## 2\. Stephan’s inverse problem.
The above mentioned boundary value problem (16),(17), is given provided
$\alpha_{k},\,\,\varphi_{k}(x_{1})$, $k=1,\,2$, $\gamma_{1}(x_{1})$,
$\alpha_{k}(x_{1})$, $k=\overline{1,\,4}$ $\varphi_{3}(x_{1})$ \- are the
known, $u(x)$, $x\in D$ $\gamma_{2}(x_{1})$, $x_{1}\in[a_{1},\,b_{1}]$ are the
unknown functions.
## Reference
1\. Courant, R. and Hilbert, D. Methods of mathematical physics, 1-2
Interscience, 1953 - 1962
2\. Bitsadze, A.V. Boundary value problems for second order elliptic
equations, North - Holland, 1968
3\. Hadamard, Tohoku, Math., J., vol. 37, pp. 133-150, 1933
4\. Hadamard, L’Enseignement Math. Vol. 35, pp. 5 -42, 1936
5\. Jurayev T.D On a boundary value problem for a composite type equation. DAN
USSR, N4,1962,p5-8(Russian)
6\. Aliev N.A., Aliev A.M. Investigation of solutions of boundary value
problems for a composite type equation on abounded plane domain. Dep.AzNIINTI,
N688-Az from 03.03.1987, 48 pp.
7\. Aliev N. and Jahanshahi M. Sufficient conditions for reduction of the BVP
including a mixed PDE with non –local boundary conditions to Fredholm integral
equations, INT. J.Math. Sci. Technol. 1997, vol. 28, N 3, 419-425
8\. Vladimirov V.S. Equations of Mathematical Physics, Mir, 1984
9\. Aliev N. and Jahanshahi M. Solation of Poisson’s equation with global,
local and non -local boundary conditions, Int. J. Math. Educ. Sci. Technol.
2002, vol.33, N2, 241-247
10\. Aliev N., Hosseini S.M. An analysis of a parabolic problem with a general
(non - local and global) supplementary linear conditions I, II Italian Journal
of Pure and Applied Mathematics N12, 2002 (143-154), N13, 2003 (115-127)
11\. Gakhov F.D. Boundary value problems, Pergamon 1966
| arxiv-papers | 2008-07-10T16:33:25 | 2024-09-04T02:48:56.681335 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "N. A. Aliev, A.M. Aliev",
"submitter": "Efendiev Rakib Feyruz",
"url": "https://arxiv.org/abs/0807.1697"
} |
0807.1752 | # A New Spiral Arm of the Galaxy: The Far 3-Kpc Arm
T. M. Dame and P. Thaddeus Harvard-Smithsonian Center for Astrophysics, 60
Garden Street, Cambridge MA 02138
tdame@cfa.harvard.edu, pthaddeus@cfa.harvard.edu
###### Abstract
We report the detection in CO of the far-side counterpart of the well-known
expanding 3-Kpc Arm in the central region of the Galaxy. In a CO longitude-
velocity map at $b=0^{\circ}$ the Far 3-Kpc Arm can be followed over at least
$20^{\circ}$ of Galactic longitude as a faint lane at positive velocities
running parallel to the Near Arm. The Far Arm crosses $l=0^{\circ}$ at +56 km
s-1, quite symmetric with the $-53$ km s-1expansion velocity of the Near Arm.
In addition to their symmetry in longitude and velocity, we find that the two
arms have linewidths ($\sim 21$ km s-1), linear scale heights ($\sim 103$ pc
FWHM), and H2 masses per unit length ($\sim 4.3$ x $10{{}^{6}}$ M⊙kpc-1) that
agree to 26% or better. Guided by the CO, we have also identified the Far Arm
in high-resolution 21 cm data and find, subject to the poorly known CO-to-H2
ratio in these objects, that both arms are predominately molecular by a factor
of 3–4. The detection of these symmetric expanding arms provides strong
support for the existence of a bar at the center of our Galaxy and should
allow better determination of the bar’s physical properties.
###### Subject headings:
Galaxy: center — Galaxy: structure — Galaxy: kinematics and dynamics — ISM:
molecules — radio lines: ISM
††slugcomment: ApJ Letters, accepted 7/9/08
##
Since its identification in 21 cm emission by van Woerden et al. (1957), the
expanding 3-Kpc Arm has remained at once one of the most obvious Galactic
spiral arms and the most puzzling. Its structure in 21 cm and CO longitude-
velocity diagrams and its absorption of continuum emission toward the Galactic
center demonstrate beyond doubt that the arm lies on the near side of the
center and is expanding away from it at a velocity of $-53$ km s-1at
$l=0^{\circ}$. The arm’s large non-circular motion has been attributed to
explosive expulsion of gas from the center (van de Kruit 1971; Sanders &
Prendergast 1974) and was central to one of the earliest arguments for the
existence of a bar at the Galactic center (de Vaucouleurs 1964).
On the basis of a reanalysis of the Columbia-CfA-Chile CO survey of the Milky
Way (Dame, Hartmann, & Thaddeus, 2001) in the vicinity of the Galactic center
(Bitran et al., 1997), we have found clear evidence on the far side of the
Galaxy for the long postulated and long sought counterpart of the expanding
3-Kpc Arm. The Far 3-Kpc Arm displays a clear symmetry with its near-side
counterpart in longitude and velocity, and once account is taken of its
greater distance, very similar physical characteristics as well (Table 1).
The Near and Far Arms appear in the CO $l$-v diagram of Figure 1 as two
inclined, parallel lanes symmetrical in velocity on either side of the
Galactic center. The Far Arm as expected is weaker than the Near, but it can
be readily followed over at least $20^{\circ}$ of longitude, starting at
$l=-12^{\circ}$, with an average intensity of $\sim 0.5$ K, more than 3 times
the instrumental noise. The linear fits indicated by the dashed lines yield
expansion velocities toward $l=0^{\circ}$ of $-53.1$ km s-1for the Near Arm
and $+56.0$ km s-1for the Far Arm, and, within the uncertainties, identical
velocity gradients with $l$ (Table 1).
Figure 1.— Longitude-velocity diagram of CO $1-0$ emission at $b=0^{\circ}$
from Bitran et al. (1997); this survey is sampled every $7.5\arcmin$ at
$|b|\leq 1^{\circ}$ (every $15\arcmin$ elsewhere) with an $8.8\arcmin$ beam to
an rms sensitivity of 0.14 K in $1.3$ km s-1channels. The dashed lines are
linear fits to the inclined, parallel lanes of emission from the Near and Far
3-kpc Arms: for the Near Arm v $=-53.1+4.16$ $l$, for the Far v $=+56.0+4.08$
$l$. The insert is a schematic showing the approximate locations of the arms
over the longitude ranges in which they can be followed clearly in CO. The
dotted lines are the limits of the present analysis.
To examine the average velocity structure of the Near and Far Arms, smoothing
out fluctuations owing to individual clouds along each arm, we averaged the
emission in Figure 1 in narrow linear strips inclined parallel to the arms
(which, as noted, are parallel to each other) and labeled each strip by its
velocity at $l=0^{\circ}$. In the resulting plot (Fig. 2), the Near and Far
Arms stand out as well-defined peaks to either side of a broad central peak,
mainly from the foreground and background disks. Note that the Far Arm is only
a factor of two fainter than the Near at $b=0^{\circ}$, and both are detected
at a level far above the instrumental noise. It is also clear from Figure 2
that the CO linewidths of the two arms are similar, with Gaussian fits
yielding values of $19.7$ km s-1(FWHM) for the Near Arm and $22.2$ km s-1for
the Far. We expect that further, more refined analyses, which will allow for
the expected curvature of the arms with longitude, will result in linewidths
that are at most 10% lower than those we find here. In spite of the difficulty
of estimating the overall noise level in Figure 2, which is dominated by the
clutter of unrelated emission over which we average, the figure forcefully
shows that the detection of the Far 3-Kpc Arm is not marginal, and a similar
exercise carried out with any moderate-resolution molecular-line survey of the
Galactic center should reveal the peak of the Far Arm.
Figure 2.— A composite CO spectrum for the Near and Far Arms obtained by
averaging the emission in Fig. 1 in narrow linear strips running parallel to
the arms, their essentially identical slopes with longitude determined by the
linear fits in Fig. 1. The blended regions indicated in Fig. 3 were excluded
from the averaging. Figure 3.— The velocity-integrated CO intensity of the
Near and Far 3-Kpc Arms. Emission was integrated over a 26 km s-1bin centered
at each longitude on the arm velocity given by the linear fits in Fig. 1.
Brackets indicate blended and other regions excluded in computing the physical
properties of the two arms (see Table 1 and the captions to other figures).
The color palette shown below is used in both maps.
If the inclined, linear emission feature at positive velocities in Figure 1
is, in fact, on the far side of the Galactic center, it should appear
systematically thinner on the sky than does the Near Arm, and in the spatial
maps in Figure 3 this is indeed the case. Excluding the blended regions
indicated, the angular thicknesses of the two arms appear to be roughly
constant along their lengths and to differ by about a factor of two. This
result is quantified by the latitude profiles in Figure 4, which yield FWHM
thicknesses of $1.1^{\circ}$ for the Near Arm and $0.52^{\circ}$ for the Far.
Because the Near 3-kpc Arm was first discovered in HI, on finding its far CO
counterpart we naturally attempted to identify it in existing 21 cm data. The
early 21 cm surveys were too poor in angular resolution and sensitivity to
distinguish the thin, faint lane of the Far Arm, but in the recent Australia
Telescope (ATNF) 21 cm survey at a resolution of $2\arcmin$ (McClure-Griffiths
et al., 2005) there is clear evidence of the Far Arm. In their Figure 6, a 21
cm $l$-v diagram at $b=0^{\circ}$, the Far Arm is seen extending from the
right at v $\sim 75$ km s-1. At $l<6^{\circ}$ this arm shows a curious
velocity bifurcation that is also seen in the Far CO Arm in this direction and
in some segments of the Near CO Arm (e.g., $l=-4^{\circ}$ to $-1^{\circ}$ in
Fig. 1). At negative longitudes the Far Arm in HI is largely blended with
emission from both distant gas beyond the solar circle and foreground gas in
the inner disk; however, a segment of the Far Arm can be traced in the ATNF
survey at $l>354^{\circ}$.
As we required of the Far CO Arm in Figure 1, the Far HI Arm evident in Figure
6 of McClure-Griffiths et al. should be narrow in latitude, and the HI $b$-v
map in Figure 5a shows that indeed it is. Comparison of this map to the CO
$b$-v map in Figure 5b shows that the Far Arm has about the same thickness in
both species (see also Table 1). In the positive longitude range included in
Figure 5a, it is the Near Arm that is masked in HI by both distant gas beyond
the solar circle and foreground gas in the inner disk. In contrast, the Near
Arm is seen clearly in CO, because there is little CO beyond the solar circle
and because the cloud-cloud velocity dispersion of the CO is lower than that
of the HI.
Figure 4.— Latitude profiles of the CO emission from (a) the Far and (b) Near
3-Kpc Arms, obtained by averaging the spatial maps in Fig. 3 over longitude.
The blended regions and two vertical extensions marked in Fig. 3 were
excluded. The dotted curves are Gaussian fits. Since the profile of the Far
Arm shows high-latitude wings presumably from unrelated foreground material,
the fit for this arm was confined to points at $|b|<0.5^{\circ}$.
To further quantify the properties of the two arms, distances are required.
The Near 3-Kpc Arm derives its name from its Galactic radius as estimated by
van Woerden et al. (1957) on the basis of an apparent southern tangent at
$l=-22^{\circ}$ and the R⊙value of 8.2 kpc then adopted. Subsequent work by
Bania (1980), Cohen & Davies (1976), and others similarly suggested a possible
northern tangent near $l=+23.5^{\circ}$. Although the locations and even the
existence of both tangents are still in doubt, here we adopt tangent
directions of $\pm 23^{\circ}$ for the Near Arm on the basis of careful study
of our composite CO survey (Dame et al. 2001). With R⊙$=8.5$ kpc, these
tangents imply a radius of 3.3 kpc for the Near Arm, and given its similarity
in so many other respects, we assume the same radius for the Far Arm. Although
the inclinations of the arms are unknown, here we assume for simplicity that
the arm distances vary little along their lengths; thus the Near Arm is at a
distance of 5.2 kpc (R⊙$-$ Rarm) and the Far at 11.8 kpc (R⊙\+ Rarm).
Figure 5.— (a) HI latitude-velocity map integrated in $l$ from $5^{\circ}$ to
$9^{\circ}$, a range in which the Far Arm is well separated in velocity from
foreground 21 cm emission. Data are from McClure-Griffiths et al. (2005).
Colors represent log intensity (K-arcdeg) from gray (1.0) to white (2.5). (b)
CO latitude-velocity map integrated in $l$ from $2^{\circ}$ to $8^{\circ}$, a
range in which both the Near and Far Arms are well defined and which overlaps
that of the HI map in (a); two of the blended regions indicated in Fig. 3 were
excluded from the integration. Colors represent log intensity (K-arcdeg) from
gray (0.0) to white (0.8).
With these distances, H2 and HI masses for the arms can be computed directly
from their integrated luminosities in CO and 21 cm. Although little is known
about the value of the CO-to-H2 mass conversion factor in these unusual arms,
it is reasonable to suppose that a similar value applies to both. Here we
assume a fairly standard Galaxy-wide average value of N(H2)/WCO = 1.8 x 1020
cm-2 K-1 km-1 s (Dame, Hartmann, & Thaddeus, 2001). Excluding the regions of
blending indicated in Figure 3, the longitude range considered here covers a
1.47-kpc length of the Near Arm and a 3.24-kpc length of the Far. Over these
ranges we find that the two arms have H2 masses per unit length that differ by
only 26% (Table 1). HI masses were similarly computed from the ATNF 21 cm
survey using a well-defined $4^{\circ}$ segment of the Far Arm at positive
longitudes and a comparable $9^{\circ}$ segment of the Near Arm at negative
longitudes (Table 1). We find that both arms are primarily molecular, the H2
mass exceeding the HI mass in the Near Arm by a factor of 3 and in the Far Arm
by a factor of 4 (a conclusion of course sensitive to the adopted CO-to-H2
ratio).
It is worth noting that the Near and Far Arms are found to have almost
identical linear thicknesses in H2 and similar thicknesses in HI (Table 1). We
speculate that the similarity of the thickness of the HI and CO in the two
3-Kpc Arms, versus the roughly 2:1 ratio found elsewhere in the Galaxy, is the
result of the evident absence of star formation in these arms (Lockman 1980),
activity that one expects would heat the gas and inflate the HI thickness.
The idea that the Milky Way is a barred spiral galaxy can be traced back at
least fifty years to Johnson (1957), de Vaucouleurs (1964), and others, and
there is now fairly wide consensus that we are indeed located in a barred
spiral (Binney et al., 1991; Blitz & Spergel, 1991; Benjamin et al., 2005).
Binney & Tremaine (2008) in their text Galactic Dynamics state unequivocally
that ”Our Galaxy is the nearest barred spiral”. The delineation here of the
predicted two symmetric, expanding 3-Kpc Arms confirms beyond a reasonable
doubt that we are located in such a galaxy; further study of these arms should
permit better determination of the orientation, size, and other properties of
the bar. Had it been possible to detect the Far Arm when van Woerden et al.
(1957) discovered the Near Arm, it seems likely that attempts over 50 years to
map the structure of the inner Galaxy would have proceeded more rapidly and
directly.
Because the properties of the Far Arm are so close to those expected for a
far-side counterpart of the expanding 3-Kpc Arm, it is natural to ask how the
arm escaped notice for so long. Hints of the Far Arm are evident in even the
first large-scale CO survey of the region (Bania 1977; see Fig. 1), but not at
a level that would have allowed a convincing case to be made. On the contrary,
Oort (1977) cited Bania’s survey as evidence ”That there is no counterpart of
the 3-kpc arm behind the center”. This view hardened over decades as the so-
called ”+135 km s-1feature” was widely adopted as the only possible far-side
counterpart of the expanding arm. The 21 cm surveys with angular resolution
adequate to resolve the thin Far Arm have become available only in the past
few years (ATNF: McClure-Griffiths et al. 2005, VLA: Stil et al. 2006), and
such data for the region within $5^{\circ}$ of the center have yet to be
published. Among the factors that masked the Far Arm in molecular line surveys
is the common practice of tracing spiral arms with $l$-v maps integrated over
latitude (e.g., Fig. 3 of Dame et al. 2001), which leaves the Far Arm
unresolved and confused with foreground emission, and the widespread use of
contour maps in the past and color maps at present—both of which can sometimes
mask weak, large-scale features. The two parallel lanes obvious in Figure 1
are less apparent in 21 cm data even at high angular resolution, because at
negative longitudes the Far Arm is badly blended with emission from the outer
and inner disks, and at positive longitudes the Near Arm is similarly blended.
Detection of the Far 3-Kpc Arm immediately suggests many avenues for follow-up
study with existing data, new observations, and theoretical studies. Tracking
both arms into the more confused regions outside the longitude range
considered here—perhaps to their origins at either end of the bar—is a high
priority which we are pursuing. The Far Arm should be detectable in other
existing spectral line surveys with adequate angular resolution; even those
with relatively low sensitivity should reveal the arm in the manner of Figure
2. CO observations with higher sensitivity and angular resolution are needed
to better define the Far Arm, in particular, its angular scale height as a
function of longitude, which could help constrain the arm’s inclination to our
line of sight. Since the Near Arm is known to be deficient in star formation,
searches for star formation in the Far Arm are of interest. In addition,
delineation of the Far Arm will provide a badly needed new constraint for the
hydrodynamical models (e.g., Fux 1999; Bissantz, Englmaier, & Gerhard 2003)
that seek to understand the overall properties of the central bar and its
influence on the gas.
In contrast to the controversy that has long characterized attempts to
determine the structure of the inner Galaxy, the 3-Kpc Arms together stand out
clearly as an unambiguous, beautifully symmetric structure. Rougoor & Oort
(1960), in one of their first papers on the expanding 3-Kpc Arm, noted that
”The arm is very well defined, and more homogeneous in density as well as
velocity than any of the outer arms.” That statement is even more true of the
twofold larger structure that we can now trace in the inner Galaxy.
We are indebted to T. Bania, R. Benjamin, J. Binney, A. Toomre, and S.
Tremaine for highly informative discussions and N. McClure-Griffiths for
providing the ATNF 21 cm survey.
## References
* Bania (1977) Bania, T. M. 1977, ApJ, 216, 381
* Bania (1980) Bania, T. M. 1980, ApJ, 242, 95
* Benjamin et al. (2005) Benjamin, R. A. et al.2005, ApJ, 630, L149
* Binney et al. (1991) Binney, J, Gerhard, O. E., Stark, A. A., Bally, J., & Uchida, K. I. 1991, MNRAS, 252, 210
* Binney & Tremaine (2008) Binney, J., & Tremaine, S. 2008, Galactic Dynamics (Princeton: Princeton Univ. Press), p. 529
* Bissantz, Englmaier, Gerhard (2003) Bissantz, N., Englmaier, P., & Gerhard, O. 2003, MNRAS, 340, 949
* Bitran et al. (1997) Bitran, M., Alvarez, H., Bronfman, L., May, J., & Thaddeus, P. 1997, A&AS, 125, 99
* Blitz & Spergel (1991) Blitz, L, & Spergel, D. N. 1991, ApJ, 379, 631
* Cohen & Davies (1976) Cohen, R. J. & Davies, R. D. 1976, MNRAS, 175, 1
* Dame, Hartmann, & Thaddeus (2001) Dame, T. M., Hartmann, Dap, & Thaddeus, P. 2001, ApJ, 547, 792
* de Vaucouleurs (1964) de Vaucouleurs, G. 1964, IAU Symposium 20: The Galaxy and the Magellanic Clouds, ed. F. J. Kerr & A.W. Rodgers (Sydney: Australian Academy of Science), 195
* Fux (1999) Fux, R. 1999, A&A, 345, 787
* Johnson (1957) Johnson, H. M. 1957, AJ, 62, 19
* Lockman (1980) Lockman, F. J. 1980, ApJ, 241, 200
* McClure-Griffiths et al. (2005) McClure-Griffiths, N. M., Dickey, J. M., Gaensler, B. M., Green, A. J., Haverkorn, M., & Strasser, S. 2005, ApJS, 158, 178
* Oort (1977) Oort, J. H. 1977, ARA&A, 15, 295
* Rougoor & Oort (1960) Rougoor, G. W., & Oort, J. H. 1960, Proc. Nat. Acad. Sci. USA, 46, 1
* Sander & Prendergast (1974) Sanders, R. H., & Prendergast, K. H. 1974, ApJ, 188, 489
* Stil et al. (2006) Stil, J. M., et al.2006, AJ, 132, 1158
* van der Kruit (1971) van der Kruit, P. C. 1971, A&A, 13, 405
* van Woerden et al. (1957) van Woerden, Rougoor, G. W., & Oort, J. H. 1957, Comp. Rend., 244, 1691
Table 1Comparison of Near and Far 3-Kpc Arms Arm | d | v0aaVelocity at $l=0^{\circ}$ and velocity gradient with longitude from linear fits to Fig. 1 | dv$/dl$aaVelocity at $l=0^{\circ}$ and velocity gradient with longitude from linear fits to Fig. 1 | $\Delta$vbbMean velocity width (FWHM) from Gaussian fits to composite profiles in Fig. 2 | $H_{2}$ dM/d$l$ccAveraged $l=-12^{\circ}$ to $8^{\circ}$, excluding blended regions: $l=-1.5^{\circ}$ to $3.25^{\circ}$ for Near Arm, $-1^{\circ}$ to $2^{\circ}$, $2.75^{\circ}$ to $3.5^{\circ}$, and $5.25^{\circ}$ to $5.75^{\circ}$ for the Far Arm (see Fig. 1) | HI dM/d$l$ddAveraged over the arm segments best defined in HI: $l=-9.5^{\circ}$ to $-1.5^{\circ}$ for the Near Arm and $l=5^{\circ}$ to $9^{\circ}$ for the Far | $\Delta$z(H2)eeVertical linear thickness (FWHM); see Fig. 4 for H2; HI profiles not shown | $\Delta$z(HI)d,ed,efootnotemark:
---|---|---|---|---|---|---|---|---
| (kpc) | (km s-1) | (km s-1deg-1) | (km s-1) | ($10{{}^{6}}$ M⊙kpc-1) | ($10{{}^{6}}$ M⊙kpc-1) | (pc) | (pc)
Near | 5.2 | –53.1 | 4.16 | 19.7 | 4.8 | 1.6 | 101 | 138
Far | 11.8 | +56.0 | 4.08 | 22.2 | 3.8 | 0.9 | 105 | 109
| arxiv-papers | 2008-07-10T22:53:38 | 2024-09-04T02:48:56.685649 | {
"license": "Public Domain",
"authors": "T. M. Dame and P. Thaddeus",
"submitter": "Thomas M. Dame",
"url": "https://arxiv.org/abs/0807.1752"
} |
0807.1755 | # Propagation of field disturbances in the Yang-Mills theory
Vitorio A. De Lorenci delorenci@unifei.edu.br Institute of Science, Federal
University of Itajubá, 37500-903 Itajubá, M. G., Brazil, PH Department, TH
Unit, CERN, 1211 Geneva 23, Switzerland Shi-Yuan Li lishy@sdu.edu.cn School
of Physics, Shandong University, Jinan, 250100, P. R. China, PH Department,
TH Unit, CERN, 1211 Geneva 23, Switzerland
###### Abstract
The propagation of field disturbances is examined in the context of the
effective Yang-Mills Lagrangian, which is intended to be applied to QCD
systems. It is shown that birefringence phenomena can occur in such systems
provided some restrictive conditions, as causality, are fulfilled. Possible
applications to phenomenology are addressed.
###### pacs:
42.25.Lc, 12.38.Mh, 24.70.+s
## I Introduction
Small disturbances on nonlinear fields propagate with velocity depending on
the polarization states. In general there will be two polarization modes,
leading to the existence of two waves propagating with different velocities.
This phenomenon is known in the literature as birefringence. In the Maxwell
theory (i.e., Abelian gauge field) it can be found when light propagates
inside certain material media landau . It can also appear in the context of
nonlinear spin-one fields birula70 ; iacopini1979 ; delorenci00 , as it occurs
in the quantum electrodynamics (QED). The effective Lagrangian for QED was
derived long ago heisenberg36 for slowly varying but arbitrary strong
electromagnetic fields. Its nonlinearities lead to effects like birefringence
and photon splitting birula70 . Some other investigations on this issue can be
found in rikken2000 ; adler2007 ; biswas2007 ; kruglov2007 .
For the case of non-Abelian gauge theories, the quantum fluctuations lead to a
vaccum state which does not coincides with the vacuum coming from the
perturbation theories. The structure of the vacuum state was discussed for
several models in savvidy1977 ; pagels78 ; nielsen1979 ; shuryak1980 ;
ambjorn1980 ; ambjorn1980b . The one-loop effective action for Yang-Mills
theory was presented and discussed in, e.g., savvidy1977 ; pagels78 . For the
asymptotically free theory in the regime of large mean fields the effective
action is controlled by perturbation theory. In this context the issue of
event horizon formation in the physical vacuum associated with color
confinement was considered in castorina07 .
The mathematical formalism (see Section II for details) to deal with the
propagation of small disturbances in nonlinear spin-one fields birula70 ;
delorenci00 depends on the Lagrangian as a general function of the field
invariant. Thus, it can be used to examine the wave propagation in systems
governed by an Yang-Mills effective Lagrangian. Particularly, it is worthwhile
to analyze if effects like birefringence can occur in this context. Quantum
Chromodynamics (QCD) could be taken as a specific application, since it
presents strong nonlinear properties.
In this manuscript the one-parameter effective Lagrangian presented in
pagels78 is used as a ‘working model’ when discussing Yang-Mills fields. Due
to the possibility of two polarization modes presenting different velocities,
as derived from our theoretical framework, it is shown that the birefringence
phenomena can occur provided that causal conditions are fulfilled. In QCD
case, we discuss how to observe the birefringence phenomena associated with
the propagation of small disturbances of the gluon field. Though the gluon is
not directly observable due to confinement, a bulk of deconfined hot (and/or
dense) quark-gluon matter is expected to exist in the ultra-relativistic heavy
ion interactions, as well as in the early phase of the universe collins1975 .
In those cases when the gluon propagates in the quark gluon matter, it is
argued that the birefringence of gluon field leads to local polarization of
gluons. The polarization correlation is suggested to be measured in Gold-Gold
collision on Relativistic Heavy Ion Collider (RHIC) at Brookhaven National
Laboratory and Lead-Lead collision on Large Hadron Collider (LHC) at CERN.
The paper is organized as follows. In the Section II the light cone conditions
associated with the propagation of small disturbances in one-parameter spin-
one theories are reviewed. In the Section III the effective Lagrangian for
Yang-Mills field pagels78 is presented and the procedure of taking volumetric
spatial average is defined. Then it is discussed the non-trivial behavior of
the phase velocity. The conditions on the causal propagation are stated and
some limiting cases from the effective Lagrangian are examined. Sections IV
and V are dedicated to the possible applications to phenomenology and the
effective geometry issue, respectively. Finally, some final remarks are
presented in the conclusion section.
The present investigation is considered under the regime of the eikonal
approximation, as addressed in delorenci00 . Latin indices run in the range
$(1,2,3)$ and Greek indices run in the range $(0,1,2,3)$. The Minkowski
spacetime is used, employing a Cartesian coordinate system. The background
metric is denoted by $\eta_{\mu\nu}$, which is defined by diag(+1,-1,-1,-1).
Heaviside non-rationalized units are used and $c=1=\hbar$. The completely
anti-symmetric tensor $\epsilon^{\alpha\beta\mu\nu}$ is defined such that
$\epsilon^{0123}=1$.
## II Light cone conditions
### II.1 Field equations for one-parameter spin-one theories
The strength tensor field $F_{\mu\nu}^{(a)}$ and the gauge field
$A_{\mu}^{(a)}$ are related by
$F_{\mu\nu}^{(a)}=\partial_{\mu}A_{\nu}^{(a)}-\partial_{\nu}A_{\mu}^{(a)}+C^{abc}A_{\mu}^{(b)}A_{\nu}^{(c)},$
(1)
where $C^{abc}$ represent the structure constant for a compact Lie group $G$.
This tensor field can be conveniently defined in terms of the (non-Abelian)
electric $E_{\mu}^{(a)}$ and magnetic $H_{\mu}^{(a)}$ fields as
$F_{\mu\nu}^{(a)}=V_{\mu}E_{\nu}^{(a)}-V_{\nu}E_{\mu}^{(a)}-\epsilon_{\mu\nu}{}^{\alpha\beta}V_{\alpha}H_{\beta}^{(a)},$
(2)
where $V_{\mu}$ represents the four-velocity of an observer at rest with
respect to the laboratory. In Cartesian coordinates it is given by
$V_{\mu}=\delta_{\mu}^{0}$. In order to alleviate the notation, the ‘color’
indices in the upper brackets will be omitted in what follows.
Let us assume the gauge invariant density of Lagrangian as a general function
of the Lorentz invariant $F\doteq F^{\mu\nu}F_{\mu\nu}$ as $L=L(F)$. From the
minimal action principle we get the equation of motion
$\left(L_{F}F^{\mu\nu}\right){}_{,\nu}=0.$ (3)
where a comma denotes partial derivatives with respect to the Cartesian
coordinates. $L_{F}$ represents the derivative of $L$ with respect to the
invariant $F$. $L_{FF}$ is the second derivative. Using the relation
$F_{,\nu}=2F^{\alpha\beta}F_{\alpha\beta,\nu}$ in equation (3) we obtain:
$\displaystyle
2L_{FF}F^{\mu\nu}F^{\alpha\beta}F_{\alpha\beta,\nu}+L_{F}F^{\mu\nu}{}_{,\nu}=0.$
(4)
The field strength $F_{\mu\nu}$ must satisfy the identity
$F_{\alpha\beta,\lambda}+F_{\beta\lambda,\alpha}+F_{\lambda\alpha,\beta}=0.$
(5)
Let us now derive the expression for the light cone conditions for this class
of theories.
### II.2 The propagation of the field disturbances
In this section we analyze the propagation of waves associated with the
discontinuities of the field Hadamard . Let us consider a surface of
discontinuity $\Sigma$ defined by ${\cal Z}(x^{\mu})=0$. Whenever $\Sigma$ is
a global surface, it divides the spacetime in two distinct regions $U^{-}$ and
$U^{+}$ (${\cal Z}<0$ and ${\cal Z}>0$, respectively). Given an arbitrary
function of the coordinates, $f(x^{\mu})$, we define its discontinuity on
$\Sigma$ as
$\left[f(x^{\alpha})\right]_{\Sigma}\doteq\lim_{\\{P^{\pm}\\}\rightarrow
P}\left[f(P^{+})-f(P^{-})\right]$ (6)
where $P^{+},\,P^{-}$ and $P$ belong to $U^{+},\,U^{-}$ and $\Sigma$
respectively. Applying the conditions Hadamard for the tensor field
$F_{\mu\nu}$ and its derivative we set
$\displaystyle\left[F_{\alpha\beta}\right]_{\Sigma}$ $\displaystyle=$
$\displaystyle 0$ (7a)
$\displaystyle\left[F_{\alpha\beta,\lambda}\right]_{\Sigma}$ $\displaystyle=$
$\displaystyle f_{\alpha\beta}k_{\lambda}$ (7b)
where $f_{\alpha\beta}$ represents the discontinuities of field on the surface
$\Sigma$ and $k_{\mu}\doteq(\omega,\vec{k})$ represents the components of the
wave 4-vector. The discontinuity of equations (4) and (5) yields,
respectively,
$\displaystyle
f_{\beta\lambda}k^{\lambda}=-\frac{2}{L_{F}}L_{FF}F_{\beta}{}^{\mu}F^{\nu\rho}f_{\nu\rho}k_{\mu},$
(8) $\displaystyle
f_{\alpha\beta}k_{\lambda}+f_{\beta\lambda}k_{\alpha}+f_{\lambda\alpha}k_{\beta}=0.$
(9)
For the case where $f_{\alpha\beta}$ is the wave propagation tensor given by
equation (7b), for which equation (9) applies, it follows that
$f_{\alpha\beta}=\sigma(\epsilon_{\alpha}k_{\beta}-\epsilon_{\beta}k_{\alpha}),$
(10)
where $\sigma$ is the strength of the wavelet and $\epsilon_{\beta}$
represents the polarization vector. Working with Eqs. (8)-(10) we obtain the
eigenvalue equation
$Z^{\mu}{}_{\nu}\epsilon^{\nu}=0,$ (11)
where we defined
$Z^{\mu}{}_{\nu}\doteq\delta^{\mu}{}_{\nu}+\frac{4}{L_{F}k^{2}}L_{FF}F^{\mu\alpha}F^{\nu\beta}k_{\alpha}k_{\beta},$
(12)
with $k^{2}\doteq k^{\mu}k_{\mu}$. The eigenvectors of $Z^{\mu}{}_{\nu}$
represent the dynamically allowed polarization modes $(e_{+},e_{-})$. The
general solution for the eigenvalue equation is formally given by
$det|Z^{\mu}{}_{\beta}|=0$, and results in the following light cone conditions
boillat70 ; birula70 ; delorenci00 :
$\displaystyle k_{+}^{2}$ $\displaystyle=$ $\displaystyle\gamma
F^{\lambda\mu}F^{\nu}{}_{\lambda}k^{+}_{\mu}k^{+}_{\nu},\qquad\gamma\doteq\frac{4L_{FF}}{L_{F}}$
(13) $\displaystyle k_{-}^{2}$ $\displaystyle=$ $\displaystyle 0,$ (14)
The $\pm$ signs are related with the two possible polarization modes
associated with the wave propagation birula70 . The existence of these two
solutions shows that birefringence effects may appear, provided that
$L_{FF}/L_{F}\neq 0$. In the formalism of geometrical optics it is usually
said that generally there will be two rays inside the medium, the ordinary ray
($o$-ray) and the extraordinary ray ($e$-ray). The former does not depend on
the direction of wave propagation and its velocity is equal to the light
velocity in the classical vacuum of electrodynamics. The latter presents an
explicit dependence on the direction of propagation. The light cone conditions
for two-parameters Lagrangians can be obtained in the same lines. For further
details see Refs. birula70 ; delorenci00 .
## III Wave propagation in the Yang-Mils field
### III.1 The effective Lagrangian
The effective Lagrangian for QCD in terms of the parameter background field
$F$ can be presented pagels78 in the form
$\displaystyle{\cal L}_{eff}=\frac{1}{4}\frac{F}{\bar{g}(t)^{2}},\qquad
t\doteq\log\frac{F}{\mu^{4}}$ (15)
where the effective coupling $\bar{g}(t)$ is implicitly given by
$t=\int_{g}^{\bar{g}(t)}\\!\\!\\!dg\;\frac{1}{\beta(g)},$ (16)
with $\beta(g)$ the Callan-Symanzik $\beta$-function and $g$ the gauge field
coupling constant appearing in the basic QCD Lagrangian.
In fact, there are many invariants of the Yang-Mills field with the number
dependent on the specific gauge group roskies1977 . The ansatz used to derive
this effective Lagrangian takes in consideration only the algebraic invariant
$F$ and imposes consistency with the trace anomaly for the energy-momentum
tensor collins1977 .
For the present proposes the system described by Eq. (15) is assumed to
satisfy the following requirements:
1. 1.
the volumetric spatial average of the color field strength is independent of
direction;
2. 2.
it is equally probable that the products ${E^{i}E^{j}}$ and ${H^{i}H^{j}}$
(with $i\neq j$), at any time, take positive or negative values;
3. 3.
there is no net flow of energy as measured by a observer at rest with respect
to the system.
The above mentioned volumetric spatial average of an arbitrary quantity $X$
for a given instant of time $t$ is defined as
$\overline{X}\stackrel{{\scriptstyle.}}{{=}}\lim_{V\rightarrow
V_{o}}\frac{1}{V}\int X\,\sqrt{-g}\,d^{3}\\!x^{i},$ (17)
with $V=\int\sqrt{-g}\,d^{3}\\!x^{i}$, and $V_{o}$ stands for the time
dependent volume of the whole space. Similar average procedure has been
already considered in the context of general relativity solutions tolman1930 ;
delorenci2002 ; kunze2008 .
In terms of the color fields, these requirements imply that:
$\displaystyle\overline{\rule{0.0pt}{8.61108pt}E_{i}}=0,\qquad\overline{\rule{0.0pt}{8.61108pt}H_{i}}$
$\displaystyle=$ $\displaystyle
0,\qquad\overline{\rule{0.0pt}{8.61108pt}E_{i}\,H_{j}-H_{i}\,E_{j}}=0,$ (18)
$\displaystyle\overline{\rule{0.0pt}{8.61108pt}E_{i}\,E_{j}}$ $\displaystyle=$
$\displaystyle-\,\frac{1}{3}E^{2}\,\eta_{ij},$ (19)
$\displaystyle\overline{\rule{0.0pt}{8.61108pt}H_{i}\,H_{j}}$ $\displaystyle=$
$\displaystyle-\,\frac{1}{3}H^{2}\,\eta_{ij},$ (20)
where we have defined $E^{2}\doteq-E^{i}E_{i}$ and $H^{2}\doteq-H^{i}H_{i}$.
The above average procedure consists in an idealization to deal with systems
like a quark-gluon plasma. Nevertheless, it has adequate elements for our
discussions. From the point of view of a statistical ensemble, we can assume
that the field average over the whole bulk is vanishing comparing to the
fluctuation, so it is isotropic as whole, while anisotropic for each local
area.
### III.2 Application to the effective Yang-Mills Lagrangian
For the special Lagrangian presented in Eq. (15), the factor $\gamma$ in Eq.
(13) is given by
$\gamma=\frac{-4}{E^{2}(Z^{2}-1)}G(\bar{g}),$ (21)
where we have defined the quantities:
$\displaystyle Z^{2}$ $\displaystyle\doteq$ $\displaystyle\frac{H^{2}}{E^{2}}$
(22) $\displaystyle G(\bar{g})$ $\displaystyle\doteq$
$\displaystyle\frac{\bar{g}\dot{\bar{g}}-3\dot{\bar{g}}^{2}+\bar{g}\ddot{\bar{g}}}{\bar{g}^{2}-2\bar{g}\dot{\bar{g}}},$
(23)
with $\dot{\bar{g}}\doteq\partial\bar{g}/\partial t$.
The phase velocity for the wave perturbation can be obtained from the
dispersion relation as $v_{e}^{2}=\omega/|\vec{k}|$, where the index $e$
stands for the $e$-ray. The $o$-ray propagates with the light velocity, as
determined by Eq. (14). Now using the previous results, we obtain
$v_{e}^{2}=1-\frac{8}{3}\frac{(Z^{2}+1)G(\bar{g})}{(Z^{2}-1)+4G(\bar{g})}.$
(24)
In order to guarantee causality, the physical solutions must satisfy the
requirement $0\leq v_{e}\leq 1$, which implies that:
$0\leq\frac{8}{3}\frac{(Z^{2}+1)G(\bar{g})}{(Z^{2}-1)+4G(\bar{g})}\leq 1.$
(25)
From the analysis of the energy density for the effective action associated
with this problem, it can be inferred that the case $E^{2}>H^{2}$ leads to a
metastability of the vaccum. The interpretation for this behavior is that if a
region in the system develops a large $E$ field, it will decay fastly to a
configuration where $H^{2}>E^{2}$ pagels78 . Therefore it is adopted here that
$H^{2}>E^{2}$ (which means $Z^{2}>1$ and $F>0$) and the condition stated by
Eq. (25) yields in
$0\leq G(\bar{g})\leq\frac{3(Z^{2}-1)}{4(2Z^{2}-1)}.$ (26)
Now we are going to examine two cases in which the explicit form of the
effective coupling was presented in the literature. The first one is obtained
when the regime of small coupling is taken in consideration, and the second
one was previously proposed in gross1973 as a suggestion for the case of
large coupling constant.
For the case of small coupling the beta function can be expanded as gross1973
; poli1973 ; jone1974 ; casw1974
$\beta(g)=-\frac{1}{2}b_{0}g^{3}+b_{1}g^{5}+\cdots$ (27)
where $b_{0}$ and $b_{1}$ are constants. Now, taking the limit of large mean
fields ($F\rightarrow\infty$) we obtain from Eq. (16) that
$\frac{1}{\bar{g}(t)}=b_{0}t-2\frac{b_{1}}{b_{0}}\log t+\cdots$ (28)
Introducing these results in the effective Lagrangian we obtain pagels78 ;
nielsen1978
$\displaystyle{\cal L}_{eff}=\frac{1}{4}b_{0}F\log\frac{F}{\mu^{4}}.$ (29)
For this case the function $G(\bar{g})$ results to be
$G(\bar{g})=-\frac{1}{2}\frac{1}{t+1}.$ (30)
Since $t>1$ at the large mean field regime we conclude that $G(\bar{g})<0$,
and there will be no propagation associated with the $e$-ray. The $o$-ray
travels with the unperturbed velocity $v_{-}=1$ and does not depend on the
direction of propagation.
The non-perturbative expression for the beta function is still unknown.
Nevertheless, a suggestion about the strong coupling form of the beta function
was presented long ago gross1973 , and is given by
$\beta(g)=-\frac{a}{2}g,$ (31)
with $a$ a positive constant. If we assume this form for the beta function, we
obtain from Eqs. (16) and (31):
$\frac{1}{\bar{g}(t)^{2}}=\frac{1}{g^{2}}\left(\frac{F}{\mu^{4}}\right)^{a}.$
(32)
In terms of the parameter $t$ it follows that
$\bar{g}(t)^{2}=g^{2}e^{-at}.$ (33)
Introducing these results in the effective Lagrangian we obtain
$\displaystyle{\cal
L}_{eff}=\frac{1}{4g^{2}}F\left(\frac{F}{\mu^{4}}\right)^{a}.$ (34)
For this case the function $G(\bar{g})$ results in
$G(\bar{g})=-\frac{1}{2}a.$ (35)
Thus since $a>0$ we conclude that $G(\bar{g})<0$, and again there will be no
propagation associated with the $e$-ray. In the both cases, once the
superluminal propagation is suppressed, the medium allows just one
polarization mode to propagate. So it seems to behave like a polarizer.
## IV Observable for birefringence
As shown in the previous section, birefringence effects can occur in the Yang-
Mills fields. Nevertheless, due to confinement phenomenon, a direct
measurement of these effects on gluons propagating in an external color field
seems to be improbable. However, deconfined quark-gluon matter, also known as
quark-gluon plasma (QGP), is expected to be produced in the Gold-Gold
interaction at RHIC or Lead-Lead interaction at LHC. This provides an
opportunity to investigate the issue of gluon propagation in QGP systems.
The different velocities in which the field disturbances can propagate are
associated with different polarization modes. It can be simply understood as a
quantum measurement on the gluon, by which the gluon fall onto the eigenstate
of the polarization mode. In the special cases where the $e$-ray gluon is
forbidden by causality, the external field works like a polarizer, and only
the $o$-ray gluon is allowed there. The polarization is assigned by the
external field. When one gluon propagates in a QGP, it is ‘measured’ time and
time by the local field. Hence, the last polarization direction before its
hadronization into hadrons is completely out of control. This will also
destroy any global polarization information 111One of the examples is that
originated from the non-central interaction, as suggested by the Shandong
Group liangw . Corresponding to those cases, the phenomenon suggested here
appears as a ‘local polarization’. Hence this local polarization destroys the
information of the global one and itself is not measurable if we simply take
the average in the experiment. The STAR Collaboration of RHIC has recently
measured the global polarization effect s1 ; s2 . From the above discussion,
it seems very natural to adopt the negative results from STAR, still with
confidence that both of the polarization effects can exist.. However, due to
the complete polarization at any local area, the polarization correlation will
be very strong. A crucial point is how to identify two particles at the same
local area with the same polarization. In order to explore this point let us
consider, for simplicity, a ‘gluon plasma’. When the hard parton propagates in
the medium, it works as a source of small disturbances on the external field,
and could lead to the emission of Cherenkov radiation. The $o$-gluon can exist
within the Cherenkov cone, but the $e$-gluon must be outside. If the gluon is
not confined it can be measured that the polarizations for two gluons inside
the cone are parallel. More specifically
$\hat{\epsilon}_{1}\cdot\hat{\epsilon}_{2}=1$. The experimental results depend
not only on the polarization transfer from the gluon to a certain kind of
hadron in hadronization process, but also on the identification of the
Cherenkov cone. So, one suggestion is to measure the polarization correlation
of two particles of the same kind (2 vector mesons, 2 hyperons, etc.) with
almost parallel momenta and within the same jet cone (when the jet can be
identified, as expected in LHC). By studying the correlation dependent on the
jet cone angle, one may have a way to measure the Cherenkov angle.
One experimental observable for the polarization correlation can be suggested.
This can be extracted from the ideal case of two $\Lambda$ particles with the
same polarization $\vec{P}$, with $P=|\vec{P}|$ representing the polarization
rate. The conventional way to measure $\vec{P}$ of a single $\Lambda$ is by
measuring the direction vector (denoted as $\hat{p}$) of the momenta of the
daughter particles, e.g., proton or pion from the $\Lambda$ decay, at the rest
frame of $\Lambda$. Then the angular distribution,
$\frac{dN}{dcos\theta}\sim 1+\alpha\hat{p}\cdot\vec{P}=1+\alpha Pcos\theta,$
(36)
can give the information on the polarization. Here $\alpha$ is the hyperon
decay parameter. From this equation we see that if the direction of $\vec{P}$
is random, the average of all $\Lambda^{\prime}s$ gives zero, then $P$ is not
able to be measured. However, for the two $\Lambda^{\prime}s$ with the same
polarization, in the rest frame of each $\Lambda$, respectively, the direction
vectors $\hat{p}_{1}$ and $\hat{p_{2}}$ can be measured. Then we calculate the
expectation value $<\hat{p}_{1}\cdot\hat{p}_{2}>$, which results in
$P=3\sqrt{<\hat{p}_{1}\cdot\hat{p}_{2}>}/\alpha.$ (37)
Because $<\hat{p}_{1}\cdot\hat{p}_{2}>$ is a SO(3) scalar, it does not depend
on the direction of polarization. Thus, it can be averaged for all the jets of
all the (QGP) events in order to get the scalar value of averaged polarization
from Eq (37). In an experiment, the $\Lambda$ pair of the same jet 222If the
jet is not defined it can be considered the near-parallel pairs of the same
kind of hyperon. are expected to have the same polarization with larger
probability, as discussed above.
## V The effective geometry issue
The effective theory approach has long been considered as a possible way to
understand confinement phenomena. One of the possible ways to investigate
these phenomena consists in the construction of analogue models in which
confinement would be related with an event horizon formation, as it occurs in
the black hole physics. Such a interpretation was considered in castorina07 ;
elbaz (see also the references therein).
The results discussed in this paper can be read in the context of the optical
geometry. In this context some comments are in order. Eqs. (13)–(14) can be
presented in the appealing form
$g_{\pm}^{\mu\nu}k_{\mu}k_{\nu}=0,$ (38)
where we defined the two symmetric contravariant tensors
$\displaystyle g^{\mu\nu}_{+}$ $\displaystyle=$
$\displaystyle\eta^{\mu\nu}-\gamma F^{\lambda\mu}F^{\nu}{}_{\lambda},$ (39)
$\displaystyle g^{\mu\nu}_{-}$ $\displaystyle=$ $\displaystyle\eta^{\mu\nu}.$
(40)
The inverse symmetric tensor $g_{\mu\nu}$ is defined in such way that
$g^{\mu\alpha}g_{\alpha\nu}=\delta^{\mu}_{\nu}$. Therefore, for each
propagation vector $k_{\nu}$, the corresponding tensor $g^{\mu\nu}$ plays the
role of an effective metric tensor. Indeed, $k_{\nu}$ is a light-like (or
null) vector with respect to the associated metric tensor. It can be shown
delorenci02 that it also satisfies a geodesic equation in terms of the
Christoffel symbols $\Gamma^{\alpha}{}_{\\!\mu\nu}$ associated with this
metric. In this way we can refer to $k_{\nu}$ as a geodesic null vector with
respect to the effective metric tensor $g^{\mu\nu}$. This geometric
interpretation could be used in order to produce an analogue model for
confinement based on the possible formation of an event horizon. In this case,
the requirement of causal propagation would assume a nontrivial role. Inside
the hadron, where quark and gluons can interact and propagate, the velocity of
the field disturbances would be expected to be smaller than 1. On the other
hand, if the vacuum outside the hadron is described by the effective
Lagrangian, the larger than 1 velocity could be interpreted as an indication
of confinement, since no physical observable could propagate there.
## VI Conclusions
In this paper the propagation of field disturbances was investigated in the
context of the effective Yang-Mills Lagrangian. The general dispersion
relations for one-parameter Lagrangians, Eqs. (13) and (14), were derived
employing the method presented in delorenci00 ; birula70 . It was shown that
birefringence phenomena can occur.
Let us remark some points. First, it should be stressed that the method
depends on the effective Lagrangian as $L(F)$, so the conclusions are quite
general. Second, the assumption of causal propagation of the signals sets non-
trivial constraint when exploring specific solutions. Finally, for the case of
a deconfined quark gluon system, which naturally provides an effective
external field, the birefringence phenomena with gluons and its local
polarization effects are expected to be observed at RHIC and LHC by measuring
the strong local spin correlations of various hadrons from QGP. The
measurements of the spin correlations suggested here, at the same time, can be
useful in assigning the details of the effective fields in QGP. These
informations provide opportunities to develop the effective Lagrangian
framework and hence the better understanding of QCD.
###### Acknowledgements.
This work is partially supported by the Brazilian research agencies CNPq and
FAPEMIG (V.A.D.L.); and National Natural Science Foundation of China (NSFC)
with grant No.10775090, and China Scholarship Council (CSC) (S.Y.L.).
## References
* (1) See, e.g., L. Landau and E. Lifchitz, Electrodynamics of Continuous Media, (Ed. Elsevier, 1984).
* (2) Z. Bialynicka-Birula and I. Bialynicki-Birula, Phys. Rev. D 2, 2341 (1970).
* (3) E. Iacopini and E. Zavattini, Phys. Lett. B 85, 151 (1979).
* (4) V. A. De Lorenci, R. Klippert, M. Novello and J. M. Salim, Phys. Lett. B 482, 134 (2000).
* (5) W. Heisenberg and H. Euler, Z. Physik 98, 714 (1936).
* (6) G. L. J. A. Rikken, and C. Rizzo, Phys. Rev. A 63, 012107 (2000).
* (7) S. L. Adler, J. Phys. A 40, F143 (2007).
* (8) S. Biswas and K. Melnikov, Phys. Rev. D 75, 053003 (2007).
* (9) S. I. Kruglov, Phys. Rev. D 75, 117301 (2007).
* (10) G. K. Savvidy, Phys. Lett. B 71, 133 (1977).
* (11) H. Pagels and E. Tomboulis, Nucl. Phys. B 143, 485 (1978).
* (12) H. B. Nielsen and M. Ninomiya, Nucl. Phys. B 156, 1 (1979).
* (13) E. V. Shuryak, Nucl. Phys. B 203, 93 (1982).
* (14) J. Ambjorn and P. Olesen, Nucl. Phys. B 170, 60 (1980).
* (15) J. Ambjorn and P. Olesen, Nucl. Phys. B 170, 265 (1980).
* (16) P. Castorina, D. Kharzeev and H. Satz, Eur. Phys. J. C 52, 187 (2007).
* (17) J. C. Collins and M. J. Perry, Phys. Rev. Lett. 34, 1353 (1975).
* (18) J. Hadamard, in Leçons sur la propagation des ondes et les équations de l’hydrodynamique (Ed. Hermann, Paris, 1903).
* (19) G. Boillat, J. Math. Phys. 11, 941 (1970).
* (20) R. Roskies, Phys. Rev. D 15, 1722 (1977).
* (21) J. C. Collins, A. Duncan, and S. D. Joglekar, Phys. Rev. D 16, 438 (1977).
* (22) R. C. Tolman and P. Ehrenfest, Phys. Rev. 15, 1791 (1930).
* (23) V. A. De Lorenci, R. Klippert, M. Novello, and J. M. Salim, Phys. Rev. D 65, 063501 (2002).
* (24) K. E. Kunze, Phys. Rev. D 77, 023530 (2008).
* (25) D. J. Gross and F. Wilczek, Phys. Rev. Lett. 30, 1343 (1973).
* (26) H. D. Politzer, Phys. Rev. Lett. 30, 1346 (1973).
* (27) D. R. T. Jones, Nucl. Phys. B 75, 531 (1974).
* (28) W. E. Caswell, Phys. Rev. Lett. 33, 244 (1974).
* (29) H. B. Nielsen, Phys. Lett. B 80, 133 (1978).
* (30) M. Novello, V.A. De Lorenci, and E. Elbaz, em Int. J. Mod. Phys. A 13, 4539 (1998).
* (31) see, e.g., Z. T. Liang and X. N. Wang, Phys. Rev. Lett. 94, 102301 (2005).
* (32) B. I. Abelev et al. [STAR Collaboration], Phys. Rev. C 76, 024915 (2007).
* (33) B. I. Abelev et al. [STAR Collaboration], arXiv:0801.1729[nucl-ex].
* (34) V. A. De Lorenci and R. Klippert, Phys. Rev. D 65, 064027 (2002).
| arxiv-papers | 2008-07-10T23:08:59 | 2024-09-04T02:48:56.689358 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "V.A. De Lorenci and S.Y. Li",
"submitter": "Shi-Yuan Li",
"url": "https://arxiv.org/abs/0807.1755"
} |
0807.1789 | 11institutetext: Institut de Physique Théorique, CEA, IPhT, CNRS, URA 2306,
F-91191 Gif-sur-Yvette, France, 22institutetext: Faculty of Engineering,
Takushoku University, Hachioji, Tokyo 193-0985, Japan, 22email:
y-suzuki@la.takushoku-u.ac.jp 33institutetext: Department of Physics, Faculty
of Science, Kyushu University, 4-2-1 Ropponmatsu, Fukuoka 810-8560, Japan
# Kinetics of water flow through polymer gel
Yasuo Y. Suzuki 1122 Masayuki Tokita 33 Sada-atsu Mukai 33
(Received: date / Revised version: date)
###### Abstract
The water flow through the poly(acrylamide) gel under a constant water
pressure is measured by newly designed apparatus. The time evolution of the
water flow in the gel, is calculated based on the collective diffusion model
of the polymer network coupled with the friction between the polymer network
and the water. The friction coefficient are determined from the equilibrium
velocity of water flow. The Young modulus and the Poisson’s ratio of the rod
shape gels are measured by the uni-axial elongation experiments, which
determine the longitudinal modulus independently from the water flow
experiments. With the values of the longitudinal modulus and of the friction
determined by the experiments, the calculated results are compared with the
time evolution of the flow experiments. We find that the time evolution of the
water flow is well described by a single characteristic relaxation time
predicted by the collective diffusion model coupled with the water friction.
###### pacs:
83.10.BbKinetics of deformation and flow and 83.80.KnPhysical gels and
microgels
## 1 Introduction
The gel is an important state of matter that is found in a wide variety of
biological, chemical, and food systems TanakaSci . The polymer gel consists of
a cross-linked polymer network and a large amount of solvent (typically
water). Since the average mesh size (the distance between neighboring
crosslinks) of the polymer network is in general large compared to small
molecules, they can pass through the gel easily. Two different transport
processes are important in the gel.
The first one is the diffusive flow of molecules Muhr . The transport of
molecules by the diffusion in the polymer network is modeled as the diffusion
of the probe objects in the fixed mesh of obstacles. The ratio between the
probe size and the mesh size plays essential role to determine the diffusion
coefficient of the molecules in the gels. The diffusion of the probe molecules
in the gel has been measured by the pulsed field gradient nuclear magnetic
resonance. The results indicate that the diffusion coefficient of the probe
molecules of various sizes in the gel is well described by a simple scaling
relationship TokitaPRE .
The second process is the convective flow of the water through the polymer
network TokitaJCP95 ; TokitaSci ; TokitaAPS . When the water flows in the gel,
it experiences the hydrodynamic friction from the polymer network, at the same
time, the polymer network of gel is deformed from the initial configuration by
the drag force of the water. The entire flowing process of the water is
determined by the balance between the viscoelastic response of the water flow
and that of the polymer network. The collective diffusion of the polymer
network coupled with water friction, therefore, plays essential roles in the
convective flow process in the gel.
As well as the transport phenomena in the gel, the water flow is also
important in the kinetics of the volume phase transition of the gel including
the pattern formations in shrinking gelsTanakaPRL ; TanakaJCP70 ; Matsuo ;
TokitaJPSJ ; MaskawaJCP ; TakigawaJCP111 ; TokitaJCP113 ; TakigawaJCP117 ;
BoudaoudPRE ; UrayamaJCP ; NosakaPoly . Kinetics of solvent flow in the gel is
the main subject of the application of the gel to control the solvent flow
YoshikawaJJAP ; SuzukiJCP . Although the significance, the solvent flow
process in the polymer gel has yet to be studied in detail.
It is the purpose of this paper to describe the kinetics of the water flow
through the polymer gel, based on the collective diffusion model coupled with
water friction. The characteristic relaxation time of the water flow process
in the gel is expressed as $\tau=L^{2}/\pi^{2}D_{c}$ with the parameters: the
typical size of the gel, $L$, and the collective diffusion coefficient of the
polymer network, $D_{c}$. The collective diffusion coefficient is expressed by
using the longitudinal modulus of the gel, $\kappa$, and the friction
coefficient between the polymer network and the water, $f$, as TanakaJCP70 ;
TanakaJCP59
$D_{c}=\frac{\kappa}{f}.$ (1)
The water flow in the polymer network is, thus, directly governed by the
viscoelastic property of the gel. The collective diffusion coefficient of
polymer gels has been extensively measured by quasi-elastic light scattering
Munch1 ; Munch2 . The friction coefficient between the polymer network of gel
and the water has been measured by flowing the water through the gel
TokitaJCP95 ; TokitaSci ; TokitaAPS .
In this paper, we present experimental results on the water flow through the
gel, and show that the time evolution of the water flow through the gel is
described satisfactorily by the collective diffusion model of polymer network
with water friction. The information on the kinetics of water flow in the gel
will help the better use of them in the applications.
## 2 Experiment
### 2.1 Water flow measurements
The experimental setting of the water flow measurements is illustrated
schematically in Figure 1.
Figure 1: Experimental setting of water flow measurements
A thin circular slab gel of thickness, $L$, is set in a tube and the rim of
the gel is glued to the tube. The surface of the gel at the lower reaches side
is mechanically fixed to the end plate through an O-ring. When the small
pressure, $P_{0}$, is applied to the water from the upper stream side of the
gel, the water flows through the gel deforming the polymer network of gel by
the drag force. The macroscopic value of the friction coefficient of the gel,
$f$, is determined from the velocity of the water flows out of the gel in the
equilibrium state, $v_{\infty}$, as
$f=\frac{P_{0}}{v_{\infty}L}.$ (2)
Even though the concept is simple, the mechanical measurement of the friction
coefficient between the polymer network of gel and the water is not easy
because the friction of the gel is huge. The difficulty of evaluating the
friction coefficient measured from the water flow through the gel is described
in the reference TokitaJCP95 . In our experiments, the difficulty is overcome
by using a well calibrated homogeneous glass capillary to amplify the velocity
of water as shown in Figure 1. The position of the meniscus of water in the
capillary is measured as a function of time after the pressure is applied to
the water.
The photographs of the apparatus used in this study are given in Figure 2.
Figure 2: Apparatus for water flow measurements: (a) the upper photograph
shows the external appearance, (b) the lower photographs shows the inside of
the end plates.
The essential structure of the apparatus is the same with the previous one but
some points are improved for the precise measurements of the friction of the
gel TokitaJCP95 . The right hand side of Figure 2 (a) is the upper stream side
and is connected to a water column to apply the hydrostatic pressure. The
glass capillary is set at the top of stainless steel pipe which can be seen at
the left hand side of Figure 2 (a). Leak of water around the gel is prevented
by setting O-rings inside the cell as shown in Figure 2 (b). All parts of the
cell is made of stainless steel to avoid the mechanical deformation due to the
applied pressure. The mechanical deformation of the apparatus is considerably
reduced compared to the previous one.
### 2.2 Mechanical response measurements
The mechanical response of the polymer gel relates the friction coefficient to
the collective diffusion coefficient of the polymer network as
$D_{c}=\kappa/f$ where $\kappa=K+\frac{4\mu}{3}$ TanakaJCP59 . Here, $\kappa$
is the longitudinal modulus of the gel and $K$ and $\mu$ are the bulk modulus
and the shear modulus of the gel, respectively. We measure two elastic moduli
of the gel simultaneously, and the longitudinal modulus of the gel is
determined from them.
The rod shape gel is elongated in water at a strain of about 10 $\%$ and the
stress is measured. The size of the gel is observed by using a microscope
during the elasticity measurements to obtain the Young’s modulus of the gel,
$E$, and the Poisson’s ratio, $\sigma$. The longitudinal modulus of the gel,
$\kappa$, is determined from the Young’s modulus and the Poisson’s ratio by
$\kappa=K+\frac{4\mu}{3}$, $K=E/3(1-2\sigma)$ and $\mu=E/2(1+\sigma)$.
In this short time experiments, the water solvent neither goes out nor comes
into the gels. Thus we determine the mechanical response of the gel
independently from the friction experiments by the water flow. We use the
results to analyze the kinetics of water flow through the polymer network.
### 2.3 Sample
The samples used in this paper are poly(acrylamide) gel that is obtained by
the radical co-polymerization of the main-chain component, acrylamide, and the
cross-linker, N,N’-methylene-bis-acrylamide. The concentration of the gel used
in this measurement is 1 M. Ammoniumpersulfate and N,N,N’,N’-tetramethyl-
ethylenediamine are used as the initiator and the accelerator. All chemicals
used here are of electrophoresis grade, purchased from BioRad and used without
further purification. The gel for the friction measurements is prepared in the
circular gel mold, where the gel-bond films (FMC) with circular opening are
glued on the both sides. For the friction coefficient measurements, the gel is
prepared in the measurement cell and kept under water at a constant room
temperature for overnight to reach the equilibrium state.
For the elastic modulus measurements, the rod shaped gels are prepared in the
capillary. The desired amounts of the main-chain component, the cross-linker,
and the accelerator are dissolved into the distilled and de-ionised water,
which is prepared by a Mili-Q system. The pre-gel solution is de-gassed for 20
min, and then the desired amount of the initiator is added to the solution to
initiate the reaction. The sample gels are taken out of the reaction bath and
then washed by distilled and de-ionized water extensively and used in each
measurement.
## 3 Experimental results
### 3.1 Friction coefficient of gel
The experimental results of the measurement of water flow through the gel are
shown in Figure 3.
Figure 3: Position of meniscus in capillary fit by equation (19) with $a$ and
$b$ as adjusted parameters.
The position of the meniscus in the capillary of $10\mu l$ glass micro-pipette
is plotted as a function of the time elapsed after the application of the
pressure to the water. The slope of the flow curve, thus obtained, represents
the water velocity in the capillary, $v^{\prime}$. The velocity of the flow is
high at first. Then it eventually decreases with time and approaches to an
equilibrium value. The equilibrium state is attained about 5 to 7 $\times$ 103
s after the pressure is applied to the water. The velocity of the meniscus in
the equilibrium state is determined by the least-squares analysis of the
linear portion of the flow curve in Figure 3, typically in the time region
more than 1 $\times$ 104 s, that yields to $v^{\prime}_{\infty}=1.9\times
10^{-7}$ m/s. The velocity of water flow in the gel, $v$, is then calculated
from the velocity of the meniscus, $v^{\prime}$, with the ratio of the radius
of the circular opening of the gel mold, $R$, and that of the capillary, $r$,
as $v=v^{\prime}(r/R)^{2}$. The values of $R$ and $r$ are measured as 1.745 mm
and 0.268 mm, respectively. The equilibirium velocity of water flow in the gel
is, thus, calculated as $v_{\infty}=4.4\times 10^{-9}$ m/s. The pressure
applied to the water is $P_{0}=2.9\times 10^{3}$ N/m2, which corresponds to
the height of water column of 30 cm, and the thickness of the gel is
$L=1.0\times 10^{-3}$ m. From these values, the friction coefficient of the
gel is calculated by equation (2) that yields to $f=6.6\times 10^{14}$ N s/m4.
This agrees with a typical value of the friction coefficient of the
transparent poly(acrylamide) gel evaluated by other experiments TokitaJCP95 .
### 3.2 Elastic modulus of gel
In Figure 4, we show the concentration dependence of the Young’s modulus and
the Poisson’s ratio of the gels used in our experiment.
Figure 4: Concentration dependence of elastic moduli: (a) the upper graph
shows Young’s modulus, $E$, (b) the lower graph shows Poisson’s ratio,
$\sigma$.
The least-squares analysis of the results yields that the concentration
dependence of the Young’s modulus of the gel is well described by a power law
relationship with a scaling exponent of 1.7, which is slightly smaller than
the value expected from the scaling argument deGennes . The Poisson’s ratio is
about 0.45 in the dilute gels. It decreases with the concentration of the gel
and reaches to about 0.35 at higher concentration region. The values of the
Poisson’s ratio of the dilute gels are close to those of the incompressible
materials. On the other hand, the values of the Poisson’s ratio of the dense
gels are rather close to those of the metals and the glasses. The Young’s
modulus and the Poisson’s ratio of the gel at a concentration of 1 M is
observed as $E=1.3\times 10^{4}$ N/m2 and $\sigma=0.44$, respectively. The
longitudinal modulus is then calculated as $\kappa=4.1\times 10^{4}$ N/m2 for
the gel at a concentration of 1 M. Our results are consistent with previously
published Poisson’s ratio of poly(acrylamide) gels and a theory TakigawaPJ1 ;
TakigawaPJ2 .
The diffusion coefficient of the polymer network is deduced from equation (1)
as $D_{c}=6.2\times 10^{-11}$ m2/s, which agrees with a typical value of the
diffusion coefficient of the transparent poly(acrylamide) gel evaluated by
other experiments TanakaJCP70 ; TanakaJCP59 ; Munch1 ; Munch2 .
## 4 Kinetics of water flow through gel
When the water flows in the gel (polymer network) by the application of the
mechanical pressure, the polymer network of gel deforms from the initial
position by the drag force of the water flow. We introduce a function,
$u(x,t)$, that represents the displacement of a point in the polymer network
from its initial position at $x$. We assume $u$ is always small compared to
the characteristic size of the gel, $L$. Under the present definition,
$u(x,t)=0$ at $t=0$. We also assume that the macroscopic velocity of the water
in the polymer network, $v(t)$, and microscopic friction coefficient, $f$, are
uniform in space.
$\frac{\partial v}{\partial x}=0.$ (3)
Then, we write down the equation of motion of the polymer network and that of
the water in the gel as
$\displaystyle\rho_{g}\ddot{u}$ $\displaystyle=$
$\displaystyle-f(\dot{u}-v)+\kappa\frac{\partial^{2}u}{\partial x^{2}},$ (4)
$\displaystyle\rho_{w}\dot{v}$ $\displaystyle=$ $\displaystyle
f(\dot{u}-v)-\frac{\partial P}{\partial x},$ (5)
where $\rho_{g}$ and $\rho_{w}$ are the density of the polymer network and
that of the water, respectively; A dot over the variable denotes the time
derivative; $P(x,t)$ represents the pressure inside the region that the gel
occupied.
By neglecting the inertia terms in equations (4) and (5), we obtain
$\kappa\frac{\partial^{2}u}{\partial x^{2}}-\frac{\partial P}{\partial x}=0,$
(6)
and
$f(v-\dot{u})=-\frac{\partial P}{\partial x}.$ (7)
A set of equations (3), (6), (7) corresponds to one dimensional case under a
limit of small polymer volume fraction $\phi\rightarrow 0$ of a general
description for 3 dimensional deformation of gelsYamaue ; Doi .
In our experimental setup of water flow measurement in Fig.1, we see
$P(x=0,t)=P_{0}$ and $P(x=L,t)=0$. Taking a spacial average of equation (7)
over the space the gel occupied, we obtain
$v-\frac{1}{L}\int_{0}^{L}\dot{u}dx=\frac{P_{0}}{Lf}.$ (8)
The time evolution of water flow in the gel $v(t)$ is determined locally from
$u(x,t)$ by
$v=\dot{u}-D_{c}\frac{\partial^{2}u}{\partial x^{2}}.$ (9)
Here, we use the collective diffusion coefficient of the gel, $D_{c}$, defined
in equation (1). The displacement of polymer network $u(x,t)$ is, hence,
determined by solving the equation:
$\dot{u}-\frac{1}{L}\int_{0}^{L}\dot{u}dx=D_{c}\frac{\partial^{2}u}{\partial
x^{2}}+\frac{P_{0}}{Lf}.$ (10)
In the equilibrium state, $\dot{u}(x,t)\rightarrow 0$ as $t\rightarrow\infty$.
The restoring force due to the deformation of the polymer network is in
balance with constant gradient of the pressure. In the equilibrium state,
therefore, equations (3), (6), (7), (8) give
$\frac{\partial^{2}u(x,t=\infty)}{\partial x^{2}}=-\frac{P_{0}}{D_{c}Lf}.$
(11)
The displacement, $u(x,t)$, satisfies the following boundary conditions:
$\displaystyle u(L,t)$ $\displaystyle=$ $\displaystyle 0,$ (12)
$\displaystyle\frac{\partial u(0,\infty)}{\partial x}$ $\displaystyle=$
$\displaystyle 0.$ (13)
Integration of equation (11) with respect to $x$ with the boundary conditons
yields
$u(x,\infty)=\frac{P_{0}}{2D_{c}Lf}(L^{2}-x^{2}).$ (14)
Solving equation (10) under the above boundary conditions, we obtain the
displacement, $u(x,t)$, as (See Appendix)
$u(x,t)=\sum_{n}\frac{2v_{\infty}\tau}{n^{2}}\left[(-1)^{n}\left(\cos{\frac{n\pi}{L}x}-1\right)\right]\exp{\left(-\frac{n^{2}}{\tau}t\right)},$
(15)
where
$\displaystyle\tau$ $\displaystyle=$
$\displaystyle\frac{L^{2}f}{\pi^{2}\kappa},$ (16) $\displaystyle v_{\infty}$
$\displaystyle=$ $\displaystyle\frac{P_{0}}{Lf}.$ (17)
The velocity of the water flow, $v(t)$, is derived from equation (7) as
$v=v_{\infty}\left[1+2\sum_{n=1}^{\infty}\exp\left(-\frac{n^{2}}{\tau}t\right)\right]=v_{\infty}\theta_{3}(0,e^{t/\tau}),$
(18)
where $\theta_{3}(u,q)$ is the elliptic theta function. The position of
meniscus is expressed as
$a\left[t-2\sum_{n=1}^{\infty}\frac{\tau}{n^{2}}\exp\left(-\frac{n^{2}}{\tau}t\right)\right]+b,$
(19)
where $b$ is the initial position, and
$a=\left(\frac{R}{r}\right)^{2}v_{\infty}.$ (20)
## 5 Discussion
Now we compare the experimental results with the theoretical predictions by
equations (18) and (19). The values obtained by the experimental measurements
are, $L=1.0\times 10^{-3}$ m, $\kappa=4.1\times 10^{4}$ N/m2, and $f=6.6\times
10^{14}$ Ns/m4, respectively.
By applying these values into equation (16), we obtain the characteristic
relaxation time $\tau=1.6\times 10^{3}$ s. The equilibrium velocity of water
flow in the gel observed by the experiments is, as is already given in the
section 3, $v_{\infty}=4.4\times 10^{-9}$ m/s. From equation (14) the
deformation of gel is estimated as 3.5$\%$ which is consistent with our
assumption of small $u$.
The time evolution of the water flow is then calculated by equation (18) with
above experimental values. The results are given in Figure 5.
Figure 5: Water flow velocity in gel: (a) the solid line corresponds to the
theoretical curve fitting all the data, (b) the dotted line corresponds to the
theoretical curve expected from the short time data ($<$200s).
The normalized velocity of the water flow in the gel is also calculated from
the slope of the flow curve in Figure 3 and shown in Figure 5. We find that
the time evolution of the velocity of water flow in the gel is well explained
by the equation (18) with a single characteristic relaxation time $\tau$.
However, there is a slight discrepancy in the time evolution curve of the
velocity at a short time range ($<$200s) between the theory and the
experimental results as one can see in figure 3 and 5. The absolute value of
the equilibrium velocity determined from the short time data ($<$200s) of the
friction experiment is smaller than that determined by including the long time
data. Figure 5 shows that the deference between those two values is roughly 20
$\%$.
It would be worth noting that the theory predicts that the water flow creates
a uni-axial deformation of polymer network, i.e., the one dimensional
concentration gradient in the gel, proportional to
$\left(1+\frac{\partial u(x,\infty)}{\partial x}\right)^{-1}\approx
1+\frac{P_{0}x}{D_{c}Lf}.$ (21)
The compression of gel is the cause of the initial rapid water flow. It is
reported that the friction of the homegeneous gel depends on the polymer
volume fraction, $\phi$, as $f\propto\phi^{1.5}$ TokitaJCP95 ; deGennes . The
friction, therefore, is expected to increase according to the compression of
the gel by the water flow. The increment of the friction may explain the
difference between the short time data and long time data. The difference
between the estimated equilibrium water velocities corresponds to 14 $\%$
homogeneous compression of the sample slab gel, which is larger than the
deformation which the present theory estimates. Unfortunately, we cannot
measure the deformation of the gel under the water flow with the current
experimental setup.
## 6 Conclusion
The kinetics of the water solvent flow through the polymer network of gel is
measured experimentally and described by a simple phenomenological theory. The
experiment shows that the equilibrium state is reached after long time and
that the friction of the gel shows a large value. The time evolution of the
velocity of the water flow in the gel is calculated on the basis of the
collective diffusion model of the polymer network coupled with the water
friction, assuming both the water velocity and the friction coefficient
between the polymer network and the water are uniform in the gel. The theory
reproduces the time evolution of the water flow of the experiments for thin
dilute gel fairly well with a single characteristic relaxation time. We are
able to deduce the collective diffusion coefficient from the water flow
experiment on gels.
However there is a difference between experimental value of the equilibrium
velocity and the value expected from the theoretical calculation by using
parameters obtained from the short time flow measurements and the mechanical
measurements. It suggests that the inhomogeneity of the friction and the
concentration of the gel under the water flow might be important for the
kinetics of the water flow. A better microscopic model is required for
complete understanding of water flow through gel beyond the phenomenological
understanding presented in this paper.
###### Acknowledgements.
This work was first suggested by the late Professor Toyoichi Tanaka in 1989
during the stay of Y.Y.S. and M.T. in Massachusetts Institute of Technology.
Y.Y.S. thanks CEA, IPhT and Takushoku University, RISE for the financial
support. M.T. thanks NEDO for the financial support.
## APPENDIX A: Solution of the equation of motion
We define a function $W(x,t)$ as a deviation from the equilibrium
displacement:
$u(x,t)-\int v(t)dt\equiv W(x,t)+u(x,\infty).$ (22)
We obtain an equation for $W(x,t)$ from equation (10) as
$\frac{\partial W}{\partial t}=D_{c}\frac{\partial^{2}W}{\partial x^{2}}.$
(23)
By assuming $W(x,t)=X(x)T(t)$, the equation (23) becomes
$\displaystyle\dot{T}(t)$ $\displaystyle=$
$\displaystyle-\lambda^{2}D_{c}T(t),$ (24)
$\displaystyle\frac{d^{2}X(x)}{dx^{2}}$ $\displaystyle=$
$\displaystyle-\lambda^{2}X(x),$ (25)
where $\lambda$ is a constant. The general solution for $W(x,t)$ would be
$\displaystyle W(x,t)$ $\displaystyle=$
$\displaystyle\sum_{n=1}^{\infty}\left[A_{n}\cos\lambda_{n}x+B_{n}\sin\lambda_{n}x\right]$
(26) $\displaystyle\times\exp(-\lambda_{n}^{2}D_{c}t)+A_{0}x+B_{0}.$
The boundary conditions for $W(x,t)$ are
$\displaystyle\left.\frac{du}{dx}\right|_{x=0}=0$
$\displaystyle\longrightarrow$
$\displaystyle\left.\frac{dW}{dx}\right|_{x=0}=0,$ (27)
$\displaystyle\left.\frac{du}{dt}\right|_{t=\infty}=0$
$\displaystyle\longrightarrow$
$\displaystyle\left.\frac{dW}{dt}\right|_{t=\infty}=v(\infty),$ (28)
$\displaystyle u(L,t)=0$ $\displaystyle\longrightarrow$ $\displaystyle
W(L,t)=-\int vdt.$ (29)
The boundary condition (27) yields
$\displaystyle\left.\frac{dW}{dx}\right|_{x=0}$ $\displaystyle=$
$\displaystyle\sum_{n=1}^{\infty}[A_{n}(-\lambda_{n}\sin\lambda_{n}x)+B_{n}(\lambda_{n}\cos\lambda_{n}x)]$
$\displaystyle\left.\times\exp(-\lambda_{n}^{2}D_{c}t)\right|_{x=0}+B_{0}$
$\displaystyle=$
$\displaystyle\left.\sum_{n=1}^{\infty}[A_{n}(-\lambda_{n}\times
0)+B_{n}(\lambda_{n}\times 1)]\exp(-\lambda_{n}^{2}D_{c}t)\right|_{x=0}$
$\displaystyle+B_{0}$ $\displaystyle=$ $\displaystyle 0.$
This indicates $B_{n}=0$.
The boundary condition (28) yields
$\displaystyle\left.\frac{dW}{dx}\right|_{t=\infty}$ $\displaystyle=$
$\displaystyle\left.\sum_{n=1}^{\infty}[A_{n}\cos\lambda_{n}x](-\lambda_{n}^{2}D_{c})\exp(-\lambda_{n}^{2}D_{c}t)\right|_{t=\infty}+A_{0}$
$\displaystyle=$ $\displaystyle v(\infty).$
This indicates that the constant, $\lambda_{n}^{2}$, should be positive,
$\lambda_{n}^{2}>0$ for $n\geq 1$ and $A_{0}=v(\infty)$. Finally, the boundary
condition (29) yields
$\displaystyle W(L,t)$ $\displaystyle=$
$\displaystyle\sum_{n=1}^{\infty}[A_{n}(\cos\lambda_{n}L+C_{nA})]\exp(-\lambda_{n}^{2}D_{c}t)+A_{0}L+B_{0}$
$\displaystyle=$ $\displaystyle-\int vdt.$
This indicates $C_{nA}=-\cos\lambda_{n}L$. The following relationships for
$B_{n}$, $\lambda_{n}^{2}$, and $C_{nA}$ are obtained:
$\displaystyle B_{n}$ $\displaystyle=$ $\displaystyle 0,$ (30)
$\displaystyle\lambda_{n}^{2}$ $\displaystyle>$ $\displaystyle 0,$ (31)
$\displaystyle C_{nA}$ $\displaystyle=$ $\displaystyle-\cos\lambda_{n}L.$ (32)
And $W(x,t)$ becomes
$W(x,t)=\sum_{n=1}^{\infty}[A_{n}(\cos\lambda_{n}x+C_{nA})]\exp(-\lambda_{n}^{2}D_{c}t).$
(33)
Substitution of this solution (33) into equation (23) yields
$\displaystyle\sum_{n=1}^{\infty}[A_{n}(\cos\lambda_{n}x+C_{nA})](-\lambda_{n}^{2}D_{c})\exp(-\lambda_{n}^{2}D_{c}t)$
$\displaystyle-\frac{1}{L}\int_{0}^{L}\sum_{n=1}^{\infty}[A_{n}(\cos\lambda_{n}x+C_{nA})]$
$\displaystyle\times(-\lambda_{n}^{2}D_{c})\exp(-\lambda_{n}^{2}D_{c}t)dx$
$\displaystyle=D_{c}\sum_{n=1}^{\infty}A_{n}\lambda_{n}^{2}(-\cos\lambda_{n}x)\exp(-\lambda_{n}^{2}D_{c}t).$
(34)
The first term of the left handed member (LHM1) is written as follows:
$\displaystyle{\rm LHM1}=$
$\displaystyle\sum_{n=1}^{\infty}(-\lambda_{n}^{2}D_{c})A_{n}(\cos\lambda_{n}x)\exp(-\lambda_{n}^{2}D_{c}t)$
$\displaystyle+$
$\displaystyle\sum_{n=1}^{\infty}(-\lambda_{n}^{2}D_{c})(A_{n}C_{nA})\exp(-\lambda_{n}^{2}D_{c}t).$
The second term of the left handed member (LHM2) is expressed as follows:
$\displaystyle{\rm LHM2}=$ $\displaystyle-$
$\displaystyle\frac{1}{L}\int_{0}^{L}\sum_{n=1}^{\infty}(-\lambda_{n}^{2}D_{c})(A_{n}\cos\lambda_{n}x)$
$\displaystyle\times\exp(-\lambda_{n}^{2}D_{c}t)dx$ $\displaystyle-$
$\displaystyle\frac{1}{L}\int_{0}^{L}\sum_{n=1}^{\infty}(-\lambda_{n}^{2}D_{c})(A_{n}C_{nA})\exp(-\lambda_{n}^{2}D_{c}t)dx.$
The first term of LHM2 is expressed as
$\displaystyle-$
$\displaystyle\frac{1}{L}\int_{0}^{L}\sum_{n=1}^{\infty}(-\lambda_{n}^{2}D_{c})(A_{n}\cos\lambda_{n}x)\exp(-\lambda_{n}^{2}D_{c}t)dx$
$\displaystyle=$ $\displaystyle-$
$\displaystyle\left.\frac{1}{L}\sum_{n=1}^{\infty}(-\lambda_{n}^{2}D_{c})\frac{1}{\lambda_{n}}(A_{n}\sin\lambda_{n}x)\exp(-\lambda_{n}^{2}D_{c}t)\right|_{0}^{L}$
$\displaystyle=$ $\displaystyle-$
$\displaystyle\frac{1}{L}\sum_{n=1}^{\infty}(-\lambda_{n}^{2}D_{c})\frac{1}{\lambda_{n}}(A_{n}\sin\lambda_{n}L)\exp(-\lambda_{n}^{2}D_{c}t).$
The second term of LHM2 becomes
$\displaystyle-$
$\displaystyle\frac{1}{L}\int_{0}^{L}\sum_{n=1}^{\infty}(-\lambda_{n}^{2}D_{c})(A_{n}C_{nA})\exp(-\lambda_{n}^{2}D_{c}t)dx$
$\displaystyle=$ $\displaystyle-$
$\displaystyle\sum_{n=1}^{\infty}(-\lambda_{n}^{2}D_{c})(A_{n}C_{nA})\exp(-\lambda_{n}^{2}D_{c}t).$
Hence, the equation (34) is reduced to
$-\frac{1}{L}\sum_{n=1}^{\infty}(-\lambda_{n}^{2}D_{c})A_{n}\frac{1}{\lambda_{n}}(\sin\lambda_{n}L)\exp(-\lambda_{n}^{2}D_{c}t)=0.$
This indicates
$\lambda_{n}=\frac{n\pi}{L},$ (35)
and
$C_{nA}=-\cos(n\pi)=-(-1)^{n}.$ (36)
The solution, $W(x,t)$, can be now written as
$W(x,t)=\sum_{n=1}^{\infty}A_{n}\left(\cos\frac{n\pi}{L}x-(-1)^{n}\right)\exp\left(-\frac{n^{2}\pi^{2}}{L^{2}}D_{c}t\right).$
(37)
Since the initial condition is
$W(x,0)=\frac{P_{0}}{2D_{c}Lf}(x^{2}-L^{2}),$ (38)
the coefficients, $A_{n}$, are determined by the Fourier series expansion:
$\sum_{n=1}^{\infty}A_{n}\left(\cos\frac{n\pi}{L}x-(-1)^{n}\right)=\frac{P_{0}}{2D_{c}Lf}(x^{2}-L^{2}).$
(39)
We find
$A_{n}=\frac{(-1)^{n}}{n^{2}\pi^{2}}\frac{2P_{0}L}{D_{c}f}.$ (40)
The displacement, $u(x,t)$, is now determined as
$u(x,t)=\sum_{n=1}^{\infty}\mathcal{A}_{n}\left[(-1)^{n}(\cos
q_{n}x-1)\right]\exp(-\Gamma_{n}t),$ (41)
where, $\mathcal{A}_{n}$, $q_{n}$, and $\Gamma_{n}$ are given by
$\displaystyle\mathcal{A}_{n}$ $\displaystyle=$
$\displaystyle\frac{2P_{0}L}{n^{2}\pi^{2}D_{c}f},$ (42) $\displaystyle q_{n}$
$\displaystyle=$ $\displaystyle\frac{n\pi}{L},$ (43) $\displaystyle\Gamma_{n}$
$\displaystyle=$ $\displaystyle\left(\frac{n\pi}{L}\right)^{2}D_{c}.$ (44)
## References
* (1) T. Tanaka, Sci. Am. 244, 124 (1981)
* (2) A.H. Muhr, J.M.B. Blanshard, Polymer 23, 1012 (1982)
* (3) M. Tokita, T. Miyoshi, K. Takegoshi, K.Hikichi, Phys. Rev. E 53, 1823 (1996)
* (4) M. Tokita, T. Tanaka, J. Chem. Phys. 95, 4613 (1991)
* (5) M. Tokita, T. Tanaka, Science 253, 1121 (1991)
* (6) M. Tokita, Adv. Polym. Sci. 110, 27 (1993)
* (7) T. Tanaka, S. Ishiwata, C. Ishimoto, Phys. Rev. Lett. 38, 771 (1977)
* (8) T. Tanaka, D.J. Fillmore, J. Chem. Phys. 70, 1214 (1979)
* (9) E.S. Matsuo, T. Tanaka, Nature 358, 482 (1992)
* (10) M. Tokita, S. Suzuki, K. Miyamoto, T. Komai, J. Phys. Soc. Jpn. 68, 330 (1999)
* (11) J. I. Maskawa, T. Takeuchi, K. Maki, K. Tsujii, T. Tanaka, J. Chem. Phys. 105, 1735 (1999)
* (12) T. Takigawa, K. Uchida, K. Takahashi, T. Masuda, J. Chem. Phys. 111, 2295 (1999)
* (13) M. Tokita, K. Miyamoto, T. Komai, J. Chem. Phys. 113, 1647 (2000)
* (14) T. Takigawa, T. Ikeda, Y. Takakura, T. Masuda, J. Chem. Phys. 117, 17306 (2002)
* (15) A. Boudaoud, S. Chaieb, Phys. Rev. E 68, 021801 (2003)
* (16) L. Urayama, S. Okada, S. Nosaka, H. Watanabe, K. Takigawa, J. Chem. Phys. 122, 024906 (2005)
* (17) S. Nosaka, S. Okada, Y. Takayama, K. Urayama, H. Watanabe, T. Takigawa, Polymer 46, 12607 (2005)
* (18) M. Yoshikawa, R. Ishii, J. Matsui, A. Suzuki, M. Tokita, Jpn. J. Appl. Phys. Part 1 44, 8196 (2005)
* (19) A. Suzuki, M. Yoshikawa, J. Chem. Phys. 125, 174901 (2006)
* (20) T. Tanaka, L.O. Hocker, G.B. Benedek, J. Chem. Phys. 59, 5151 (1973)
* (21) P-G. de Gennes, Scaling Concepts in Polymer Physics (Cornell University Press, Ithaca, London, 1979)
* (22) J.P. Munch, S. Candau, J. Hertz, G. Hild, J. Phys. (Paris) 38, 971 (1977)
* (23) J.P. Munch, P. Lemarechal, S. Candau, J. Phys. (Paris) 38, 1499 (1977)
* (24) T. Takigawa, K. Urayama, and T. Masuda, Polymer J. 26, 225 (1994)
* (25) T. Takkigawa, Y. Morino, K. Urayama, and T. Masuda, Polymer J. 28, 1012 (1996)
* (26) T. Yamaue, M. Doi, Phys. Rev. E69, 041402 (2004)
* (27) M. Doi, Dynamics and Patterns in Complex Fluids, edited by A. Onuki and K. Kawasaki (Springer, New York, 1990), p.100
| arxiv-papers | 2008-07-11T08:16:43 | 2024-09-04T02:48:56.694254 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Yasuo Y. Suzuki, Masayuki Tokita, Sata-atsu Mukai",
"submitter": "Yasuo Y. Suzuki",
"url": "https://arxiv.org/abs/0807.1789"
} |
0807.1847 | # Form factors of integrable higher-spin XXZ chains and the affine quantum-
group symmetry
Tetsuo Deguchi1111e-mail deguchi@phys.ocha.ac.jp and Chihiro Matsui2,3
222e-mail matsui@spin.phys.s.u-tokyo.ac.jp
###### Abstract
We derive exactly scalar products and form factors for integrable higher-spin
XXZ chains through the algebraic Bethe-ansatz method. Here spin values are
arbitrary and different spins can be mixed. We show the affine quantum-group
symmetry, $U_{q}(\widehat{sl_{2}})$, for the monodromy matrix of the XXZ spin
chain, and then obtain the exact expressions. Furthermore, through the
quantum-group symmetry we explicitly derive the diagonalized forms of the $B$
and $C$ operators in the $F$-basis for the spin-1/2 XXZ spin chain, which was
conjectured in the algebraic Bethe-ansatz calculation of the XXZ correlation
functions. The results should be fundamental in studying form factors and
correlation functions systematically for various solvable models associated
with the integrable XXZ spin chains.
1 Department of Physics, Graduate School of Humanities and Sciences,
Ochanomizu University
2-1-1 Ohtsuka, Bunkyo-ku, Tokyo 112-8610, Japan
2 Department of Physics, Graduate School of Science, University of Tokyo
7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan
3 CREST, JST, 4-1-8 Honcho Kawaguchi, Saitama, 332-0012, Japan
## 1 Introduction
Correlation functions of the spin-1/2 XXZ spin chain have attracted much
attention in mathematical physics for more than a decade [1, 2, 3]. The
multiple-integral representations of XXZ correlation functions were first
derived in terms of the $q$-vertex operators [4]. Based on the algebraic Bethe
ansatz method, the determinant expressions [5] of the scalar products and the
norms of Bethe ansatz eigenstates were reconstructed in terms of the $F$-basis
[6], and then the XXZ correlation functions are derived under an external
magnetic field [7, 8]. The multiple-integral representations at zero
temperature have been extended into those at finite temperature [9].
Furthermore, dynamical structure factors have been evaluated by solving the
Bethe ansatz equations numerically [10, 11, 12, 13].
The Hamiltonian of the spin-1/2 XXZ chain under the periodic boundary
conditions is given by
${\cal H}_{\rm
XXZ}={\frac{1}{2}}\sum_{j=1}^{L}\left(\sigma_{j}^{X}\sigma_{j+1}^{X}+\sigma_{j}^{Y}\sigma_{j+1}^{Y}+\Delta\sigma_{j}^{Z}\sigma_{j+1}^{Z}\right)\,.$
(1.1)
Here $\sigma_{j}^{a}$ ($a=X,Y,Z$) are the Pauli matrices defined on the $j$th
site, and $\Delta$ the XXZ coupling. By $\Delta=(q+q^{-1})/2$ we define
parameter $q$, which plays a significant role in the paper. We note that $L$
denotes the number of the one-dimensional lattice sites. Solvable higher-spin
generalizations of the XXX and XXZ chains have been constructed by the fusion
method in several references [14, 15, 16, 17, 18, 19]. For instance, the
Hamiltonian of the solvable spin-$s$ XXX chain is given by the following [15]:
$H_{s}=J\sum_{j=1}^{L}Q_{2s}({\vec{S}}_{j}\cdot{\vec{S}}_{j+1})\,,$ (1.2)
where ${\vec{S}}_{j}$ are operators of spin $s$ acting on the $j$th site and
$Q_{2s}(x)$ is a polynomial of degree $2s$
$Q_{2s}(x)=\sum_{p=1}^{2s}\left(\sum_{k=1}^{p}{\frac{1}{k}}\right)\prod_{\ell=0,\ell\neq
j}^{2s}{\frac{x-x_{\ell}}{x_{j}-x_{\ell}}}\,.$ (1.3)
Here $x_{\ell}=[\ell(\ell+1)-2s(s+1)]/2$. At $T=0$ in the critical regime, it
is discussed that the low-excitation spectrum of the spin-$s$ XXX chain is
described in terms of the level-$k$ $SU(2)$ WZWN model where $k=2s$ [20].
In the present paper we derive exact expressions of scalar products and form
factors for the integrable higher-spin XXZ spin chains. Here different spins
can be mixed. We first show that the monodromy matrix of the XXZ spin chain
has the symmetry of the affine quantum group $U_{q}({\widehat{sl_{2}}})$.
Then, we derive the exact expressions taking advantage of the quantum group
symmetry. By a similarity transformation [21], we transform the symmetric
$R$-matrix into an asymmetric one, which is directly connected to the quantum
group $U_{q}(sl_{2})$. We derive projection operators from the asymmetric
$R$-matrices [19], and construct integrable higher-spin XXZ spin chains by the
fusion method similarly as the case of the XXX spin chain [14]. Here we make
an extensive use of the $q$-analogues of Young’s projection operators, which
play a central role in the $q$-analogue of the Schur-Weyl reciprocity of the
quantum group $U_{q}(sl_{2})$ [22, 23]. Hereafter, we call transformations on
the $R$-matrix gauge transformations. After we construct integrable higher-
spin models, applying the inverse transformation to them, we reduce the
asymmetric monodromy matrices into those of the symmetric $L$-operators
constructed from the symmetric $R$-matrices, and thus obtain scalar products
and form factors for the integrable higher-spin models in the standard
formulation.
Form factors and correlation functions have been discussed for integrable
higher-spin XXX models in previous researches [24, 25, 26, 27]. In the
approach of the so-called quantum inverse scattering problem for the
integrable $N$-state models, one has to construct the $N$-by-$N$ monodromy
matrix in order to express local operators in terms of the global operators.
However, it seems to be technically nontrivial to construct the $N$-by-$N$
monodromy matrix for the integrable higher-spin systems (see also [28]). By
the approach of the present paper, the calculational task is much reduced into
the minimal level. In fact, as a consequence of the $q$-analogue of the Schur-
Weyl duality, the exact expressions of scalar products and form factors are
derived from the formulas of the spin-1/2 XXZ chain by setting the
inhomogeneous parameters in the form of “complete $\ell$-strings”[29], and the
most general results are straightforwardly obtained.
The affine quantum-group invariance has another important consequence. In the
XXZ case we can explicitly prove the pseudo-diagonalized forms of the $B$ and
$C$ operators of the algebraic Bethe-ansatz method through the quantum-group
symmetry. Here we remark that the XXZ spin chain has no spin SU(2) symmetry.
In the pseudo-diagonal basis, the $B$ and $C$ operators are expressed as sums
of local spin operators $\sigma_{i}^{-}$ and $\sigma_{i}^{+}$ multiplied by
diagonal matrices, respectively, where each of the local spin operators
$\sigma_{i}^{\pm}$ are defined on one lattice-site, i.e. on the $i$th site.
The pseudo-diagonalized forms of the $B$ and $C$ operators were conjectured in
the algebraic Bethe-ansatz derivation of the XXZ correlation functions [6, 7,
8]. In fact, the $B$ operators create the Bethe states where the $C$ operators
are conjugate to them, and the pseudo-diagonalized forms play a central role
in the calculation of the scalar products and the norms of Bethe states. They
are also fundamental in the quantum inverse scattering problem, by which local
operators are expressed in terms of the global operators such as the $B$ and
$C$ operators. In the XXZ case, however, an explicit derivation of the pseudo-
diagonalized forms of the $B$ and $C$ operators has not been shown previously,
yet. Thus, the explicit derivation in the paper completes the algebraic Bethe-
ansatz formulation of the form factors and correlation functions of the
integrable XXZ spin chains. Here we remark that for the XXX case (i.e. the
isotropic case), the diagonalized forms have been shown by Maillet and Sanchez
de Santos in Ref. [6] by making an explicit use of the rotational
SU(2)-symmetry. However, the method for the XXX case does not hold for the XXZ
spin chain which has no SU(2) symmetry.
The derivation of the affine quantum-group symmetry of the monodromy matrix
should be not only theoretically interesting but also practically useful for
calculation. Here we remark that the infinite-dimensional symmetry,
$U_{q}({\widehat{sl_{2}}})$, was realized for the infinite XXZ spin chain with
the $q$-vertex operators [2, 4]. Thus, it should be an interesting open
problem how the affine quantum-group symmetry of the finite XXZ spin chain can
be related to that of the infinite XXZ spin chain. Furthermore, there are
several advantages in the present formulation of the affine quantum-group
symmetry. We derive the symmetry through gauge transformations. The
transformed asymmetric $R$-matrix is directly related to the quantum group
$U_{q}(sl_{2})$ so that we can systematically construct higher-spin
representations of the $R$-matrices through the $q$-analogue of the Young
symmetrizers. Here we should note that the gauge transformation connects the
$R$ matrix in the different gradings of $U_{q}(\widehat{sl_{2}})$. The
symmetric and asymmetric $R$-matrices are equivalent to that of the principal
and homogeneous gradings, respectively (see for instance §5.4 of [2]).
Moreover, we thus avoid technical difficulties appearing when we directly
derive the matrix representation of the universal $R$-matrix of the affine
quantum group, which is given by a product of infinite series of generators
[30]. Although one can construct matrix representations of the modified
universal $R$-matrix with its derivation $d$ dropped [30, 22](see A.2 of
[31]), it seems that the calculation is not quite straightforward when we
construct higher-spin representations.
The results of the present paper should be useful for calculating exact
expressions of correlation functions for various integrable models associated
with higher-spin XXZ chains. For instance, the $\tau_{2}$ model in the
$N$-dimensional nilpotent representation corresponds to the integrable
spin-$(N-1)/2$ XXZ spin chain with $q$ being a primitive $N$th root of unity
[32]. Here we remark that the $\tau_{2}$ model is closely related to the
$N$-state superintegrable chiral Potts model. Furthermore, there are several
possible physical applications, such as calculating form factors of quantum
impurity models through the result in the case of mixed spins. As an
illustrative example, we have calculated exact expressions for the emptiness
formation probability of the higher-spin XXZ spin chains, which we shall
discuss in a subsequent paper.
The content of the paper consists of the following: In section 2, we introduce
the symmetric $R$-matrix of the spin-1/2 XXZ spin chain, and define the
monodromy matrix [33]. We also define the action of the symmetric group on
products of $R$-matrices. In section 3 we derive the symmetry of the quantum
affine algebra for the monodromy matrix of the XXZ spin chain. We introduce
the asymmetric $R$-matrix and then derive it from the symmetric one by a gauge
transformation. We decompose the asymmetric $R$-matrix in terms of the
generators of the Temperley-Lieb algebra, and show the affine quantum-group
symmetry. We also derive it by a systematic method for expressing products of
$R$-matrices formulated in definition 2.21. In fact, all the fundamental
relations of the quantum inverse-scattering problem can be derived much more
simply without using the ${\hat{R}}$-matrix of Ref. [6], as shown in
Appendices A and B. In section 4 we construct the $R$ matrices of integrable
higher-spin XXZ spin chains with projection operators of $U_{q}(sl_{2})$ by
the fusion method. We also discuss the case of mixed spins. In section 5 we
formulate an explicit derivation of the pseudo-diagonalized forms of the
$B$-operators. We also show it for the $C$-operator in Appendix E. In section
6 we derive determinant expressions of scalar products for the higher-spin XXZ
spin chains. In section 7, for the higher-spin cases we show the method by
which we can express local operators in terms the global operators. We give
some useful formulas of the quantum inverse scattering problem for the higher
spin case. Finally, we derive some examples of form factors for the integrable
higher-spin XXZ spin chains.
## 2 $R$-matrices and $L$-operators
### 2.1 Symmetric $R$-matrix
We shall introduce the $R$-matrix for the XXZ spin chain [33]. We consider two
types of $R$-matrices, $R_{ab}(u)$ and $R_{ab}(\lambda,\mu)$. The $R$-matrix
with a single rapidity argument, $R_{ab}(u)$, acts on the tensor product of
two vector spaces $V_{a}$ and $V_{b}$, i.e. $R_{ab}(u)\in End(V_{a}\otimes
V_{b})$, where parameter $u$ is independent of $V_{a}$ or $V_{b}$. The
$R$-matrix with two rapidity arguments, $R_{ab}(\lambda,\mu)$, acts on the
tensor product of vector spaces with parameters, $V_{a}(\lambda)$ and
$V_{b}(\mu)$, i.e. $R_{ab}(\lambda,\mu)\in End(V_{a}(\lambda)\otimes
V_{b}(\mu))$.
Let us denote by $e^{a,b}$ such a matrix that has only one nonzero element
equal to 1 at entry $(a,b)$. We denote by $V$ the two-dimensional vector
space. We define the $R$-matrix acting on the tensor product $V\otimes V$ by
$R(u)=\sum_{a,b,c,d=1,2}R^{ab}_{cd}(u)e^{a,c}\otimes e^{b,d}\,.$ (2.1)
Here matrix elements $R^{ab}_{cd}(u)$ satisfy the charge conservation, i.e.
$R^{ab}_{cd}(u)=0$ unless $a+b=c+d$, and all the nonzero elements are given by
the following:
$\displaystyle R^{11}_{11}(u)$ $\displaystyle=$ $\displaystyle
R^{22}_{22}(u)=1\,,\quad R^{12}_{12}(u)=R^{21}_{21}(u)=b(u),$ $\displaystyle
R^{12}_{21}(u)$ $\displaystyle=$ $\displaystyle R^{21}_{12}(u)=c(u)\,,$ (2.2)
where functions $b(u)$ and $c(u)$ are given by
$b(u)={\frac{\sinh(u)}{\sinh(u+\eta)}}\,,\quad
c(u)={\frac{\sinh(\eta)}{\sinh(u+\eta)}}\,.$ (2.3)
Here, parameter $\eta$ is related to $q$ of $\Delta=(q+q^{-1})/2$ by
$q=\exp(\eta)$.
We now introduce operators acting on the $L$th power of tensor product of
vector spaces with parameters, $V(\lambda_{1})\otimes\cdots\otimes
V(\lambda_{L})$. We generalize the notation of (2.1). Let us take a pair of
integers $j$ and $k$ satisfying $1\leq j<k\leq L$. For a given set of matrix
elements $A^{a,\,b}_{c,\,d}(\lambda_{j},\lambda_{k})$ ($a,b,c,d=1,2$) we
define operators $A_{j,k}(\lambda_{j},\lambda_{k})$ and
$A_{k,j}(\lambda_{k},\lambda_{j})$ by
$\displaystyle A_{j,k}(\lambda_{j},\lambda_{k})$ $\displaystyle=$
$\displaystyle\sum_{a,b,\alpha,\beta=1,2}A^{a,\,\alpha}_{b,\,\beta}(\lambda_{j},\lambda_{k})I_{1}\otimes\cdots\otimes
I_{j-1}$ $\displaystyle\quad\otimes e^{a,b}_{j}\otimes
I_{j+1}\otimes\cdots\otimes I_{k-1}\otimes e_{k}^{\alpha,\beta}\otimes
I_{k+1}\otimes\cdots\otimes I_{L}\,,$ $\displaystyle
A_{k,j}(\lambda_{k},\lambda_{j})$ $\displaystyle=$
$\displaystyle\sum_{a,b,\alpha,\beta=1,2}A^{\alpha,\,a}_{\beta,\,b}(\lambda_{k},\lambda_{j})I_{1}\otimes\cdots\otimes
I_{j-1}$ (2.4) $\displaystyle\quad\otimes e^{a,b}_{j}\otimes
I_{j+1}\otimes\cdots\otimes I_{k-1}\otimes e_{k}^{\alpha,\beta}\otimes
I_{k+1}\otimes\cdots\otimes I_{L}\,.$
Here $I$ is the two-by-two unit matrix, and $I_{j}$ and $e_{j}^{a,b}$ act on
the $j$th vector space $V(\lambda_{j})$ of $V(\lambda_{1})\otimes\cdots\otimes
V(\lambda_{L})$. In terms of matrices we express operators $A_{j,k}$ and
$A_{k,j}$ for $j<k$ by
$A_{j,k}=\left(\begin{array}[]{cccc}A^{11}_{11}&A^{11}_{12}&A^{11}_{21}&A^{11}_{22}\\\
A^{12}_{11}&A^{12}_{12}&A^{12}_{21}&A^{12}_{22}\\\
A^{21}_{11}&A^{21}_{12}&A^{21}_{21}&A^{21}_{22}\\\
A^{22}_{11}&A^{22}_{12}&A^{22}_{21}&A^{22}_{22}\end{array}\right)_{[j,k]}\,,\quad
A_{k,j}=\left(\begin{array}[]{cccc}A^{11}_{11}&A^{11}_{21}&A^{11}_{12}&A^{11}_{22}\\\
A^{21}_{11}&A^{21}_{21}&A^{21}_{12}&A^{21}_{22}\\\
A^{12}_{11}&A^{12}_{21}&A^{12}_{12}&A^{12}_{22}\\\
A^{22}_{11}&A^{22}_{21}&A^{22}_{12}&A^{22}_{22}\end{array}\right)_{[j,k]}$
(2.5)
Here by the symbol $[j,k]$ we express that matrix element $A^{ab}_{cd}$
corresponds to $e_{j}^{a,c}\otimes e_{k}^{b,d}$ for $A_{j,k}$, and to
$e_{j}^{b,d}\otimes e_{k}^{a,c}$ for $A_{k,j}$.
Let us now introduce operators $R_{j,k}(\lambda_{j},\lambda_{k})$ and
$R_{k,j}(\lambda_{k},\lambda_{j})$ acting on the tensor product
$V(\lambda_{1})\otimes\cdots\otimes V(\lambda_{L})$. We define them by putting
$A^{ab}_{cd}(\lambda_{j},\lambda_{k})=R^{ab}_{cd}(\lambda_{j}-\lambda_{k})$ in
(2.4). Here the matrix elements $R^{ab}_{cd}(u)$ are given in (2.2). For
instance, setting $u=\lambda_{1}-\lambda_{2}$, we have explicitly
$R_{12}(\lambda_{1},\lambda_{2})=\left(\begin{array}[]{cccc}1&0&0&0\\\
0&b(u)&c(u)&0\\\ 0&c(u)&b(u)&0\\\ 0&0&0&1\end{array}\right)_{[1,2]}\,.$ (2.6)
The $R$-matrices satisfy the Yang-Baxter equations:
$R_{12}(\lambda_{1},\lambda_{2})R_{13}(\lambda_{1},\lambda_{3})R_{23}(\lambda_{2},\lambda_{3})=R_{23}(\lambda_{2},\lambda_{3})R_{13}(\lambda_{1},\lambda_{3})R_{12}(\lambda_{1},\lambda_{2})$
(2.7)
They also satisfy the inversion relations (unitarity conditions):
$R_{jk}(\lambda_{j},\lambda_{k})R_{kj}(\lambda_{k},\lambda_{j})=I^{\otimes
L}\qquad{\rm for}\,\,1\leq j,k\leq L\,.$ (2.8)
Here $I^{\otimes L}$ denotes the $L$th power of tensor product of $I$.
Hereafter we often abbreviate $R_{jk}(\lambda_{j},\lambda_{k})$ simply by
$R_{jk}$.
### 2.2 $L$-operators and the monodromy matrix
Let us introduce parameters $\xi_{1},\xi_{2},\ldots,\xi_{L}$, which we call
the inhomogeneous parameters. In the case of the monodromy matrix, we assume
that parameters $\lambda_{j}$ of the tensor product
$V(\lambda_{1})\otimes\cdots\otimes V(\lambda_{L})$ are given by the
inhomogeneous parameters, i.e. $\lambda_{j}=\xi_{j}$ for $j=1,2,\ldots,L$. Let
us denote by $0$ the suffix for the auxiliary space. We define $L$-operators
acting on the $m$th site for $m=1,2,\ldots,L$, by
$L_{m}(\lambda,\xi_{m})=R_{0m}(\lambda,\xi_{m})\,.$ (2.9)
We define the monodromy matrix acting on the $L$ lattice-sites in one
dimension by
$T_{0,12\cdots
L}(\lambda;\xi_{1},\ldots,\xi_{L})=L_{L}(\lambda,\xi_{L})L_{L-1}(\lambda,\xi_{L-1})\cdots
L_{2}(\lambda,\xi_{2})L_{1}(\lambda,\xi_{1})\,.$ (2.10)
We shall also denote it by $R_{0,12\cdots
L}(\lambda_{0};\xi_{1},\ldots,\xi_{L})$ in §2.4. Hereafter we often suppress
the symbols of inhomogeneous parameters and express the monodromy matrix
$T_{0,12\cdots L}(\lambda;\xi_{1},\ldots,\xi_{L})$ simply as $R_{0,12\cdots
L}(\lambda;\\{\xi_{j}\\})$ or $T_{0}(\lambda)$.
Let us consider two auxiliary spaces with suffices $a$ and $b$. We define
monodromy matrices $T_{a}(\lambda_{a})$ and $T_{b}(\lambda_{b})$ similarly as
(2.10) with $0$ replaced by $a$ and $b$, respectively. It is clear that they
satisfy the following Yang-Baxter equations.
$R_{ab}(\lambda_{a},\lambda_{b})T_{a}(\lambda_{a})T_{b}(\lambda_{b})=T_{b}(\lambda_{b})T_{a}(\lambda_{a})R_{ab}(\lambda_{a},\lambda_{b})\,.$
(2.11)
Let us introduce operator $A_{j}$ acting on the $j$th site by
$A_{j}=\sum_{a,b=1,2}A^{a}_{b}I_{0}\otimes\cdots\otimes I_{j-1}\otimes
e^{a,\,b}_{j}\otimes\cdots\otimes I_{L}$ (2.12)
We express it in terms of the matrix notation as follows:
$A_{j}=\left(\begin{array}[]{cc}A^{1}_{1}&A^{1}_{2}\\\
A^{2}_{1}&A^{2}_{2}\end{array}\right)_{[j]}$ (2.13)
We express the matrix elements of the monodromy matrix by
$T_{0,12\cdots
L}(u;\xi_{1},\ldots,\xi_{L})=\left(\begin{array}[]{cc}A_{12\cdots
L}(u;\xi_{1},\ldots,\xi_{L})&B_{12\cdots L}(u;\xi_{1},\ldots,\xi_{L})\\\
C_{12\cdots L}(u;\xi_{1},\ldots,\xi_{L})&D_{12\cdots
L}(u;\xi_{1},\ldots,\xi_{L})\end{array}\right)_{[0]}$ (2.14)
The transfer matrix, $t(u)$, is given by the trace of the monodromy matrix
with respect to the 0th space:
$\displaystyle t(u;\xi_{1},\ldots,\xi_{L})$ $\displaystyle=$
$\displaystyle{\rm{tr}_{0}}\left(T_{0,12\cdots
L}(u;\xi_{1},\ldots,\xi_{L})\right)$ (2.15) $\displaystyle=$ $\displaystyle
A_{12\cdots L}(u;\xi_{1},\ldots,\xi_{L})+D_{12\cdots
L}(u;\xi_{1},\ldots,\xi_{L})\,.$
Here we note that the transfer matrix $t(u)$ is nothing but the transfer
matrix of the six-vertex model defined on the two-dimensional square lattice
[34].
Hereafter, we shall often denote $B_{12\cdots L}(u;\xi_{1},\ldots,\xi_{L})$ by
$B(u;\\{\xi_{j}\\})$ or $B(u)$, briefly.
### 2.3 Products of $R$-matrices and the symmetric group
Let us consider the symmetric group ${\cal S}_{n}$ of $n$ integers,
$1,2,\ldots,n$. We denote by $\sigma$ an element of ${\cal S}_{n}$. Then
$\sigma$ maps $j$ to $\sigma(j)$ for $j=1,2,\ldots,n$
###### Definition 1.
Let $p$ be a sequence of $n$ integers, $1,2,\ldots,n$, and $\sigma$ an element
of the symmetric group ${\cal S}_{n}$. We define the action of $\sigma$ on $p$
by
$\sigma(p)=(p_{\sigma(1)},\ldots,p_{\sigma(n)})\,.$ (2.16)
Here we remark that $(\sigma_{A}\sigma_{B})\,p=\sigma_{B}(\sigma_{A}p)$ for
$\sigma_{A},\sigma_{B}\in{\cal S}_{n}$. We shall show it in Appendix A.
Let us recall that $R_{jk}$ denote $R_{jk}(\lambda_{j},\lambda_{k})$.
###### Definition 2.
Let $p=(p_{1},p_{2},\ldots,p_{n})$ be a sequence of $n$ integers,
$1,2,\ldots,n$. We define $R_{p_{1},\,p_{2}p_{3}\cdots p_{n}}$ and
$R_{p_{1}p_{2}\cdots p_{n-1},\,p_{n}}$ by
$\displaystyle R_{p_{1},\,p_{2}p_{3}\cdots p_{n}}$ $\displaystyle=$
$\displaystyle R_{p_{1}p_{n}}R_{p_{1}p_{n-1}}\cdots R_{p_{1}p_{2}}\,,$
$\displaystyle R_{p_{1}p_{2}\cdots p_{n-1},\,p_{n}}$ $\displaystyle=$
$\displaystyle R_{p_{1}p_{n}}R_{p_{2}p_{n}}\cdots R_{p_{n-1}p_{n}}\,.$ (2.17)
For $p=(1,2,\ldots,n)$ we have
$R_{1,23\cdots n}=R_{1n}R_{1n-1}\cdots R_{12}\,,\quad R_{12\cdots
n-1,n}=R_{1n}R_{2n}\cdots R_{n-1n}.$ (2.18)
We thus express the monodromy matrix as follows
$T_{0,12\cdots L}(\lambda_{0};\\{\xi_{k}\\})=R_{0,12\cdots
L}(\lambda_{0};\\{\xi_{k}\\})\,.$ (2.19)
Here we have assumed that $\lambda_{k}=\xi_{k}$ for $k=1,2,\ldots,n$.
Let us express by $s_{j}=(j\,\,j+1)$ such a permutation that maps $j$ to $j+1$
and $j+1$ to $j$ and does not change other integers.
###### Definition 3.
Let $p$ be a sequence of $n$ integers, $1,2,\ldots,n$. We define
$R^{s_{j}}_{p}$ by
$R^{s_{j}}_{p}=R_{p_{j},p_{j+1}}(\lambda_{p_{j}},\lambda_{p_{j+1}})\,.$ (2.20)
For the unit element $e$ of ${\cal S}_{n}$, we define $R^{e}_{p}$ by
$R^{e}_{p}=1$. For a given element $\sigma$ of ${\cal S}_{n}$, we define
$R^{\sigma}_{p}$ recursively by the following:
$R_{p}^{\sigma_{A}\sigma_{B}}=R_{\sigma_{A}(p)}^{\sigma_{B}}R_{p}^{\sigma_{A}}\,.$
(2.21)
We remark that every permutation $\sigma$ is expressed as a product of some
$s_{j}=(j\,j+1)$ with $j=1,2,\ldots,n-1$. We thus obtain $R_{p}^{\sigma}$ as a
product of $R_{p}^{s_{j}}$ for some $j$ s. For an illustration, let us
calculate $R^{(123)}_{(1,2,3)}$. Noting $(123)=(1\,2)(2\,3)$, we have
$\displaystyle R^{(12)(23)}_{(1,2,3)}$ $\displaystyle=$ $\displaystyle
R^{(23)}_{(2,1,3)}R^{(12)}_{(1,2,3)}$ (2.22) $\displaystyle=$ $\displaystyle
R_{13}R_{12}=R_{1,23}\,.$
Through the defining relations of the symmetric group ${\cal S}_{n}$ [35], we
can show that definition 2.21 is well defined. The proof is given in
proposition A.1 of Appendix A.
Let us denote by $\sigma_{c}$ such a cyclic permutation that maps $j$ to $j+1$
for $j=1,\ldots,n-1$ and $n$ to 1. We also express it as $\sigma_{c}=(12\cdots
n)$. Noting $(12\cdots n)=(1\,\,2)\cdots(n-1\,\,n)=s_{1}s_{2}\cdots s_{n-1}$,
we can show the following lemma
###### Lemma 4.
Let us denote by $p_{q}$ the sequence $p_{q}=(1,2,\ldots,n)$. For
$\sigma_{c}=(12\cdots n)$ we have
$R^{\sigma_{c}}_{p_{q}}=R_{1,2\cdots n}\,.$ (2.23)
The proof of lemma 2.23 is given in lemma A.6 of Appendix A.
### 2.4 ${\check{R}}$-matrices and permutation operators
Let us consider two-dimensional vector spaces $V_{a}$ and $V_{b}$. We define
permutation operator $\Pi_{ab}$ which maps elements of $V_{a}\otimes V_{b}$ to
those of $V_{b}\otimes V_{a}$ as follows.
$\Pi_{ab}\,v_{a}\otimes v_{b}=v_{b}\otimes v_{a}\,,\quad v_{a}\in
V_{a}\,,\quad v_{b}\in V_{b}\,.$ (2.24)
We define ${\check{R}}_{ab}(u)$ by
${\check{R}}_{ab}(u)=\Pi_{ab}R_{ab}(u)$ (2.25)
The operators ${\check{R}}_{ab}$ satisfy the Yang-Baxter equations
${\check{R}}_{12}(u){\check{R}}_{23}(u+v){\check{R}}_{12}(v)={\check{R}}_{23}(v){\check{R}}_{12}(u+v){\check{R}}_{23}(u)$
(2.26)
The operator ${\check{R}}_{ab}$ gives a linear map from $V_{a}\otimes V_{b}$
to $V_{b}\otimes V_{a}$. If $V_{a}$ and $V_{b}$ are equivalent, then we may
regard ${\check{R}}_{ab}$ as a map from $V_{a}^{\otimes 2}$ to $V_{a}^{\otimes
2}$.
We add 0 to the $n$ integers. For a given element $\sigma$ of the symmetric
group ${\cal S}_{n+1}$, we define $\Pi^{\sigma}$ acting on integers
$0,1,\ldots,n$, as follows. We first express $\sigma$ in terms of
$s_{j}=(j\,\,j+1)$ such as $\sigma=s_{j_{1}}s_{j_{2}}\cdots s_{j_{r}}$, and
then we define $\Pi^{\sigma}=\Pi^{(j_{1}\,j_{1}+1)\cdots(j_{r}\,j_{r}+1)}$ by
$\Pi^{\sigma}=\Pi_{j_{1},j_{1}+1}\Pi_{j_{2},j_{2}+1}\cdots\Pi_{j_{r},j_{r}+1}\,.$
(2.27)
###### Lemma 5.
We have the following relation between $R$-matrices and operators
${\check{R}}_{ab}$:
$R_{0,12\cdots n}=\Pi^{(01\cdots
n)}\,{\check{R}}_{n-1\,n}(\lambda_{0}-\xi_{n-1})\cdots{\check{R}}_{12}(\lambda_{0}-\xi_{2}){\check{R}}_{01}(\lambda_{0}-\xi_{1})$
(2.28)
## 3 The quantum group invariance
We shall show that the monodromy matrix, $T_{0,12\cdots
L}(\lambda;\xi_{1},\ldots,\xi_{L})$, has the symmetry of the affine quantum
group, $U_{q}(\widehat{sl_{2}})$.
### 3.1 Quantum group $U_{q}(sl_{2})$ and the asymmetric $R$-matrices
The quantum algebra $U_{q}(sl_{2})$ is an associative algebra over ${\bf C}$
generated by $X^{\pm},K^{\pm}$ with the following relations: [31]
$\displaystyle KK^{-1}$ $\displaystyle=$ $\displaystyle KK^{-1}=1\,,\quad
KX^{\pm}K^{-1}=q^{\pm 2}X^{\pm}\,,\quad\,,$ $\displaystyle{[}X^{+},X^{-}{]}$
$\displaystyle=$ $\displaystyle{\frac{K-K^{-1}}{q-q^{-1}}}\,.$ (3.1)
The algebra $U_{q}(sl_{2})$ is also a Hopf algebra over ${\bf C}$ with
comultiplication
$\displaystyle\Delta(X^{+})$ $\displaystyle=$ $\displaystyle X^{+}\otimes
1+K\otimes X^{+}\,,\quad\Delta(X^{-})=X^{-}\otimes K^{-1}+1\otimes X^{-}\,,$
$\displaystyle\Delta(K)$ $\displaystyle=$ $\displaystyle K\otimes K\,,$ (3.2)
and antipode: $S(K)=K^{-1}\,,S(X^{+})=-K^{-1}X^{+}\,,S(X^{-})=-X^{-}K$, and
coproduct: $\epsilon(X^{\pm})=0$ and $\epsilon(K)=1$.
In association with the quantum group, we define the $q$-integer of an integer
$n$ by $[n]_{q}=(q^{n}-q^{-n})/(q-q^{-1})$.
The universal $R$-matrix, ${\cal R}$, of $U_{q}(sl_{2})$ satisfies the
following relations:
${\cal R}\Delta(x)=\tau\circ\Delta(x){\cal R}\quad{\rm for\,all}\quad x\in
U_{q}(sl_{2})\,.$ (3.3)
Here $\tau$ denotes a permutation such that $\tau\,a\otimes b=b\otimes a$ for
$a,b\in U_{q}(sl_{2})$.
We now introduce some notation of a Hopf algebra. Let $x(1),\ldots,x(n)$ be
elements of Hopf algebra ${\cal A}$. For a given permutation $\sigma$ of
${\cal S}_{n}$, we define its action on the tensor product
$x(1)\otimes\cdots\otimes x(n)$ as follows:
$\sigma\circ\left(x(1)\otimes\cdots\otimes
x(n)\right)=x(\sigma^{-1}1)\otimes\cdots x(\sigma^{-1}n)\,.$ (3.4)
We note that ${\cal A}$ has coassociativity: $\left(\Delta\otimes
id\right)\Delta(x)=\left(id\otimes\Delta\right)\Delta(x)$ for any element $x$
of ${\cal A}$. We therefore denote it by $\Delta^{(2)}(x)$. We define
$\Delta^{(n)}(x)$ recursively by
$\Delta^{(n)}(x)=\left(\Delta^{(n-1)}\otimes id\right)\Delta(x)\quad{\rm
for}\,\,x\in{\cal A}.$ (3.5)
Let us now introduce the following asymmetric $R$-matrices:
$R^{\pm}(u)=\left(\begin{array}[]{cccc}1&0&0&0\\\ 0&b(u)&c^{\mp}(u)&0\\\
0&c^{\pm}(u)&b(u)&0\\\ 0&0&0&1\end{array}\right)\,,$ (3.6)
where $c^{\pm}(u)$ are defined by
$c^{\pm}(u)={\frac{e^{\pm u}\sinh(\eta)}{\sinh(u+\eta)}}\,.$ (3.7)
In the spin-1/2 representation of $U_{q}(sl_{2})$, we have the following
relations:
$R_{12}^{+}(u)\Delta(x)=\tau\circ\Delta(x)R^{+}_{12}(u)\quad{\rm for\,\,}\quad
x=X^{\pm},K\,.$ (3.8)
Here we remark that spectral parameter $u$ is arbitrary and independent of
$X^{\pm}$ or $K$. Similarly as in the symmetric case, we define the monodromy
matrix $R^{+}_{0,1\cdots n}$ by $R^{+}_{0,1\cdots n}=R^{+}_{0,n}\cdots
R^{+}_{0,1}$, and ${\check{R}}^{+}$ by
${\check{R}}^{+}_{12}(u)=\Pi_{12}R^{+}(u)$.
###### Lemma 6.
The monodromy matrix expressed in terms of ${\check{R}}$’s commutes with the
action of the quantum group $U_{q}(sl_{2})$:
${[}{\check{R}}^{+}_{L-1,L}(\lambda-\xi_{L})\cdots{\check{R}}^{+}_{1,2}(\lambda-\xi_{2}){\check{R}}^{+}_{0,1}(\lambda-\xi_{1}),\quad\Delta^{(L)}(x){]}=0\,,\quad{\rm
for}\,{\rm all}\,x\in U_{q}(sl_{2})\,.$ (3.9)
Here parameters $\lambda$, $\xi_{1},\ldots,\xi_{L}$ are independent of element
$x$ of $U_{q}(sl_{2})$.
We shall show lemma 6 and eqs. (3.8) through the Temperley-Lieb algebra in
§3.2.
Lemma 6 leads to the following symmetry relations of the monodromy matrix
$R^{+}_{0,12\cdots L}$ with respect to the quantum group $U_{q}(sl_{2})$:
###### Proposition 7.
Let $\sigma_{c}$ be a cyclic permutation: $\sigma_{c}=(01\cdots L)$. Then we
have
$R^{+}_{0,12\cdots
L}(\lambda;\xi_{1},\ldots,\xi_{L})\,\Delta^{(L)}(x)=\sigma_{c}\circ\Delta^{(L)}(x)\,R^{+}_{0,12\cdots
L}(\lambda;\xi_{1},\ldots,\xi_{L})\quad{\rm for\,\,all}\,\,x\in U_{q}(sl_{2})$
(3.10)
Here parameters $\lambda$, $\xi_{1},\ldots,\xi_{L}$ are independent of element
$x$ of $U_{q}(sl_{2})$.
###### Proof.
Making use of lemma 6 we show (3.10) from (2.28) and the following relation:
$\sigma_{c}\circ\Delta^{(L)}(x)=\Pi^{\sigma_{c}}\,\Delta^{(L)}(x)\,\left(\Pi^{\sigma_{c}}\right)^{-1}\,.$
(3.11)
∎
### 3.2 Derivation in terms of the Temperley-Lieb algebra
Let us define $U_{j}^{\pm}$ for $j=0,1,\ldots,L-1$, by
$U^{\pm}_{j}=\left(\begin{array}[]{cccc}0&0&0&0\\\ 0&q^{\mp}&-1&0\\\
0&-1&q^{\pm}&0\\\ 0&0&0&0\end{array}\right)_{[j,j+1]}\,.$ (3.12)
They satisfy the defining relations of the Temperley-Lieb algebra: [34]
$\displaystyle U^{\pm}_{j}U^{\pm}_{j+1}U^{\pm}_{j}$ $\displaystyle=$
$\displaystyle U^{\pm}_{j},$ $\displaystyle
U^{\pm}_{j+1}U^{\pm}_{j}U^{\pm}_{j+1}$ $\displaystyle=$ $\displaystyle
U^{\pm}_{j},\quad{\rm for}\,j=0,1,\ldots,L-2,$
$\displaystyle\left(U^{\pm}_{j}\right)^{2}$ $\displaystyle=$
$\displaystyle(q+q^{-1})\,U^{\pm}_{j}\quad{\rm for}\,j=0,1,\ldots,L-1,$
$\displaystyle U^{\pm}_{j}U^{\pm}_{k}$ $\displaystyle=$ $\displaystyle
U^{\pm}_{k}U^{\pm}_{j}\quad{\rm for}\,\,|j-k|>1\,.$ (3.13)
Let us now show commutation relations (3.10), making use of the Temperley-Lieb
algebra. The operator ${\check{R}}_{j,j+1}^{+}(u)$ is decomposed in terms of
the generators of the Temperley-Lieb algebra as follows [36].
${\check{R}}_{j,j+1}^{+}(u)=I-b(u)U_{j}^{+}\,.$ (3.14)
###### Lemma 8.
The monodromy matrix of the six-vertex model is expressed in terms of the
generators of the Temperley-Lieb algebra as follows.
$\displaystyle{\check{R}}^{+}_{L-1,L}(\lambda-\xi_{L})\cdots{\check{R}}^{+}_{1,2}(\lambda-\xi_{2}){\check{R}}^{+}_{0,1}(\lambda-\xi_{1})$
$\displaystyle=\sum_{k=0}^{L}(-1)^{k}\sum_{0\leq
i_{1}<\cdots<i_{k}<L}\,\left(\prod_{j=1}^{k}b(\lambda-\xi_{i_{j}})\right)U^{+}_{i_{k}}\cdots
U^{+}_{i_{2}}U^{+}_{i_{1}}\,.$ (3.15)
###### Lemma 9.
The generators $U^{+}_{j}$ commute with the generators of $U_{q}(sl_{2})$. For
$x=X^{\pm},K$ and for $j=0,1,\ldots,L-1$, we have in the tensor-product
representation
${\big{[}}U^{+}_{j},\Delta^{(L)}(x){\big{]}}=0\,.$ (3.16)
Proof of proposition 7. From lemmas 8 and 3.16 we have lemma 6, which is
equivalent to proposition 7.
We now show that in the limit of taking $u$ to $-\infty$, ${\hat{R}}^{+}(u)$
is equivalent to the spin-1/2 matrix representation of the universal
$R$-matrix ${\cal R}$ of $U_{q}(sl(2))$. An explicit expression of ${\cal R}$
is given by
${\cal R}=q^{-{\frac{1}{2}}H\otimes
H}\exp_{q}\left(-(q-q^{-1})K^{-1}X^{+}\otimes X^{-}K\right)$ (3.17)
where $\exp_{q}x$ denotes the following series:
$\exp_{q}x=\sum_{n=0}^{\infty}{\frac{q^{-n(n-1)/2}}{[n]_{q}!}}x^{n}\,.$ (3.18)
Here $q$ is generic. We recall that $[n]_{q}$ denotes the $q$-integer of an
integer $n$: $[n]_{q}=(q^{n}-q^{-n})/(q-q^{-1})$. Putting $X^{+}=e^{1,2}$,
$X^{-}=e^{2,1}$ and $K={\rm diag}(q,q^{-1})$ in the series (3.17), we have the
following matrix representation.
$R_{{\frac{1}{2}},{\frac{1}{2}}}=\left(\begin{array}[]{cccc}q^{-1/2}&0&0&0\\\
0&q^{1/2}&-q^{1/2}(q-q^{-1})&0\\\ 0&0&q^{1/2}&0\\\
0&0&0&q^{-1/2}\end{array}\right)\,.$ (3.19)
Thus, for arbitrary $u$, we have the following:
$R^{+}_{12}(u)=c^{+}(u)\Pi_{12}+q^{-1/2}b(u)R_{{\frac{1}{2}},{\frac{1}{2}}}$
(3.20)
Therefore, we have ${R}^{+}(-\infty)=q^{1/2}R_{{\frac{1}{2}},{\frac{1}{2}}}$.
We remark that some relations equivalent to (3.16) have been shown in
association with the $sl(2)$ loop algebra symmetry of the XXZ spin chain at
roots of unity [37].
### 3.3 Gauge transformations
Let us introduce operators ${\Phi}_{j}$ with arbitrary parameters $\phi_{j}$
for $j=0,1,\ldots,L$ as follows:
${\Phi}_{j}=\left(\begin{array}[]{cc}1&0\\\
0&e^{\phi_{j}}\end{array}\right)_{[j]}=I^{\otimes(j)}\otimes\left(\begin{array}[]{cc}1&0\\\
0&e^{\phi_{j}}\end{array}\right)\otimes I^{\otimes(L-j)}.$ (3.21)
In terms of ${\chi}_{jk}={\Phi}_{j}{\Phi}_{k}$, we define a similarity
transformation on the $R$-matrix by
$R_{jk}^{\chi}={\chi}_{jk}R_{jk}{\chi}_{jk}^{-1}$ (3.22)
Explicitly, the following two matrix elements are transformed.
$\left(R_{jk}^{\chi}\right)^{21}_{12}=c(\lambda_{j},\lambda_{k})e^{\phi_{j}-\phi_{k}}\,,\quad\left(R_{jk}^{\chi}\right)^{12}_{21}=c(\lambda_{j},\lambda_{k})e^{-\phi_{j}+\phi_{k}}\,.$
(3.23)
We now put $\phi_{j}=\lambda_{j}$ in eq. (3.21) for $j=0,1,\ldots,L$. For
$j,k=0,1,\ldots,L$, we have
$R_{jk}^{\pm}(\lambda_{j},\lambda_{k})=\left(\chi_{jk}\right)^{\pm
1}\,R_{jk}(\lambda_{j},\lambda_{k})\,\left(\chi_{jk}\right)^{\mp 1}\,.$ (3.24)
Thus, the asymmetric $R$-matrices $R_{12}^{\pm}(\lambda_{1},\lambda_{2})$ are
derived from the symmetric one through the gauge transformation $\chi_{jk}$.
For the monodromy matrix, in terms of the inhomogeneous parameters,
$\xi_{1},\ldots,\xi_{L}$, we put $\lambda_{j}=\xi_{j}$ for $j=1,\ldots,L$. We
define ${\chi}_{012\cdots L}$ by ${\chi}_{012\cdots
L}=\Phi_{0}\Phi_{1}\cdots\Phi_{L}$. Then, the asymmetric monodromy matrices
are transformed into the symmetric one as follows.
$R_{0,12\cdots L}^{\pm}=\left(\chi_{012\cdots L}\right)^{\pm 1}\,R_{0,12\cdots
L}\left(\chi_{012\cdots L}\right)^{\mp 1}\,.$ (3.25)
We note that the asymmetric $R$-matrices ${\check{R}}^{\pm}_{j,j+1}(u)$ are
derived from the symmetric $R$-matrix through the gauge transformations, and
they are related to the Jones polynomial. [21]
### 3.4 Affine quantum group symmetry
The affine quantum algebra $U_{q}(\widehat{sl_{2}})$ is an associative algebra
over ${\bf C}$ generated by $X_{i}^{\pm},K_{i}^{\pm}$ for $i=0,1$ with the
following relations:
$\displaystyle K_{i}K_{i}^{-1}$ $\displaystyle=$ $\displaystyle
K^{-1}_{i}K_{i}=1\,,\quad K_{i}X_{i}^{\pm}K_{i}^{-1}=q^{\pm
2}X_{i}^{\pm}\,,\quad K_{i}X_{j}^{\pm}K_{i}^{-1}=q^{\mp
2}X_{j}^{\pm}\quad(i\neq j)\,,$ $\displaystyle{[}X_{i}^{+},X_{j}^{-}{]}$
$\displaystyle=$
$\displaystyle\delta_{i,j}\,{\frac{K_{i}-K_{i}^{-1}}{q-q^{-1}}}\,,$
$\displaystyle(X_{i}^{\pm})^{3}X_{j}^{\pm}$ $\displaystyle-$
$\displaystyle[3]_{q}\,(X_{i}^{\pm})^{2}X_{j}^{\pm}X_{i}^{\pm}+[3]_{q}\,X_{i}^{\pm}X_{j}^{\pm}(X_{i}^{\pm})^{2}-X_{j}^{\pm}(X_{i}^{\pm})^{3}=0\quad(i\neq
j)\,.$ (3.26)
The algebra $U_{q}(\widehat{sl_{2}})$ is also a Hopf algebra over ${\bf C}$
with comultiplication
$\displaystyle\Delta(X_{i}^{+})$ $\displaystyle=$ $\displaystyle
X_{i}^{+}\otimes 1+K_{i}\otimes
X_{i}^{+}\,,\quad\Delta(X_{i}^{-})=X_{i}^{-}\otimes K_{i}^{-1}+1\otimes
X_{i}^{-}\,,$ $\displaystyle\Delta(K_{i})$ $\displaystyle=$ $\displaystyle
K_{i}\otimes K_{i}\,,$ (3.27)
and antipode:
$S(K_{i})=K_{i}^{-1}\,,S(X_{i})=-K_{i}^{-1}X_{i}^{+}\,,S(X_{i}^{-})=-X_{i}^{-}K_{i}$.
We now introduce evaluation representations for $U_{q}(\widehat{sl_{2}})$
[22]. For a given complex number $a$ there is a homomorphism of algebras
$\varphi_{a}$: $U_{q}(\widehat{sl_{2}})\rightarrow U_{q}({sl_{2}})$ such that
$\displaystyle\varphi_{a}(X_{0}^{\pm})$ $\displaystyle=$
$\displaystyle\exp(\pm a)\,X^{\mp}\,,\quad\varphi_{a}(K_{0})=K^{-1}\,,$
$\displaystyle\varphi_{a}(X_{1}^{\pm})$ $\displaystyle=$ $\displaystyle
X^{\pm}\,,\quad\varphi_{a}(K_{1})=K\,.$ (3.28)
We denote by $(\pi,V)$ a representation of an algebra ${\cal A}$ such that
$\pi(x)$ give linear maps on vector space $V$ for $x\in{\cal A}$. For a given
finite-dimensional representation $(\pi_{V},V)$ of $U_{q}(sl_{2})$ we have a
finite-dimensional representation $(\pi_{V(a)},V(a))$ of
$U_{q}(\widehat{sl_{2}})$ through homomorphism $\varphi_{a}$, i.e.
$\pi_{V(a)}(x)=\pi_{V}(\varphi_{a}(x))$ for $x\in U_{q}(\widehat{sl_{2}})$. We
call $(\pi_{V(a)},V(a))$ or $V(a)$ the evaluation representation of $V$ and
nonzero parameter $a$ the evaluation parameter of $V(a)$. If $V$ is
$(2s+1)$-dimensional, then we also denote it by $V^{(2s)}(a)$. Hereafter we
express $2s$ by an integer $\ell$.
Similarly as (3.14), we have the following decomposition:
${\check{R}}_{j,j+1}^{-}(u)=I-b(u)U_{j}^{-}\,,\quad{\rm
for}\,\,j=0,1,\ldots,L-1.$ (3.29)
###### Lemma 10.
Generators $U^{-}_{j}$ commute with $X_{0}^{\pm}$ and $K_{0}$ of
$U_{q}(\widehat{sl_{2}})$ in the tensor-product representation
$V^{(1)}(a_{0})\otimes\cdots\otimes V^{(1)}(a_{L})$ with
$a_{0}=a_{1}=\cdots=a_{L}=a$. For $j=0,1,\ldots,L-1$, we have
${\big{[}}U^{-}_{j},\,\,\varphi_{a}^{\otimes(L+1)}\left(\Delta^{(L)}(x)\right){\big{]}}=0\,\quad\left(x=X_{0}^{\pm},\,K_{0}.\right)$
(3.30)
###### Proof.
Let us denote by $U^{-}$ the $4\times 4$ matrix given by the decomposition:
${\check{R}}(u)=I-b(u)U^{-}$. Through an explicit calculation we show that
$U^{-}$ and $\varphi_{a}\otimes\varphi_{b}\left(\Delta(X_{0}^{\pm})\right)$
commute if $a=b$:
${[}U^{-},\,\varphi_{a}\otimes\varphi_{a}\left(\Delta(X_{0}^{\pm})\right){]}=0\,.$
(3.31)
We derive (3.30) through (3.31). ∎
In the spin-1/2 representation of $U_{q}(sl_{2})$, we thus have the following
relations:
$R_{12}^{-}(u)\,\,\varphi_{a}^{\otimes 2}\Delta(x)=\varphi_{a}^{\otimes
2}\left(\tau\circ\Delta(x)\right)\,\,R^{-}_{12}(u)\quad{\rm for\,\,}\quad
x=X_{0}^{\pm},K_{0}\,.$ (3.32)
Here we note that $u$ is arbitrary and independent of $X_{0}^{\pm},K_{0}$.
Similarly as lemmas 8 and 3.16, we have from lemma 3.30 the quantum-group
symmetry of the monodromy matrix $R^{-}_{0,12\cdots L}$:
###### Proposition 11.
Let $\sigma_{c}$ be a cyclic permutation: $\sigma_{c}=(01\cdots L)$. In the
evaluation representation (3.28) we have, for $x=X_{0}^{\pm},K_{0}$, the
following:
$R^{-}_{0,12\cdots
L}(\lambda;\xi_{1},\ldots,\xi_{L})\,\varphi_{a}^{\otimes(L+1)}\left(\Delta^{(L)}(x)\right)=\varphi_{a}^{\otimes(L+1)}\left(\sigma_{c}\circ\Delta^{(L)}(x)\right)\,R^{-}_{0,12\cdots
L}(\lambda;\xi_{1},\ldots,\xi_{L})\,.$ (3.33)
Here parameters $\lambda$, $\xi_{1},\ldots,\xi_{L}$ are arbitrary and
independent of $x=X_{0}^{\pm},K_{0}$.
Let us now make a summary of the symmetry relations of $R^{+}_{12}$. Here we
recall that $R^{+}_{12}(\lambda_{1},\lambda_{2})\in End(V(\lambda_{1})\otimes
V(\lambda_{2}))$. For simplicity, we put $a=0$ in (3.33). Combining (3.8) and
(3.32) We have the following relations:
$\displaystyle R^{+}_{12}(\lambda_{1},\lambda_{2})\,\,\varphi_{0}^{\otimes
2}\left(\Delta(X_{1}^{\pm})\right)$ $\displaystyle=$
$\displaystyle\varphi_{0}^{\otimes
2}\left(\tau\circ\Delta(X_{1}^{\pm})\right)\,\,R^{+}_{12}(\lambda_{1},\lambda_{2})\,,$
$\displaystyle
R^{+}_{12}(\lambda_{1},\lambda_{2})\,(\chi_{12})^{2}\varphi_{0}^{\otimes
2}\left(\Delta(X_{0}^{\pm})\right)(\chi_{12})^{-2}$ $\displaystyle=$
$\displaystyle(\chi_{12})^{2}\varphi_{0}^{\otimes
2}\left(\tau\circ\Delta(X_{0}^{\pm})\right)(\chi_{12})^{-2}\,R^{+}_{12}(\lambda_{1},\lambda_{2})\,,$
$\displaystyle R^{+}_{12}(\lambda_{1},\lambda_{2})\,\varphi_{0}^{\otimes
2}\left(\Delta(K_{i}^{\pm})\right)$ $\displaystyle=$
$\displaystyle\varphi_{0}^{\otimes
2}\left(\tau\circ\Delta(K_{i}^{\pm})\right)\,R^{+}_{12}(\lambda_{1},\lambda_{2})\,{\rm
for}\,i=0,1\,.$
Let us now consider $\varphi_{a_{1}}\otimes\varphi_{a_{2}}$ with
$a_{j}=2\lambda_{j}$ for $j=1,2$. We have
$\varphi_{2\lambda_{1}}\otimes\varphi_{2\lambda_{2}}\left(\Delta(X_{0}^{\pm})\right)=(\chi_{12})^{2}\,\,\varphi_{0}\otimes\varphi_{0}\left(\Delta(X_{0}^{\pm})\right)\,\,(\chi_{12})^{-2}$
(3.35)
Thus, relations (LABEL:eq:affineCR) are now expressed as follows.
$R^{+}_{12}(\lambda_{1},\lambda_{2})\,\varphi_{2\lambda_{1}}\otimes\varphi_{2\lambda_{2}}\left(\Delta(x)\right)=\varphi_{2\lambda_{1}}\otimes\varphi_{2\lambda_{2}}\left(\tau\circ\Delta(x)\right)\,R^{+}_{12}(\lambda_{1},\lambda_{2})\,,\,\,{\rm
for}\,\,x=X_{0}^{\pm},X_{1}^{\pm},K_{0},K_{1}\,.$ (3.36)
In (3.36) all the parameters are now associated with the evaluation parameters
of the tensor product $V(2\lambda_{1})\otimes V(2\lambda_{2})$. Therefore, we
conclude that the asymmetric $R$-matrix $R^{+}_{12}(\lambda_{1},\lambda_{2})$
satisfies the affine quantum-group symmetry.
The fundamental commutation relations (3.10) and (3.33) are summarized as
follows.
###### Proposition 12.
Let $\sigma_{c}$ be a cyclic permutation: $\sigma_{c}=(01\cdots L)$. The
asymmetric $R$-matrix $R^{+}$ satisfies the commutation relations for the
affine-quantum group:
$R^{+}_{0,12\cdots L}(\lambda_{0})\left(\Delta^{(n)}(x)\right)_{01\cdots
L}=\left(\sigma_{c}\circ\Delta^{(L)}(x)\right)_{01\cdots L}R^{+}_{0,12\cdots
L}(\lambda_{0})\quad{\rm for\,\,all}\,\,x\in U_{q}({\widehat{sl_{2}}})\,.$
(3.37)
Here the symbol $(x)_{01\cdots n}$ denotes the matrix representation of $x$ in
the tensor product of evaluation representations, $V({2\lambda_{0}})\otimes
V({2\xi_{1}})\otimes\cdots\otimes V({2\xi_{L}})$.
### 3.5 Symmetry relations of $U_{q}(\widehat{sl_{2}})$ for all permutations
Let us generalize relations (3.37). Making an extensive use of relations
(3.36), we can show commutation relations for $R_{p}^{\sigma}$ for all
permutations $\sigma$. In fact, we can prove (3.37) also by the method for
showing proposition A.2. In Appendix A we shall show in proposition A.2 how we
generalize the symmetry relation of $R_{12}$ such as (3.36) into those of
$R_{p}^{\sigma}$ for any permutation $\sigma$.
We now formulate the symmetry relations in terms of the symmetric
$R$-matrices. Let us denote by ${\bar{\chi}}$ the inverse of the gauge
transformation $\chi$. We express by $\left(\Delta^{(n)}(x)\right)_{01\cdots
n}^{\bar{\chi}}$ the following:
$\left(\Delta^{(n)}(x)\right)_{01\cdots n}^{\bar{\chi}}=\left(\chi_{01\cdots
n}\right)^{-1}\,\varphi_{2\lambda_{0}}\otimes\varphi_{2\xi_{1}}\otimes\cdots\otimes\varphi_{2\xi_{n}}\left(\Delta^{(n)}(x)\right)\,\chi_{01\cdots
n}$ (3.38)
###### Proposition 13.
Let $p_{q}$ be an increasing sequence of $n+1$ integers:
$p_{q}=(0,1,2,\ldots,n)$, and $\sigma$ a permutation on $n+1$ integers,
$0,1,\ldots,n$. With the symmetric $R$-matrices, we have
$R^{\sigma}_{p_{q}}(\lambda_{0})\,\left(\Delta^{(n)}(x)\right)_{01\cdots
n}^{\bar{\chi}}=\left(\sigma\circ\Delta^{(n)}(x)\right)^{\bar{\chi}}_{01\cdots
n}\,R^{\sigma}_{p_{q}}(\lambda_{0})\quad{\rm for\,\,all}\,\,x\in
U_{q}({\widehat{sl_{2}}})\,.$ (3.39)
## 4 Projection operators and the fusion procedure
### 4.1 Projection operators
Let us recall that ${\check{R}}_{12}^{+}(u)$ has been defined by
${\check{R}}_{12}^{+}(u)=\Pi_{12}R_{12}^{+}(u)\,.$
We define operator $P_{12}^{+}$ by
$P_{12}^{+}={\check{R}}_{12}^{+}(\eta)$ (4.1)
Explicitly we have
$P_{12}^{+}=\left(\begin{array}[]{cccc}1&0&0&0\\\
0&{\frac{q}{[2]}}&{\frac{1}{[2]}}&0\\\
0&{\frac{1}{[2]}}&{\frac{q^{-1}}{[2]}}&0\\\
0&0&0&1\end{array}\right)_{[12]}\,.$ (4.2)
By making use of the matrix representation (4.2) it is easy to show that
operator $P_{12}^{+}$ is idempotent:
$\left(P_{12}^{+}\right)^{2}=P_{12}^{+}$ (4.3)
Thus, we may consider that the operator $P_{12}^{+}$ is a projection operator.
In fact, $P_{12}^{+}$ is nothing but the $q$-analogue of the Young operator
which projects out the spin-1 representation of $U_{q}(sl_{2})$ from of the
tensor product of two spin-1/2 representations. It should be noticed that in
the case of symmetric $R$-matrix, $R(\eta)^{2}$ is not equal to $R(\eta)$.
We now introduce projection operators for the spin-$s$ irreducible
representations of $U_{q}(sl_{2})$. Hereafter we set $\ell=2s$. We define
projection operator $P_{12\cdots\ell}^{(\ell)}$ acting on the tensor product
$V_{12\cdots\ell}$ of spin-1/2 representations $V$, i.e.
$V_{12\cdots\ell}=V^{\otimes\ell}$, by the following recursive relations:
$P^{(\ell)}_{12\cdots\ell}=P^{(\ell-1)}_{12\cdots\ell-1}{\check{R}}_{\ell-1,\ell}^{+}((\ell-1)\eta)P^{(\ell-1)}_{12\cdots\ell-1}$
(4.4)
Here $P_{12}^{(1)}=P_{12}^{+}$.
Making use of the Yang-Baxter equations (2.26) and through induction on $\ell$
we can show
$\left(P^{(\ell)}_{12\cdots\ell}\right)^{2}=P^{(\ell)}_{12\cdots\ell}$ (4.5)
Thus, operator $P^{(\ell)}_{12\cdots\ell}$ gives a projection operator.
Similarly, we define projection operators acting on
$V_{j\,j+1\cdots\,j+\ell-1}$ recursively by
$P^{(\ell)}_{j\,j+1\cdots j+\ell-1}=P^{(\ell-1)}_{j\,j+1\cdots
j+\ell-2}{\check{R}}_{j+\ell-2,\,j+\ell-1}^{+}((\ell-1)\eta)P^{(\ell-1)}_{j\,j+1\cdots
j+\ell-2}$ (4.6)
Hereafter we shall abbreviate $P^{(\ell)}_{j\,j+1\cdots j+\ell-1}$ by
$P^{(\ell)}_{j}$ .
From idempotency (4.5) and by the Yang-Baxter equations we can show the
following:
###### Lemma 14.
Suppose that inhomogeneous parameters
$\xi_{j},\xi_{j+1},\ldots,\xi_{j+\ell-1}$ are given by
$\xi_{j+i-1}=z-(i-1)\eta$ for $i=1,2,\ldots,\ell$ with a constant $z$. Then,
the monodromy matrix $R^{+}_{0,12\cdots L}$ satisfies the following property:
$P_{j}^{(\ell)}\,R^{+}_{0,12\cdots L}=P_{j}^{(\ell)}\,R^{+}_{0,12\cdots
L}\,P_{j}^{(\ell)}\,.$ (4.7)
Making use of (2.28) we can express projection operators in terms of
$R$-matrices.
$P^{(\ell)}_{1}=\left(\prod_{j=1}^{\ell}\Pi^{(j\,\,\ell-j+1)}\right)\,\,R^{+}_{\ell-1,\,\ell}\cdots
R^{+}_{2,\,3\cdots\ell}R^{+}_{1,\,2\cdots\ell}$ (4.8)
### 4.2 Fusion of monodromy matrices
#### 4.2.1 The case of tensor product of spin-$s$ representations
We first consider the case of tensor product of spin-$s$ representations. We
set $L=N_{s}\ell$. We introduce a set of parameters, $\xi_{1}^{(\ell)}$,
$\xi_{2}^{(\ell)}$, …, $\xi_{L}^{(\ell)}$, as follows:
$\xi_{(k-1)\ell+j}^{(\ell)}={\zeta}_{k}-(j-1)\eta+(\ell-1)\eta/2\qquad\mbox{for}\quad
j=1,\ldots\ell,\,\,\mbox{and}\,\,k=1,\ldots,N_{s}.$ (4.9)
Let us set inhomogeneous parameters $\xi_{1},\xi_{2},\ldots,\xi_{L}$, by
$\xi_{j}=\xi_{j}^{(\ell)}$ for $j=1,2,\ldots,L$. We define the monodromy
matrix $T^{(\ell+)}_{0}(u;\zeta_{1},\ldots,\zeta_{N_{s}})$ acting on the
tensor product of spin-$s$ representations, $V^{(2s)}(\zeta_{1})\otimes
V^{(2s)}(\zeta_{N_{s}})$ by
$T^{(\ell+)}_{0}(\lambda_{0};\zeta_{1},\ldots,\zeta_{N_{s}})=\prod_{k=0}^{N_{s}-1}P^{(\ell)}_{\ell
k+1}\,\cdot\,R^{+}_{0,12\cdots
L}(\lambda_{0};\xi_{1}^{(\ell)},\ldots,\xi_{L}^{(\ell)})\,\cdot\,\prod_{k=0}^{N_{s}-1}P^{(\ell)}_{\ell
k+1}$ (4.10)
Making use of properties of projection operators (4.5) and (4.7), we can show
the Yang-Baxter equation [fusionXXX, 19]
$\displaystyle
R^{+}_{ab}(\lambda_{a}-\lambda_{b})T^{(\ell+)}_{a}(\lambda_{a};\zeta_{1},\ldots,\zeta_{N_{s}})T^{(\ell+)}_{b}(\lambda_{b};\zeta_{1},\ldots,\zeta_{N_{s}})$
(4.11) $\displaystyle=$ $\displaystyle
T^{(\ell+)}_{b}(\lambda_{b};\zeta_{1},\ldots,\zeta_{N_{s}})T^{(\ell+)}_{a}(\lambda_{a};\zeta_{1},\ldots,\zeta_{N_{s}})R^{+}_{ab}(\lambda_{a}-\lambda_{b})$
Through the inverse of the gauge transformation, we derive the symmetric
spin-$s$ monodromy matrix as follows:
$T^{(\ell)}_{0}(\lambda_{0};\zeta_{1},\ldots,\zeta_{N_{s}})=\prod_{k=0}^{N_{s}-1}\left(P^{(\ell)}_{\ell
k+1}\right)^{\bar{\chi}}\,\cdot\,R_{0,12\cdots
L}(\lambda_{0};\xi_{1}^{(\ell)},\ldots,\xi_{L}^{(\ell)})\,\cdot\,\prod_{k=0}^{N_{s}-1}\left(P^{(\ell)}_{\ell
k+1}\right)^{\bar{\chi}}$ (4.12)
where we have defined the transformed projectors by
$\left(P^{(\ell)}_{\ell k+1}\right)^{\bar{\chi}}=\left(\chi_{01\cdots
L}\right)^{-1}\,P^{(\ell)}_{\ell k+1}\,\left(\chi_{01\cdots L}\right)$ (4.13)
#### 4.2.2 The case of mixed spins
Let us consider the tensor product of representations with different spins,
$s_{1},s_{2},\ldots,s_{r}$. Here we introduce $\ell_{j}$ by $\ell_{j}=2s_{j}$
for $j=1,2,\ldots,r$, and we assume that
$\ell_{1}+\ell_{2}+\cdots+\ell_{r}=L$. @ Let us introduce a set of parameters,
$\xi_{1}^{({\mbox{\boldmath$\ell$}})}$,
$\xi_{2}^{({\mbox{\boldmath$\ell$}})}$, …,
$\xi_{L}^{({\mbox{\boldmath$\ell$}})}$, as follows:
$\xi_{\ell_{1}+\cdots+\ell_{k-1}+j}^{({\mbox{\boldmath$\ell$}})}=\zeta_{k}-(j-1)\eta+(\ell_{k}-1)\eta/2\qquad\mbox{for}\quad
j=1,\ldots\ell_{k},\,\,\mbox{and}\,\,k=1,\ldots,r.$ (4.14)
We define the asymmetric monodromy matrix for the mixed spin case
$T^{({\mbox{\boldmath$\ell$}}+)}_{0}(u;\zeta_{1},\ldots,\zeta_{r})$ acting on
the tensor product representation
$V_{\ell_{1}}({\zeta_{1}})\otimes\cdots\otimes V_{\ell_{r}}({\zeta_{r}})$ by
$T^{({\mbox{\boldmath$\ell$}}+)}_{0}(\lambda_{0};\zeta_{1},\ldots,\zeta_{r})=\prod_{k=1}^{r}P^{(\ell_{k})}_{\ell_{1}+\cdots+\ell_{k-1}+1}\,\cdot\,R_{0,12\cdots
L}(\lambda_{0};\xi_{1}^{(\mbox{\boldmath$\ell$})},\ldots,\xi_{L}^{(\mbox{\boldmath$\ell$})})\,\cdot\,\prod_{k=1}^{r}P^{(\ell_{k})}_{\ell_{1}+\cdots+\ell_{k-1}+1}\,.$
(4.15)
It is easy to show that they satisfy the Yang-Baxter equations.
$\displaystyle
R^{+}_{ab}(\lambda_{a}-\lambda_{b})T^{({\mbox{\boldmath$\ell$}}+)}_{a}(\lambda_{a};\zeta_{1},\ldots,\zeta_{N_{s}})T^{({\mbox{\boldmath$\ell$}}+)}_{b}(\lambda_{b};\zeta_{1},\ldots,\zeta_{N_{s}})$
(4.16) $\displaystyle=$ $\displaystyle
T^{({\mbox{\boldmath$\ell$}}+)}_{b}(\lambda_{b};\zeta_{1},\ldots,\zeta_{N_{s}})T^{({\mbox{\boldmath$\ell$}}+)}_{a}(\lambda_{a};\zeta_{1},\ldots,\zeta_{N_{s}})R^{+}_{ab}(\lambda_{a}-\lambda_{b})$
We also define the symmetric monodromy matrix for the mixed spin case
$T^{({\mbox{\boldmath$\ell$}})}_{0}(u;\zeta_{1},\ldots,\zeta_{r})$ as follows.
$T^{({\mbox{\boldmath$\ell$}})}_{0}(\lambda_{0};\zeta_{1},\ldots,\zeta_{r})=\prod_{k=1}^{r}\left(P^{(\ell_{k})}_{\ell_{1}+\cdots+\ell_{k-1}+1}\right)^{\bar{\chi}}\,\cdot\,R_{0,12\cdots
L}(\lambda_{0};\xi_{1}^{(\mbox{\boldmath$\ell$})},\ldots,\xi_{L}^{(\mbox{\boldmath$\ell$})})\,\cdot\,\prod_{k=1}^{r}\left(P^{(\ell_{k})}_{\ell_{1}+\cdots+\ell_{k-1}+1}\right)^{\bar{\chi}}$
(4.17)
### 4.3 Higher-spin $L$-operators
We now define the basis vectors of the $(\ell+1)$-dimensional irreducible
representation of $U_{q}(sl_{2})$, $||\ell,n\rangle$ for $n=0,1,\ldots,\ell$
as follows. We define $||\ell,0\rangle$ by
$||\ell,0\rangle=|1\rangle_{1}\otimes|1\rangle_{2}\otimes\cdots|1\rangle_{\ell}$
(4.18)
Here $|\alpha\rangle_{j}$ for $\alpha=1,2$ denote the basis vectors of the
spin-1/2 representation defined on the $j$th position in the tensor product.
We define $||\ell,n\rangle$ for $n\geq 1$ by
$||\ell,n\rangle=\left(\Delta^{(\ell-1)}(X^{-})\right)^{n}||\ell,0\rangle\,{\frac{1}{[n]_{q}!}}\,.$
(4.19)
Then we have
$||\ell,n\rangle=\sum_{1\leq
i_{1}<\cdots<i_{n}\leq\ell}\sigma_{i_{1}}^{-}\cdots\sigma_{i_{n}}^{-}|0\rangle\,q^{i_{1}+i_{2}+\cdots+i_{n}-n\ell+n(n-1)/2}$
(4.20)
It is easy to show the following:
$P^{(\ell)}_{12\cdots\ell}||\ell,n\rangle=||\ell,n\rangle$ (4.21)
We define the conjugate vectors by the following conditions:
$\langle\ell,n||P^{(\ell)}_{12\cdots\ell}=\langle\ell,n||$ (4.22)
with the normalization condition: $\langle\ell,n||\,||\ell,n\rangle=1$. Let us
define the $q$-factorial, $[n]_{q}!$, by
$[n]_{q}!=[n]_{q}[n-1]_{q}\cdots[1]_{q}\,.$ (4.23)
For integers $m$ and $n$ satisfying $m\geq n$ we define the $q$-binomial
coefficients as follows
$\left[\begin{array}[]{c}m\\\
n\end{array}\right]_{q}={\frac{[m]_{q}!}{[m-n]_{q}![n]_{q}!}}$ (4.24)
Then we have the following expression of the conjugate vectors
$\langle\ell,n||=\left[\begin{array}[]{c}\ell\\\
n\end{array}\right]_{q}^{-1}\,q^{n(\ell-n)}\,\sum_{1\leq
i_{1}<\cdots<i_{n}\leq\ell}\langle
0|\sigma_{i_{1}}^{+}\cdots\sigma_{i_{n}}^{+}\,q^{i_{1}+\cdots+i_{n}-n\ell+n(n-1)/2}$
(4.25)
The projection operators are given explicitly as follows.
$P^{(\ell)}_{12\cdots\ell}=\sum_{n=0}^{\ell}||\ell,n\rangle\,\langle\ell,n||$
(4.26)
We define the $L$-operator of the spin-$\ell/2$ XXZ model by
$L^{(\ell+)}(\lambda_{0})=P^{(\ell)}_{1}\,R^{+}_{0,12\cdots\ell}\,P^{(\ell)}_{1}\,.$
(4.27)
and then by the inverse gauge transformation we have
$L^{(\ell)}(\lambda_{0})=\left(P^{(\ell)}_{1}\right)^{\bar{\chi}}\,R_{0,12\cdots\ell}\,\left(P^{(\ell)}_{1}\right)^{\bar{\chi}}\,.$
(4.28)
Let us define $|\ell,n\rangle$ and their conjugates $\langle\ell,n|$ by
$\displaystyle|\ell,n\rangle$ $\displaystyle=$ $\displaystyle
N(\ell,n)\,||\ell,n\rangle$ $\displaystyle\langle\ell,n|$ $\displaystyle=$
$\displaystyle\langle\ell,n||\,{\frac{1}{N(\ell,n)}}\,.$ (4.29)
The matrix elements of the $L$-operator are given by
$\langle\ell,a|\,L^{(\ell)}(\lambda)\,|\ell,b\rangle=\left(\begin{array}[]{cc}\langle\ell,a|\,L^{(\ell)}_{11}(\lambda)\,|\ell,b\rangle&\langle\ell,a|\,L^{(\ell)}_{12}(\lambda)\,|\ell,b\rangle\\\
\langle\ell,a|\,L^{(\ell)}_{21}(\lambda)\,|\ell,b\rangle&\langle\ell,a|\,L^{(\ell)}_{22}(\lambda)\,|\ell,b\rangle\end{array}\right)_{[0]}$
(4.30)
for $a,b=0,1,\ldots,\ell$. Choosing the normalization factors $N(\ell,n)$, we
can derive the following symmetric expression of the $L$-operator:
$L^{(\ell)}(\lambda)={\frac{1}{2\sinh(u+\ell\eta/2)}}\left(\begin{array}[]{cc}zK^{1/2}-z^{-1}K^{-1/2}&2\sinh\eta
X^{-}\\\ 2\sinh\eta X^{+}&zK^{-1/2}-z^{-1}K^{1/2}\end{array}\right)_{[0]}$
(4.31)
Here $X^{\pm}$ and $K$ are in the $(\ell+1)$-dimensional representation of
$U_{q}(sl_{2})$, and $u=\lambda-\xi_{1}+\ell\eta/2$ and $z=\exp u$. Explicitly
they are given by
$\displaystyle\langle\ell,a|\,X^{+}\,|\ell,b\rangle$ $\displaystyle=$
$\displaystyle\delta_{a,b-1}\,[\ell-a]_{q}\,,$
$\displaystyle\langle\ell,a|\,X^{-}\,|\ell,b\rangle$ $\displaystyle=$
$\displaystyle\delta_{a,b+1}\,[a]_{q}\,,$
$\displaystyle\langle\ell,a|\,K\,|\ell,b\rangle$ $\displaystyle=$
$\displaystyle\delta_{a,b}\,q^{\ell-2a}\quad{\rm
for}\,\,a,b=0,1,\ldots,\ell,.$ (4.32)
### 4.4 Algebraic Bethe-ansatz method for higher-spin cases
We now discuss the eigenvalues of the transfer matrix of an integrable higher-
spin XXZ spin chain constructed by the fusion method. We consider the case of
mixed spins, where we define the transfer matrix on the tensor product of
spin-$s_{j}$ representations for $j=1,2,\ldots,r$.
We define $A$, $B$, $C$, and $D$ operators of the algebraic Bethe ansatz for
higher-spin cases by the following matrix elements of the monodromy matrix:
$\left(\begin{array}[]{cc}A^{({\mbox{\boldmath$\ell$}}+)}(\lambda_{0};\zeta_{1},\ldots,\zeta_{r})&B^{({\mbox{\boldmath$\ell$}}+)}(\lambda_{0};\zeta_{1},\ldots,\zeta_{r})\\\
C^{({\mbox{\boldmath$\ell$}}+)}(\lambda_{0};\zeta_{1},\ldots,\zeta_{r})&D^{({\mbox{\boldmath$\ell$}}+)}(\lambda_{0};\zeta_{1},\ldots,\zeta_{r})\end{array}\right)=T^{({\mbox{\boldmath$\ell$}}+)}_{0}(\lambda_{0};\zeta_{1},\ldots,\zeta_{r})\,.$
(4.33)
In terms of projection operators we have
$\displaystyle B^{({\mbox{\boldmath$\ell$}}+)}_{1\cdots
r}(\lambda_{0};\zeta_{1},\ldots,\zeta_{r})=\prod_{k=1}^{r}P^{(\ell_{k})}_{\ell(k-1)+1}\,\cdot\
B^{+}_{12\cdots
L}(\lambda_{0};\xi_{1}^{({\mbox{\boldmath$\ell$}})},\ldots,\xi_{L}^{({\mbox{\boldmath$\ell$}})})\prod_{k=1}^{r}P^{(\ell_{k})}_{\ell(k-1)+1}\,.$
(4.34)
We define $A^{({\mbox{\boldmath$\ell$}})},B^{({\mbox{\boldmath$\ell$}})}$,
$C^{({\mbox{\boldmath$\ell$}})}$ and $D^{({\mbox{\boldmath$\ell$}})}$
similarly for the monodromy matrix
$T^{({\mbox{\boldmath$\ell$}})}_{0}(\lambda_{0};\zeta_{1},\ldots,\zeta_{r})$.
The operators $A^{({\mbox{\boldmath$\ell$}}+)}$s of the asymmetric monodromy
matrix $R^{+}_{0,12\cdots n}$ are related to the symmetric ones
$A^{({\mbox{\boldmath$\ell$}})}$ s as follows.
$\displaystyle R^{+}_{0,12\cdots
n}(\lambda_{0},\\{\zeta_{i}\\})=\left(\begin{array}[]{cc}A^{({\mbox{\boldmath$\ell$}}+)}(\lambda_{0};\\{\zeta_{i}\\})&B^{({\mbox{\boldmath$\ell$}}+)}(\lambda_{0};\\{\zeta_{i}\\})\\\
C^{({\mbox{\boldmath$\ell$}}+)}(\lambda_{0};\\{\zeta_{i}\\})&D^{({\mbox{\boldmath$\ell$}}+)}(\lambda_{0};\\{\zeta_{i}\\})\end{array}\right)$
(4.37) $\displaystyle=$ $\displaystyle\left(\begin{array}[]{cc}\chi_{12\cdots
L}A^{({\mbox{\boldmath$\ell$}})}(\lambda_{0};\\{\zeta_{i}\\})\chi_{12\cdots
L}^{-1}&e^{-\lambda_{0}}\,\chi_{12\cdots
L}B^{({\mbox{\boldmath$\ell$}})}(\lambda_{0};\\{\zeta_{i}\\})\chi_{12\cdots
L}^{-1}\\\ e^{\lambda_{0}}\,\chi_{12\cdots
L}C^{({\mbox{\boldmath$\ell$}})}(\lambda_{0};\\{\zeta_{i}\\})\chi_{12\cdots
L}^{-1}&\chi_{12\cdots
L}D^{({\mbox{\boldmath$\ell$}})}(\lambda_{0};\\{\zeta_{i}\\})\chi_{12\cdots
L}^{-1}\end{array}\right)$ (4.40)
It follows from the Yang-Baxter equations (4.16) that the $A,B,C,D$ operators
in the higher-spin case also satisfy the standard commutation relations.
$A^{({\mbox{\boldmath$\ell$}}+)}(\lambda_{1})B^{({\mbox{\boldmath$\ell$}}+)}(\lambda_{2})={\frac{1}{b(\lambda_{2}-\lambda_{1})}}B^{({\mbox{\boldmath$\ell$}}+)}(\lambda_{2})A^{({\mbox{\boldmath$\ell$}}+)}(\lambda_{1})-{\frac{c^{-}(\lambda_{2}-\lambda_{1})}{b(\lambda_{2}-\lambda_{1})}}B^{({\mbox{\boldmath$\ell$}}+)}(\lambda_{1})A^{({\mbox{\boldmath$\ell$}}+)}(\lambda_{2})$
(4.41)
Through the inverse gauge transformation ${\bar{\chi}}$ we have
$A^{({\mbox{\boldmath$\ell$}})}(\lambda_{1})B^{({\mbox{\boldmath$\ell$}})}(\lambda_{2})={\frac{1}{b(\lambda_{2}-\lambda_{1})}}B^{({\mbox{\boldmath$\ell$}})}(\lambda_{2})A^{({\mbox{\boldmath$\ell$}})}(\lambda_{1})-{\frac{c(\lambda_{2}-\lambda_{1})}{b(\lambda_{2}-\lambda_{1})}}B^{({\mbox{\boldmath$\ell$}})}(\lambda_{1})A^{({\mbox{\boldmath$\ell$}})}(\lambda_{2})$
(4.42)
Therefore, we derive Bethe ansatz eigenvectors of the higher-spin transfer
matrix by the same method as the case of spin-1/2.
Let us denote by $|0\rangle$ the vacuum state where all spins are up. Noting
$\prod_{k=1}^{r}P^{(\ell_{k})}_{\ell(k-1)+1}|0\rangle=|0\rangle\,,$ (4.43)
it is easy to show the following relations:
$\displaystyle A^{({\mbox{\boldmath$\ell$}})}(\lambda)|0\rangle$
$\displaystyle=$ $\displaystyle
a^{({\mbox{\boldmath$\ell$}})}(\lambda;\\{\zeta_{k}\\})|0\rangle\,,$
$\displaystyle D^{({\mbox{\boldmath$\ell$}})}(\lambda)|0\rangle$
$\displaystyle=$ $\displaystyle
d^{({\mbox{\boldmath$\ell$}})}(\lambda;\\{\zeta_{k}\\})|0\rangle\,,$ (4.44)
where $a^{({\mbox{\boldmath$\ell$}})}(\lambda;\\{\zeta_{k}\\})$ and
$d^{({\mbox{\boldmath$\ell$}})}(\lambda;\\{\zeta_{k}\\})$ are given by
$\displaystyle a^{({\mbox{\boldmath$\ell$}})}(\lambda;\\{\zeta_{k}\\})$
$\displaystyle=$ $\displaystyle
a^{({\mbox{\boldmath$\ell$}})}(\lambda;\\{\xi_{j}^{({\mbox{\boldmath$\ell$}})}\\})=1\,,$
$\displaystyle d^{({\mbox{\boldmath$\ell$}})}(\lambda;\\{\zeta_{k}\\})$
$\displaystyle=$ $\displaystyle
d^{({\mbox{\boldmath$\ell$}})}(\lambda;\\{\xi_{j}^{({\mbox{\boldmath$\ell$}})}\\})=\prod_{j=1}^{L}b(\lambda-\xi_{j}^{({\mbox{\boldmath$\ell$}})})=\prod_{k=1}^{r}{\frac{\sinh(\lambda-\zeta_{k}-(\ell_{k}-1)\eta/2)}{\sinh(\lambda-\zeta_{k}+(\ell_{k}+1)\eta/2)}}$
Thus, the vector $B^{{(\mbox{\boldmath$\ell$})}}(\lambda_{1})\cdots
B^{(\mbox{\boldmath$\ell$})}(\lambda_{n})|0\rangle$ becomes an eigenvector of
the transfer matrix
$A^{({\mbox{\boldmath$\ell$}})}(\lambda)+D^{({\mbox{\boldmath$\ell$}})}(\lambda)$
with the following eigenvalue
$\tau^{({\mbox{\boldmath$\ell$}})}(\mu)=\prod_{j=1}^{n}{\frac{\sinh(\lambda_{j}-\mu+\eta)}{\sinh(\lambda_{j}-\mu)}}+\prod_{k=1}^{r}{\frac{\sinh(\mu-\zeta_{k}-(\ell_{k}-1)\eta/2)}{\sinh(\mu-\zeta_{k}+(\ell_{k}+1)\eta/2)}}\,\cdot\,\prod_{j=1}^{n}{\frac{\sinh(\mu-\lambda_{j}+\eta)}{\sinh(\mu-\lambda_{j})}}$
(4.46)
if rapidities $\tilde{\lambda}_{j}=\lambda_{j}+\eta/2$ satisfy the Bethe
ansatz equations
$\prod_{k=1}^{r}{\frac{\sinh(\tilde{\lambda}_{\alpha}-\zeta_{k}+\ell_{k}\eta/2)}{\sinh(\tilde{\lambda}_{\alpha}-\zeta_{k}-\ell_{k}\eta/2)}}=\prod_{\beta=1;\beta\neq\alpha}^{n}{\frac{\sinh(\tilde{\lambda}_{\alpha}-\tilde{\lambda}_{\beta}+\eta)}{\sinh(\tilde{\lambda}_{\alpha}-\tilde{\lambda}_{\beta}-\eta)}}$
(4.47)
## 5 Pseudo-diagonalization of the $B$ and $C$ operators
### 5.1 Diagonalizing the $A$ and $D$ operators
#### 5.1.1 The $F$-basis
In order to formulate the derivation of the pseudo-diagonalized forms of $B$
and $C$ operators for the XXZ case, we briefly formulate some symbols and
review some useful formulas shown in Ref. [6] in SS5.1. First, we introduce
the $F$-basis .
###### Definition 15.
(Partial $F$ and total $F$) We define partial $F$ by
$\displaystyle F_{1,\,2\cdots n}$ $\displaystyle=$ $\displaystyle
e^{11}_{1}+e_{1}^{22}R_{1,\,2\cdots n}$ $\displaystyle F_{12\cdots n-1,\,n}$
$\displaystyle=$ $\displaystyle e^{22}_{n}+e_{n}^{11}R_{12\cdots n-1,\,n}$
(5.1)
We define total $F$ recursively with respect to $n$ by
$F_{12\cdots n}=F_{12\cdots n-1}F_{12\cdots n-1,\,n}$ (5.2)
###### Lemma 16.
(Cocycle conditions)
$\displaystyle F_{1,\,2}F_{12,\,3}$ $\displaystyle=$ $\displaystyle
F_{23}F_{1,\,23}$ $\displaystyle F_{1,\,2\cdots n-1}F_{12\cdots n-1\,n}$
$\displaystyle=$ $\displaystyle F_{2\cdots n-1,\,n}F_{1,\,2\cdots n}$ (5.3)
###### Proof.
Expressing the $F$-basis in terms of $R$-matrices through (5.1), we show that
the cocycle conditions of the $F$-basis are reduced to those of the
$R$-matrices, which are shown in Appendix B. ∎
From the cocycle conditions we have the following:
###### Lemma 17.
$F_{12\cdots n}=F_{2\cdots n}F_{1,2\cdots n}$ (5.4)
#### 5.1.2 Basic properties of the $R$-matrix
Let us introduce some important properties of the $R$-matrix of the XXZ spin-
chain.
The $R$-matrix is invariant under the charge conjugation. For the symmetric
$R$-matrix, we define the charge conjugation operator ${\cal C}$ by
${\cal C}_{12\cdots n}=\sigma^{x}_{1}\cdots\sigma_{n}^{x}$ (5.5)
For a given operator $A\in End(V(\lambda_{1})\otimes\cdots\otimes
V(\lambda_{n}))$ we define ${\bar{A}}$ by ${\bar{A}}={\cal C}_{1\cdots
n}A{\cal C}_{1\cdots n}$. For instance, we define ${\bar{F}}_{0,1\cdots n}$ by
${\bar{F}}_{0,1\cdots n}={\cal C}_{01\cdots n}F_{0,1\cdots n}{\cal
C}_{01\cdots n}$ (5.6)
###### Proposition 18.
The charge conjugation operator ${\cal C}$ commutes with the monodromy matrix
of the symmetric $R$-matrix:
${[}{\cal C}_{01\cdots n},\,R_{0,1\cdots n}{]}=0$ (5.7)
We thus have ${\bar{A}}_{1\cdots n}(\lambda_{0})=D_{1\cdots n}(\lambda_{0})$
and ${\bar{B}}_{1\cdots n}(\lambda_{0})=C_{1\cdots n}(\lambda_{0})$.
###### Lemma 19 (Crossing symmetry).
The $R$-matrix has the crossing symmetry relation:
$\left(\gamma\otimes
I\right)R_{12}(\lambda_{1}-\eta,\lambda_{2})\left(\gamma\otimes
I\right)=b_{21}^{-1}R_{21}^{t_{1}}(\lambda_{2},\lambda_{1})$ (5.8)
where $b_{21}=b(\lambda_{2}-\lambda_{1})$ and $\gamma$ is given by
$\gamma=\sigma^{y}=\left(\begin{array}[]{cc}0&-i\\\ i&0\end{array}\right)$
(5.9)
Here the crossing symmetry is slightly different from [6].
###### Lemma 20.
(Crossing symmetry of the monodromy matrix)
$\gamma_{0}R_{0,1\cdots
n}(\lambda_{0}-\eta;\xi_{1},\ldots,\xi_{n})\gamma_{0}=\left(\prod_{i=1}^{n}b^{-1}(\xi_{i}-\lambda_{0})\right)R^{t_{0}}_{1\cdots
n,0}(\lambda_{0};\xi_{1},\ldots,\xi_{n})$ (5.10)
Let us introduce $\dagger$ operation. We shall use it when we pseudo-
diagonalize the $B$ operators.
###### Definition 21.
For $X_{1\cdots n}(\lambda_{1},\cdots,\lambda_{n})\in
End(V(\lambda_{1})\otimes\cdots\otimes V(\lambda_{n}))$ we define
$X^{\dagger}_{1\cdots n}$ by
$X^{\dagger}_{1\cdots n}(\lambda_{1},\ldots,\lambda_{n})=X^{t_{1}\cdots
t_{n}}_{1\cdots n}(-\lambda_{1},\cdots,-\lambda_{n})$ (5.11)
Here we note $(X^{\dagger})^{\dagger}=X$, and
$(XY)^{\dagger}=Y^{\dagger}X^{\dagger}$. It is easy to show
$R_{12}^{\dagger}=R_{21}$.
###### Lemma 22.
Under the $\dagger$ operation the monodromy matrix is given by the following:
$R^{\dagger}_{0,1\cdots n}=R_{0,1\cdots n}^{-1}=R_{1\cdots n,\,0}\\\ $ (5.12)
We define operators $A^{\dagger}_{1\cdots n}$, $B^{\dagger}_{1\cdots n}$,
$C^{\dagger}_{1\cdots n}$ and $D^{\dagger}_{1\cdots n}$ by
$R^{\dagger}_{0,1\cdots n}=\left(\begin{array}[]{cc}A^{\dagger}_{1\cdots
n}(\lambda_{0})&C^{\dagger}_{1\cdots n}(\lambda_{0})\\\ B^{\dagger}_{1\cdots
n}(\lambda_{0})&D^{\dagger}_{1\cdots n}(\lambda_{0})\end{array}\right)_{[0]}$
(5.13)
###### Proposition 23.
Under the $\dagger$ operation the monodromy matrix is given by
$\displaystyle R_{0,12\cdots
n}^{\dagger}=\left(\begin{array}[]{cc}A^{\dagger}_{1\cdots
n}(\lambda_{0})&C^{\dagger}_{1\cdots n}(\lambda_{0})\\\ B^{\dagger}_{1\cdots
n}(\lambda_{0})&D^{\dagger}_{1\cdots n}(\lambda_{0})\end{array}\right)_{[0]}$
(5.16)
$\displaystyle\quad=\left(\prod_{i=1}^{n}b(\xi_{i}-\lambda_{0})\right)\left(\begin{array}[]{cc}A_{1\cdots
n}(\lambda_{0}-\eta)&-B_{1\cdots n}(\lambda_{0}-\eta)\\\ -C_{1\cdots
n}(\lambda_{0}-\eta)&D_{1\cdots n}(\lambda_{0}-\eta)\end{array}\right)_{[0]}$
(5.19)
#### 5.1.3 The diagonalized forms of operators $A$ and $D$
Let us give the diagonalized forms of the $A$ and $D$ operators [6].
The following criterion for the $F$-basis to be non-singular should be useful.
###### Proposition 24.
The determinants of the partial and total $F$ matrices are given by
${\rm det}F_{0,1\cdots n}=\prod_{j=1}^{n}b(\lambda_{0}-\xi_{j})\,,\quad{\rm
det}F_{1\cdots n}=\prod_{1\leq i<j\leq n}b(\xi_{i}-\xi_{j})$ (5.20)
Proposition 5.20 follows from lemma D.1 of Appendix D.
We can show the diagonalized forms of operators $A$ and $D$ as follows [6].
###### Proposition 25 (Diagonalization of $A$ and $D$).
$\displaystyle F_{1\cdots n}D_{1\cdots n}(\lambda_{0})F_{1\cdots n}^{-1}$
$\displaystyle=$
$\displaystyle\bigotimes_{i=1}^{n}\left(\begin{array}[]{cc}b_{0i}&0\\\
0&1\end{array}\right)_{[i]}$ (5.23) $\displaystyle{\bar{F}}_{1\cdots
n}A_{1\cdots n}(\lambda_{0}){\bar{F}}_{1\cdots n}^{-1}$ $\displaystyle=$
$\displaystyle\bigotimes_{i=1}^{n}\left(\begin{array}[]{cc}1&0\\\
0&b_{0i}\end{array}\right)_{[i]}\,,$ (5.26)
where $b_{0i}=b(\lambda_{0}-\xi_{i})$
###### Proposition 26 (Diagonalization of $A^{\dagger}$ and $D^{\dagger}$).
$\displaystyle F_{1\cdots n}A^{\dagger}_{1\cdots n}(\lambda_{0})F_{1\cdots
n}^{-1}$ $\displaystyle=$
$\displaystyle\bigotimes_{i=1}^{n}\left(\begin{array}[]{cc}1&0\\\
0&b_{i0}\end{array}\right)_{[i]}$ (5.29) $\displaystyle{\bar{F}}_{1\cdots
n}D^{\dagger}_{1\cdots n}(\lambda_{0}){\bar{F}}_{1\cdots n}^{-1}$
$\displaystyle=$
$\displaystyle\bigotimes_{i=1}^{n}\left(\begin{array}[]{cc}b_{i0}&0\\\
0&1\end{array}\right)_{[i]}$ (5.32)
where $b_{i0}=b(\xi_{i}-\lambda_{0})$
The derivation of the diagonalized forms and some useful formulas are briefly
reviewed in Appendix D.
For a given operator $X_{1\cdots n}(\lambda_{1},\cdots,\lambda_{n})\in
End(V(\lambda_{1})\otimes\cdots\otimes V(\lambda_{n}))$ we denote $FAF^{-1}$
by ${\widetilde{F}}$:
${\widetilde{X}}_{1\cdots n}=F_{12\cdots n}X_{12\cdots n}F_{12\cdots
n}^{-1}\,.$ (5.33)
For instance we have ${\widetilde{D}}_{1\cdots n}(\lambda_{0})={F}_{1\cdots
n}D_{1\cdots n}(\lambda_{0}){F}_{1\cdots n}^{-1}$.
### 5.2 Pseudo-diagonalization of the $B$ operator
Let us recall that the matrix elements of the monodromy matrix
$R^{+}_{0,1\cdots L}$ are related to the symmetric ones as follows.
$\displaystyle R^{+}_{0,12\cdots
L}(u;\xi_{1},\ldots,\xi_{L})=\left(\begin{array}[]{cc}A^{+}_{12\cdots
L}(u;\xi_{1},\ldots,\xi_{L})&B^{+}_{12\cdots L}(u;\xi_{1},\ldots,\xi_{L})\\\
C^{+}_{12\cdots L}(u;\xi_{1},\ldots,\xi_{L})&D^{+}_{12\cdots
L}(u;\xi_{1},\ldots,\xi_{L})\end{array}\right)_{[0]}$ (5.36) $\displaystyle=$
$\displaystyle\left(\begin{array}[]{cc}\chi_{12\cdots L}A_{12\cdots
L}(u;\\{\xi_{j}\\})\left(\chi_{12\cdots
L}\right)^{-1}&e^{-\lambda_{0}}\chi_{12\cdots L}B_{12\cdots
L}(u;\\{\xi_{j}\\})\left(\chi_{12\cdots L}\right)^{-1}\\\
e^{\lambda_{0}}\chi_{12\cdots L}C_{12\cdots
L}(u;\\{\xi_{j}\\})\left(\chi_{12\cdots L}\right)^{-1}&\chi_{12\cdots
L}D_{12\cdots L}(u;\\{\xi_{j}\\})\left(\chi_{12\cdots
L}\right)^{-1}\end{array}\right)_{[0]}$ (5.39)
Then, from the quantum-group invariance (3.10) we have the following
commutation relations:
$\displaystyle B^{+}_{1\cdots n}(\lambda)$ $\displaystyle=$ $\displaystyle
D^{+}_{1\cdots
n}(\lambda)\Delta^{(n-1)}(X^{-})-q\Delta^{(n-1)}(X^{-})D^{+}_{1\cdots
n}(\lambda)$ (5.41) $\displaystyle C^{+}_{1\cdots n}(\lambda)$
$\displaystyle=$ $\displaystyle\Delta^{(n-1)}(X^{+})D^{+}_{1\cdots
n}(\lambda)-q^{-1}D^{+}_{1\cdots n}(\lambda)\Delta^{(n-1)}(X^{+})$ (5.42)
Here $X^{\pm}$ are generators of $U_{q}(sl_{2})$, and $\Delta^{(n-1)}(X^{-})$
denote the tensor-product representation of $\Delta^{(n-1)}(X^{-})$ acting on
the $n$ sites from the 1st to $n$th. We remark that more generally, we have
commutation relations (3.39) for the affine quantum group
$U_{q}({\widehat{sl_{2}}})$.
In this subsection we abbreviate the superscript $+$ for the asymmetric
monodromy matrix, for simplicity. In fact, the essential parts of formulas
such as the fundamental commutation relations are invariant under gauge
transformations if we express them in terms of the generators of the quantum
affine algebra $U_{q}(\widehat{sl_{2}})$ in the evaluation representation
(3.28). Here we remark that the matrix representation of the evaluation
representation of $U_{q}(\widehat{sl_{2}})$ can be changed through gauge
transformations.
Let us now introduce some symbols.
###### Definition 27.
We define operators ${\widehat{\delta}}_{jk}(\lambda_{j},\lambda_{k})$ for
$j,k$ satisfying $0\leq j<k\leq L$ by
${\widehat{\delta}}_{jk}(\lambda_{j},\lambda_{k})=\left(\begin{array}[]{cccc}1&0&0&0\\\
0&b^{-1}_{kj}&0&0\\\ 1&0&b^{-1}_{jk}&0\\\ 1&0&0&1\\\
\end{array}\right)_{[jk]}\,,$ (5.43)
where $b_{jk}=b(\lambda_{j}-\lambda_{k})$ and
$b_{kj}=b(\lambda_{k}-\lambda_{j})$. We define ${\widehat{\delta}}_{1\cdots
n}$ and ${\widehat{\delta}}_{0,1\cdots n}$ by
$\displaystyle{\widehat{\delta}}_{1\cdots n}$ $\displaystyle=$
$\displaystyle\prod_{1\leq j<k\leq
n}{\widehat{\delta}}_{jk}(\lambda_{j},\lambda_{k})$
$\displaystyle{\widehat{\delta}}_{0,1\cdots n}$ $\displaystyle=$
$\displaystyle{\widehat{\delta}}_{01\cdots n}{\widehat{\delta}}^{-1}_{1\cdots
n}=\prod_{j=1}^{n}{\widehat{\delta}}_{0j}(\lambda_{0},\lambda_{j})$ (5.44)
We define ${\widehat{\delta}}^{1\cdots n}_{i}$ by
${\widehat{\delta}}^{1\cdots n}_{i}=\widehat{\delta}_{i,i+1\cdots n1\cdots
i-1}=\prod_{j=1;j\neq i}^{n}{\widehat{\delta}}_{ij}$ (5.45)
Some useful formulas are given in Appendix D.
Let us denote $I^{\otimes m}\otimes\Delta^{(\ell-1)}(x)\otimes
I^{\otimes(n-\ell-m)}$ by $\Delta^{(\ell-1)}_{m+1\,m+2\,\cdots\,m+\ell}(x)$ or
$\Delta_{m+1\,m+2\,\cdots\,m+\ell}(x)$ for $x\in U_{q}(sl_{2})$ in the tensor-
product representation.
###### Lemma 28.
Let $X^{-}$ denote the generator of the quantum group $U_{q}(sl_{2})$ and
$X_{j}^{-}$ the spin-1/2 representation of $X^{-}$ acting on the $j$th site in
the tensor product representation $(V^{(1)})^{\otimes n}$. We have
$\widetilde{\Delta}_{1\cdots
n}(X^{-})=\left(X_{1}^{-}+e_{1}^{11}\widetilde{\Delta}_{2\cdots
n}(X^{-})\widetilde{A^{\dagger}}_{2\cdots
n}(\xi_{1})+e_{1}^{22}\widetilde{D}_{2\cdots
n}(\xi_{1})\widetilde{\Delta}_{2\cdots
n}(X^{-})\right)\widehat{\delta}_{1,2\cdots n}$ (5.46)
###### Proof.
Making use of (D.39) we show $F^{-1}_{1\cdots n}=F^{t_{1}\cdots
t_{n}}_{n\cdots 2,1}F_{2\cdots n}^{-1}\widehat{\delta}_{1,2\cdots n}$. We have
$\displaystyle\widetilde{\Delta}_{1\cdots n}(X^{-})$ $\displaystyle=$
$\displaystyle F_{2\cdots n}F_{1,2\cdots
n}{\Delta}^{(n-1)}(X^{-})F^{t_{1}\cdots t_{n}}_{n\cdots 2,1}F_{2\cdots
n}^{-1}\widehat{\delta}_{1,2\cdots n}$ (5.47)
Putting $F_{1,2\cdots n}=e_{1}^{11}+e_{1}^{22}R_{1,2\cdots n}$ and
$F^{t_{1}\cdots t_{n}}_{n\cdots 2,1}=e_{1}^{22}+R_{2\cdots n,1}e_{1}^{11}$, we
have
$\displaystyle F_{1,2\cdots n}{\Delta}^{(n-1)}(X^{-})F^{t_{1}\cdots
t_{n}}_{n\cdots 2,1}=\left(e_{1}^{11}+e_{1}^{22}R_{1,2\cdots
n}\right)\,{\Delta}^{(n-1)}(X^{-})\,\left(e_{1}^{22}+R_{2\cdots
n,1}e_{1}^{11}\right)$
$\displaystyle=e_{1}^{11}{\Delta}^{(n-1)}(X^{-})e_{1}^{22}+e_{1}^{11}{\Delta}^{(n-1)}(X^{-})R_{2\cdots
n,1}e_{1}^{11}$ $\displaystyle\quad+e_{1}^{22}R_{1,2\cdots
n}{\Delta}^{(n-1)}(X^{-})e_{1}^{22}+e_{1}^{22}R_{1,2\cdots
n}{\Delta}^{(n-1)}(X^{-})R_{2\cdots n,1}e_{1}^{11}$
$\displaystyle=0+e_{1}^{11}{\Delta}_{2\cdots n}(X^{-})R_{2\cdots
n,1}e_{1}^{11}+e_{1}^{22}R_{1,2\cdots n}{\Delta}_{2\cdots
n}(X^{-})e_{1}^{22}+X_{1}^{-}\,.$ (5.48)
Here we have made use of the following:
${\Delta}^{(n-1)}(X^{-})={\Delta}^{(n-2)}\Delta(X^{-})=X^{-}\otimes\Delta^{(n-2)}(K^{-1})+I\otimes\Delta^{(n-2)}(X^{-})\,.$
(5.49)
We thus have
$\widetilde{\Delta}_{1\cdots
n}(X^{-})=\left(X_{1}^{-}+e_{1}^{11}\widetilde{\Delta}_{2\cdots
n}(X^{-})\widetilde{R}_{2\cdots
n,1}e_{1}^{11}+e_{1}^{22}\widetilde{R}_{1,2\cdots
n}\widetilde{\Delta}_{2\cdots
n}(X^{-})e_{1}^{11}\right)\widehat{\delta}_{1,2\cdots n}\,.$ (5.50)
We obtain the case of $n$ from (5.12). ∎
###### Lemma 29.
In the tensor-product representation $(V^{(1)})^{\otimes n}$ we have
${\widetilde{\Delta}}_{1\cdots
n}(X^{-})=\sum_{i=1}^{n}X_{i}^{-}{\widehat{\delta}}^{1\cdots n}_{i}\,.$ (5.51)
###### Proof.
We show it by induction on $n$. The case of $n=1$ is trivial. Let us assume
the case of $n-1$. In eq. (5.46), the first term gives the following:
$X_{1}^{-}\widehat{\delta}_{1,2\cdots
n}=X_{1}^{-}\widehat{\delta}_{1}^{1\cdots n}$. Assuming (5.51) for
$\widehat{\Delta}_{2\cdots n}$ and putting it into the second term of (5.46),
we have
$\displaystyle e_{1}^{11}\widetilde{\Delta}_{2\cdots
n}(X^{-})\widetilde{A^{\dagger}}_{2\cdots
n}(\xi_{1})\widehat{\delta}_{1,2\cdots n}$ $\displaystyle=$ $\displaystyle
e_{1}^{11}\sum_{i=2}^{n}X_{i}^{-}\widehat{\delta}_{i}^{2\cdots
n}\bigotimes_{k=2}^{n}\left(\begin{array}[]{cc}1&0\\\
0&b_{k1}\end{array}\right)_{[k]}\widehat{\delta}_{1,2\cdots n}$
$\displaystyle=$
$\displaystyle\sum_{i=2}^{n}X_{i}^{-}\widehat{\delta}_{i}^{1\cdots
n}e_{1}^{11}$ $\displaystyle=$ $\displaystyle
e_{1}^{11}\sum_{i=2}^{n}X_{i}^{-}\widehat{\delta}_{i}^{1\cdots n}$ (5.55)
Here we have made use of (5.29). Similarly, we have
$e_{1}^{22}\widetilde{D}_{2\cdots n}(\xi_{1})\widetilde{\Delta}_{2\cdots
n}(X^{-})e_{1}^{22}\widehat{\delta}_{1,2\cdots
n}=e_{1}^{22}\sum_{i=2}^{n}X_{i}^{-}\widehat{\delta}_{i}^{1\cdots n}$ (5.56)
Thus, we have the case of $n$ as follows.
$\displaystyle{\widetilde{\Delta}}_{1\cdots n}(X^{-})$ $\displaystyle=$
$\displaystyle X_{1}^{-}{\widehat{\delta}}_{1}^{1\cdots
n}+(e_{1}^{11}+e_{1}^{22})\sum_{i=2}^{n}X_{1}^{-}{\widehat{\delta}}_{i}^{1\cdots
n}$ (5.57) $\displaystyle=$
$\displaystyle\sum_{i=1}^{n}X_{i}^{-}{\widehat{\delta}}_{i}^{1\cdots n}\,.$
∎
From the fundamental commutation relation (5.41) we have the following:
###### Lemma 30.
In the tensor product $(V^{(1)})^{\otimes n}$ we have
${\widetilde{B}}_{1\cdots
n}(\lambda)=\sum_{i=1}^{n}X_{i}^{-}({\widetilde{D}}_{1\cdots i-1,i+1\cdots
n}(\lambda)-q{\widetilde{D}}_{1\cdots
n}(\lambda))\widehat{\delta}_{i}^{1\cdots n}\,.$ (5.58)
###### Proof.
We transform the both sides of the fundamental commutation relation (5.41) by
$F_{1\cdots n}$, and put (5.23) and (5.51) into it, we have the following:
$\displaystyle{\widetilde{B}}_{1\cdots n}(\lambda)$ $\displaystyle=$
$\displaystyle{\widetilde{D}}_{1\cdots
n}(\lambda){\widetilde{\Delta}}_{1\cdots
n}(X^{-})-q{\widetilde{\Delta}_{1\cdots n}}(X^{-}){\widetilde{D}}_{1\cdots
n}(\lambda)$ (5.63) $\displaystyle=$
$\displaystyle\bigotimes_{j=1}^{n}\left(\begin{array}[]{cc}b_{0j}&0\\\
0&1\end{array}\right)_{[j]}\sum_{i=1}^{n}X_{i}^{-}{\widehat{\delta}}^{1\cdots
n}_{i}-q\sum_{i=1}^{n}X_{i}^{-}{\widehat{\delta}}^{1\cdots
n}_{i}\bigotimes_{j=1}^{n}\left(\begin{array}[]{cc}b_{0j}&0\\\
0&1\end{array}\right)_{[j]}$ $\displaystyle=$
$\displaystyle\sum_{i=1}^{n}X_{i}^{-}({\widetilde{D}}_{1\cdots i-1,i+1\cdots
n}(\lambda)-q{\widetilde{D}}_{1\cdots
n}(\lambda))\widehat{\delta}_{i}^{1\cdots n}\,.$ (5.64)
∎
###### Proposition 31 (Pseudo-diagonalization of $B$ operator).
We have
${\widetilde{B}}_{1\cdots
n}(\lambda)=\sum_{i=1}^{n}c_{0i}^{-}\,X^{-}_{i}\bigotimes_{j=1;j\neq
i}^{n}\left(\begin{array}[]{cc}b_{0j}&0\\\
0&b_{ji}^{-1}\end{array}\right)_{[j]}\,,$ (5.65)
where $b_{0i}=b(\lambda_{0}-\xi_{i})$, $b_{ji}=b(\xi_{j}-\xi_{i})$ and
$c_{0i}^{-}=c^{-}(\lambda_{0}-\xi_{i})=\exp(-(\lambda_{0}-\xi_{i}))c(\lambda_{0}-\xi_{i})$.
###### Proof.
Let us denote $b_{0i}=\sinh(\lambda-\xi_{i})/\sinh(\lambda-\xi_{i}+\eta)$ and
$c_{0i}=\sinh(\eta)/\sinh(\lambda-\xi_{i}+\eta)$, by $b_{0i}$ and $c_{0i}$
respectively, . Putting $1-qb_{01}=c_{0i}^{-}$ in (5.58) we show
${\widetilde{B}}_{1\cdots
n}(\lambda)=\sum_{i=1}^{n}c_{0i}^{-}X_{i}^{-}{\widetilde{D}}_{1\cdots
i-1\,i+1\cdots n}{\widehat{\delta}}_{i}^{1\cdots n}\,.$ (5.66)
After some calculation, we have (5.65). ∎
Similarly, making use of lemmas E.1, E.2 and E.3, we can show the diagonalized
form of operator $C$.
###### Proposition 32.
Let $X_{i}^{+}$ denote the spin-1/2 representation of $X^{+}$ acting on the
$i$th site in the tensor product representation. We have
${\widetilde{C}}_{1\cdots
n}(\lambda_{0})=\sum_{i=1}^{n}c_{0i}^{+}\,X^{+}_{i}\bigotimes_{j=1;j\neq
i}^{n}\left(\begin{array}[]{cc}b_{0j}b_{ij}^{-1}&0\\\
0&1\end{array}\right)_{[j]}\,,$ (5.67)
where $b_{0i}=b(\lambda_{0}-\xi_{i})$, $b_{ji}=b(\xi_{j}-\xi_{i})$ and
$c_{0i}^{+}=c^{+}(\lambda_{0}-\xi_{i})$.
### 5.3 Pseudo-diagonalized forms of the symmetric $B$ and $C$ operators
Let us show the pseudo-diagonalized forms of the $B$ and $C$ operators of the
symmetric monodromy matrix $R_{0,1\cdots n}$. Here we recall that expressions
(5.65) and (5.67) are for $\widetilde{B}^{+}_{12\cdots n}(\lambda)$ and
$\widetilde{C}^{+}_{12\cdots n}(\lambda)$, respectively. They are matrix
elements of the asymmetric monodromy matrix $R^{+}_{0,1\cdots
n}=\chi_{01\cdots n}R_{0,1\cdots n}\chi^{-1}_{01\cdots n}$. We have the
following relations:
$\displaystyle B^{+}_{12\cdots n}(\lambda)$ $\displaystyle=$ $\displaystyle
e^{-\lambda}\chi_{01\cdots n}R_{0,1\cdots n}\chi^{-1}_{01\cdots n}$
$\displaystyle C^{+}_{12\cdots n}(\lambda)$ $\displaystyle=$ $\displaystyle
e^{\lambda}\chi_{01\cdots n}R_{0,1\cdots n}\chi^{-1}_{01\cdots n}$ (5.68)
Therefore, applying the inverse gauge transformation ${\bar{\chi}}$ to (5.65)
and (5.67), we obtain
${\widetilde{B}}_{1\cdots
n}(\lambda)=\sum_{i=1}^{n}c_{0i}\,\sigma^{-}_{i}\bigotimes_{j=1;j\neq
i}^{n}\left(\begin{array}[]{cc}b_{0j}&0\\\
0&b_{ji}^{-1}\end{array}\right)_{[j]}\,,$ (5.69)
and
${\widetilde{C}}_{1\cdots
n}(\lambda_{0})=\sum_{i=1}^{n}c_{0i}\,\sigma^{+}_{i}\bigotimes_{j=1;j\neq
i}^{n}\left(\begin{array}[]{cc}b_{0j}b_{ij}^{-1}&0\\\
0&1\end{array}\right)_{[j]}\,.$ (5.70)
Here we recall $c_{0i}=\sinh(\eta)/\sinh(\lambda-\xi_{i}+\eta)$.
We should remark that expressions (5.69) and (5.70) coincide with eq. (2.29)
and (2.30) of Ref. [7], respectively.
## 6 Scalar products formulas
### 6.1 Formula for higher-spin scalar products
Let us consider the case of tensor product of spin-$s$ representations. We
recall that $\ell=2s$ and $L=\ell N_{s}$. We introduce parameters
$\xi_{j}^{(\ell;\epsilon)}$ for $j=1,2,\ldots,L$, as follows:
$\xi_{(k-1)\ell+j}^{(\ell;\epsilon)}=\zeta_{k}-(j-1)\eta+\ell\eta/2+\epsilon
r_{j}\qquad j=1,\ldots\ell;k=1,\ldots,N_{s}\,.$ (6.1)
Here $r_{j}$ $(j=1,2,\ldots,\ell)$ are distinct and nonzero parameters, and
$\epsilon$ is an arbitrary small number. We also introduce the following
symbol:
$P^{(\ell)}_{1\cdots L}=\prod_{j=1}^{N_{s}}P^{(\ell)}_{(j-1)\ell+1}\,,\quad
P^{(\ell)\,{\bar{\chi}}}_{1\cdots
L}=\prod_{j=1}^{N_{s}}\left(P^{(\ell)}_{(j-1)\ell+1}\right)^{\bar{\chi}}$
(6.2)
Here we recall that $B$ operator acting on the tensor product of spin-$s$
representations, $\left(V^{(2s)}\right)^{\otimes N_{s}}$, is given by $B$
operator acting on the tensor product of spin-1/2 representations
$\left(V^{(1)}\right)^{\otimes L}$ with $L=N_{s}\ell$ and multiplied by the
projection operators:
$B^{(\ell)}_{1\cdots N_{s}}(u;\zeta_{1},\ldots,\zeta_{N_{s}})=P_{1\cdots
L}^{(\ell)}\,B_{1\cdots
L}^{(1)}(u;\xi_{1}^{(\ell)},\ldots,\xi_{L}^{(\ell)})\,P_{1\cdots
L}^{(\ell)}\,.$
We now define the scalar product for the spin-$\ell/2$ case as follows.
###### Definition 33.
Let $\\{\lambda_{\alpha}\\}$ ($\alpha=1,2,\ldots,n$) be a set of solutions of
the Bethe ansatz equations and $\\{\mu_{j}\\}$ ($j=1,2,\ldots,n$) be arbitrary
numbers. We define the scalar product
$S^{(\ell)}_{n}(\\{\mu_{j}\\},\\{\lambda_{\alpha}\\};\\{\zeta_{k}\\})$ by the
following:
$S^{(\ell)}_{n}(\\{\mu_{j}\\},\\{\lambda_{\alpha}\\};\\{\zeta_{k}\\})=\langle
0|\,C^{(\ell)}(\mu_{1})\cdots C^{(\ell)}(\mu_{n})B^{(\ell)}(\lambda_{1})\cdots
B^{(\ell)}(\lambda_{n})\,|0\rangle$ (6.3)
Here $C^{(\ell)}(\mu_{j})$ and $B^{(\ell)}(\lambda_{\alpha})$ abbreviate
$C^{(\ell)}_{1\cdots N_{s}}(\mu_{j};\\{\zeta_{j}\\})$ and $B^{(\ell)}_{1\cdots
N_{s}}(\lambda_{\alpha};\\{\zeta_{j}\\})$, respectively, and $\zeta_{k}$
denote the centers of $\ell_{k}$-strings of the inhomogeneous parameters
$\\{\xi_{k}^{\ell}\\}$.
We calculate the scalar product for the higher-spin XXZ chains by the formula
in the next proposition.
###### Proposition 34.
Let $\\{\lambda_{\alpha}\\}$ satisfy the Bethe ansatz equations for the
spin-$\ell/2$ case. The scalar product of the spin-$\ell/2$ XXZ spin chain is
reduced into that of the spin-1/2 XXZ spin chain as follows:
$S^{(\ell)}_{n}(\\{\mu_{j}\\},\\{\lambda_{\alpha}\\};\\{\zeta_{k}\\})=\lim_{\epsilon\rightarrow
0}\Bigg{[}S^{(1)}_{n}(\\{\mu_{j}\\},\\{\lambda_{\alpha}\\};\\{\xi^{(\ell;\epsilon)}_{k}\\})\Bigg{]}$
(6.4)
###### Proof.
We now calculate the scalar product making use of eq. (4.7) of lemma 4.7 as
follows.
$\displaystyle\langle 0|\,C^{(\ell)}_{1\cdots
N_{s}}(\mu_{1};\\{\zeta_{j}\\})\cdots C^{(\ell)}_{1\cdots
N_{s}}(\mu_{n};\\{\zeta_{j}\\})B^{(\ell)}_{1\cdots
N_{s}}(\lambda_{1};\\{\zeta_{j}\\})\cdots B^{(\ell)}_{1\cdots
N_{s}}(\lambda_{n};\\{\zeta_{j}\\})\,|0\rangle$ (6.5) $\displaystyle=$
$\displaystyle\langle 0|\left(P_{1\cdots
L}^{(\ell)\,{\bar{\chi}}}C^{(1)}_{1\cdots
L}(\mu_{1};\\{\xi_{j}^{(\ell)}\\})P_{1\cdots
L}^{(\ell)\,{\bar{\chi}}}\right)\,\cdots\,\left(P_{1\cdots
L}^{(\ell)\,{\bar{\chi}}}C^{(1)}_{1\cdots
L}(\mu_{n};\\{\xi_{j}^{(\ell)}\\})P_{1\cdots L}^{(\ell)\,{\bar{\chi}}}\right)$
$\displaystyle\quad\cdot\,\left(P_{1\cdots
L}^{(\ell)\,{\bar{\chi}}}B^{(1)}_{1\cdots
L}(\lambda_{1};\\{\xi_{j}^{(\ell)}\\})P_{1\cdots
L}^{(\ell)}\right)\,\cdots\left(P_{1\cdots
L}^{(\ell)\,{\bar{\chi}}}B^{(1)}_{1\cdots
L}(\lambda_{n};\\{\xi_{j}^{(\ell)}\\})P_{1\cdots
L}^{(\ell)}\right)\,|0\rangle$ $\displaystyle=$ $\displaystyle\langle
0|P_{1\cdots L}^{(\ell)\,{\bar{\chi}}}\,\times\,C^{(1)}_{1\cdots
L}(\mu_{1};\\{\xi_{j}^{(\ell)}\\})\,\cdots\,C^{(1)}_{1\cdots
L}(\mu_{n};\\{\xi_{j}^{(\ell)}\\})$ $\displaystyle\quad\cdot\,B^{(1)}_{1\cdots
L}(\lambda_{1};\\{\xi_{j}^{(\ell)}\\})\,\cdots B^{(1)}_{1\cdots
L}(\lambda_{n};\\{\xi_{j}^{(\ell)}\\})\,\times\,P_{1\cdots
L}^{(\ell)\,{\bar{\chi}}}\,|0\rangle$
Here we note that we have $\langle 0|P_{1\cdots L}^{(\ell)}=\langle 0|$ and
$P_{1\cdots L}^{(\ell)}|0\rangle=|0\rangle$. Moreover, we have $\langle
0|P_{1\cdots L}^{(\ell)\,{\bar{\chi}}}=\langle 0|$ and $P_{1\cdots
L}^{(\ell)\,{\bar{\chi}}}|0\rangle=|0\rangle$. We thus have
$\displaystyle\langle 0|\,C^{(\ell)}_{1\cdots
N_{s}}(\mu_{1};\\{\zeta_{j}\\})\cdots C^{(\ell)}_{1\cdots
N_{s}}(\mu_{n};\\{\zeta_{j}\\})B^{(\ell)}_{1\cdots
N_{s}}(\lambda_{1};\\{\zeta_{j}\\})\cdots B^{(\ell)}_{1\cdots
N_{s}}(\lambda_{n};\\{\zeta_{j}\\})\,|0\rangle$ $\displaystyle=$
$\displaystyle\langle 0|C^{(1)}_{1\cdots
L}(\mu_{1};\\{\xi_{j}^{(\ell)}\\})\,\cdots\,C^{(1)}_{1\cdots
L}(\mu_{n};\\{\xi_{j}^{(\ell)}\\})\,\cdot\,B^{(1)}_{1\cdots
L}(\lambda_{1};\\{\xi_{j}^{(\ell)}\\})\,\cdots B^{(1)}_{1\cdots
L}(\lambda_{n};\\{\xi_{j}^{(\ell)}\\})\,|0\rangle\,$
We evaluate the last line through the following limit of sending $\epsilon$ to
zero:
$\displaystyle\langle 0|C^{(1)}_{1\cdots
L}(\mu_{1};\\{\xi_{j}^{(\ell)}\\})\,\cdots\,C^{(1)}_{1\cdots
L}(\mu_{n};\\{\xi_{j}^{(\ell)}\\})\,\cdot\,B^{(1)}_{1\cdots
L}(\lambda_{1};\\{\xi_{j}^{(\ell)}\\})\,\cdots B^{(1)}_{1\cdots
L}(\lambda_{n};\\{\xi_{j}^{(\ell)}\\})\,|0\rangle\,$ (6.7) $\displaystyle=$
$\displaystyle\lim_{\epsilon\rightarrow 0}\langle 0|C^{(1)}_{1\cdots
L}(\mu_{1};\\{\xi_{j}^{(\ell;\epsilon)}\\})\cdots C^{(1)}_{1\cdots
L}(\mu_{n};\\{\xi_{j}^{(\ell;\epsilon)}\\})$ $\displaystyle\qquad\times
B^{(1)}_{1\cdots L}(\lambda_{1};\\{\xi_{j}^{(\ell;\epsilon)}\\})\,\cdots
B^{(1)}_{1\cdots L}(\lambda_{n};\\{\xi_{j}^{(\ell;\epsilon)}\\})\,|0\rangle$
$\displaystyle=$ $\displaystyle\lim_{\epsilon\rightarrow 0}\Bigg{[}\langle
0|{\widetilde{C}}^{(1)}_{1\cdots
L}(\mu_{1};\\{\xi_{j}^{(\ell;\epsilon)}\\})\cdots{\widetilde{C}}^{(1)}_{1\cdots
L}(\mu_{n};\\{\xi_{j}^{(\ell;\epsilon)}\\})$
$\displaystyle\qquad\quad\times\,{\widetilde{B}}^{(1)}_{1\cdots
L}(\lambda_{1};\\{\xi_{j}^{(\ell;\epsilon)}\\})\,\cdots\,{\widetilde{B}}^{(1)}_{1\cdots
L}(\lambda_{n};\\{\xi_{j}^{(\ell;\epsilon)}\\})\Bigg{]}\,|0\rangle\,$
∎
We evaluate the spin-1/2 scalar product taking the limit of sending $\epsilon$
to 0, so that we can make the determinant of $F_{L\cdots 21}$ being nonzero.
Here we remark that the operator $F_{L\cdots 21}$ appears in the pseudo-
diagonalization process of the $B$ and $C$ operators, as shown in Section 5,
and also that the determinant of $F_{L\cdots 21}$ vanishes at $\epsilon=0$,
when parameters $\xi_{j}$ are given by eq. (6.1). In fact, if we put some
inhomogeneous parameters $\xi_{j}$ in the form of a “complete $\ell$-string”
[29], that is, for some integers $\ell$, $m$ and a constant $z$, we have
$\xi_{m+j}=z-j\eta$ for $j=1,2,\ldots,\ell$, then the determinant of
$F_{L\cdots 21}$ vanishes. Here we also note that ${\rm det}F_{12\cdots L}\neq
0$ even at $\epsilon=0$.
Let us discuss the mixed spin case. We set the inhomogeneous parameters as
follows:
$\xi_{\ell_{1}+\cdots\ell_{k-1}+j}^{({\mbox{\boldmath$\ell$}};\epsilon)}=\zeta_{k}-(j-1)\eta+(\ell-1)\eta/2+\epsilon
r_{j}\qquad j=1,\ldots\ell;\,k=1,\ldots,r\,.$ (6.8)
Let us define $P^{({\mbox{\boldmath$\ell$}})}_{12\cdots L}$ and
$P^{({\mbox{\boldmath$\ell$}})\,{\bar{\chi}}}_{12\cdots L}$ by
$\displaystyle P^{({\mbox{\boldmath$\ell$}})}_{12\cdots L}$ $\displaystyle=$
$\displaystyle\prod_{k=1}^{r}P^{(\ell_{k})}_{\ell(k-1)+1}$ $\displaystyle
P^{({\mbox{\boldmath$\ell$}})\,{\bar{\chi}}}_{12\cdots L}$ $\displaystyle=$
$\displaystyle\prod_{k=1}^{r}\left(P^{(\ell_{k})}_{\ell(k-1)+1}\right)^{\bar{\chi}}$
(6.9)
It is easy to see the following:
$\displaystyle\langle 0|P^{({\mbox{\boldmath$\ell$}})}_{12\cdots L}$
$\displaystyle=$ $\displaystyle\langle 0|\,,\quad
P^{({\mbox{\boldmath$\ell$}})}_{12\cdots L}|0\rangle=|0\rangle\,,$
$\displaystyle\langle 0|P^{({\mbox{\boldmath$\ell$}})\,{\bar{\chi}}}_{12\cdots
L}$ $\displaystyle=$ $\displaystyle\langle 0|\,,\quad
P^{({\mbox{\boldmath$\ell$}})\,{\bar{\chi}}}_{12\cdots L}|0\rangle=|0\rangle$
(6.10)
We now define the scalar product for the mixed spin case as follows.
$S^{({\mbox{\boldmath$\ell$}})}_{n}(\\{\mu_{j}\\},\\{\lambda_{\alpha}\\};\\{\zeta_{k}\\})=\langle
0|\,C^{({\mbox{\boldmath$\ell$}})}(\mu_{1})\cdots
C^{({\mbox{\boldmath$\ell$}})}(\mu_{n})B^{({\mbox{\boldmath$\ell$}})}(\lambda_{1})\cdots
B^{({\mbox{\boldmath$\ell$}})}(\lambda_{n})\,|0\rangle$ (6.11)
Here $C^{({\mbox{\boldmath$\ell$}})}(\mu_{j})$ and
$B^{({\mbox{\boldmath$\ell$}})}(\lambda_{\alpha})$ abbreviate
$C^{({\mbox{\boldmath$\ell$}})}_{1\cdots r}(\mu_{j};\\{\zeta_{j}\\})$ and
$B^{({\mbox{\boldmath$\ell$}})}_{1\cdots
r}(\lambda_{\alpha};\\{\zeta_{j}\\})$, respectively.
###### Proposition 35.
Let $\\{\lambda_{\alpha}\\}$ satisfy the Bethe ansatz equations for the mixed
spin case. The scalar product of the mixed-spin XXZ spin chain is reduced into
that of the spin-1/2 XXZ spin chain as follows:
$S^{({\mbox{\boldmath$\ell$}})}_{n}(\\{\mu_{j}\\},\\{\lambda_{\alpha}\\};\\{\zeta_{k}\\})=\lim_{\epsilon\rightarrow
0}\Bigg{[}S^{(1)}_{n}(\\{\mu_{j}\\},\\{\lambda_{\alpha}\\};\\{\xi^{({\mbox{\boldmath$\ell$}};\epsilon)}_{k}\\})\Bigg{]}$
(6.12)
### 6.2 Determinant expressions of the scalar products
Let us review the result of the spin-1/2 case [7]. Suppose that
$\lambda_{\alpha}$ for $\alpha=1,2,\ldots,n$, are solutions of the Bethe
ansatz equations with in homogeneous parameters $\xi_{j}$ for
$j=1,2,\ldots,L$, the scalar product is defined by
$S_{n}(\\{\mu_{j}\\},\,\\{\lambda_{\alpha}\\};\\{\xi_{k}\\})=\langle
0|\prod_{j=1}^{n}C(\mu_{j};\\{\xi_{k}\\})\prod_{\alpha=1}^{n}B(\lambda_{\alpha};\\{\xi_{k}\\})|0\rangle$
(6.13)
Here $\mu_{j}$ for $j=1,2,\ldots,n$ are arbitrary. We note that
$S_{n}(\\{\mu_{j}\\},\,\\{\lambda_{\alpha}\\};\\{\xi_{k}\\})$ has been denoted
by $S_{n}^{(1)}(\\{\mu_{j}\\},\,\\{\lambda_{\alpha}\\};\\{\xi_{k}\\})$ in the
last subsection. Then, the exact expression of the scalar product has been
shown through the pseudo-diagonalized forms of the $B$ and $C$ operators as
follows [7]:
$\displaystyle
S_{n}(\\{\mu_{j}\\},\,\\{\lambda_{\alpha}\\};\\{\xi_{k}\\})=\langle
0|\prod_{j=1}^{n}\widetilde{C}(\mu_{j})\prod_{\alpha=1}^{n}\widetilde{B}(\lambda_{\alpha})|0\rangle$
(6.14) $\displaystyle=$
$\displaystyle{\frac{\prod_{\alpha=1}^{n}\prod_{j=1}^{n}\sinh(\mu_{j}-\lambda_{\alpha})}{\prod_{j>k}\sinh(\mu_{k}-\mu_{j})\prod_{\alpha<\beta}\sinh(\lambda_{\beta}-\lambda_{\alpha})}}\,{\rm
det}T(\\{\mu_{j}\\},\,\\{\lambda_{\alpha}\\};\\{\xi_{k}\\})$
Here the matrix elements $T_{ab}$ for $a,b=1,\ldots,n$, are given by
$T_{ab}={\frac{\partial}{\partial\lambda_{\alpha}}}\tau(\mu_{b},\\{\lambda_{k}\\};\\{\xi_{k}\\})$
(6.15)
where
$\displaystyle\tau(\mu,\\{\lambda_{k}\\};\\{\xi_{k}\\})$ $\displaystyle=$
$\displaystyle
a(\mu)\prod_{k=1}^{n}b^{-1}(\lambda_{k}-\mu)+d(\mu;\\{\xi_{j}\\};\\{\xi_{k}\\})\prod_{k=1}^{n}b^{-1}(\mu-\lambda_{k})$
and
$a(\mu)=1\,,\quad d(\mu;\\{\xi_{j}\\})=\prod_{j=1}^{L}b(\mu-\xi_{j})\,.$
(6.17)
Let us express the scalar product of the higher-spin case in terms of the
determinant of the matrix $T$. In the tensor product of spin-$\ell/2$
representations we have
$\displaystyle
S^{(\ell)}_{n}(\\{\mu_{j}\\},\,\\{\lambda_{\alpha}\\};\\{\zeta_{k}\\})$ (6.18)
$\displaystyle=$
$\displaystyle{\frac{\prod_{\alpha=1}^{n}\prod_{j=1}^{n}\sinh(\mu_{j}-\lambda_{\alpha})}{\prod_{j>k}\sinh(\mu_{k}-\mu_{j})\prod_{\alpha<\beta}\sinh(\lambda_{\beta}-\lambda_{\alpha})}}\,{\rm
det}T(\\{\mu_{j}\\},\,\\{\lambda_{\alpha}\\};\\{\xi_{k}^{(\ell)}\\})$
In the mixed-spin case we have
$\displaystyle
S^{({\mbox{\boldmath$\ell$}})}_{n}(\\{\mu_{j}\\},\,\\{\lambda_{\alpha}\\};\\{\zeta_{k}\\})$
(6.19) $\displaystyle=$
$\displaystyle{\frac{\prod_{\alpha=1}^{n}\prod_{j=1}^{n}\sinh(\mu_{j}-\lambda_{\alpha})}{\prod_{j>k}\sinh(\mu_{k}-\mu_{j})\prod_{\alpha<\beta}\sinh(\lambda_{\beta}-\lambda_{\alpha})}}\,{\rm
det}T(\\{\mu_{j}\\},\,\\{\lambda_{\alpha}\\};\\{\xi_{k}^{({\mbox{\boldmath$\ell$}})}\\})$
### 6.3 Norms of the Bethe states in the higher-spin case
For two sets of $n$ parameters, $\mu_{1},\ldots,\mu_{n}$ and
$\lambda_{1},\ldots,\lambda_{n}$, we define matrix elements $H_{a}b$ by
$H_{ab}(\\{\mu_{j}\\},\\{\lambda_{\alpha};\\{\xi_{k}\\})={\frac{\sinh\eta}{\sinh(\lambda_{a}-\mu_{b})}}\left({\frac{a(\mu_{b})}{d(\mu_{b})}}\prod_{k=1;\neq
a}^{n}\sinh(\lambda_{k}-\mu_{b}+\eta)-\prod_{k=1;\neq
a}^{n}\sinh(\lambda_{k}-\mu_{b}-\eta)\right)$ (6.20)
Let us assume that $\lambda_{1},\ldots,\lambda_{n}$ are solutions of the Bethe
ansatz equations. We have
${\rm det}H(\\{\lambda_{\alpha}\\},\\{\mu_{j}\\};\\{\xi_{k}\\})={\rm
det}T(\\{\lambda_{\alpha}\\},\\{\mu_{j}\\};\\{\xi_{k}\\})\prod_{\alpha=1}^{n}\prod_{j=1}^{n}\sinh(\mu_{j}-\lambda_{\alpha})\left(\prod_{j=1}^{n}d(\mu_{j})\right)^{-1}$
(6.21)
Let us now take the limit of sending $\mu_{j}$ to $\lambda_{j}$ for each $j$.
Then we have
$\lim_{\mu_{j}\rightarrow\lambda_{j}}detH(\\{\lambda_{\alpha}\\},\\{\mu_{j}\\};\\{\xi_{k}\\})=\sinh^{n}\eta\prod_{\beta=1}^{n}\prod_{m=1;m\neq\beta}^{n}\sinh(\lambda_{m}-\lambda_{\beta}-\eta)\cdot
det\Phi^{{}^{\prime}}(\\{\lambda_{\alpha}\\})$ (6.22)
where matrix elements $\Phi^{{}^{\prime}}_{ab}$ for $a,b=1,\ldots,n$, are
given by
$\Phi^{{}^{\prime}}_{ab}(\\{\lambda_{\alpha}\\};\\{\xi\\})=-{\frac{\partial}{\partial\lambda_{b}}}\left({\frac{a(\lambda_{a};\\{\xi_{j}\\})}{d(\lambda_{b};\\{\xi_{j}\\})}}\prod_{k=1;k\neq
a}^{n}{\frac{b(\lambda_{a}-\lambda_{k})}{b(\lambda_{k}-\lambda_{a})}}\right)$
(6.23)
Suppose that $\lambda_{\alpha}$ for $n=1,2,\ldots,n$ are solutions of the
Bethe ansatz equations. Gaudin’s formula for the square of he norm of the
Bethe state is given by
$\displaystyle{\mbox{\boldmath$N$}}_{n}(\\{\lambda_{\alpha}\\};\\{\xi_{j}\\})$
$\displaystyle=$ $\displaystyle\langle
0|\prod_{j=1}^{n}C(\lambda_{j})\prod_{j=1}^{n}B(\lambda_{k})|0\rangle$ (6.24)
$\displaystyle=$
$\displaystyle\sinh^{n}\eta\prod_{\alpha,\beta=1;\alpha\neq\beta}^{n}b^{-1}(\lambda_{\alpha}-\lambda_{\beta})\cdot{\rm
det}\Phi^{{}^{\prime}}({\lambda_{\alpha}};\\{\xi_{j}\\})$
Let us define the norm of the Bethe state for the mixed spin case of $\ell$ as
follows.
${\mbox{\boldmath$N$}}_{n}^{({\mbox{\boldmath$\ell$}})}(\\{\lambda_{\alpha}\\};\\{\xi^{({\mbox{\boldmath$\ell$}})}_{k}\\})=\langle
0|\prod_{j=1}^{n}C^{({\mbox{\boldmath$\ell$}})}(\lambda_{j})\prod_{j=1}^{n}B^{({\mbox{\boldmath$\ell$}})}(\lambda_{k})|0\rangle$
(6.25)
Then we have
${\mbox{\boldmath$N$}}_{n}^{({\mbox{\boldmath$\ell$}})}(\\{\lambda_{\alpha}\\};\\{\xi^{({\mbox{\boldmath$\ell$}})}_{k}\\})=\sinh^{n}\eta\prod_{\alpha,\beta=1;\alpha\neq\beta}^{n}b^{-1}(\lambda_{\alpha}-\lambda_{\beta})\cdot{\rm
det}\Phi^{{}^{\prime}}({\lambda_{\alpha}};\\{\xi_{j}^{\mbox{\boldmath$\ell$}}\\})$
(6.26)
## 7 Form factors and the inverse-scattering for the higher-spin case
### 7.1 Formulas of the quantum inverse scattering problem
Let us briefly review the derivation of the fundamental lemma of the quantum-
inverse scattering problem for the spin-1/2 XXZ spin chain [6]. It will thus
becomes clear how the pseudo-diagonalization of $B$ and $C$ are important.
###### Proposition 36.
Let us denote by $p_{q}$ sequence $p_{q}=(1,2,\ldots,n)$. Recall the notation
$\widetilde{R}_{0,p_{q}}=F_{12\cdots n}R_{0,12\cdots n}F^{-1}_{12\cdots n}$.
Then, $\widetilde{R}_{0,p_{q}}$ is invariant under any permutation. We have
$\widetilde{R}_{0,p_{q}}=\widetilde{R}_{0,\sigma(p_{q})}\,,\quad{\rm
for}\,\,\sigma\in{\cal S}_{n}\,.$ (7.1)
We shall show (7.1) in Appendix B.
The following lemma plays a central role in the quantum inverse-scattering
problem [7].
###### Lemma 37.
For arbitrary inhomogeneous parameters $\xi_{1},\xi_{2},\ldots,\xi_{L}$ we
have
$x_{i}=\prod_{\alpha=1}^{i-1}(A+D)(\xi_{\alpha})\,{\rm
tr}_{0}(x_{0}R_{0,1\cdots
L}(\xi_{i}))\,\prod_{\alpha=i+1}^{L}(A+D)(\xi_{\alpha})$ (7.2)
###### Proof.
For any operator $x_{0}$ defined on the auxiliary space, we have
$\displaystyle{\rm tr}_{0}(x_{0}R_{0,12\cdots L}(\lambda=\xi_{i}))$
$\displaystyle=$ $\displaystyle F_{12\cdots L}^{-1}{\rm
tr}_{0}(x_{0}\widetilde{R}_{0,12\cdots L}(\lambda=\xi_{i}))F_{12\cdots L}$
$\displaystyle=$ $\displaystyle F_{12\cdots L}^{-1}{\rm
tr}_{0}(x_{0}\widetilde{R}_{i,i+1\cdots L1\cdots
i-1}(\lambda=\xi_{i}))F_{12\cdots L}$ $\displaystyle=$ $\displaystyle
F_{12\cdots L}^{-1}F_{i\cdots L1\cdots i-1}\,{\rm tr}_{0}(x_{0}R_{i,i+1\cdots
L1\cdots i-1}(\lambda=\xi_{i}))\,F_{i\cdots L1\cdots i-1}^{-1}F_{12\cdots L}$
$\displaystyle=$ $\displaystyle\left(F_{i\cdots L1\cdots i-1}^{-1}F_{12\cdots
L}\right)^{-1}x_{i}(A(\xi_{i})+D(\xi_{i}))\,\left(F_{i\cdots L1\cdots
i-1}^{-1}F_{12\cdots L}\right)$
Here we have used (7.1), i.e. $\widetilde{R}_{0,12\cdots
L}=\widetilde{R}_{i,i+1\cdots L1\cdots i-1}$. From the expression of
$F^{-1}_{i\cdots L\,1\cdots i-1}F_{1\cdots L}$ we now have
${\rm tr}_{0}(x_{0}R_{0,1\cdots
L})=\prod_{\alpha=1}^{i-1}\left((A+D)(\xi_{\alpha})\right)^{-1}\cdot
x_{i}\cdot\prod_{\alpha=1}^{i}(A+D)(\xi_{\alpha})\,.$ (7.3)
∎
### 7.2 Quantum inverse-scattering problem for the higher-spin operators
Let us consider monodromy matrix $T^{+}_{0,1\cdots\ell N_{s}}$. Here we recall
$L=\ell N_{s}$. For simplicity, we shall suppress the superscript ‘$+$’ for
$A,B,C$ and $D$ operators through this subsection.
We recall the following: $\Delta^{(n-1)}(K)=K^{\otimes n}$ and
$\displaystyle\Delta^{(n-1)}(X^{+})$ $\displaystyle=$
$\displaystyle\sum_{j=1}^{n}K^{\otimes(j-1)}\otimes X^{+}_{j}\otimes
I^{\otimes(n-j)}\,,$ $\displaystyle\Delta^{(n-1)}(X^{-})$ $\displaystyle=$
$\displaystyle\sum_{j=1}^{n}I^{\otimes(j-1)}\otimes
X^{-}_{j}\otimes\left(K^{-1}\right)^{\otimes(n-j)}\,.$ (7.4)
It is useful to note that for $i=1,2,\ldots,\ell N_{s}$ we have
$K_{i}=\prod_{\alpha=1}^{i-1}(A+D)(\xi_{\alpha})(qA+q^{-1}D)(\xi_{i})\prod_{\alpha=i+1}^{\ell
N_{s}}(A+D)(\xi_{\alpha})\,.$ (7.5)
In the tensor product of $2N_{s}$ spin-1/2 representations,
$\left(V_{1}^{(1)}\otimes V_{2}^{(1)}\right)^{\otimes N_{s}}$, we have
$\displaystyle\Delta_{12}(X^{-})$ $\displaystyle=$
$\displaystyle\left(X^{-}_{1}\otimes K^{-1}_{2}+I_{1}\otimes
X^{-}_{2}\right)\otimes I^{\otimes 2(N_{s}-1)}$ (7.6) $\displaystyle=$
$\displaystyle\Bigg{\\{}\left(B(\xi_{1})\cdot\prod_{\alpha=2}^{2N_{s}}(A+D)(\xi_{\alpha})\right)\cdot\bigg{(}(A+D)(\xi_{1})(q^{-1}A+qD)(\xi_{2})\prod_{\alpha=3}^{2N_{s}}(A+D)(\xi_{\alpha})\bigg{)}$
$\displaystyle\quad+(A+D)(\xi_{1})\cdot
B(\xi_{2})\cdot\prod_{\alpha=3}^{2N_{s}}(A+D)(\xi_{\alpha})\Bigg{\\}}$
In the $\ell$th tensor product of spin-1/2 representations,
$V_{1}^{(1)}\otimes\cdots\otimes V_{\ell}^{(1)}$, we have
$P_{1\cdots\ell}^{(\ell)}\,\cdot\,\Delta^{(\ell-1)}(X^{\pm})\,\cdot\,P_{1\cdots\ell}^{(\ell)}=P_{1\cdots\ell}^{(\ell)}\,\cdot\,\Delta^{(\ell-1)}(X^{\pm})=\Delta^{(\ell-1)}(X^{\pm})\,\cdot\,P_{1\cdots\ell}^{(\ell)}$
(7.7)
In the tensor product of spin-$\ell/2$ representations,
$V_{1}^{(\ell)}\otimes\cdots\otimes V_{N_{s}}^{(\ell)}$, we have for
$i=1,2,\ldots,N_{s}$ the following relations:
$\displaystyle X_{i}^{-(\ell+)}=\Delta^{(\ell-1)}_{(i-1)\ell+1\,\cdots
i\ell}(X^{-})$
$\displaystyle\quad=\sum_{k=1}^{\ell}\prod_{\alpha=1}^{(i-1)\ell+k-1}(A+D)(\xi_{\alpha})\cdot
B(\xi_{(i-1)\ell+k})\cdot\prod_{\alpha=(i-1)\ell+k+1}^{\ell
N_{s}}(A+D)(\xi_{\alpha})$
$\displaystyle\quad\times\prod_{j=k+1}^{\ell}\prod_{\alpha=1}^{(i-1)\ell+j-1}(A+D)(\xi_{\alpha})\cdot(q^{-1}A+qD)(\xi_{(i-1)\ell+j})\cdot\prod_{\alpha=(i-1)\ell+j+1}^{\ell
N_{s}}(A+D)(\xi_{\alpha})$
$\displaystyle=\sum_{k=1}^{\ell}\prod_{\alpha=1}^{(i-1)\ell+k-1}(A+D)(\xi_{\alpha})\cdot
B(\xi_{(i-1)\ell+k})\cdot$
$\displaystyle\quad\times\prod_{j=k+1}^{\ell}(q^{-1}A+qD)(\xi_{(i-1)\ell+j})\cdot\prod_{\alpha=i\ell+1}^{\ell
N_{s}}(A+D)(\xi_{\alpha})\,.$ (7.8)
Here we have made use of the following:
$\prod_{\alpha=1}^{\ell N_{s}}(A+D)(\xi_{\alpha})=I^{\otimes\ell N_{s}}\,.$
(7.9)
Similarly, we can express $X_{i}^{+(\ell+)}$ and $K_{i}^{(\ell+)}$ as follows.
$\displaystyle X_{i}^{+(\ell+)}=\Delta^{(\ell-1)}_{(i-1)\ell+1\,\cdots
i\ell}(X^{+})$
$\displaystyle=\sum_{k=1}^{\ell}\prod_{\alpha=1}^{(i-1)\ell}(A+D)(\xi_{\alpha})\cdot\prod_{j=1}^{k-1}(qA+q^{-1}D)(\xi_{(i-1)\ell+j})\cdot
C(\xi_{(i-1)\ell+k})$ $\displaystyle\quad\prod_{\alpha=(i-1)\ell+k+1}^{\ell
N_{s}}(A+D)(\xi_{\alpha})\,.$ (7.10)
$\displaystyle K_{i}^{(\ell+)}=\Delta^{(\ell-1)}_{(i-1)\ell+1\,\cdots
i\ell}(K)$
$\displaystyle=\prod_{\alpha=1}^{(i-1)\ell}(A+D)(\xi_{\alpha})\,\prod_{j=1}^{\ell}(qA+q^{-1}D)(\xi_{(i-1)\ell+j})\,\prod_{\alpha=i\ell+1}^{\ell
N_{s}}(A+D)(\xi_{\alpha})\,.$ (7.11)
### 7.3 Useful formulas in the higher-spin case
Let us denote by $X^{\pm(\ell)}$ the matrix representations of generators
$X^{\pm}$ in the spin-$\ell/2$ representation of $U_{q}(sl_{2})$. Here we
recall that the matrix representations of $X^{\pm(\ell)}$ are obtained by
calculating the action of $\Delta^{(\ell-1)}(X^{\pm})$ on the basis
$\\{||\ell,n\rangle\\}$.
We explicitly calculate the actions of $\sigma_{1}^{-}=\sigma^{-}\otimes
I^{\otimes(\ell-1)}$ and
$\sigma_{\ell}^{+}=I^{\otimes(\ell-1)}\otimes\sigma^{+}$ on the basis
$\\{||\ell,n\rangle\\}$ in the spin-$\ell/2$ representation. Multiplying
projection operators to them, we obtain the following formulas:
$\displaystyle
P^{(\ell)}_{1\cdots\ell}\,\sigma_{1}^{-}\,P^{(\ell)}_{1\cdots\ell}$
$\displaystyle=$ $\displaystyle{\frac{1}{[\ell]_{q}}}\,X^{-(\ell+)}$
$\displaystyle
P^{(\ell)}_{1\cdots\ell}\,\sigma_{\ell}^{+}\,P^{(\ell)}_{1\cdots\ell}$
$\displaystyle=$ $\displaystyle{\frac{1}{[\ell]_{q}}}\,X^{-(\ell+)}$ (7.12)
Therefore, we have for $i=1,2,\ldots,N_{s}$, the following formulas:
$\displaystyle
P^{(\ell)}_{(i-1)\ell+1}\,X_{i}^{-(\ell+)}\,P^{(\ell)}_{(i-1)\ell+1}$
$\displaystyle={[}\ell{]}_{q}P^{(\ell)}_{(i-1)\ell+1}\,\prod_{\alpha=1}^{(i-1)\ell}(A^{+}+D^{+})(\xi_{\alpha})\cdot
B^{+}(\xi_{(i-1)\ell+1})\cdot\prod_{\alpha=(i-1)\ell+2}^{\ell
N_{s}}(A^{+}+D^{+})(\xi_{\alpha})\,P^{(\ell)}_{(i-1)\ell+1}$ (7.13)
$\displaystyle
P^{(\ell)}_{(i-1)\ell+1}\,X_{i}^{+(\ell+)}\,P^{(\ell)}_{(i-1)\ell+1}$
$\displaystyle=[\ell]_{q}P^{(\ell)}_{(i-1)\ell+1}\,\prod_{\alpha=1}^{i\ell-2}(A^{+}+D^{+})(\xi_{\alpha})\cdot
C^{+}(\xi_{i\ell-1})\cdot\prod_{\alpha=i\ell}^{\ell
N_{s}}(A^{+}+D^{+})(\xi_{\alpha})\,P^{(\ell)}_{(i-1)\ell+1}\,.$ (7.14)
Taking advantage of projection operators, we thus have shown that the
summation over $k$ arising from the $(\ell-1)$th comultiplication operation
can be calculated by a single term. This reduces the calculational task very
much.
In the derivation of (7.13), we first note
$\chi_{1\cdots L}\sigma_{(i-1)\ell+1}^{-}\chi_{1\cdots
L}^{-1}\exp(-\xi_{(i-1)\ell+1})\,,$ (7.15)
and then we show the following:
$\sigma_{(i-1)\ell+1}^{-}=\prod_{\alpha=1}^{(i-1)\ell}(A^{+}+D^{+})(\xi_{\alpha})\,B^{+}(\xi_{(i-1)\ell+1})\,\prod_{\alpha=(i-1)\ell+2}^{\ell
N_{s}}(A^{+}+D^{+})(\xi_{\alpha})\,.$ (7.16)
We shall show relations (7.12) in Appendix C.
We now introduce useful formulas expressing any given operator in the
spin-$\ell$ representation. Let us take two sets of integers
$i_{1},\ldots,i_{m}$ and $j_{1},\ldots,j_{n}$ satisfying $1\leq
i_{1}<\cdots<i_{m}\leq\ell$ and $1\leq j_{1}<\cdots<j_{n}\leq\ell$,
respectively. We can show the following:
$\displaystyle||\ell,m\rangle\langle\ell,n||$ $\displaystyle=$
$\displaystyle\left[\begin{array}[]{c}\ell\\\
m\end{array}\right]_{q}\,q^{m(m+1)/2-n(n-1)/2+n\ell-(i_{1}+\cdots+i_{m}+j_{1}+\cdots+j_{n})}$
(7.20)
$\displaystyle\times\,P_{1\cdots\ell}^{(\ell)}\left(\prod_{k=1}^{m}e_{i_{k}}^{21}\cdot\prod_{p=1;p\neq
i_{k},j_{q}}^{\ell}e_{p}^{22}\cdot\prod_{q=1}^{n}e_{j_{q}}^{12}\right)P_{1\cdots\ell}^{(\ell)}$
Setting $i_{1}=1,i_{2}=2$, …, $i_{m}=m$ and $j_{1}=1,j_{2}=2$, …, $j_{n}=n$,
we have for $m>n$
$||\ell,m\rangle\langle\ell,n||=\left[\begin{array}[]{c}\ell\\\
n\end{array}\right]_{q}\,q^{(\ell-n)}\,P_{1\cdots\ell}^{(\ell)}\left(\prod_{k=1}^{n}e_{k}^{22}\,\prod_{k=n+1}^{m}e_{k}^{21}\,\prod_{k=m+1}^{\ell}e_{k}^{11}\right)P_{1\cdots\ell}^{(\ell)}$
(7.21)
and for $m<n$
$||\ell,m\rangle\langle\ell,n||=\left[\begin{array}[]{c}\ell\\\
n\end{array}\right]_{q}\,q^{(\ell-n)}\,P_{1\cdots\ell}^{(\ell)}\left(\prod_{k=1}^{m}e_{k}^{22}\,\prod_{k=m+1}^{n}e_{k}^{12}\,\prod_{k=n+1}^{\ell}e_{k}^{11}\right)P_{1\cdots\ell}^{(\ell)}$
(7.22)
and for $m=n$
$||\ell,n\rangle\langle\ell,n||=\left[\begin{array}[]{c}\ell\\\
n\end{array}\right]_{q}\,q^{(\ell-n)}\,P_{1\cdots\ell}^{(\ell)}\left(\prod_{k=1}^{n}e_{k}^{22}\,\prod_{k=n+1}^{\ell}e_{k}^{11}\right)P_{1\cdots\ell}^{(\ell)}$
(7.23)
Let us denote by $E^{mn(\ell+)}$ the unit matrices acting on spin-$\ell$
representation $V^{(\ell)}$ for $m,n=0,1,\ldots,\ell$. We now define
$E_{i}^{mn(\ell+)}$ by the unit matrices acting on the $i$th component of the
tensor product $(V^{(\ell)})^{\otimes N_{s}}$. Explicitly, we have
$E_{i}^{mn\,(\ell+)}=(I^{(\ell)})^{\otimes(i-1)}\otimes
E^{mn}\otimes(I^{(\ell)})^{\otimes(N_{s}-i)}$ (7.24)
where $I^{(\ell)}$ denotes the $(\ell+1)\times(\ell+1)$ identity matrix. Then,
we derive the following formulas from (7.21), (7.22) and (7.23), respectively.
For $m>n$ we have
$\displaystyle E_{i}^{mn\,(\ell+)}$ $\displaystyle=$
$\displaystyle\left[\begin{array}[]{c}\ell\\\
m\end{array}\right]_{q}\,q^{n(\ell-n)}\,P^{(\ell)}_{1\cdots
L}\,\prod_{\alpha=1}^{(i-1)\ell}(A+D)(\xi_{\alpha})\prod_{k=1}^{n}D(\xi_{(i-1)\ell+k})\prod_{k=n+1}^{m}B(\xi_{(i-1)\ell+k})$
(7.28)
$\displaystyle\quad\times\,\prod_{k=m+1}^{\ell}A(\xi_{(i-1)\ell+k})\prod_{\alpha=i\ell+1}^{\ell
N_{s}}(A+D)(\xi_{\alpha})\,\,P^{(\ell)}_{1\cdots L}\,.$
For $m<n$ we have
$\displaystyle E_{i}^{mn\,(\ell+)}$ $\displaystyle=$
$\displaystyle\left[\begin{array}[]{c}\ell\\\
m\end{array}\right]_{q}\,q^{n(\ell-n)}\,P^{(\ell)}_{1\cdots
L}\,\prod_{\alpha=1}^{(i-1)\ell}(A+D)(\xi_{\alpha})\prod_{k=1}^{m}D(\xi_{(i-1)\ell+k})\prod_{k=m+1}^{n}C(\xi_{(i-1)\ell+k})$
(7.32)
$\displaystyle\quad\times\,\prod_{k=n+1}^{\ell}A(\xi_{(i-1)\ell+k})\prod_{\alpha=i\ell+1}^{\ell
N_{s}}(A+D)(\xi_{\alpha})\,\,P^{(\ell)}_{1\cdots L}\,.$
For $m=n$ we have
$\displaystyle E_{i}^{nn\,(\ell+)}$ $\displaystyle=$
$\displaystyle\left[\begin{array}[]{c}\ell\\\
n\end{array}\right]_{q}\,q^{n(\ell-n)}\,P^{(\ell)}_{1\cdots
L}\,\prod_{\alpha=1}^{(i-1)\ell}(A+D)(\xi_{\alpha})\prod_{k=1}^{n}D(\xi_{(i-1)\ell+k})$
(7.36)
$\displaystyle\quad\times\,\prod_{k=n+1}^{\ell}A(\xi_{(i-1)\ell+k})\prod_{\alpha=i\ell+1}^{\ell
N_{s}}(A+D)(\xi_{\alpha})\,\,P^{(\ell)}_{1\cdots L}\,.$
Let us now discuss the derivation of formula (7.20). It is easy to show the
following:
$\sigma_{i_{1}}^{-}\cdots\sigma_{i_{m}}^{-}|0\rangle\langle
0|\sigma_{j_{1}}^{+}\cdots\sigma_{j_{n}}^{+}=e_{i_{1}}^{21}\cdots
e_{i_{n}}^{21}\prod_{p=1;p\neq
i_{k},j_{q}}^{\ell}e_{p}^{11}e_{i_{1}}^{12}\cdots e_{i_{n}}^{12}$ (7.37)
Then, making use of expressions (4.20) and (4.25), we obtain (7.20).
### 7.4 Form factors for higher-spin operators
Making use of the fundamental lemma of the quantum inverse-scattering problem,
lemma 7.2, together with the useful formulas given in §7.3 such as (7.13) and
(7.14), and (7.28), (7.32) and (7.36), we can systematically calculate form
factors for the higher-spin cases. Here we note that the form factors
associated with generators $S^{\pm}$ of the spin $SU(2)$ have been derived for
the higher-spin XXX chains [27]. They are derived through the relations
corresponding to (7.8) and (7.10) in the limit of $q=1$.
For an illustration, let us calculate the following form factor:
$F_{n}^{-(\ell+)}(i;\,\\{\mu_{j}\\},\\{\lambda_{k}\\})=\langle
0|\prod_{j=1}^{n+1}C^{(\ell+)}(\mu_{j})\cdot
X_{i}^{-(\ell+)}\prod_{k=1}^{n}B^{(\ell+)}(\lambda_{k})|0\rangle\,,$ (7.38)
where $\\{\mu_{j}\\}$ and $\\{\lambda_{k}\\}$ are solutions of the Bethe
ansatz equations. Putting (7.13) into (7.38) and making use of the fact that
projector $P_{1\cdots L}^{(\ell)}$ commutes with the matrix elements of
$R_{0,1\cdots L}^{+}$, we have
$\displaystyle\langle 0|\prod_{j=1}^{n+1}C^{+}(\mu_{j})\cdot
X_{i}^{-(\ell+)}\prod_{k=1}^{n}B^{+}(\lambda_{k})|0\rangle$ (7.39)
$\displaystyle=$ $\displaystyle[\ell]_{q}\,\langle
0|\prod_{j=1}^{n+1}C^{+}(\mu_{j})\cdot
P_{(i-1)\ell+1}^{(\ell)}\prod_{\alpha=1}^{(i-1)\ell}(A^{+}+D^{+})(\xi_{\alpha})\cdot
B^{+}(\xi_{(i-1)\ell+1})\cdot$
$\displaystyle\quad\times\prod_{\alpha=(i-1)\ell+2}^{\ell
N_{s}}(A^{+}+D^{+})(\xi_{\alpha})P_{(i-1)\ell+1}^{(\ell)}\prod_{k=1}^{n}B^{+}(\lambda_{k})|0\rangle$
$\displaystyle=$
$\displaystyle[\ell]_{q}\,e^{\sum_{j=1}^{n+1}\mu_{j}-\sum_{k=1}^{n}\lambda_{k}-\xi_{(i-1)\ell+1}}{\frac{\phi_{(i-1)\ell}(\\{\mu_{j}\\}_{n+1})}{\phi_{(i-1)\ell+1}(\\{\lambda\\}_{n})}}\langle
0|\prod_{j=1}^{n+1}C(\mu_{j})\cdot
B(\xi_{(i-1)\ell+1})\prod_{k=1}^{n}B(\lambda_{k})|0\rangle$ $\displaystyle=$
$\displaystyle[\ell]_{q}\,e^{\sum_{j=1}^{n+1}\mu_{j}-\sum_{k=1}^{n}\lambda_{k}-\xi_{(i-1)\ell+1}}{\frac{\phi_{(i-1)\ell}(\\{\mu_{j}\\}_{n+1})}{\phi_{(i-1)\ell+1}(\\{\lambda\\}_{n})}}$
$\displaystyle\quad\times
S_{n+1}\left(\\{\mu_{j}\\}_{n+1},\\{\xi_{(i-1)\ell+1},\lambda_{1},\ldots,\lambda_{n}\\};\\{\xi_{k}^{(\ell)}\\}\right)$
Let us define the form factor in the symmetric case as follows.
$F_{n}^{-(\ell)}(i;\,\\{\mu_{j}\\},\\{\lambda_{k}\\})=\langle
0|\prod_{j=1}^{n+1}C(\mu_{j})\cdot\left(X_{i}^{-(\ell+)}\right)^{\bar{\chi}}\prod_{k=1}^{n}B(\lambda_{k})|0\rangle$
(7.40)
Here we recall that $\\{\mu_{j}\\}$ and $\\{\lambda_{k}\\}$ are solutions of
the Bethe ansatz equations. Then, we obtain the following expression.
$\displaystyle
F_{n}^{-(\ell)}(i;\,\\{\mu_{j}\\},\\{\lambda_{k}\\})=[\ell]_{q}\,{\frac{\phi_{(i-1)\ell}(\\{\mu_{j}\\}_{n+1})}{\phi_{(i-1)\ell+1}(\\{\lambda\\}_{n})}}$
$\displaystyle\qquad\times
S_{n+1}\left(\\{\mu_{j}\\}_{n+1},\\{\xi_{(i-1)\ell+1},\lambda_{1},\ldots,\lambda_{n}\\};\\{\xi_{k}^{(\ell)}\\}\right)\,.$
(7.41)
Let us define the form factor for $K$ by
$F^{K(\ell+)}_{n}(i,\,\\{\mu_{j}\\},\\{\lambda_{k}\\})=\langle
0|\prod_{j=1}^{n}C^{(\ell+)}(\mu_{j})\cdot\left(K_{i}^{(\ell+)}\right)^{\bar{\chi}}\prod_{k=1}^{n}B^{(\ell+)}(\lambda_{k})|0\rangle\,.$
(7.42)
Here we recall that $\\{\mu_{j}\\}$ and $\\{\lambda_{k}\\}$ are solutions of
the Bethe ansatz equations. Similarly, we can calculate as follows.
$\displaystyle
F^{K(\ell+)}_{n}(i,\,\\{\mu_{j}\\},\\{\lambda_{k}\\})=\exp\left({\sum_{j=1}^{n}\mu_{j}-\sum_{k=1}^{n}\lambda_{k}}\right){\frac{\phi_{(i-1)\ell}(\\{\mu_{j}\\}_{n})}{\phi_{i\ell}(\\{\lambda\\}_{n})}}$
$\displaystyle\quad\times\sum_{n=0}^{\ell}\left[\begin{array}[]{c}\ell\\\
n\end{array}\right]_{q}\,q^{\ell-2n+n(\ell-n)}\langle
0|\prod_{j=1}^{n}C(\mu_{j})\prod_{k=1}^{n}D(\xi_{(i-1)\ell+k})\prod_{k=n+1}^{\ell}A(\xi_{(i-1)\ell+k})\prod_{k=1}^{n}B(\lambda_{k})|0\rangle$
(7.45)
Through the commutation relation between $A$ and $B$ and that between $C$ and
$D$, the vacuum expectation of the product of $C$, $D$, $A$ and $B$ operators
can be expressed in terms of the sum of scalar products.
## Appendix A Derivation of symmetry relations for $R_{p}^{\sigma}$
###### Lemma A.1.
Let $p$ be a sequence of $n$ integers, $1,2,\ldots,n$. For any
$\sigma_{A},\sigma_{B}\in{\cal S}_{n}$ we have
$(\sigma_{A}\sigma_{B})\,p=\sigma_{B}(\sigma_{A}p)\,.$ (A.1)
###### Proof.
Let us denote $p_{\sigma_{A}i}$ by $q_{i}$ for $i=1,2,\ldots,n$. We thus have
$\displaystyle\sigma_{B}(q_{1},\ldots,q_{n})$ $\displaystyle=$
$\displaystyle(q_{\sigma_{B}1},\ldots,q_{\sigma_{B}n})$ (A.2) $\displaystyle=$
$\displaystyle(p_{\sigma_{A}(\sigma_{B}1)},\ldots,p_{\sigma_{A}(\sigma_{B}n)})$
$\displaystyle=$
$\displaystyle(p_{\sigma_{A}\sigma_{B}\,1},\ldots,p_{\sigma_{A}\sigma_{B}\,n})$
$\displaystyle=$ $\displaystyle(\sigma_{A}\sigma_{B})\,p$
∎
###### Proposition A.1.
Definition 2.21 is well defined. That is, for
$R_{j,k}=R_{j,k}(\lambda_{j},\lambda_{k})$ the following relations hold:
$\displaystyle R_{p}^{(\sigma_{A}\sigma_{B})\sigma_{C}}$ $\displaystyle=$
$\displaystyle R_{p}^{\sigma_{A}(\sigma_{B}\sigma_{C})}$ (A.3) $\displaystyle
R_{p}^{s_{j}^{2}}$ $\displaystyle=$ $\displaystyle
R_{p}^{e}=1\quad{\mbox{f}or}\,j=1,2,\ldots,n\,,$ (A.4) $\displaystyle
R_{p}^{s_{i}s_{i+1}s_{i}}$ $\displaystyle=$ $\displaystyle
R_{p}^{s_{i+1}s_{i}s_{i+1}}\quad{\mbox{f}or}\,i=1,2,\ldots,n-1\,.$ (A.5)
###### Proof.
We recall that in terms of generators $s_{j}$ the defining relations of the
symmetric group ${\cal S}_{n}$ are given by
$s_{i}s_{i+1}s_{i}=s_{i+1}s_{i}s_{i+1}$ for $i=1,2,\ldots,n-1$ and
$s_{j}^{2}=1$ for $j=1,2,\ldots,n$ [35]. It thus follows that definition 2.21
is well defined if and only if conditions (A.3), (A.4) and (A.5) hold. We now
show them. Lemma A.1 leads to conditions (A.3). Conditions (A.4) are derived
from the inversion relations (unitarity conditions) (2.8). Finally we show
conditions (A.5) by the Yang-Baxter equations. ∎
###### Lemma A.2.
Let $p$ be a sequence of integers $1,2,\ldots,n$, and $R_{j,k}$ denote
$R_{j,k}(\lambda_{j},\lambda_{k})$ for $j,k=1,2,\ldots,n$. For
$\sigma_{c}=(12\cdots n)$ we have
$R^{\sigma_{c}}_{p}=R_{p_{1},p_{2}\cdots p_{n}}\,.$ (A.6)
###### Proof.
Noting $(12\cdots n)=(12)(23)\cdots(n-1\,n)=s_{1}s_{2}\cdots s_{n-1}$. we have
$\displaystyle R_{p}^{\sigma_{c}}$ $\displaystyle=$ $\displaystyle
R_{p}^{s_{1}s_{2}\cdots s_{n-1}}=R_{s_{1}p}^{s_{2}\cdots
s_{n-1}}R_{p}^{s_{1}}$ (A.7) $\displaystyle=$ $\displaystyle
R_{s_{2}(s_{1}p)}^{s_{3}\cdots
s_{n-1}}R_{s_{1}p}^{s_{2}}R_{p}^{s_{1}}=R_{(s_{1}s_{2})p}^{s_{3}\cdots
s_{n-1}}R_{s_{1}p}^{s_{2}}R_{p}^{s_{1}}$ $\displaystyle=$ $\displaystyle
R_{(s_{1}\cdots s_{n-2})p}^{s_{n-1}}\cdots
R_{(s_{1}s_{2})p}^{s_{3}}R_{s_{1}p}^{s_{2}}R_{p}^{s_{1}}$ $\displaystyle=$
$\displaystyle R_{(12\cdots n-1)p}^{s_{n-1}}\cdots
R_{(123)p}^{s_{3}}R_{(12)p}^{s_{2}}R_{p}^{s_{1}}$ $\displaystyle=$
$\displaystyle R_{p_{1},p_{n}}\cdots R_{p_{1},p_{3}}R_{p_{1},p_{2}}$
$\displaystyle=$ $\displaystyle R_{p_{1},p_{2}\cdots p_{n}}\,.$
∎
###### Proposition A.2.
Let ${\cal A}$ be a Hopf algebra and ${\cal R}$ is an element of ${\cal
A}\otimes{\cal A}$ such that ${\cal R}\Delta(x)=\tau\circ\Delta(x){\cal R}$
for all $x\in{\cal A}$. Suppose that ${\cal R}$ is given by ${\cal
R}=\sum_{a}r^{(a,1)}\otimes r^{(a,2)}$, where $r^{(a,1)}$, $r^{(a,2)}\in{\cal
A}$. We define $R_{j,k}$ by
$R_{j,k}=\sum_{a}id_{1}\otimes\cdots\otimes r_{j}^{(a,1)}\otimes\cdots\otimes
r_{k}^{(a,2)}\otimes\cdots\otimes id_{n}\quad\in{\cal A}^{\otimes n}\,.$ (A.8)
If $R_{j,k}$ satisfy the inversion relations and the Yang-Baxter equations:
$\displaystyle R_{12}R_{21}$ $\displaystyle=$ $\displaystyle id\,,$
$\displaystyle R_{12}R_{13}R_{23}$ $\displaystyle=$ $\displaystyle
R_{23}R_{13}R_{12}\,,$ (A.9)
then we have the following symmetry relations:
$R_{p_{q}}^{\sigma}\Delta^{(n-1)}(x)=\sigma\circ\Delta^{(n-1)}(x)R_{p_{q}}^{\sigma}\quad
x\in{\cal A}\,.$ (A.10)
Here $p_{q}$ denotes $p_{q}=(1,2,\ldots,n)$, a sequence of $n$ integers, and
we have defined $R_{p_{q}}^{\sigma}$ as in definition 2.21.
###### Proof.
Recall that any given permutation $\sigma$ is expressed as a product of
generators $s_{j}=(j\,j+1)$. We show symmetry relations (A.10) by induction on
the number of generators $s_{j}$ whose product gives permutation $\sigma$.
Suppose that (A.10) holds for $\sigma=\sigma_{A}$. We now show that (A.10)
holds for $\sigma=\sigma_{A}s_{j}$. Making use of eq. (2.21) in definition
2.21 we have
$R_{p_{q}}^{\sigma_{A}s_{j}}\Delta^{(n-1)}(x)=R_{\sigma_{A}(p_{q})}^{s_{j}}\,R_{p_{q}}^{\sigma_{A}}\Delta^{(n-1)}(x)=R_{\sigma_{A}(p_{q})}^{s_{j}}\,\sigma_{A}\circ\Delta^{(n-1)}(x)\,R_{p_{q}}^{\sigma_{A}}$
(A.11)
Expressing $\Delta^{(n-1)}(x)$ as $\Delta^{(n-1)}(x)=\sum x^{(1)}\otimes
x^{(2)}\otimes\cdots\otimes x^{(n)}$, we have
$\sigma_{A}\circ\Delta^{(n-1)}(x)=\sum x^{(\sigma_{A}^{-1}1)}\otimes
x^{(\sigma_{A}^{-1}2)}\otimes\cdots\otimes x^{(\sigma_{A}^{-1}n)}$ (A.12)
It now follows from definition 2.21 that we have
$R_{\sigma_{A}(p_{q})}^{s_{j}}=R_{\sigma_{A}j,\,\sigma_{A}(j+1)}$. Let us
denote $\sigma_{A}j$ and $\sigma_{A}(j+1)$ by $a$ and $b$, respectively. Then
we have $\sigma_{A}^{-1}a=j$ and $\sigma_{A}^{-1}b=j+1$. Assuming $a<b$ we
have
$\displaystyle
R_{\sigma_{A}j,\,\sigma_{A}(j+1)}\,\,\sigma_{A}\circ\Delta^{(n-1)}(x)$ (A.13)
$\displaystyle=$ $\displaystyle R_{a,b}\,\sum
x_{1}^{(\sigma_{A}^{-1}1)}\otimes
x_{2}^{(\sigma_{A}^{-1}2)}\otimes\cdots\otimes
x^{(\sigma_{A}^{-1}a)}_{a}\otimes\cdots\otimes
x^{(\sigma_{A}^{-1}b)}_{b}\otimes\cdots\otimes x^{(\sigma_{A}^{-1}n)}_{n}$
$\displaystyle=$ $\displaystyle\sum x^{(\sigma_{A}^{-1}1)}_{1}\otimes
x^{(\sigma_{A}^{-1}2)}_{2}\otimes\cdots\otimes
x^{(\sigma_{A}^{-1}b)}_{a}\otimes\cdots\otimes
x^{(\sigma_{A}^{-1}a)}_{b}\otimes\cdots\otimes
x^{(\sigma_{A}^{-1}n)}_{n}\,\,R_{a,b}$ $\displaystyle=$
$\displaystyle(\sigma_{A}s_{j})\circ\Delta^{(n-1)}(x)\,\,R_{\sigma_{A}j,\,\sigma_{A}(j+1)}$
Here we note that
$R_{a,b}\,x^{(\sigma_{A}^{-1}a)}_{a}\otimes
x^{(\sigma_{A}^{-1}b)}_{b}=x^{(\sigma_{A}^{-1}b)}_{a}\otimes
x^{(\sigma_{A}^{-1}a)}_{b}\,R_{a,b}$ (A.14)
since we have
$x^{(\sigma_{A}^{-1}a)}\otimes x^{(\sigma_{A}^{-1}b)}=x^{(j)}\otimes
x^{(j+1)}=\Delta({\bar{x}}^{(j)})$
where ${\bar{x}}^{(j)}$ are defined by
$\Delta^{(n-2)}(x)=\sum{\bar{x}}^{(1)}\otimes\cdots\otimes{\bar{x}}^{(n-1)}$
We therefore have
$\displaystyle
R_{\sigma_{A}(p_{q})}^{s_{j}}\,\sigma_{A}\circ\Delta^{(n-1)}(x)\,R_{p_{q}}^{\sigma_{A}}=(\sigma_{A}s_{j})\circ\Delta^{(n-1)}(x)\,R_{\sigma_{A}(p_{q})}^{s_{j}}\,R_{p_{q}}^{\sigma_{A}}$
$\displaystyle=(\sigma_{A}s_{j})\circ\Delta^{(n-1)}(x)\,R_{p_{q}}^{\sigma_{A}s_{j}}$
(A.15)
∎
Symmetry relations similar to (A.10) hold for products of monodromy matrices.
Let us consider $m$ auxiliary spaces with suffices $a(1),a(2),\ldots,a(m)$,
respectively. We denote the monodromy matrix
$T_{a(j)}(\lambda_{a(j)};\xi_{1},\ldots,\xi_{L})$ simply by $T_{a(j)}$. We
denote by $\Delta^{(m-1)}(T)$ the following operator:
$\Delta^{(m-1)}(T)=T_{a(1)}T_{a(2)}\cdots T_{a(m)}$ (A.16)
Let $\sigma$ an element of ${\cal S}_{m}$. We define the action of $\sigma$ on
$\Delta^{(m-1)}(T)$ by the following:
$\sigma\circ\Delta^{(m-1)}(T)=T_{a({\bar{\sigma}}1)}T_{a({\bar{\sigma}}2)}\cdots
T_{a({\bar{\sigma}}m)}$ (A.17)
Here ${\bar{\sigma}}$ denotes the inverse of $\sigma$:
${\bar{\sigma}}=\sigma^{-1}$. Then we have the following.
###### Proposition A.3.
Let $p_{q}$ be $p_{q}=(1,2,\ldots,m)$. For any $\sigma\in{\cal S}_{m}$ we have
$R^{\sigma}_{p_{q}}\Delta^{(m-1)}(T)=\sigma\circ\Delta^{(m-1)}(T)R^{\sigma}_{p_{q}}$
(A.18)
## Appendix B Symmetric-group action on products of $R$-matrices and the
$F$-basis
###### Lemma B.1.
([6]) (i) Cocycle conditions hold for $n\leq L$.
$R_{2\cdots n-1,\,n}R_{1,\,2\cdots n}=R_{1,\,2\cdots n-1}R_{12\cdots n-1,\,n}$
(B.1)
(ii) The unitarity relations hold for $n\leq L$.
$R_{1,\,2\cdots n}R_{23\cdots n,\,1}=I^{\otimes L}$ (B.2)
###### Proof.
Cocycle conditions (B.1) are derived from the Yang-Baxter equations. ∎
Let us denote by the symbol $(p_{0},p_{1},p_{2},\ldots,p_{n})$ a sequence of
$n+1$ integers, $0,1,2,\ldots,n$, and we express it as $(p_{0},p)$ where $p$
denotes the subsequence $(p_{1},p_{2},\ldots,p_{n})$.
###### Lemma B.2.
Let $p$ be a sequence of $n$ integers, $1,2,\ldots,n$. For
$s_{j}=(j\,j+1)\in{\cal S}_{n}$ we have
$\displaystyle R_{p}^{s_{j}}R_{0,\,p}$ $\displaystyle=$ $\displaystyle
R_{0,\,s_{j}(p)}R_{p}^{s_{j}}$ $\displaystyle R_{p}^{s_{j}}R_{p,\,0}$
$\displaystyle=$ $\displaystyle R_{s_{j}(p),\,0}R_{p}^{s_{j}}$ (B.3)
###### Proof.
We first note $R_{p}^{s_{j}}=R_{p_{j},p_{j+1}}$. We have
$\displaystyle R^{s_{j}}_{p}R_{0,p}$ $\displaystyle=$ $\displaystyle
R_{p_{j},p_{j+1}}\,R_{0,p_{n}}\cdots R_{0,p_{1}}$ (B.4) $\displaystyle=$
$\displaystyle R_{0,p_{n}}\cdots R_{0,p_{j+2}}\cdot
R_{p_{j},p_{j+1}}R_{0,p_{j+1}}R_{0,p_{j}}\cdot R_{0,p_{j+2}}\cdots
R_{0,p_{1}}$
Applying the Yang-Baxter equations
$R_{p_{j},p_{j+1}}R_{0,p_{j+1}}R_{0,p_{j}}=R_{0,p_{j}}R_{0,p_{j+1}}R_{p_{j},p_{j+1}}$
we have
$\displaystyle R^{s_{j}}_{p}R_{0,p}$ $\displaystyle=$ $\displaystyle
R_{0,p_{n}}\cdots R_{0,p_{j+2}}\cdot R_{0,p_{j}}R_{0,p_{j+1}}\cdot
R_{0,p_{j-1}}\cdots R_{0,p_{1}}\cdot R_{p_{j},p_{j+1}}$ (B.5) $\displaystyle=$
$\displaystyle R_{0,s_{j}(p)}R_{p_{j},p_{j+1}}\,.$
∎
###### Proposition B.1.
Let $p$ be a sequence of $n$ integers, $1,2,\ldots,n$, and $\sigma$ a
permutation of the $n$ integers. We have
$\displaystyle R_{p}^{\sigma}R_{0,\,p}$ $\displaystyle=$ $\displaystyle
R_{0,\,\sigma(p)}R_{p}^{\sigma}$ $\displaystyle R_{p}^{\sigma}R_{p,\,0}$
$\displaystyle=$ $\displaystyle R_{\sigma(p),\,0}R_{p}^{\sigma}$ (B.6)
###### Proof.
Expressing permutation $\sigma$ as a product of generators $s_{j}$, and
applying (B.3) many times, we can show the symmetry relations. ∎
We define the action of $\sigma$ on the $F$-basis as follows.
$F_{\sigma(p)}=F_{p_{\sigma(1)}p_{\sigma(2)}\cdots p_{\sigma(n)}}$ (B.7)
###### Proposition B.2.
Let $p$ be a sequence of integers, $1,2,\ldots,n$. For $\sigma\in{\cal S}_{n}$
we have
$R^{\sigma}_{p}F_{0,p}=F_{0,\sigma(p)}R_{p}^{\sigma}$ (B.8)
We also have
$F_{p}=F_{\sigma(p)}R_{p}^{\sigma}$ (B.9)
###### Proof.
Expressing the $F$-basis in terms of the $R$-matrices we have
$\displaystyle R^{\sigma}_{p}F_{0,p}$ $\displaystyle=$ $\displaystyle
R^{\sigma}_{p}\left(e_{0}^{11}+e_{0}^{22}R^{\sigma}_{0,p}\right)$ (B.10)
$\displaystyle=$ $\displaystyle
e_{0}^{11}R^{\sigma}_{p}+e_{0}^{22}R^{\sigma}_{p}R^{\sigma}_{0,p}$
$\displaystyle=$ $\displaystyle
e_{0}^{11}R^{\sigma}_{p}+e_{0}^{22}R^{\sigma}_{0,\sigma(p)}R^{\sigma}_{p}$
$\displaystyle=$
$\displaystyle\left(e_{0}^{11}+e_{0}^{22}R^{\sigma}_{0,\sigma(p)}\right)R^{\sigma}_{p}$
$\displaystyle=$ $\displaystyle F_{0,\sigma(p)}R^{\sigma}_{p}$
We first show (B.9) with $\sigma=s_{j}$ for $j=1,2,\ldots,n-1$, and then we
derive it for all permutations $\sigma$. ∎
###### Lemma B.3.
The propagator $F^{-1}_{i\cdots L\,1\cdots i-1}F_{1\cdots L}$ is given by the
following:
$F^{-1}_{i\cdots L\,1\cdots i-1}F_{1\cdots
L}=\prod_{\alpha=1}^{i-1}\left(A_{1\cdots L}(\xi_{\alpha})+D_{1\cdots
L}(\xi_{\alpha})\right)$ (B.11)
###### Proof.
Let $\sigma_{c}$ be a cyclic permutation: $\sigma_{c}=(12\cdots L)$, and
$p_{q}$ the sequence $p_{q}=(1,2,\ldots,n)$. We have
$F_{i\cdots L\,1\cdots i-1}=F_{\sigma_{c}^{i-1}(p_{q})}=F_{1\cdots
L}R_{p_{q}}^{\sigma_{c}^{i-1}}$ (B.12)
and hence we have
$\displaystyle F^{-1}_{i\cdots L\,1\cdots i-1}F_{1\cdots L}$ $\displaystyle=$
$\displaystyle\left(F_{1\cdots
L}R_{p_{q}}^{\sigma_{c}^{i-1}}\right)^{-1}F_{1\cdots L}$ (B.13)
$\displaystyle=$ $\displaystyle R_{p_{q}}^{\sigma_{c}^{i-1}}\,F_{1\cdots
L}^{-1}F_{1\cdots L}$ $\displaystyle=$ $\displaystyle
R_{p_{q}}^{\sigma_{c}^{i-1}}$
We thus obtain
$\displaystyle R_{p_{q}}^{\sigma_{c}^{i-1}}$ $\displaystyle=$ $\displaystyle
R_{\sigma(p_{q})}^{\sigma_{c}^{i-2}}R_{p_{q}}^{\sigma_{c}}$ $\displaystyle=$
$\displaystyle R_{\sigma_{c}^{i-2}(p_{q})}^{\sigma_{c}}\cdots
R_{p_{q}}^{\sigma_{c}}$ $\displaystyle=$ $\displaystyle R_{i-1\cdots
L\,1\cdots i-2}^{\sigma_{c}}\cdots R_{2\cdots L1}^{\sigma_{c}}R_{12\cdots
L}^{\sigma_{c}}$ $\displaystyle=$ $\displaystyle R_{i-1,i\cdots L\,1\cdots
i-2}\cdots R_{2,3\cdots L1}R_{1,2\cdots L}$ $\displaystyle=$
$\displaystyle\prod_{\alpha=1}^{i-1}\left(A_{1\cdots
L}(\xi_{\alpha})+D_{1\cdots L}(\xi_{\alpha})\right)$
∎
## Appendix C Formulas of the $q$-analogue
###### Lemma C.1.
For two integers $\ell$ and $n$ satisfying $0\leq n\leq\ell$ we have
$\sum_{1\leq
i_{1}<\cdots<i_{n}\leq\ell}q^{2i_{1}+\cdots+2i_{n}}=q^{n(\ell+1)}\,\left[\begin{array}[]{c}\ell\\\
n\end{array}\right]_{q}$ (C.1)
###### Proof.
We can show by induction on $\ell$ the $q$-binomial expansion as follows.
$\prod_{k=0}^{\ell-1}\left(1-zq^{2k}\right)=\sum_{n=0}^{\ell}(-z)^{n}q^{n(\ell-1)}\left[\begin{array}[]{c}\ell\\\
n\end{array}\right]_{q}$ (C.2)
It is now easy to show the following:
$\prod_{k=1}^{\ell}\left(1-zq^{2k}\right)=\sum_{n=0}^{\ell}(-z)^{n}\sum_{1\leq
i_{1}<\cdots<i_{n}\leq\ell}q^{2i_{1}+\cdots+2i_{n}}$ (C.3)
Comparing (C.3) with (C.2) we obtain formula (C.1). ∎
###### Lemma C.2.
The spin-$\ell$ matrix representations $X^{\pm(\ell+)}$ of the generators
$X^{\pm}$ of $U_{q}(sl_{2})$ are related to $\sigma_{\ell}^{+}$ and
$\sigma_{1}^{-}$, respectively, as follows.
$\displaystyle
P^{(\ell)}_{1\cdots\ell}\,\sigma_{1}^{-}\,P^{(\ell)}_{1\cdots\ell}$
$\displaystyle=$ $\displaystyle{\frac{1}{[\ell]_{q}}}\,X^{-(\ell+)}$ (C.4)
$\displaystyle
P^{(\ell)}_{1\cdots\ell}\,\sigma_{\ell}^{+}\,P^{(\ell)}_{1\cdots\ell}$
$\displaystyle=$ $\displaystyle{\frac{1}{[\ell]_{q}}}\,X^{+(\ell+)}$ (C.5)
###### Proof.
Expressing the projection operator $P^{(\ell)}_{1\cdots\ell}$ as
$P^{(\ell)}_{1\cdots\ell}=\sum_{n=0}^{\ell}||\ell,n\rangle\langle\ell,n||$
we have
$P^{(\ell)}_{1\cdots\ell}\sigma_{1}^{-}P^{(\ell)}_{1\cdots\ell}=\sum_{n=0}^{\ell-1}||\ell,n+1\rangle\langle\ell,n||\quad\langle\ell,n+1||\sigma_{1}^{-}||\ell,n\rangle$
(C.6)
Making use of (C.1) we have
$\langle\ell,n+1||\sigma_{1}^{-}||\ell,n\rangle=\frac{[n+1]}{[\ell]},$ (C.7)
and then we obtain (C.4). Similarly, we can show (C.5). ∎
## Appendix D Formulas of the $F$-basis useful for the diagonalization.
Let us review some points of the diagonalization process of the $A$ and $D$
operators [6].
###### Lemma D.1.
Operators $A$ and $D$ are upper- and lower-triangular matrices, respectively.
Moreover, the eigenvalues of operators $A$ and $D$ are given by
$\displaystyle{\rm diag}\left(D_{12\cdots n}(\lambda_{0})\right)$
$\displaystyle=$
$\displaystyle\bigotimes_{i=1}^{n}\left(\begin{array}[]{cc}b_{0i}&0\\\
0&1\end{array}\right)_{[i]}\,,$ (D.3) $\displaystyle{\rm
diag}\left(A_{12\cdots n}(\lambda_{0})\right)$ $\displaystyle=$
$\displaystyle\bigotimes_{i=1}^{n}\left(\begin{array}[]{cc}1&0\\\
0&b_{0i}\end{array}\right)_{[i]}\,,$ (D.6)
where $b_{0i}=b(\lambda_{0}-\xi_{i})$.
###### Proof.
We can show it by induction on $n$. Noting $D_{12\cdots n}=C_{n}B_{1\cdots
n-1}+D_{n}D_{1\cdots n-1}$, we show ${\rm diag}(D_{1\cdots n}={\rm
diag}(D_{1\cdots n-1}(\lambda_{0})\otimes{\rm diag}(b_{0n},1)_{[n]}$. We can
show (D.6) similarly. ∎
###### Lemma D.2.
The partial $F_{0,1\cdots n}$ and ${\bar{F}}_{0,1\cdots n}$ are expressed as
follows.
$\displaystyle{F}_{0,1\cdots n}$ $\displaystyle=$
$\displaystyle\left(\begin{array}[]{cc}I_{1\cdots n}&0\\\ C_{1\cdots
n}(\lambda_{0})&D_{1\cdots n}(\lambda_{0})\end{array}\right)_{[0]}\,,$ (D.9)
$\displaystyle{\bar{F}}_{0,1\cdots n}$ $\displaystyle=$
$\displaystyle\left(\begin{array}[]{cc}A_{1\cdots n}(\lambda_{0})&B_{1\cdots
n}(\lambda_{0})\\\ 0&I_{1\cdots n}\end{array}\right)_{[0]}\,.$ (D.12)
###### Proof.
It is clear from definition 5.2 of the $F$-basis. ∎
Proof of propositions 25 and 26
For an illustration, we now derive the diagonalized form of the $D$ operator.
From $R_{0,1\cdots n}=R_{0,2\cdots n}R_{0,1}$ we have
$\displaystyle D_{12\cdots n}$ $\displaystyle=$ $\displaystyle C_{2\cdots
n}B_{1}+D_{2\cdots n}D_{1}$ (D.15) $\displaystyle=$
$\displaystyle\left(\begin{array}[]{cc}b_{01}D_{2\cdots n}(\lambda_{0})&0\\\
c_{01}C_{2\cdots n}(\lambda_{0})&D_{2\cdots
n}(\lambda_{0})\end{array}\right)_{[0]}\,.$
We thus calculate
$\displaystyle F_{1\cdots n}D_{1\cdots n}(\lambda_{0})$ $\displaystyle=$
$\displaystyle F_{2\cdots n}\,F_{1,2\cdots n}\,D_{1\cdots n}(\lambda_{0})$
(D.20) $\displaystyle=$ $\displaystyle F_{2\cdots
n}\,\left(\begin{array}[]{cc}I_{2\cdots n}&0\\\ C_{2\cdots
n}(\lambda_{0})&D_{2\cdots
n}(\lambda_{0})\end{array}\right)_{[0]}\left(\begin{array}[]{cc}b_{01}D_{2\cdots
n}(\lambda_{0})&0\\\ c_{01}C_{2\cdots n}(\lambda_{0})&D_{2\cdots
n}(\lambda_{0})\end{array}\right)_{[0]}$ $\displaystyle=$ $\displaystyle{\rm
diag}(b_{01},1)_{[1]}\widetilde{D}_{2\cdots n}(\lambda_{0})F_{1\cdots n}$
(D.21)
Therefore, by induction we have the diagonalized form of operator $D$.
Similarly, we can diagonalize $A$, $A^{\dagger}$ ] and $D^{\dagger}$.
###### Lemma D.3.
The diagonalized form of $F_{0,1\cdots n}{\bar{F}}^{\dagger}_{0,1\cdots n}$ is
given by the following:
$F_{1\cdots n}\left(F_{0,1\cdots n}{\bar{F}}^{\dagger}_{0,1\cdots
n}\right)F_{1\cdots n}^{-1}={\hat{\delta}}^{-1}_{0,1\cdots n}$ (D.22)
###### Proof.
Making use of (5.12) we show
$\displaystyle F_{0,1\cdots n}{\bar{F}}_{0,1\cdots n}^{\dagger}=F_{0,1\cdots
n}{\cal C}_{01\cdots n}{F}_{0,1\cdots n}^{\dagger}{\cal C}_{01\cdots n}$
(D.31) $\displaystyle=$ $\displaystyle\left(\begin{array}[]{cc}1&0\\\
C_{1\cdots n}&D_{1\cdots
n}\end{array}\right)_{[0]}\left(\begin{array}[]{cc}0&{\cal C}_{1\cdots n}\\\
{\cal C}_{1\cdots
n}&0\end{array}\right)_{[0]}\left(\begin{array}[]{cc}1&C^{\dagger}_{1\cdots
n}\\\ 0&D_{1\cdots n}\end{array}\right)_{[0]}\left(\begin{array}[]{cc}0&{\cal
C}_{1\cdots n}\\\ {\cal C}_{1\cdots n}&0\end{array}\right)_{[0]}$
$\displaystyle=$ $\displaystyle\left(\begin{array}[]{cc}A^{\dagger}&0\\\
0&D_{1\cdots n}\end{array}\right)_{[0]}\,.$ (D.34)
Calculating $F_{12\cdots n}\cdot F_{0,1\cdots n}{\bar{F}}_{0,1\cdots
n}^{\dagger}\cdot F^{-1}_{12\cdots n}$ we have (D.22). ∎
###### Corollary D.1.
The inverse of the total $F$ is given as follows.
$F^{-1}_{1\cdots n}={\bar{F}}_{1\cdots n}^{\dagger}{\hat{\delta}}_{1\cdots n}$
(D.35)
###### Proof.
We show it by induction on $n$. Let us assume (D.35) for the case of $n$. We
have
${\bar{F}}^{\dagger}_{01\cdots n}\widehat{\delta}_{01\cdots
n}={\bar{F}}^{\dagger}_{01\cdots n}\left({\bar{F}}^{\dagger}_{1\cdots
n}\widehat{\delta}_{1\cdots n}\right)\widehat{\delta}_{0,1\cdots
n}={\bar{F}}^{\dagger}_{01\cdots n}{F}^{-1}_{1\cdots
n}\widehat{\delta}_{0,1\cdots n}$
From (D.22) we have ${\bar{F}}^{\dagger}_{01\cdots n}{F}^{-1}_{1\cdots
n}\widehat{\delta}_{0,1\cdots n}=F_{01\cdots n}^{-1}$, which corresponds to
(D.35) for the case of $n+1$. ∎
###### Lemma D.4.
The dagger of total $F$ is given by the charge conjugation of transposed total
$F$.
$F_{1\cdots n}^{\dagger}={\cal C}_{1\cdots n}F^{t_{1}\cdots t_{n}}_{n\cdots
21}{\cal C}_{1\cdots n}$ (D.36)
Or equivalently we have
${\bar{F}}_{1\cdots n}^{\dagger}=F^{t_{1}\cdots t_{n}}_{1\cdots n}$ (D.37)
###### Proof.
We show it by induction on $n$. Let us assume (D.36) for the case of $n-1$. We
first show
${\bar{F}}_{1,2\cdots n}^{\dagger}={\cal C}_{12\cdots
n}\left(e_{1}^{11}+e_{1}^{22}R_{1,2\cdots n}\right)^{\dagger}{\cal
C}_{12\cdots n}=F_{n\cdots 2,1}^{t_{1}\cdots t_{n}}\,.$
Making use of the induction assumption we show
$\displaystyle F_{12\cdots n}^{\dagger}$ $\displaystyle=$
$\displaystyle(F_{2\cdots n}F_{1,2\cdots n})^{\dagger}=F_{1,2\cdots
n}^{\dagger}F_{2\cdots n}^{\dagger}$ $\displaystyle=$ $\displaystyle{\cal
C}_{1\cdots n}F_{n\cdots 2,1}^{t_{1}\cdots t_{n}}{\cal C}_{1\cdots n}\,{\cal
C}_{1\cdots n}F_{n\cdots 2}^{t_{2}\cdots t_{n}}{\cal C}_{1\cdots n}$
$\displaystyle=$ $\displaystyle{\cal C}_{1\cdots n}F_{n\cdots 21}^{t_{1}\cdots
t_{n}}{\cal C}_{1\cdots n}\,.$
∎
###### Corollary D.2.
The inverse of total $F$ is given as follows.
$F_{1\cdots n}F_{n\cdots 21}^{t_{1}\cdots
t_{n}}=\widehat{\delta}^{-1}_{1\cdots n}$ (D.38)
Equivalently, we have
$F_{1\cdots n}^{-1}=F_{n\cdots 21}^{t_{1}\cdots
t_{n}}\,{\widehat{\delta}}_{1\cdots n}$ (D.39)
###### Proof.
It follows from (D.35) that ${\bar{F}}^{\dagger}=F_{n\cdots 1}^{t_{1}\cdots
t_{n}}\widehat{\delta}_{12\cdots n}$. From (D.37) we have (D.38). ∎
## Appendix E Lemmas for Diagonalizing the $C$ operator
###### Lemma E.1.
Let $X^{+}$ be the generator of the quantum group $U_{q}(sl_{2})$ and
$X_{i}^{+}$ the matrix representation of $X^{+}$ acting on the $i$th site. We
have
${\widetilde{\Delta}}_{1\cdots
n}(X^{+})=\left(X_{n}^{+}+e_{n}^{11}{\widetilde{A^{+}}}^{\dagger}_{1\cdots
n-1}(\xi_{n}){\widetilde{\Delta}}_{1\cdots
n-1}(X^{+})+e_{n}^{22}{\widetilde{\Delta}}_{1\cdots
n-1}(X^{+}){\widetilde{D^{+}}}_{1\cdots
n-1}(\xi_{n})\right)\widehat{\delta}^{12\cdots n}_{n}\,.$ (E.1)
###### Lemma E.2.
The diagonalized form of $\Delta^{(n-1)}(X^{+})$ is expressed in terms of
local operators $X_{i}^{+}$ as follows.
${\widetilde{\Delta}}_{1\cdots
n}(X^{+})=\sum_{i=1}^{n}X_{i}^{+}\widehat{\delta}^{1\cdots n}_{i}$ (E.2)
###### Lemma E.3.
Operator $C^{+}$ in the $F$-basis is expressed in terms of operator $D^{+}$
and $X_{i}^{+}$ as follows.
${\widetilde{C^{+}}}_{1\cdots
n}(\lambda)=\sum_{i=1}^{n}({\widetilde{D^{+}}}_{1\cdots i-1,i+1\cdots
n}(\lambda)-q^{-1}{\widetilde{D^{+}}}_{1\cdots
n}(\lambda))\,X_{i}^{+}\widehat{\delta}_{i}^{1\cdots n}\,.$ (E.3)
## Acknowledgment
The authors would like to thank Prof. S. Miyashita for encouragement and keen
interest in this work. One of the authors (C.M.) would like to thank Dr. K.
Shigechi for his introduction to the mathematical physics of integrable
models. This work is partially supported by Grant-in-Aid for Scientific
Research (C) No. 20540365.
## References
* [1] V.E. Korepin, N.M. Bogoliubov and A.G. Izergin, Quantum Inverse Scattering Method and Correlation Functions (Cambridge University Press, Cambridge, 1993)
* [2] M. Jimbo and T. Miwa, Algebraic Analysis of Solvable Lattice Models (AMS, Providence, RI, 1995).
* [3] N. Kitanine, J. M. Maillet, N. A. Slavnov and V. Terras, On the algebraic Bethe ansatz approach to the correlation functions of the XXZ spin-1/2 Heisenberg chain, hep-th/0505006
* [4] M. Jimbo, K. Miki, T. Miwa and A. Nakayashiki, Correlation functions of the XXZ model for $\Delta<-1$, Phys. Lett. A 168 (1992) 256–263.
* [5] N.A. Slavnov, Calculation of scalar products of wave functions and form factors in the framework of the algebraic Bethe ansatz, Theor. Math. Phys. 79 (1989) 502–508.
* [6] J.M. Maillet and J. Sanchez de Santos, Drinfel’d twists and algebraic Bethe ansatz, ed. M. Semenov-Tian-Shansky, Amer. Math. Soc. Transl. 201 Ser. 2, (Providence, R.I.: Ameri. Math. Soc., 2000) pp. 137–178.
* [7] N. Kitanine, J.M. Maillet and V. Terras, Form factors of the XXZ Heisenberg spin-1/2 finite chain, Nucl. Phys. B 554 [FS] (1999) 647–678
* [8] N. Kitanine, J.M. Maillet and V. Terras, Correlation functions of the XXZ Heisenberg spin-1/2 chain in a magnetic field, Nucl. Phys. B 567 [FS] (2000) 554–582.
* [9] F. Göhmann, A. Klümper and A. Seel, Integral representations for correlation functions of the XXZ chain at finite temperature, J. Phys. A: Math. Gen. 37 (2004) 7625–7651.
* [10] D. Biegel, M. Karbach, G. Müller and K. Wiele, Spectrum of transition rates of the XX chain analyzed via Bethe ansatz, Phys. Rev. B 69, 174404 (2004) .
* [11] J. Sato, M. Shiroishi and M. Takahashi, Evaluation of Dynamic Spin Structure Factor for the Spin-1/2 XXZ Chain in a Magnetic Field, J. Phys. Soc. Jpn. 73 (2004) 3008–3014.
* [12] J.-S. Caux and J. M. Maillet, Phys. Rev. Lett. 95, 077201 (2005).
* [13] R.G. Pereira, J. Sirker, J.-S. Caux, R. Hagemans, J.M. Maillet, S.R. White, and I. Affleck, Dynamical structure factor at small $q$ for the XXZ spin-1/2 chain, J. Stat. Mech. (2007) P08022.
* [14] P.P. Kulish, N. Yu. Reshetikhin and E.K. Sklyanin, Yang-Baxter equation and representation theory: I, Lett. Math. Phys. 5 (1981) 393–403
* [15] H.M. Babujan, Exact solution of the isotropic Heisenberg chain with arbutrary spins: thermodynamics of the model, Nucl. Phys. B 215 [FS7] (1983) 317–336.
* [16] A.B. Zamolodchikov and V.A. Fateev, Sov. J. Nucl. Phys. 32 (1980) 298.
* [17] K. Sogo, Y. Akutsu and T. Abe, Prog. Theor. Phys. 70 (1983) 730; 739.
* [18] A. N. Kirillov and N. Yu. Reshetikhin, Exact solution of the integrable XXZ Heisenberg model with arbitrary spin. I. the ground state and the excitation spectrum, J. Phys. A: Math. Gen. 20 (1987) 1565–1585.
* [19] T. Deguchi, M. Wadati and Y. Akutsu, Exactly Solvable Models and New Link polynomials. V. Yang-Baxter Operator and Braid-Monoid Algebra, J. Phys. Soc. Jpn. 57 (1988) 1905-1923.
* [20] J. Suzuki, Spinons in magnetic chains of arbitrary spins at finite temperatures, J. Phys. A: Math. Gen. 32 (1999) 2341–2359.
* [21] Y. Akutsu and M. Wadati, Exactly Solvable Models and New Link polynomials. I. $N$-State Vertex Models, J. Phys. Soc. Jpn. 56 (1987) 3039–3051.
* [22] M. Jimbo, A $q$-Difference Analogue of $U({\ calg})$ and the Yang-Baxter Equation, Lett. Math. Phys. 10 (1985) 63–69.
* [23] M. Jimbo, A $q$-analogue of $U(gl(N+1))$, Hecke algebra and the Yang-Baxter equation, Lett. Math. Phys. 11 (1986) 247–252.
* [24] V. Terras, Drinfel’d Twists and Functional Bethe Ansatz, Lett. Math. Phys. 48 (1999) 263–276.
* [25] J.M. Maillet and V. Terras, On the quantum inverse scattering problem, Nucl. Phys. B 575 [FS] (2000) 627–644.
* [26] N. Kitanine, Correlation functions of the higher spin XXX chains, J. Phys. A: Math. Gen. 34(2001) 8151–8169.
* [27] O.A. Castro-Alvaredo and J.M. Maillet, Form factors of integrable Heisenberg (higher) spin chains, J. Phys. A: Math. Theor. 40 (2007) 7451–7471.
* [28] C.S. Melo and M.J. Martins, Algebraic Bethe Ansatz for $U(1)$ Invariant Integrable Models: The Method and General Results, arXiv:0806.2404 [math-ph].
* [29] K. Fabricius and B. M. McCoy, Evaluation Parameters and Bethe Roots for the Six-Vertex Model at Roots of Unity, in Progress in Mathematical Physics Vol. 23 (MathPhys Odyssey 2001), edited by M. Kashiwara and T. Miwa, (Birkhäuser, Boston, 2002) 119–144.
* [30] V.G. Drinfeld, Quantum groups, Proc. ICM Berkeley 1986, pp. 798–820.
* [31] M. Jimbo, Topics from representations of $U_{q}(g)$– An Introductory Guide to Physicists, in Nankai Lectures on Mathematical Physics (World Scientific, Singapore, 1992) pp. 1–61.
* [32] A. Nishino and T. Deguchi, The $L(sl_{2})$ symmetry of the Bazhanov-Stroganov model associated with the superintegrable chiral Potts model, Phys. Lett. A 356 (2006) 366–370.
* [33] L. Takhtajan and L. Faddeev, Spectrum and Scattering of Excitations in the One-Dimensional Isotropic Heisenberg Model, J. Sov. Math. 24 (1984) 241–267.
* [34] R. J. Baxter, Exactly Solved Models in Statistical Mechanics (Academic Press, 1982, London).
* [35] W. Magnus, A. Karrass and D. Solitar, Combinatorial Group Theory: Presentations of groups in terms of generators and relations, (Dover Publications Inc., 1976, New York).
* [36] R. Baxter, The Inversion Relation Method for Some Two-Dimensional Exactly Solved Models in Lattice Statistics, J. Stat. Phys. 28 (1982) 1–41.
* [37] T. Deguchi, K. Fabricius and B. M. McCoy, The $sl_{2}$ Loop Algebra Symmetry of the Six-Vertex Model at Roots of Unity, J. Stat. Phys. 102 (2001) 701–736.
| arxiv-papers | 2008-07-11T13:14:55 | 2024-09-04T02:48:56.702788 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Tetsuo Deguchi and Chihiro Matsui",
"submitter": "Tetsuo Deguchi",
"url": "https://arxiv.org/abs/0807.1847"
} |
0807.1857 | # Magnesium doped helium nanodroplets
Alberto Hernando Departament ECM, Facultat de Física, and IN2UB, Universitat
de Barcelona. Diagonal 647, 08028 Barcelona, Spain Manuel Barranco
Departament ECM, Facultat de Física, and IN2UB, Universitat de Barcelona.
Diagonal 647, 08028 Barcelona, Spain Ricardo Mayol Departament ECM, Facultat
de Física, and IN2UB, Universitat de Barcelona. Diagonal 647, 08028 Barcelona,
Spain Martí Pi Departament ECM, Facultat de Física, and IN2UB, Universitat
de Barcelona. Diagonal 647, 08028 Barcelona, Spain Francesco Ancilotto
Dipartimento di Fisica “G. Galilei”, Università di Padova, via Marzolo 8,
I-35131 Padova, and CNR-INFM-DEMOCRITOS National Simulation Center, Trieste,
Italy
###### Abstract
We have studied the structure of 4He droplets doped with magnesium atoms using
density functional theory. We have found that the solvation properties of this
system strongly depend on the size of the 4He droplet. For small drops, Mg
resides in a deep surface state, whereas for large size drops it is fully
solvated but radially delocalized in their interior. We have studied the
$3s3p$ 1P${}_{1}\leftarrow 3s^{2}$ 1S0 transition of the dopant, and have
compared our results with experimental data from laser induced fluorescence
(LIF). Line broadening effects due to the coupling of dynamical deformations
of the surrounding helium with the dipole excitation of the impurity are
explicitly taken into account. We show that the Mg radial delocalization
inside large droplets may help reconcile the apparently contradictory
solvation properties of magnesium as provided by LIF and electron-impact
ionization experiments. The structure of 4He drops doped with two magnesium
atoms is also studied and used to interpret the results of resonant two-
photon-ionization (R2PI) and LIF experiments. We have found that the two
solvated Mg atoms do not easily merge into a dimer, but rather form a weakly-
bound state due to the presence of an energy barrier caused by the helium
environment that keep them some $9.5$ Å apart, preventing the formation of the
Mg2 molecule. From this observation, we suggest that Mg atoms in 4He drops may
form, under suitable conditions, a soft “foam”-like aggregate rather than
coalesce into a compact metallic cluster. Our findings are in qualitative
agreement with recent R2PI experimental evidences. We predict that,
contrarily, Mg atoms adsorbed in 3He droplets do not form such metastable
aggregates.
###### pacs:
36.40.-c, 32.30.Jc, 78.40.-q, 47.55.D-, 71.15.Mb, 67.40.Yv
## I Introduction
Optical investigations of atomic impurities in superfluid helium nanodroplets
have drawn considerable attention in recent years,Sti01 ; Sti06 as the shifts
of the electronic transition lines (atomic shifts) are a very useful
observable to determine the location of the foreign atom attached to a helium
drop.Bar06 In this context, the study of magnesium atoms attached to helium
drops has unraveled an interesting and somewhat unexpected solvation behaviour
as a function of the number ($N$) of helium atoms in the drop. Diffusion Monte
Carlo (DMC) calculationsMel05 carried out for small drops containing a number
of helium atoms up to $N=50$ indicate that a Mg atom is not fully solvated in
drops of sizes below $N\sim 30$. More recent quantum Monte Carlo calculations
Elh07 ; Elh08 suggest a surface Mg state for 4He clusters with up to $\sim
200$ atoms. Density Functional Theory (DFT) calculationsHer07 for 3HeN and
4HeN nanodroplets with $N\geq 300$ doped with alkaline earth atoms have shown
that Mg atoms are solvated in their interior, in agreement with the analysis
of Laser Induced Flourescence (LIF)Reh00 and Resonant Two-Photon-Ionization
(R2PI) experiments.Prz07 LIF experiments on the absorption and emission
spectra of Mg atoms in liquid 3He and 4He have been reported and successfully
analyzed within a vibrating atomic bubble model, where full solvation of the
impurity atom is assumed.Mor99 ; Mor06 A more recent experimentRen07 in
which electron-impact ionization data from Mg doped 4He drops with about 104
atoms seems to indicate that magnesium is instead at the surface of the
droplet, in disagreement with the above mentioned LIF and R2PI experiments.
There is some ambiguity associated with the notion of solvation in a helium
droplet. For not too small droplets, one may consider that Mg is fully
solvated when its position inside the droplet is such that its solvation
energy or atomic shift do not appreciably differ from their asymptotic values
in bulk liquid helium, as both quantities approach such limit fairly
alongside. However, for very small drops the energy or atomic shifts of an
impurity atom at the center of the drop may still differ appreciably from the
bulk liquid values because there is not enough helium to saturate these
quantities. This is the case, e.g., of Mg@4He50 studied in Ref. Mel05, .
Mella et al.Mel05 have discussed how the solvation properties of magnesium
are affected by the number of helium atoms in small 4He drops. Since DMC
calculations cannot be extended to the very large drops involved in LIF
experiments,Reh00 they could not carry out a detailed comparison between
their calculated atomic shifts and the experiments. They also pointed out the
sensitivity of the Mg solvation properties on apparently small differences in
the He-Mg pair potentials available in the literature.
The aim of the present work is to re-examine the solvation of magnesium in 4He
nanodroplets from the DFT perspective, extending our previous
calculationsHer07 down to drops with $N\sim 50$ atoms and improving the DFT
approach (i) by treating the dopant as a quantal particle instead of as an
external field, and (ii) by fully taking into account the coupling of the
dipole excitation of the impurity with the dynamical deformations of the
helium around the Mg atom. Our results confirm that full solvation of a Mg
atom in 4He drops requires a minimum number of helium atoms, and disclose some
unusual results for small drops. We calculate the absorption spectrum of a Mg
atom attached to small and large drops, finding a good agreement with the
experiments for the latter. We discuss in a qualitative way the effect of the
impurity angular momentum on the electron-impact ionization yield and on the
absorption spectrum. We also address the structure of a two-magnesium doped
drop; the results are used to discuss the scenario proposed by Meiwes-Broer
and collaborators to interpret their experimental results on R2PI.Prz07
This work is organized as follows. In Sec. II we briefly present our density
functional approach, as well as results for the structure of Mg@4HeN drops.
The method we have employed to obtain the atomic shifts is discussed in Sec.
III, and applied to the case of Mg-doped 4He droplets in Sec. IV. In Sec. V we
study how the presence of a second magnesium atom in the droplet may alter
both the helium drop structure and the calculated atomic shift. A summary is
presented in Sec. VI, and some technical details of our calculations are
described in the appendices.
## II DFT description of helium nanodroplets
### II.1 Theoretical approach
In recent years, static and time-dependent density functional methodsDal95 ;
Gia03 ; Leh04 have become increasingly popular to study inhomogeneous liquid
helium systems because they provide an excellent compromise between accuracy
and computational effort, allowing to address problems inaccessible to more
fundamental approaches. Although DFT cannot take into account the atomic,
discrete nature of these systems, it can nevertheless address highly
inhomogenous helium systems at the nanoscale,Anc05 including the anisotropic
deformations induced by atomic dopants in helium drops (see Ref. Bar06, for a
recent overview on the physics of helium nanodroplets).
Our starting point is the Orsay-Trento density functional,Dal95 together with
the Mg-He adiabatic ground-state potential $X^{1}\Sigma$ of Ref. Hin03, , here
denoted as $V_{Mg-He}$. To check the sensitivity of our results to the details
of different available pair potential describing the Mg-He interaction, we
also use the sligthly less attractive potential computed in Ref. Par01, . For
the sake of comparison, we plot both potentials in Fig. 1. Despite the
apparently minor differences between these two potential curves, they cause
very different solvation properties of Mg in small 4He drops, as we will show
in the following.
The energy of the Mg-helium system is written as a functional of the Mg wave
function $\Phi(\mathbf{r})$ and the 4He “order parameter”
$\Psi(\mathbf{r})=\sqrt{\rho(\mathbf{r})}$, where $\rho(\mathbf{r})$ is the
4He atomic density:
$\displaystyle E[\Psi,\Phi]$ $\displaystyle=$
$\displaystyle\frac{\hbar^{2}}{2\,m_{He}}\int\mathrm{d}^{3}\mathbf{r}\,|\nabla\Psi(\mathbf{r}\,)|^{2}+\int\mathrm{d}^{3}\mathbf{r}\,{\cal
E}(\rho)$ (1) $\displaystyle+$
$\displaystyle\frac{\hbar^{2}}{2\,m_{Mg}}\int\mathrm{d}^{3}\mathbf{r}\,|\nabla\Phi(\mathbf{r}\,)|^{2}+\int\int\mathrm{d}^{3}\mathbf{r}\,\mathrm{d}^{3}\mathbf{r^{\prime}}\,|\Phi(\mathbf{r})|^{2}\,V_{Mg-
He}(|\mathbf{r}-\mathbf{r^{\prime}}|)\,\rho(\mathbf{r^{\prime}})\;.$
In this expression, ${\cal E}(\rho)$ is the 4He “potential energy
density”.Dal95 Minimizing $E$ under the constraints of a given $N$ and a
normalized Mg wave function, yields the ground state of the drop-impurity
complex. To address the solvation of the Mg atom, we have found it convenient
to minimize $E$ subjected to the additional constraint of a fixed distance
${\cal Z}_{0}$ between the centers of mass of the helium moiety and of the
impurity atom which, due to the symmetry of the problem, can both be taken on
the $z$ axis. This is done following a method -borrowed from Nuclear Physics-
similar to that used to describe the fission of rotating 3He drops.Gui93
Specifically, we minimize the expression
$E+\frac{\lambda_{C}}{2}[{\cal Z}-{\cal Z}_{0}]^{2}\;,$ (2)
where ${\cal Z}$ is the average distance between the impurity and the center
of mass of the helium droplet
${\cal Z}=\int d\mathbf{r}^{3}\,z\,|\Phi(\mathbf{r})|^{2}-\frac{1}{N}\int
d\mathbf{r}^{3}\,z\,\rho(\mathbf{r})\;.$ (3)
$\lambda_{C}$ is an arbitrary constant. The value $\lambda_{C}\sim$ 1000 K Å-2
has been used in our calculations, which ensures that the desired ${\cal
Z}_{0}$ value is obtained within a 0.1 % accuracy.
We have solved the Euler-Lagrange equations which result from the variations
with respect to $\Psi^{*}$ and $\Phi^{*}$ of the constrained energy Eq. (2),
namely
$-\frac{\hbar^{2}}{2\,m_{He}}\Delta\Psi+\left\\{\frac{\delta{\cal
E}}{\delta\rho}+U_{He}-\lambda_{C}({\cal Z}-{\cal
Z}_{0})\frac{z}{N}\right\\}\Psi=\mu\Psi$ (4)
$-\frac{\hbar^{2}}{2\,m_{Mg}}\Delta\Phi+\left\\{U_{Mg}+\lambda_{C}({\cal
Z}-{\cal Z}_{0})\,z\right\\}\Phi=\varepsilon\Phi\;,$ (5)
where $\mu$ is the helium chemical potential and $\varepsilon$ is the lowest
eigenvalue of the Schrödinger equation obeyed by the Mg atom. The above
coupled equations have to be solved selfconsistently, starting from an
arbitrary but reasonable choice of the unknown functions $\Psi$ and $\Phi$.
The fields $U_{He}$ and $U_{Mg}$ are defined as
$\displaystyle U_{He}(\mathbf{r})$ $\displaystyle=$
$\displaystyle\int\mathrm{d}^{3}\mathbf{r^{\prime}}\,|\Phi(\mathbf{r^{\prime}})|^{2}\,V_{Mg-
He}(|\mathbf{r}-\mathbf{r^{\prime}}|)$ $\displaystyle U_{Mg}(\mathbf{r})$
$\displaystyle=$
$\displaystyle\int\mathrm{d}^{3}\mathbf{r^{\prime}}\,\rho(\mathbf{r^{\prime}})\,V_{Mg-
He}(|\mathbf{r}-\mathbf{r^{\prime}}|)\;.$ (6)
In spite of the axial symmetry of the problem, we have solved the above
equations in three-dimensional (3D) cartesian coordinates. The main reason is
that these coordinates allow to use fast Fourier transformation techniquesFFT
to efficiently compute the convolution integrals entering the definition of
${\cal E}(\rho)$, i.e. the mean field helium potential and the coarse-grained
density needed to compute the correlation term in the He density functional,
Dal95 as well as the fields defined in Eq. (6).
The differential operators in Eqs. (1,4,5) have been discretized using
13-point formulas for the derivatives. Eqs. (4-5) have been solved employing
an imaginary time method;Pre92 some technical details of our procedure are
given in Ref. Anc03, . Typical calculations have been performed using a
spatial mesh step of 0.5 Å. We have checked the stability of the solutions
against reasonable changes in the spatial mesh step.
### II.2 Structure and energetics of Mg-doped helium nanodroplets
Equations (4-5) have been solved for $\lambda_{C}=0$ and several $N$ values,
namely $N=30$, 50, 100, 200, 300, 500, 1000, and 2000\. This yields the ground
state of the drop-impurity complex, and will allow us to study the atomic
shift for selected cluster sizes.
Figure 2 shows the energy of a magnesium atom in a drop, defined as the energy
difference
$S_{N}(Mg)=E(Mg@^{4}He_{N})-E(^{4}He_{N})\;.$ (7)
We compare in Fig. 2 the $S_{N}(Mg)$ values here obtained with those of Ref.
Her07, , where the Mg atom was treated as an infinitely massive particle
-i.e., as a fixed external field acting on the 4He drop. The neglect of the
quantum kinetic energy of the impurity overestimates the impurity solvation
energies by quite a large amount, about 19.7 K for $N=50$ and 18.8 K for
$N=2000$. The value we have found for $S_{50}(Mg)$, $-18.4$ K, compares well
with the DMC result of Ref. Elh08, ($-21$ K), showing that DFT performs quite
well for small clusters,Bar06 far from the regime for which it was
parametrized. The DMC energy found for the same system is $\sim-168.2$ K,Mel05
whereas our DFT result is $\sim-157.0$ K, and the “asymptotic” DMC value for
$S_{200}(Mg)$, $-33.1$ K, compares well with the DFT value, still far from the
limit value corresponding to a very large helium drop (see Fig. 2).
The solvation properties of the Mg atom are determined by a delicate balance
between the different energy terms –surface, bulk and helium-impurity– in Eq.
(1), whose contribution is hard to disentangle and depends on the number of
atoms in the drop, as shown by DMC and DFT calculations. To gain more insight
into the solvation process of magnesium in small 4He droplets, we have
computed the energy of the doped droplet as a function of the impurity
position ${\cal Z}_{0}$.
The bottom panel in Fig. 3 shows $E({\cal Z}_{0})$ for $N=50$, as computed
using the two different Mg-He pair interactions shown in Fig. 1 (the
discussion of the top panel is postponed until Sec. IV.A). It can be seen that
$E({\cal Z}_{0})$ in both cases displays two local minima. In one case (i.e.
for the sligtly less attractive potential in Fig. 1) a “surface” state for the
Mg atom is energetically preferred, while in the other case the impurity
prefers to sit in the interior to the droplet (although not exactly at its
center). These results are in agreement with the DMC calculations of Ref.
Mel05, , where it has been shown that a bimodal distributions for the Mg
radial probability density function with respect to the center-of-mass of the
helium moiety appears for $N\leq 30$.Mel05 More recent DMC calculations
carried out up to $N\sim 200$ dropsElh08 seem to point out that Mg is always
in a surface state, although somewhat beneath the drop surface. Our DFT
calculations yield that Mg is already solvated for $N=200$.
For both pair potentials, the bottom panel of Fig. 3 shows that the two local
minima in $E({\cal Z}_{0})$ are separated by an energy barrier of about 1 K
height, allowing the impurity to temporarily visit, even at the experimental
temperature $T\sim 0.4$ K, the less energetically favored site. This causes
changes in the total energy of the system by less that 1 %, but has a large
effect on the value of the atomic shift. We will address this important issue,
and its consequencies on the computed spectral properties of Mg@HeN in the
next Sections.
Figure 4 shows the helium configurations for the four stationary points
displayed in Fig. 3, namely those corresponding to ${\cal Z}_{0}=0$, 2, 4, and
6 Å. Although the preference for the surface or the solvated state depends on
the He-Mg pair potential used in the calculations, similarly to the case of
other alkali atoms,Her07 this does not seem to be the case for the stationary
points at ${\cal Z}_{0}=2$ and 4 Å that are present in both curves. We have
compared in Fig. 5 the density profiles along the $z$ axis for the ${\cal
Z}_{0}=2(4)$ Å configuration with the profile of the pure 4He drop, finding
that the appearence of these stationary points is related to the position of
the density peak in the first helium solvation shell with respect to a
maximum(minimum) of the density of the pure drop. We see that a
minimum(maximum) in the energy is associated with a constructive(destructive)
interference in the oscillation pattern of the He density. We are lead to
conclude that the interplay between the density oscillations already present
in pure drops, and the solvation shells generated by the impurity plays an
important role in the solvation properties of the Mg atom. This effect is also
present, although to a lesser extent, in Ca-doped 4He nanodroplets.Her08
Eventually, for larger drops Mg becomes fully solvated. We have found that
this is the case whichever of these two potentials we use (see for instance
the bottom panel of Fig. 6). For this reason, the results we discuss in the
following have been obtained with the Mg-He pair potential of Ref. Hin03, ,
unless differently stated.
When the Mg atom is fully solvated, e.g. for $N=1000$, we have found that
$E({\cal Z}_{0})$ grows monotonously as ${\cal Z}_{0}$ increases (i.e. as the
Mg atom approaches the droplet surface). This is shown in the bottom panel of
Fig. 6. To better understand how $E({\cal Z}_{0})$ depends on $N$, we have
plotted in the top panel of Fig. 6 the energy of the $N=1000$ and 2000 doped
drops, referred to their equilibrium value, as a function of the distance from
the dividing surface, i.e. the radius $R_{1/2}$ at which the density of the
pure drop equals $\rho_{b}/2$, $\rho_{b}$ being the liquid density value
($R_{1/2}=r_{0}N^{1/3}$, with $r_{0}=$ 2.22 Å). These radii are 22.2 and 28.0
Å for $N=1000$ and 2000, respectively.
The top panel of Fig. 6 shows that most of the change in $\Delta E$ takes
place in the 15 Å outer region of the drop, irrespective of its size. As a
consequence of the flatness of $E({\cal Z}_{0})$, the magnesium atom is very
delocalized in the radial direction even in relatively small drops. This
delocalization might affect the absorption spectrum of the attached Mg atom,
and also must be explicitly considered in the interpretation of the electron-
impact yield experiments.Ren07 We address this issue in Sec. IV.
## III Excitation spectrum of a Mg atom attached to a 4He drop
Lax methodLax52 offers a realistic way to study the absorption spectrum of a
foreign atom embedded in liquid drops. It makes use of the Franck-Condon
principle within a semiclassical approach, and it has been widely employed to
study the absorption spectrum of atomic dopants attached to 4He drops, see
e.g. Ref. Her08, and references therein. The method is usually applied in
conjunction with the diatomics-in-molecules theory,Ell63 which means that the
atom-drop complex is treated as a diatomic molecule, where the helium moiety
plays a role of the other atom.
In the original formalism, to obtain the line shape one has to carry out an
average on the possible initial states of the system that may be thermally
populated. Usually, this average is not needed for helium drops, as their
temperature, about $0.4$ K,Toe04 is much smaller than the vibrational
excitation energies of the Mg atom in the mean field represented by the second
of Eqs. (6). In small helium droplets thermal effects can show up in the Mg
absorption spectrum due to the high mobility of the atom. For large drops,
thermal motion plays a minor role, as the Mg atom hardly gets close enough to
the drop surface to have some effect on the line shape. In this case, however,
dynamical deformations of the cavity around the impurity may be relevant.Ler93
; Kin96
The line shape for electronic transitions from the ground state $(gs)$ to the
excited state $(ex)$ in a condensed phase system can be written as
$I(\omega)\propto\int\mathrm{d}t~{}\mathrm{e}^{-i\omega
t}\langle\Psi^{gs}|D_{ge}^{\dagger}~{}\mathrm{e}^{\frac{it}{\hbar}\mathcal{H}_{ex}}~{}D_{ge}~{}\mathrm{e}^{-\frac{it}{\hbar}\mathcal{H}_{gs}}|\Psi^{gs}\rangle\;,$
(8)
where $D_{\mathrm{ge}}$ is the matrix element of the electric dipole operator,
$|\Psi^{gs}\rangle$ is the ground state of the system, and $\mathcal{H}_{gs}$
and $\mathcal{H}_{ex}$ are the Hamiltonians that describe the ground and
excited states of the system respectively.
$I(\omega)$ is evaluated using the Born-Oppenheimer approximation and the
Franck-Condon principle, whereby the heavy nuclei do not change their
positions or momenta during the electronic transition. If the excited electron
belongs to the impurity, the helium cluster remains frozen, so that the
relevant coordinate is the relative position $\mathbf{r}$ between the cluster
and the impurity. This principle amounts to assuming that $D_{\mathrm{ge}}$ is
independent of the nuclear coordinates. Taking into account that
$\mathrm{e}^{-\frac{it}{\hbar}\mathcal{H}_{gs}}|\Psi^{gs}\rangle=\mathrm{e}^{-it\omega_{gs}}|\Psi^{gs}\rangle$
and projecting on eigenstates of the orbital angular momentum of the excited
electron $|m\rangle$ one obtains
$I(\omega)\propto|D_{\mathrm{ge}}|^{2}\sum_{m}\int\mathrm{d}t~{}\mathrm{e}^{-i(\omega+\omega^{\mathrm{gs}}_{X})t}\int\mathrm{d}^{n}[\alpha]\int\mathrm{d}^{3}\mathbf{r}~{}{\psi_{X}^{\mathrm{gs}}(\mathbf{r},[\alpha])}^{*}~{}\mathrm{e}^{\frac{it}{\hbar}H^{\mathrm{ex}}_{m}(\mathbf{r},[\alpha])}~{}\psi_{X}^{\mathrm{gs}}(\mathbf{r},[\alpha])\;,$
(9)
where $\hbar\omega^{\mathrm{gs}}_{X}$ and
$\psi_{X}^{\mathrm{gs}}(\mathbf{r},[\alpha])$ are the energy and the wave
function of the ro-vibrational ground state of the frozen helium-impurity
system, and $H^{\mathrm{ex}}_{m}(\mathbf{r},[\alpha])$ is the ro-vibrational
excited Hamiltonian with potential energy
$V^{\mathrm{ex}}_{m}(\mathbf{r},[\alpha])$ determined by the electronic energy
eigenvalue, as obtained for a $p\leftarrow s$ transition. At this point, we
have introduced the variables $[\alpha]$ to represent the degrees of freedom
needed to describe possible deformations of the system, corresponding to the
zero point oscillations of the helium bubble around the impurity. If this
effect is neglected, the deformation parameters $[\alpha]$ are dropped and the
ground state wave function $\psi_{X}^{\mathrm{gs}}$ coincides with the Mg
wavefunction $\Phi$ found by solving Eq. (5).
If the relevant excited states for the transition have large quantum numbers,
they can be treated as approximately classical using the averaged energy
$\hbar\omega^{m}_{\nu}\approx V^{\mathrm{ex}}_{m}(\mathbf{r},[\alpha])$ which
is independent of $\nu$. In this case we obtain the expression
$\displaystyle I(\omega)$ $\displaystyle\propto$
$\displaystyle\sum_{m}\int\mathrm{d}^{n}[\alpha]\int\mathrm{d}^{3}\mathbf{r}~{}|\psi_{X}^{\mathrm{gs}}(\mathbf{r},[\alpha])|^{2}\delta(\omega+\omega^{\mathrm{gs}}_{X}-V^{\mathrm{ex}}_{m}(\mathbf{r},[\alpha])/\hbar)$
(10) $\displaystyle=$
$\displaystyle\hbar\int\mathrm{d}^{n}[\alpha]\int_{\Omega_{m}(\omega)}\;\mathrm{d}^{2}\mathbf{r}~{}\frac{|\psi_{X}^{\mathrm{gs}}(\mathbf{r},[\alpha])|^{2}}{|\mathbf{\nabla}V^{\mathrm{ex}}_{m}(\mathbf{r},[\alpha])|}\,,$
where $\Omega_{m}(\omega)$ is the surface defined by the equation
$\omega+\omega^{\mathrm{gs}}_{X}-V^{\mathrm{ex}}_{m}(\mathbf{r},[\alpha])/\hbar=0$.
If the atom is in bulk liquid helium, or at the center of the drop, the
problem has spherical symmetry and the above equation reduces to
$\displaystyle I(\omega)$ $\displaystyle\propto$ $\displaystyle
4\pi\sum_{m}\int\mathrm{d}^{n}[\alpha]\int\mathrm{d}r~{}|r~{}\psi_{X}^{\mathrm{gs}}(r,[\alpha])|^{2}\delta(\omega+\omega^{\mathrm{gs}}_{X}-V^{\mathrm{ex}}_{m}(r,[\alpha])/\hbar)$
(11) $\displaystyle=$ $\displaystyle
4\pi\hbar\sum_{m}\int\mathrm{d}^{n}[\alpha]\left|\frac{[r~{}\psi_{X}^{\mathrm{gs}}(r,[\alpha])]^{2}}{d{V}^{~{}\mathrm{ex}}_{m}(r,[\alpha])/dr}\right|_{r=r_{m}(\omega)}\;,$
where $r_{m}(\omega)$ is the root of the equation
$\omega+\omega^{\mathrm{gs}}_{X}-V^{\mathrm{ex}}_{m}(r,[\alpha])/\hbar=0$. In
the non-spherical case, we have evaluated $I(\omega)$ using the first
expression in Eq. (10).
The potential energy surfaces $V^{\mathrm{ex}}_{m}(\mathbf{r},[\alpha])$
needed to carry out the calculation of the atomic shifts have been obtained in
the pairwise sum approximation, using the $V_{\Pi}(r)$ and $V_{\Sigma}(r)$ Mg-
He adiabatic potentials from Ref. Mel05, . In cartesian coordinates, and
assuming that the He-impurity spin-orbit interaction is negligible for
magnesium, the eigenvalues of the symmetric matrix
$U_{ij}(\mathbf{r},[\alpha])=\int\mathrm{d}^{3}\mathbf{r^{\prime}}\rho(\mathbf{r^{\prime}}+\mathbf{r},[\alpha])\left\\{V_{\Pi}(r^{\prime})\delta_{ij}+[V_{\Sigma}(r^{\prime})-V_{\Pi}(r^{\prime})]\frac{x^{\prime
i}~{}x^{\prime j}}{r^{\prime 2}}\right\\}$ (12)
are the $V^{\mathrm{ex}}_{m}(\mathbf{r},[\alpha])$ potentials which define the
potential energy surfaces (PES) as a function of the distance between the
center-of-mass of the droplet and that of the impurity.
For spherical geometries, Eq. (12) is diagonal with matrix elements (in
spherical coordinates)
$\displaystyle\lambda_{i}(r,[\alpha])$ $\displaystyle\equiv$ $\displaystyle
U_{ii}(r,[\alpha])=2\pi\int\int r^{\prime
2}\sin\theta^{\prime}\mathrm{d}\theta^{\prime}\mathrm{d}r^{\prime}\rho(\sqrt{r^{\prime
2}+r^{2}+2r^{\prime}r\cos\theta^{\prime}},[\alpha])$ (13)
$\displaystyle\times$
$\displaystyle\left\\{V_{\Pi}(r^{\prime})+[V_{\Sigma}(r^{\prime})-V_{\Pi}(r^{\prime})]\left[\frac{1}{2}(\delta_{i1}+\delta_{i2})\sin^{2}\theta^{\prime}+\delta_{i3}\cos^{2}\theta^{\prime}\right]\right\\}$
In this case, two of the PES are degenerate, namely
$\lambda_{1}(r,[\alpha])=\lambda_{2}(r,[\alpha])\neq\lambda_{3}(r,[\alpha])$.notaCa
This holds true for $r\neq 0$, and it is relevant when we take into account
the delocalization of the impurity inside the bubble due to its quantum
motion. Otherwise, since at $r=0$ all the $\lambda_{i}$ coincide, they are
threefold degenerate.
## IV Results for the absorption spectrum of magnesium atoms
The problem of obtaining the atomic shift of magnesium in a helium drop has
been thus reduced to that of the dopant in the 3D trapping potentials
corresponding to the ground state and $P$ excited states. We consider first
the $[\alpha]=0$ case (i.e. no zero-point deformations of the helium cavity
hosting the impurity). The general situation, in particular the homogeneous
width calculation, is presented later on in this Section.
If $[\alpha]=0$, the model is expected to yield at most the energies of the
atomic transitions, but not the line shapes since the impurity-droplet
excitation interactions as well as inhomogeneous broadening resulting from
droplet size distributions, laser line width and similar effects are not
included. These limitations are often overcome by introducing line shape
functions or convoluting the calculated lines with some effective line
profiles.Sti96 ; Bue07 We discuss here two illustrative examples, namely the
atomic shift of magnesium in $N=50$ and 1000 nanodroplets. The homogeneous
width is calculated in Subsection B for the $N=1000$ droplet.
For Mg@4He50, the calculated shifts at ${\cal Z}_{0}=0$ (spherical
configuration), 2 and 6 Å are 500, 450 and 281 cm-1, respectively (281.3,
281.7 and 283.0 nm wavelengths). No experimental information is available for
such a small drop. Contrarily, the absorption spectrum of the $3s3p$
1P${}_{1}\leftarrow 3s^{2}$ 1S0 transition of Mg atoms attached to large
helium drops has been measured,Reh00 displaying a broad peak strongly blue-
shifted from its position in the gas-phase. This spectrum is remarkably close
to the one obtained by Moriwaki and MoritaMor99 in bulk liquid helium, and
hence it has been concluded that Mg is in the interior of the 4He droplet. We
want to point out that, while the absorption line in liquid helium was
attributed to a single broad peak of energy 281.5 nm (35 524 cm-1), i.e., a
shift of 474 cm-1,Mor99 in large drops a similar line profileNote1 was
fitted by two Gaussians centered at 35 358 and 35 693 cm-1 (282.8 and 280.2 nm
wavelengths) respectively, i.e., shifted 307 cm-1 and 642 cm-1 from the gas-
phase line.Reh00 The origin of the two peaks was attributed to the splitting
of the degenerate $\Pi$ state by dynamical quadrupole deformations of the
cavity surrounding the dopant, since this argument had qualitatively explained
similar doubly-shaped $D_{2}$ excitation spectra of Rb and Cs atoms in liquid
4He due to a quadrupole oscillation of the helium bubble (dynamic Jahn-Teller
effect).Kin96
LIF experiments on the heavier alkaline earth Ca and Sr in large 4He
dropletsSti97 have disclosed the existence of strong blue-shifted, broad
peaks with no apparent structure, although it cannot be discarded that this
broad line could be a superposition of unresolved peaks. The same happens for
Ba.Sti99 The surface location of Ca, Sr, and Ba in these drops has been
further confirmed by DFT calculations.Her07 ; Her08 It is also interesting to
recall that LIF experiments on Ca atoms in liquid 4He and 3He have found a
broad line in the region of the $4s4p$ 1P${}_{1}\leftarrow 4s^{2}$ 1S0
transition with no apparent splitting,Mor05 contrarily to the case of Mg.
Since Mg is fully solvated in the $N=1000$ drop, the calculated atomic shift
$\Delta\omega$ may be sensibly compared with the experimental data where drops
with $N$ in the $10^{3}-10^{4}$ range are studied. We have obtained
$\Delta\omega=659.0$ cm-1 (280.0 nm wavelength); this peak nearly corresponds
to the Gaussian that takes most of the intensity of the absorption line (about
87 %).Reh00 We have carried out a detailed analysis for this drop,
determining the equilibrium structure of Mg@4He1000 as a function of ${\cal
Z}_{0}$, and have used it to evaluate $\Delta\omega$. The results are
displayed in Table 1, showing the actual sensitivity of the absorption
spectrum to the Mg atom environment.
The impurity-drop excitations will determine the homogeneous width of the
spectral line, and the population of excited states may be relevant given the
limit temperature attained by the droplets.Sti06 ; Toe04 ; Bri90 In this
context, the relevant excitation modes of the helium bubble are radial
oscillations of monopole type (breathing modes), and multipole shape
oscillations about the equilibrium configuration, as well as displacements of
the helium bubble inside the droplet. We will address these issues in the
following Subsections.
### IV.1 Thermal motion and angular momentum effects
To describe the displacement of the helium bubble inside the droplet, we have
fitted the $E({\cal Z}_{0})$ curve of the Mg@4He1000 system to a parabola, and
have obtained the excitation energy $\hbar\omega$ for this 3D isotropic
harmonic oscillator. The hydrodynamic mass of the impurity atom has been
estimated by its bulk liquid helium value, $M^{*}\sim 40$ a.u., obtained by
the method outlined in Appendix A, Eq. (52). We find $\hbar\omega=0.1$ K,
indicating that thermal motion, i.e., the population of the excited states of
the “mean field” $E({\cal Z}_{0})$ is important at the experimental
temperature $T=0.4$ K, and may produce observable effects in the absorption
spectrum and the electron-impact ionization yield.
To describe in more detail the delocalization of the Mg atom inside the drop,
we have used an effective Hamiltonian where we interpret ${\cal Z}_{0}$ as the
radial distance $R$ between the impurity and center of mass of the helium
moiety, and $E({\cal Z}_{0})$ as the “potential energy” $V(R)$ associated with
this new degree of freedom of the impurity in the drop. Namely,
${\cal
H}=\frac{\hat{P}^{2}}{2M^{*}}+V(R)=\frac{\hat{P}_{R}^{2}}{2M^{*}}+\frac{\hat{L}^{2}}{2M^{*}R^{2}}+V(R)\;,$
(14)
where $M^{*}$ is the Mg hydrodynamic mass. In the canonical ensemble, the
total probability distribution $W$ as a function of $R$ can be written as
$\begin{array}[]{rl}\displaystyle
W(R)&\displaystyle=Q^{-1}\int_{0}^{R}\mathrm{d}R^{\prime}R^{\prime
2}\int\mathrm{d}\Omega^{\prime}\sum_{n\ell m}\langle\psi_{n\ell
m}|\mathbf{R}^{\prime}\rangle e^{-{\cal
H}(R^{\prime})/k_{B}T}\langle\mathbf{R}^{\prime}|\psi_{n\ell m}\rangle\\\
&\displaystyle=Q^{-1}\int_{0}^{R}R^{\prime
2}\mathrm{d}R^{\prime}\sum_{n\ell}(2\ell+1)e^{-E_{n\ell}/k_{B}T}|\phi_{n\ell}(R^{\prime})|^{2}\;,\end{array}$
(15)
where the partition function is defined as $Q=Tr\left(e^{-{\cal
H}/k_{B}T}\right)=\sum_{n\ell}(2\ell+1)e^{-E_{n\ell}/k_{B}T}$, being $k_{B}$
the Boltzmann constant. The radial probability density $w$ is
$w(R)=\frac{\mathrm{d}W}{\mathrm{d}R}=Q^{-1}R^{2}\sum_{n\ell}(2\ell+1)e^{-E_{n\ell}/k_{B}T}|\phi_{n\ell}(R)|^{2}\;.$
(16)
This expression has been evaluated for $N=50$ and 1000 by solving the
Schrödinger equation for the Hamiltonian Eq. (14) to obtain the orbitals
$\phi_{n\ell}$ and eigenenergies $E_{n\ell}$. For larger $N$ values, we have
used the semiclassical approximation
$E_{n\ell}\rightarrow\frac{\hat{p}_{R}^{2}}{2M^{*}}+V_{\mathrm{eff}}(R)$,
where the effective potential is
$V_{\mathrm{eff}}(R)=\frac{\hbar^{2}\ell(\ell+1)}{2M^{*}R^{2}}+V(R)\;.$ (17)
Integrating $p_{R}$ in phase space, we obtain for the probability density
$w(R)=Q^{-1}R^{2}\exp\left(-\frac{V(R)}{k_{B}T}\right)\sum_{\ell}(2\ell+1)\exp\left[-\frac{1}{k_{B}T}\frac{\hbar^{2}\ell(\ell+1)}{2M^{*}R^{2}}\right]$
(18)
with the normalization
$Q=\int_{0}^{\infty}dRR^{2}\exp\left(-\frac{V(R)}{k_{B}T}\right)\sum_{\ell}(2\ell+1)\exp\left[-\frac{1}{k_{B}T}\frac{\hbar^{2}\ell(\ell+1)}{2M^{*}R^{2}}\right]$.
Lacking a better choice, we have weighted any possible angular momentum value
with a Boltzmann energy factor. It has been shownLehm04 that some of the
angular momentum deposited in the droplet during the pickup process may be
kept in the impurity atom, resulting in a different angular momentum
distribution than the Boltzmann one. This could yield that some Mg atoms are
actually closer to the drop surface.
If the angular momentum associated with the motion of the magnesium atom
-whose “radius” is $\sim 5$ Å, see Fig. 1\- is such that Mg can be some 10 Å
beneath the drop surface, the shift of the absorption line would be hardly
distinguishable from that of the totally solvated case -as seen in Table 1. At
the same time, the electron-energy dependence of the Mg+ yield observed in
electron impact ionization experimentsRen07 (and which was considered as an
evidence of a surface location of Mg atoms on 4He droplets) could indeed be
due to Penning ionization of the impurity in a collision with a metastable He∗
atom that occupies a surface bubble state in the drop, instead of being due to
the transfer of a positive hole (He+) to the Mg atom, which is the primary
ionization mechanism when the impurity is very attractive and resides in the
deep bulk of the droplet.
We show in the top panel of Fig. 3 the probability densities $w(R)$ at $T=0.4$
K corresponding to the configurations displayed in the bottom panel.notem
Similarly, the top panel of Fig. 6 shows that for Mg@4He1000 and Mg@4He2000,
if thermal motion is taken into account and the impurity retains some of the
pick-up angular momentum, the maximum density probability of Mg is at $\sim
15$ Å beneath the drop surface in both cases. To obtain it, we have taken for
$M^{*}$ the bulk value 40 a.u. As seen from Table 1, the absorption line shift
changes by a small 2% with respect to the $R=0$ configuration. The values of
the angular momentum corresponding to these maximum density probabilities are
$\langle L^{2}\rangle^{1/2}\sim 9\hbar$ and $\sim 10\hbar$, respectively.NoteL
For a $N=10\,000$ drop,Ren07 whose radius is $R_{1/2}=47.8$ Å, we have
extrapolated inwards the $\Delta E$ curves of the calculated $N=1000$ and 2000
drops, and have obtained from it the probability distribution displayed in
Fig. 6. Its maximum is at $\sim 18.5$ Å beneath the surface, with $\langle
L^{2}\rangle^{1/2}\sim 28\hbar$.
Finally, using the effective potential of Eq. (17) we have determined that for
$N=1000(2000)$, a Mg atom with $\ell\sim 24\hbar(44\hbar)$ is in an
“equilibrium position” some 10 Å beneath the drop surface. For the $N=10\,000$
drop, this value is $\ell\sim 148\hbar$. These values look reasonable, and Mg
atoms holding this angular momentum or larger might thus be the origin of the
primary electron-collision ionization yield by the Penning process, without
questioning the conclusion drawn from LIF experiments that magnesium is fully
solvated in 4He drops.
### IV.2 Homogeneous width from shape deformations of the helium bubble
We have shown that for large drops, the Mg atom is fully solvated and its
thermal motion only produces small changes in the absorption shift. This
allows us to decouple the effect of the translational motion of the helium
bubble on the absorption line, from that of its shape fluctuation. Moreover,
we can address shape fluctuations in the much simpler spherically symetric
ground state, when magnesium is located at the center of the drop.
To quantify the effect of these fluctuations, we have first used the spherical
cap modelAnc95 to estimate the excitation energies of the helium bubble
around the impurity in liquid helium. To this end, we have fitted the
$V_{\text{Mg-He}}$ potential to a Lennard-Jones potential with depth
$\varepsilon=7$ K at a minimum distance $r_{min}=5$ Å. Minimizing the total
energy within this model yields a configuration with an equilibrium radius of
$R_{0}=0.97\times 2^{-1/6}r_{min}$, that we approximate by
$R_{0}=2^{-1/6}r_{min}$ to obtain the excitation energies.
Deformations of the 4He around the Mg atom are modeled asWil64 ; Rin80
$R(\Omega)=R_{\alpha}\left[1+\alpha_{0}+\sum_{\lambda=2}^{\infty}\sum_{\mu=-\lambda}^{\lambda}\alpha_{\lambda\mu}Z_{\lambda\mu}(\Omega)\right]\;,$
(19)
where $R$ is the radius of the bubble cavity hosting the solvated Mg atom,
$\Omega$ represents the solid angle variables $(\theta,\phi)$,
$Z_{\lambda\mu}(\Omega)$ is a real spherical harmonic, and
$\alpha_{\lambda\mu}$ is the amplitude of the $\lambda\mu-$multipole
deformation. The condition
$R_{\alpha}=R_{0}(1-\frac{1}{4\pi}\sum_{\lambda\mu}|\alpha_{\lambda\mu}|^{2})$
ensures the conservation of the number of particles up to second order in
$\alpha_{\lambda\mu}$. The dipole mode amplitude $\alpha_{1\mu}$ is absent
since, for an incompressible fluid, it corresponds to a translation of the
bubble, and this has been considered in the previous Subsection.
If $S[R(\Omega)]$ represents the bubble surface and $\sigma$ the surface
tension, the energy for a large drop can be written as
$\displaystyle E$ $\displaystyle=$ $\displaystyle\sigma
S[R(\Omega)]+\rho_{b}\,\varepsilon\int_{R(\Omega)}^{\infty}\mathrm{d}^{3}\mathbf{r}\left[\left(\frac{r_{min}}{r}\right)^{12}-2\left(\frac{r_{min}}{r}\right)^{6}\right]$
(20) $\displaystyle\simeq$ $\displaystyle
E_{eq}+\frac{1}{2}C_{0}|\alpha_{0}|^{2}+\frac{1}{2}\sum_{\lambda=2}^{\infty}C_{\lambda}\sum_{\mu=-\lambda}^{\lambda}|\alpha_{\lambda\mu}|^{2}\;,$
where $C_{0}=8\pi\sigma R_{0}^{2}\left(1+12\lambda_{A}\right)$ and
$C_{\lambda}=\sigma R_{0}^{2}[(\lambda-1)(\lambda+2)+6\lambda_{A}]$ are the
stiffness parameters, and $\lambda_{A}=\rho_{b}\varepsilon
2^{-1/6}r_{min}\sigma^{-1}$ is the impurity-He solvation parameter.Anc95 The
mass parameters are $B_{0}=4\pi\rho_{b}m_{He}R_{0}^{5}$ and
$B_{\lambda}=\rho_{b}m_{He}R_{0}^{5}/(\lambda+1)$,Fow68 and the excitation
energies are determined from
$\hbar\omega_{\lambda}=\hbar\sqrt{C_{\lambda}/B_{\lambda}}$, yielding
$\hbar\omega_{0}=10.2$ K for the breathing mode, and $\hbar\omega_{2}=9.8$ K
for the quadrupole mode. Given the droplet temperature of 0.4 K, we conclude
that only the ground state is populated. The mean amplitude of the shape
oscillations is estimated from the variance
$\gamma_{\lambda}=\hbar^{1/2}\left(B_{\lambda}C_{\lambda}\right)^{-1/4}/2$,
giving $\gamma_{0}=0.03$ and $\gamma_{2}=0.15$. This model thus yields that
the bubble can experience monopole oscillations of $\sim$3% amplitude, and
quadrupole deformations of $\gamma_{2}\sqrt{3/4\pi}\sim 8$% amplitude.
Amplitudes of this order have been determined within the atomic bubble model
for Cesium atoms in liquid helium.Kin96 Since their effect in the absorption
spectrum is expected to be relevant, we have undertanken a more refined
calculation within DFT taking Mg@4He1000 as a case study.
For helium droplets, we have described bubble deformations in a way similar as
in Refs. Wil64, ; Rin80, ; Kin96, ; Fow68, , namely, if $\rho_{0}(r)$ is the
helium spherical ground state density, deformations are introduced as
$\rho(\mathbf{r},t)=\rho_{0}[R(\mathbf{r},t)]\,{\cal K}^{-1}$, with
$R(\mathbf{r},t)=r+\alpha_{0}(t)+\sum_{\lambda=2}^{\infty}\sum_{\mu=-\lambda}^{\lambda}\alpha_{\lambda\mu}(t)Z_{\lambda\mu}(\widehat{r})\;,$
(21)
where the real spherical harmonics are normalized as $\langle
Z_{\lambda\mu}|Z_{\lambda^{\prime}\mu^{\prime}}\rangle=\frac{4\pi}{2\lambda+1}\delta_{\lambda\lambda^{\prime}}\delta_{\mu\mu^{\prime}}$
for convenience, and the normalization ${\cal
K}=N^{-1}\int\mathrm{d}^{3}\mathbf{r}\rho_{0}[R(\mathbf{r},t)]$ ensures
particle number conservation. If the Mg wave function follows adiabatically
the helium density deformation, it can be shown that to second order in
$\alpha_{\lambda\mu}$, the total energy of the system can be written as
$\displaystyle E([\dot{\alpha}],[\alpha])$ $\displaystyle\simeq$
$\displaystyle E_{gs}+\frac{1}{2}M^{*}_{0}\dot{\alpha}_{0}^{2}+2\pi
E^{(2)}_{0}\alpha_{0}^{2}$ (22)
$\displaystyle+\sum_{\lambda=2}^{\infty}\sum_{\mu=-\lambda}^{\lambda}\left\\{\frac{1}{2}M^{*}_{\lambda}\dot{\alpha}_{\lambda\mu}^{2}+\frac{2\pi}{2\lambda+1}E^{(2)}_{\lambda}\alpha_{\lambda\mu}^{2}\right\\}\;,$
being $E_{gs}$ the ground state energy, $M^{*}_{\lambda}$ the hydrodynamic
mass asociated with the $\lambda$ mode, and $E^{(2)}_{\lambda}$ the second
derivative of the total energy with respect to $\alpha_{\lambda\mu}$. This
equation represents the Hamiltonian of a set of uncoupled harmonic
oscillators, whose quantization yields a ground state to whose energy each
mode contributes with
$\varepsilon_{\lambda\mu}=\frac{1}{2}\hbar\omega_{\lambda}$, with
$\omega_{\lambda}=\sqrt{\frac{4\pi
E^{(2)}_{\lambda}}{(2\lambda+1)M^{*}_{\lambda}}}$, and a ground state wave
function
$\psi([\alpha])=\left(\frac{M^{*}_{0}\omega_{0}}{\pi\hbar}\right)^{1/4}e^{-\frac{M^{*}_{0}\omega_{0}}{2\hbar}\alpha_{0}^{2}}\prod_{\lambda=2}^{\infty}\left(\frac{M^{*}_{\lambda}\omega_{\lambda}}{\pi\hbar}\right)^{1/4}e^{-\frac{M^{*}_{\lambda}\omega_{\lambda}}{2\hbar}\alpha_{\lambda}^{2}}\;,$
(23)
where
$\alpha_{\lambda}^{2}\equiv\sum_{\mu=-\lambda}^{\lambda}\alpha_{\lambda\mu}^{2}$.
Details are given in Appendix A.
We have computed the hydrodynamic masses assuming that the drop is large
enough to use Eq. (52). We have obtained $M^{*}_{0}=15.0m_{He}+0.28m_{Mg}\sim
66.7$ a.u. and $M^{*}_{2}=1.9m_{He}+0.56m_{Mg}\sim 21.0$ a.u. In actual
calculations, instead of using Eq. (39), the energies $E^{(2)}_{0}$ and
$E^{(2)}_{2}$ have been numerically obtained by computing the total energy of
the system for different small values of $\alpha_{0}$ and $\alpha_{2}$. This
has been carried out by numerically introducing the desired deformation
parameter into the ground state density and renormalizating it, solving next
the Schrödinger equation for the Mg atom [Eq. (5)] to determine the ground
state of the impurity, and computing the total energy of the system from Eq.
(1). Fitting these curves to a parabola, we have obtained $E^{(2)}_{0}=49.7$
KÅ-2 and $E^{(2)}_{2}=16.8$ KÅ-2. We have then calculated the ground state
energies $\hbar\omega_{\lambda}/2$ and deformation mean amplitudes
$\gamma_{\lambda}$, obtaining $\hbar\omega_{0}/2=10.6$ K and
$\hbar\omega_{2}/2=6.3$ K, with mean amplitudes $\gamma_{0}=0.18$ Å
($\sim$3.7%) and $\gamma_{2}=0.42$ Å ($\sim$8.5%).
To quantitatively determine the effect of these deformations on the absorption
spectrum, we have developed Eq. (10) to first order in the deformation
parameters, and have explicitly shown that to this order, only the breathing
and quadrupole modes affect the dipole absorption spectrum. The details are
given in Appendix B, where we show that the breathing mode affects the shift
and shape of the line, whereas quadrupole modes only affect the shift.
Consequently, we restrict in Eq. (11) the deformation parameters needed to
properly describe the homogeneous broadening of the absorption dipole line,
namely
$\mathrm{d}[\alpha]\rightarrow\mathrm{d}\alpha_{0}\,\mathrm{d}^{5}\alpha_{2}$
and
$\psi_{X}^{\mathrm{gs}}(r,[\alpha])\rightarrow\psi(\alpha_{0},\alpha_{2})\Phi(r,\alpha_{0})$,
being $\Phi(r,\alpha_{0})$ the wave function of the Mg atom for a given
$\alpha_{0}$ value, and compute the spectrum as
$\displaystyle I(\omega)$ $\displaystyle\propto$ $\displaystyle
4\pi\int\mathrm{d}\alpha_{0}\,\mathrm{d}^{5}\alpha_{2}|\psi(\alpha_{0},\alpha_{2})|^{2}\sum_{m}\int\mathrm{d}r~{}|r~{}\Phi(r,\alpha_{0})|^{2}$
(24)
$\displaystyle\times\delta\left[\omega+\omega^{\mathrm{gs}}_{X}(\alpha_{0})-\frac{1}{\hbar}V^{\mathrm{ex}}_{m}(r,\alpha_{0},[\alpha_{2}])\right]$
$\displaystyle=$ $\displaystyle
4\pi\int\mathrm{d}\alpha_{0}\,\mathrm{d}^{5}\alpha_{2}|\psi(\alpha_{0},\alpha_{2})|^{2}\hbar\sum_{m}\left|\frac{[r~{}\Phi(r,\alpha_{0})]^{2}}{d{V}^{\mathrm{ex}}_{m}(r,\alpha_{0},[\alpha_{2}])/dr}\right|_{r=r_{m}(\omega)}\;,$
where
$[\alpha_{2}]=\\{\alpha_{2-2},\alpha_{2-1},\alpha_{20},\alpha_{21},\alpha_{22}\\}$,
$V^{\mathrm{ex}}_{m}(r,\alpha_{0},[\alpha_{2}])$ are the eigenvalues
$\lambda_{m}(r,[\alpha])$ of the excited potential matrix of Eq. (80), and
$r_{m}(\omega)$ is the root of the equation
$\omega+\omega^{\mathrm{gs}}_{X}(\alpha_{0})-V^{\mathrm{ex}}_{m}(r,\alpha_{0},[\alpha_{2}])/\hbar=0$.
Expression (24) has been integrated using a Monte Carlo method. We have sorted
$M=10^{6}$ sets of values
$[\alpha]^{i}=\\{\alpha_{0}^{i},\alpha_{2-2}^{i},\alpha_{2-1}^{i},\alpha_{20}^{i},\alpha_{21}^{i},\alpha_{22}^{i}\\}$
using the square of the wave function of Eq. (23) as probability density.
Next, for each set we have found the eigenvalues
$V^{\mathrm{ex}}_{m}(r,\alpha_{0}^{i},[\alpha_{2}]^{i})$ of the $U_{i,j}$
matrix that define the potential energy surfaces and have used a trapezoidal
rule to evaluate the integrals
$I_{m}(\omega,[\alpha]^{i})=4\pi\int\mathrm{d}r~{}|r~{}\Phi(r,\alpha_{0}^{i})|^{2}\delta\left[\omega+\omega^{\mathrm{gs}}_{X}(\alpha_{0}^{i})-\frac{1}{\hbar}V^{\mathrm{ex}}_{m}(r,\alpha_{0}^{i},[\alpha_{2}]^{i})\right]$
(25)
using a discretized representation of the delta function.Her08 Finally, we
have obtained the spectrum as
$I(\omega)\propto\frac{1}{M}\,\sum_{i=1}^{M}\sum_{m}I_{m}(\omega,[\alpha]^{i}).$
(26)
Figure 7 shows the absorption spectrum of one Mg atom attached to 4He1000 in
the vicinity of the $3s3p$ 1P${}_{1}\leftarrow 3s^{2}$ 1S0 transition when
homegeneous broadening is considered. We have decomposed the absorption line
into its three components, the higher frequency component being the $\Sigma$
one. The starred vertical line represents the gas-phase transition, and the
experimental curve, adapted from Ref. Reh00, , has been vertically offset for
clarity. Also shown is the absorption spectrum obtained by neglecting
homogeneous broadening (hatched region). This figure shows that the both the
energy and width of the absorption peak that takes most of the experimental
intensity are correctly described by our calculations.
## V Two magnesium atoms attached to a 4He drop
The attachment of magnesium atoms in 4He droplets has been recently addressed
using resonant two-photon-ionization.Prz07 In particular, the authors of Ref.
Prz07, have obtained the absorption spectrum for drops doped with different,
selected numbers of Mg atoms. From their measurements it appears that two main
features contribute to the observed line shapes, one peaked at about 279 nm,
and another at about 282 nm. This is in agreement with the results of Refs.
Mor99, ; Reh00, (we recall that actually, the two peaks were not resolved by
the authors of the bulk liquid experiment). The structure at 282 nm, however,
only appears if the droplet contains more than one Mg atom. Thus the two-peak
structure cannot be due to the splitting of the absorption line due to
dynamical quadrupole deformations of the helium bubble around the impurity, as
previously believed. We have indeed shown in the previous Section that this
coupling only produces a broad peak, in good agreement with the results of
Ref. Prz07, for helium drops containing just one Mg atom.
Another interesting observation reported in Ref. Prz07, is that their
experimental results for multi-atom doped 4He droplets are not consistent with
the formation of compact, metallic Mg clusters inside the 4He droplet. The
magnesium atoms in the droplet appear instead to be relatively isolated from
each other, showing only a weak interaction and leading to the 282 nm shift in
the observation.
To confirm this scenario and find an explanation for the origin of the low
energy component in the absorption peak, we have carried out DFT calculations
to determine the structure of a two-magnesium doped 4He drop. Our goal is to
verify whether the helium density oscillation around a magnesium atom may
result in an energy barrier preventing the Mg atoms from merging into a Mg2
dimer, as suggested by Przystawik et al.Prz07 To obtain the structure of two
Mg atom in a 4He drop, we have minimized the energy of the system written as
$\displaystyle E[\Psi,\Phi_{1},\Phi_{2}]$ $\displaystyle=$
$\displaystyle\frac{\hbar^{2}}{2\,m_{He}}\int\mathrm{d}^{3}\mathbf{r}\,|\nabla\Psi(\mathbf{r}\,)|^{2}+\int\mathrm{d}^{3}\mathbf{r}\,{\cal
E}(\rho)$ (27) $\displaystyle+$
$\displaystyle\frac{\hbar^{2}}{2\,m_{Mg}}\int\mathrm{d}^{3}\mathbf{r}\,\\{|\nabla\Phi_{1}(\mathbf{r}\,)|^{2}+|\nabla\Phi_{2}(\mathbf{r}\,)|^{2}\\}$
$\displaystyle+$
$\displaystyle\int\int\mathrm{d}^{3}\mathbf{r}\,\mathrm{d}^{3}\mathbf{r^{\prime}}\,\\{|\Phi_{1}(\mathbf{r})|^{2}+|\Phi_{2}(\mathbf{r})|^{2}\\}\,V_{Mg-
He}(|\mathbf{r}-\mathbf{r^{\prime}}|)\,\rho(\mathbf{r^{\prime}})$
$\displaystyle+$
$\displaystyle\int\int\mathrm{d}^{3}\mathbf{r}\,\mathrm{d}^{3}\mathbf{r^{\prime}}\,|\Phi_{1}(\mathbf{r})|^{2}\,V_{Mg-
Mg}(|\mathbf{r}-\mathbf{r^{\prime}}|)\,|\Phi_{2}(\mathbf{r^{\prime}})|^{2}\;,$
where $V_{Mg-Mg}(|\mathbf{r}-\mathbf{r^{\prime}}|)$ is the Mg-Mg pair
potential of Ref. Gue92, , and the other ingredients have the same meaning as
in Eq. (1).
There are at least two additional effects which are not considered when
modeling the Mg-Mg interaction via the pair-potential in vacuum, as implied in
the above expression. The first is due to three-body (and higher) correlation
effects involving the 4He atoms surrounding the Mg pair: these should exert an
additional, albeit small, screening effect due to He polarization, which is
expected to reduce the absolute value of the dispersion coefficients in the
long-range part of the Mg-Mg interaction. The second is a possible reduction
of the Mg atom polarizability due to the presence of the surrounding 4He
cavity which, because of the repulsive character of the electron-He
interaction, should make the electronic distribution of the impurity atom
slightly “stiffer”, thus reducing further the values of the dispersion
coefficients in the Mg-Mg pair interactions. Although in principle these
effects might reduce the net interaction between a pair of Mg atoms embedded
in liquid 4He, in practice in the present system they are indeed very small.
The correction to the leading term of the long-range dispersion interaction,
$-C_{6}/r^{6}$, due to three-body correlation effects can be written to first
orderrenne as $-C_{6}(1-2\pi n\alpha/3)/r^{6}$, $\alpha$ being the static
polarizability of the host fluid ($\alpha_{He}=1.39\,a_{0}^{3}$). Such
correction is of the order of only 1% in our case. To estimate the change in
the Mg atomic polarizability due to the surrounding He, we computed, using ab-
initio pseudopotential calculations, the (static) polarizability of a Mg atom
in the presence of an effective (mainly repulsive) potential acting on the Mg
valence electrons due to the presence of the surrounding He. The effective
interaction is derived from the equilibrium shape of the 4He bubble hosting
the Mg atom, as predicted by our DFT calculations, and assuming a (local)
electron-He density-dependent interaction which was proposed by Cole et
al.Che94 We find a very small change in the static atomic polarizability
$\alpha$ of Mg. Assuming that, roughly, $C_{6}\propto\alpha^{2}$ we find a
reduction of the $C_{6}$ coefficient of about 1-2%.
The minimisation of the total energy functional written above under the
constraint of a given number of helium atoms and normalized Mg ground state
wave functions should in principle yield the equilibrium configuration of the
system. In practice, depending on the initial configuration, we have found
several local minima, whose origin is again the “interference” of the He
solvation shells around the Mg atoms. We have found three such metastable
configurations for (Mg+Mg)@4He1000 if we start the minimization procedure with
one Mg atom near the center of the droplet, and the other placed off center,
at some distance from the first. They are displayed in Fig. 8. The energy
difference between the innermost (Mg-Mg distance $d=9.3$ Å) and the outermost
($d=18.5$ Å) configurations is 12.5 K. The energy of the $d=9.3$ Å
configuration is sensibly that of the configuration specularly symmetric about
the $z=0$ plane (-5581.4 K) also shown in the figure. It is worth noticing
that, since $R_{1/2}=22.2$ Å and the “radius” of Mg is $\sim 5$ Å, only the
upper left corner configuration has the Mg impurity in a surface state.
Figure 9 shows two density profiles of the symmetric configuration obtained
along the $z$-axis (solid line) and the $x$\- or $y$-axis (dashed line). It
shows a relatively high density helium ring around the two-bubble waist,
clearly visible in Fig. 8, where the local density is almost three times the
bulk liquid 4He density, and that prevents the collapse of the two-bubble
configuration. The Mg wave functions are peaked at $\sim\pm 4.75$ Å, and very
narrow. This justifies a posteriori the assumptions we have made to write the
total energy of the system, Eq. (27).
Our results confirm the existence of the energy barrier suggested in Ref.
Prz07, , that prevents the two Mg atoms from coming closer than some 9 Å, and
thus hindering, at least temporarily, the formation of the Mg2 dimer. This
barrier is shown in Fig. 10 as a function of the Mg-Mg distance $d$, which is
kept fixed in a constrained energy minimization. Note that the energy of the
Mg+Mg system increases as $d$ does because the two Mg atoms in a drop form a
state more bound than that of a pure drop with the impurities well apart.
Since the height of the barrier is larger than the experimental temperatures,
$T\sim 0.4\,K$, two solvated Mg atoms will not easily merge into a dimer, but
rather form of a metastable weakly bound state. Based on these finding, we
suggest that several Mg atoms solvated inside 4He drops might form a sparse,
weakly-bound “foam”-like aggregate rather than coalesce into a more tightly
bound metallic cluster. Partial coagulation of impurities was already invoked
by Toennies and coworkers to explain their experimental findings for the
successive capture of foreing atoms and molecules in helium clustersLew95
(see also Ref. Gor04, for the case of bulk liquid helium). Very recently, a
kind of “quantum gel” has been predicted to be formed in 4He drops doped with
neon atoms.Elo08 Although some degree of mutual isolation between foreign
atoms is expected in the case of strongly attractive impurities (like those
studied in the two cases mentioned above), where they are kept apart by the
presence a solid 4He layer coating the impurity,Gor04 our calculations show
that this effect is possible even for relatively weakly attractive impurities
like Mg, where such solid-like 4He layer is absent.
One may estimate the mean life of the metastable state as
$\tau=2\pi\sqrt{\frac{\mu^{*}_{Mg}}{U^{\prime\prime}(d_{eq})}}\,\exp{[\Delta
U/(k_{B}T)]}\;,$ (28)
where $\mu^{*}_{Mg}=M^{*}_{Mg}/2\sim 20$ a.u. is the hydrodynamic reduced mass
of the Mg+Mg system and $\Delta U$ is the barrier height. From Fig. 10 we have
that $U^{\prime\prime}(d_{eq})\sim 40$ K Å-2. This yields a mean life of a few
nanoseconds, which is about five to six orders of magnitude smaller than the
time needed for its experimental detection.Prz07 ; Notabarrier The mean life
becomes increasingly large as the relative angular momentum $L$ deposited into
the two Mg system increases. Writing
$\Delta E=\Delta
E(L=0)+\frac{\hbar^{2}}{2\mu^{*}_{Mg}}\frac{L(L+1)}{d^{2}}\;,$ (29)
one obtains the $L-$dependent energy barriers displayed in Fig. 10. For $L=30$
we have $\tau\sim 0.6$ $\mu$s, and for $L=40$, $\tau\sim 0.1$ milliseconds.
Thus, there is an angular momentum window that may yield mean lifes compatible
with the experimental findings. Increasing $L$ much further would produce too
a distant Mg+Mg system which would correspond to a two independent Mg
impurities in a drop.
To check whether this foam-like structure of the Mg aggregate also appears in
3He drops, we have carried out calculations for (Mg+Mg)@3He1000 using the same
density functional as in Ref. Sti04, , and the method presented in this
Section. Figure 11 shows the energy of the (Mg+Mg)@3He1000 complex as a
function of $d$. For distances smaller than some 8.6 Å we have found that the
system has a tendency to collapse into a dimer -physically unreachable from
our starting point, Eq. (27). We are led to conclude that there is no barrier
in the case of liquid 3He. The configuration corresponding to the closest $d$
we have calculated is shown in Fig. 12.
We are now in the position to determine the effect of these weakly-bound
systems on the LIF and R2PI experiments on 4He droplets containing more than
one Mg atom. Notice that the bottom right panel of Fig. 8 shows that the
helium bubbles have a non-zero static quadrupole moment, whereas they are
spherically symmetric for one single Mg atom in the drop. It is precisely the
existence of this static quadrupole moment that causes an additional
separation between the $\Sigma$ and $\Pi$ spectral components in the
absorption spectra, which results, as a consequence of the broadening of each
line, in an double-peak structure of the computed spectra, in semi-
quantitative agreement with the experimental data. Details of our calculation
are given in Appendix C, see Eq. (82). Figure 13 shows such two-peak structure
corresponding to the specularly symmetric configuration displayed in Fig. 8,
and indicates that the 282 nm structure observed in the experiments may be
attributed to the distortion produced by neighbour Mg bubbles; these bubbles
contribute incoherently to the absorption spectrum, and the relative intensity
of the 282 and 279 nm peaks might reflect the different population of drops
doped with one and two Mg atoms, since those hosting a compact cluster (dimer,
trimer, etc), would not yield an absorption signal in the neighborhood of the
monomer $3s3p$ 1P${}_{1}\leftarrow 3s^{2}$ 1S0 transition. Notice that a large
static distortion of the helium bubble could also arise if the Mg atom were in
a shallow dimple at the drop surface, but Fig. 6 discards this possibility.
We finally note that we have not considered in our work another source for an
additional splitting of the spectral lines of a Mg atom in the field produced
by a neighboring one, i.e. the resonant dipole-dipole interaction occurring
during the electronic excitation. This effect could in principle lead to an
additional (but probably small, compared with the effect discussed here)
splitting of the calculated lines. An accurate determination of the dipole
moment is required for a proper inclusion of this effect, which is beyond the
scope of the present paper.
## VI Summary
We have obtained, within DFT, the structure of 4He droplets doped with Mg
atoms and have discussed in detail the magnesium solvation properties. In
agreement with previous DMC calculations,Mel05 ; Elh08 we have found that Mg
is not fully solvated in small 4He drops, whereas it becomes fully solvated in
large droplet.
As a consequence of its interaction with the helium environment, it turns out
that magnesium is radially quite delocalized inside the droplets. This large
delocalization provides a way to reconcile two contradictory results on the
solution of one Mg atom in a 4He drop, namely center localization (LIF and
R2PI experimentsReh00 ; Prz07 ), and surface localization (electron-impact
ionization experimentsRen07 ).
We have calculated the absorption spectrum of magnesium in the vicinity of the
$3s3p$ 1P${}_{1}\leftarrow 3s^{2}$ 1S0 transition. For the large Mg@4He1000
droplet, where Mg is fully solvated, we reproduce the more intense component
of the absorption line found by LIF and R2PI experiments in large drops and in
liquid helium. This agreement is only achieved when homegeneous broadening due
to the coupling of the dipole excitation with the quadrupole deformations of
the helium bubble are fully taken into account. This coupling is naturally
included in Quantum Monte Carlo simulations of the absorption spectrum,Oga99 ;
Mel02 whereby one takes advantage of the inherent fluctuations present in
these simulations. These fluctuations are the full quantal equivalent of the
dynamical distortions of the helium bubble we have introduced for the
description of homogeneous broadening. An alternative method to include shape
fluctuations within DFT has been proposed and applied to the case of Cs in
bulk liquid helium.Nak02 It would be interesting to adapt this method to the
drop geometry, since it is not simple to handle dynamical bubble distortions
in a non-spherical environment, or in 3He drops.
To explain the origin of the low energy peak in the absorption line and
confirm the likely existence of soft, “foam”-like structure of Mg aggregates
in 4He drops as proposed by Przystawik et al,Prz07 we have addressed the
properties of two Mg atoms in 4He1000 and have found that indeed, Mg atoms are
kept apart by the presence of helium atoms that prevent the formation of a
compact Mg cluster. We have estimated that the height of the energy barrier
for the formation of the Mg dimer in 4He drops is $\sim 2-3$ K, which should
be enough, at the droplet experimental temperature of $0.4$ K, to guarantee a
relatively long lifetime to these weakly-bound Mg aggregates. We predict that,
contrarily, Mg atoms adsorbed in 3He droplets do not form such metastable
states.
The presence of neighboring Mg atoms in these structures induces a static
quadrupole deformation in the helium bubble accomodating a Mg atom. As a
consequence, the dipole absorption line around the $3s3p$ 1P${}_{1}\leftarrow
3s^{2}$ 1S0 transition splits. We attribute to this static quadrupole moment
the origin of the low energy peak in the absorption line, and confirm the
suggestion made by the Rostock group Prz07 that the splitting of the
absorption line, rather than being due to a dynamical (Jahn-Teller)
deformation of the helium bubble, is due to the presence of more than one
magnesium atom in the same droplet.
Our previous study on Ca doped helium dropsHer08 and the present work show
that DFT is able to quantitatively address the dipole absorption of dopants in
4He drops within the diatomics-in-molecules approach, provided the impurity-
helium pair potentials are accurately determined. However, we want to point
out that, while we have a consistent scenario that explains the results of LIF
and R2PI experiments, the understanding of the electron-impact ionization
experiment reported in Ref. Ren07, still requires further analysis. Indeed,
since Mg atoms may be in the bulk of the drop or just beneath the drop
surface, the experimental ion yield curve should reflect both possibilities,
whereas apparently it does not (see Fig. 2 of Ref. Ren07, ). One possible
explanation may be that for electron-impact experiments, a $N=10\,000$ drop is
still small, so that the impurity is always close enough to the drop surface
to make the Penning ionization process to prevail on the direct formation of a
He+ ion.
## Acknowledgments
We would like to thank Massimo Mella and Fausto Cargnoni for useful comments
and for providing us with their Mg-He excited pair potentials, and Marius
Lewerenz, Vitaly Kresin, Karl-Heinz Meiwes-Broer, Josef Tiggesbäumker, Andreas
Przystawik, Carlo Callegari and Kevin Lehmann for useful comments and
discussions. This work has been performed under Grants No. FIS2005-01414 from
DGI, Spain (FEDER), No. 2005SGR00343 from Generalitat de Catalunya, and under
the HPC-EUROPA project (RII3-CT-2003-506079), with the support of the European
Community - Research Infrastructure Action under the FP6 “Structuring the
European Research Area” Programme.
## Appendix A
In this Appendix we obtain the energy of the doped drop up to second order in
the deformation parameters and the hydrodynamic mass of the helium bubble. To
this end, the helium order parameter and Mg wave function are expressed as
$\Psi(\mathbf{r},t)=\sqrt{\rho(\mathbf{r},t)}\exp[i\frac{m_{He}}{\hbar}S(\mathbf{r},t)]$
and
$\Phi(\mathbf{r},t)=|\Phi(\mathbf{r},t)|\exp[i\frac{m_{Mg}}{\hbar}\varphi(\mathbf{r},t)]$,
respectively. Neglecting the velocity-dependent terms of the Orsay-Trento
functional that mimic backflow effects,Dal95 the total energy of the system
is written as
$\displaystyle E$
$\displaystyle=\frac{1}{2}m_{He}\int\mathrm{d}^{3}\mathbf{r}\,\rho(\mathbf{r},t)|\nabla
S(\mathbf{r},t)|^{2}+\frac{1}{2}m_{Mg}\int\mathrm{d}^{3}\mathbf{r}\,|\Phi(\mathbf{r},t)|^{2}|\nabla\varphi(\mathbf{r},t)|^{2}$
(30) $\displaystyle+$
$\displaystyle\frac{\hbar^{2}}{2\,m_{He}}\int\mathrm{d}^{3}\mathbf{r}\,|\nabla\sqrt{\rho(\mathbf{r},t)}|^{2}+\frac{\hbar^{2}}{2\,m_{Mg}}\int\mathrm{d}^{3}\mathbf{r}\,\left|\nabla|\Phi(\mathbf{r},t)|\right|^{2}+\int\mathrm{d}^{3}\mathbf{r}\,{\cal
E}(\rho)$ $\displaystyle+$
$\displaystyle\int\int\mathrm{d}^{3}\mathbf{r}\,\mathrm{d}^{3}\mathbf{r^{\prime}}\,|\Phi(\mathbf{r},t)|^{2}\,V_{Mg-
He}(|\mathbf{r}-\mathbf{r^{\prime}}|)\,\rho(\mathbf{r^{\prime}},t)\;,$
where the functions $\rho$($|\Phi|$) and $S$($\varphi$) fulfill the continuity
equations
$\displaystyle-\frac{\partial}{\partial t}\rho(\mathbf{r},t)$ $\displaystyle=$
$\displaystyle\nabla\left[\rho(\mathbf{r},t)\nabla S(\mathbf{r},t)\right]$
(31) $\displaystyle-\frac{\partial}{\partial t}|\Phi(\mathbf{r},t)|^{2}$
$\displaystyle=$
$\displaystyle\nabla\left[|\Phi(\mathbf{r},t)|^{2}\nabla\varphi(\mathbf{r},t)\right]$
(32)
that allow to identify $S(\mathbf{r},t)$ and $\varphi(\mathbf{r},t)$ as
velocity field potentials, and the first two terms in Eq. (30) as a collective
kinetic energy, whose density we denote as $t[\rho,S,|\Phi|,\varphi]$. Thus,
$E=T+V=\int\mathrm{d}^{3}\mathbf{r}\left\\{t[\rho,S,|\Phi|,\varphi]+v[\rho,|\Phi|]\right\\}\;.$
(33)
In the adiabatic approximation, the dynamics of the system requires the
following steps: i) introduce a set of collective variables (or deformation
parameters) $[\alpha(t)])$ that define the helium density,
$\rho(\mathbf{r},t)=\rho(\mathbf{r},[\alpha(t)])$; ii) for each helium
configuration defined by $[\alpha]$, solve the time-independent Schrödinger
equation obeyed by $|\Phi(\mathbf{r},[\alpha])|$; iii) obtain the potential
surface $V[\rho,|\Phi|]$ by computing the static energy for each
configuration; iv) determine the velocity field potentials
$S(\mathbf{r},[\alpha(t)])$ and $\varphi(\mathbf{r},[\alpha(t)])$ by solving
the continuity equations; v) compute the collective kinetic energy to obtain
the hydrodynamic mass, and vi) solve the equation of motion associated with
the effective Hamiltonian written as a function the deformation parameters.
We aim to describe harmonic deformations of a spherical helium bubble created
by an impurity in the ground state, and have to determine the helium density
$\rho(\mathbf{r},[\alpha(t)])$ resulting from a change in the radial distance
to the center of the spherical bubble induced by the $[\alpha(t)]$ parameters:
$r\longrightarrow
r+\sum_{\lambda=0}^{\infty}\sum_{\mu=-\lambda}^{\lambda}\alpha_{\lambda\mu}(t)Z_{\lambda\mu}(\widehat{r})\;,$
(34)
where the $\lambda=1$ deformation is now introduced to allow for displacements
of the bubble. We recall that the real spherical harmonics have been
normalized as $\langle
Z_{\lambda\mu}|Z_{\lambda^{\prime}\mu^{\prime}}\rangle=\frac{4\pi}{2\lambda+1}\delta_{\lambda\lambda^{\prime}}\delta_{\mu\mu^{\prime}}$,
with $Z_{00}(\widehat{r})=1$, $Z_{10}(\widehat{r})=\cos\theta$, etc. The
breathing mode corresponds to $\lambda=0$, an infinitesimal translation to
$\lambda=1$ (provided the fluid is incompressible), and a quadrupolar
deformation to $\lambda=2$. To first order, the density can be written as
$\rho(\mathbf{r},t)\simeq\rho_{0}(r)+\rho_{0}^{\prime}(r)\sum_{\lambda=0}^{\infty}\sum_{\mu=-\lambda}^{\lambda}\alpha_{\lambda\mu}(t)Z_{\lambda\mu}(\widehat{r})\;,$
(35)
where from now on, the prime will denote the derivative of the function with
respect to its argument.
### A.1 Impurity wave function
To first order, the wave function $|\Phi(\mathbf{r},[\alpha])|$ is written as
$|\Phi(\mathbf{r},t)|\simeq\Phi_{0}(r)+\sum_{\lambda=0}^{\infty}\sum_{\mu=-\lambda}^{\lambda}\alpha_{\lambda\mu}(t)\Phi^{(1)}_{\lambda\mu}(\mathbf{r})\;.$
(36)
The amplitudes $\Phi^{(1)}_{\lambda\mu}(\mathbf{r})$ are determined in first
order perturbation theory from the multipole expansion of the impurity-helium
pair potential
$\displaystyle U^{(1)}_{\lambda\mu}(\mathbf{r})$ $\displaystyle=$
$\displaystyle\int\mathrm{d}^{3}\mathbf{r}^{\prime}\rho_{0}^{\prime}(r^{\prime})Z_{\lambda\mu}(\widehat{r}^{\prime})V_{X-He}(|\mathbf{r}-\mathbf{r^{\prime}}|)$
(37) $\displaystyle=$
$\displaystyle\int\mathrm{d}^{3}\mathbf{r}^{\prime}\rho_{0}^{\prime}(r^{\prime})Z_{\lambda\mu}(\widehat{r}^{\prime})\sum_{\lambda\mu}V_{X-He}^{\lambda}(r,r^{\prime})Z_{\lambda\mu}(\widehat{r})Z_{\lambda\mu}(\widehat{r}^{\prime})$
$\displaystyle\equiv$ $\displaystyle
U^{(1)}_{\lambda}(r)Z_{\lambda\mu}(\widehat{r})\;,$
which defines $U^{(1)}_{\lambda}(r)$. We obtain
$\displaystyle\Phi^{(1)}_{\lambda\mu}(\mathbf{r})$ $\displaystyle=$
$\displaystyle\sum_{n\ell
m}\frac{\langle\Phi_{0}|U^{(1)}_{\lambda}Z_{\lambda\mu}|\Phi_{n\ell}Z_{\ell
m}\rangle}{\varepsilon_{0}-\varepsilon_{n\ell m}}\Phi_{n\ell}(r)Z_{\ell
m}(\widehat{r})$ (38) $\displaystyle=$
$\displaystyle\left[\sum_{n}\frac{4\pi}{2\lambda+1}\frac{\langle\Phi_{0}|U^{(1)}_{\lambda}|\Phi_{n\lambda}\rangle}{\varepsilon_{0}-\varepsilon_{n\lambda\mu}}\Phi_{n\lambda}(r)\right]Z_{\lambda\mu}(\widehat{r})$
$\displaystyle\equiv$
$\displaystyle\Phi^{(1)}_{\lambda}(r)Z_{\lambda\mu}(\widehat{r})$
that shows that actually, $\Phi^{(1)}_{\lambda\mu}$ is $\mu$ independent.
Once we have obtained the wave function, we can compute the energy surface
$V[\rho,|\Phi|]$. Since we describe deformations around a spherically
symmetric ground state, the first order term vanishes, and the derivative
$g_{\lambda}\equiv\partial^{2}v/\partial\alpha_{\lambda\mu}\partial\alpha_{\lambda^{\prime}\mu^{\prime}}$
is also spherically symmetric. We can evaluate the second order contribution
to the collective potential energy as
$\displaystyle V^{(2)}$ $\displaystyle=$
$\displaystyle\sum_{\lambda\mu}\sum_{\lambda^{\prime}\mu^{\prime}}\frac{1}{2}\int\mathrm{d}^{3}\mathbf{r}\left.\frac{\partial^{2}v}{\partial\alpha_{\lambda\mu}\partial\alpha_{\lambda^{\prime}\mu^{\prime}}}\right|_{\rho_{0},\varphi_{0}}\alpha_{\lambda\mu}Z_{\lambda\mu}(\widehat{r})\alpha_{\lambda^{\prime}\mu^{\prime}}Z_{\lambda^{\prime}\mu^{\prime}}(\widehat{r})$
(39) $\displaystyle=$
$\displaystyle\sum_{\lambda\mu}\frac{2\pi}{2\lambda+1}\int_{0}^{\infty}\mathrm{d}r\,r^{2}g_{\lambda}(r)\alpha_{\lambda\mu}^{2}$
$\displaystyle\equiv$
$\displaystyle\sum_{\lambda=0}^{\infty}\frac{2\pi}{2\lambda+1}E^{(2)}_{\lambda}\sum_{\mu=-\lambda}^{\lambda}\alpha_{\lambda\mu}^{2}\;,$
which defines $E^{(2)}_{\lambda}$. The parameters $[\alpha]$ are the dynamical
variables that describe the evolution of the system.
### A.2 Velocity field potentials
Introducing the expansion
$S(\mathbf{r},t)\equiv\sum_{\lambda\mu}\dot{\alpha}_{\lambda\mu}(t)\tilde{S}_{\lambda}(r)Z_{\lambda\mu}(\widehat{r})$,
where the dot denotes the time-derivative, the continuity equation for the
liquid helium is, to first order,
$-\sum_{\lambda\mu}\dot{\alpha}_{\lambda\mu}~{}Z_{\lambda\mu}~{}\frac{\mathrm{d}\rho_{0}}{\mathrm{d}r}=\sum_{\lambda\mu}\dot{\alpha}_{\lambda\mu}~{}Z_{\lambda\mu}\left\\{\frac{\mathrm{d}\rho_{0}}{\mathrm{d}r}\frac{\mathrm{d}\tilde{S}_{\lambda}}{\mathrm{d}r}+\rho_{0}\left[\frac{\mathrm{d}^{2}\tilde{S}_{\lambda}}{\mathrm{d}r^{2}}+\frac{2}{r}\frac{\mathrm{d}\tilde{S}_{\lambda}}{\mathrm{d}r}-\frac{\lambda(\lambda+1)}{r^{2}}\tilde{S}_{\lambda}\right]\right\\}\;.$
(40)
Hence,
$-\frac{\mathrm{d}\rho_{0}}{\mathrm{d}r}=\frac{\mathrm{d}\rho_{0}}{\mathrm{d}r}\frac{\mathrm{d}\tilde{S}_{\lambda}}{\mathrm{d}r}+\rho_{0}\left[\frac{\mathrm{d}^{2}\tilde{S}_{\lambda}}{\mathrm{d}r^{2}}+\frac{2}{r}\frac{\mathrm{d}\tilde{S}_{\lambda}}{\mathrm{d}r}-\frac{\lambda(\lambda+1)}{r^{2}}\tilde{S}_{\lambda}\right].$
(41)
When $r\rightarrow\infty$, the density vanishes for a drop, and approaches
$\rho_{b}$ for the liquid. In the later case, Eq. (41) reduces to the radial
part of the Laplace equation
$0=\frac{\mathrm{d}^{2}\tilde{S}_{\lambda}}{\mathrm{d}r^{2}}+\frac{2}{r}\frac{\mathrm{d}\tilde{S}_{\lambda}}{\mathrm{d}r}-\frac{\lambda(\lambda+1)}{r^{2}}\tilde{S}_{\lambda}$
(42)
whose general solution is
$\tilde{S}_{\lambda}(r)=A_{\lambda}~{}r^{\lambda}+\frac{B_{\lambda}}{r^{\lambda+1}}\;.$
(43)
We have solved Eq. (41) adapting the method proposed in Ref. Leh02, . Let
$r_{i}$ be the first point where $\rho_{0}(r)$ is significantly different from
zero $[\rho_{0}(r)=0\,\,{\rm for}\,\,r\leq r_{i}]$. At this point, Eq. (41)
implies that
$\left.-\frac{\mathrm{d}\rho_{0}}{\mathrm{d}r}\right|_{r_{i}}=\left.\frac{\mathrm{d}\rho_{0}}{\mathrm{d}r}\frac{\mathrm{d}\tilde{S}_{\lambda}}{\mathrm{d}r}\right|_{r_{i}},$
(44)
which determines the boundary condition at $r_{i}$:
$\left.\frac{\mathrm{d}\tilde{S}_{\lambda}}{\mathrm{d}r}\right|_{r_{i}}=-1\;.$
(45)
For the liquid, the other boundary condition is that when
$r\rightarrow\infty$, the solution behaves as in Eq. (43) with $A_{\lambda}=0$
to have a physically acceptable solution. If the bubble has a sharp surface of
radius $r_{i}$ and the liquid is uniform, $\tilde{S}_{\lambda}(r)$ is
completely determined by Eq. (43) and the velocity field potential that
fulfills Eq. (45) corresponds to the coefficients
$\displaystyle A_{\lambda}$ $\displaystyle=$ $\displaystyle 0$ $\displaystyle
B_{\lambda}$ $\displaystyle=$
$\displaystyle\frac{r_{i}^{\lambda+2}}{\lambda+1}\;.$ (46)
To find the velocity field potential in a large drop, we have defined a radial
distance $r_{b}$, far from the bubble and from the drop surface at $R_{1/2}$,
around which on one may consider that the density is that of the liquid.
Starting from $r=r_{b}$ with the liquid solution fixed by the coefficients
given in Eq. (46), we have integrated inwards Eq. (41), finding the solutions
$\tilde{S}_{\lambda}^{\mathrm{inh}}(r)$ and
$\tilde{S}_{\lambda}^{\mathrm{h}}(r)$ that correspond, respectively, to the
non-homegeneous and to the homegeneous differential equation that results by
setting to zero the left hand side of Eq. (41). The general solution that
satisfy the boundary condition Eq. (45) is obtained as
$\tilde{S}_{\ell}(r)=\tilde{S}_{\lambda}^{\mathrm{inh}}(r)+C\tilde{S}_{\lambda}^{\mathrm{h}}(r)$
(47)
with
$C=-\frac{1+\left.\frac{\mathrm{d}\tilde{S}_{\lambda}^{\mathrm{inh}}}{\mathrm{d}r}\right|_{r_{i}}}{\left.\frac{\mathrm{d}\tilde{S}_{\lambda}^{\mathrm{h}}}{\mathrm{d}r}\right|_{r_{i}}}\;.$
(48)
The field $\varphi$ is analogously obtained after introducing the expansion
$\varphi(\mathbf{r},t)=\sum_{\lambda\mu}\dot{\alpha}_{\lambda\mu}(t)\tilde{\varphi}_{\lambda}(r)\mathcal{Z}_{\lambda\mu}(\widehat{r})$.
We have assumed that the wave function of the impurity in the ground state is
a Gaussian $\Phi_{0}(r)=A\exp\left(-\beta r^{2}\right)$ whose shape has been
determined by fitting it to the actual wave function, and have introduce a
cutoff distance $r_{g}$ such that safely $\Phi_{0}(r)=0$ if $r\geq r_{g}$. We
then obtain the following differential equation to determine
$\tilde{\varphi}_{\lambda}$:
$4\beta r=-4\beta
r\frac{\mathrm{d}\tilde{\varphi}_{\lambda}}{\mathrm{d}r}+\frac{\mathrm{d}^{2}\tilde{\varphi}_{\lambda}}{\mathrm{d}r^{2}}+\frac{2}{r}\frac{\mathrm{d}\tilde{\varphi}_{\lambda}}{\mathrm{d}r}-\frac{\lambda(\lambda+1)}{r^{2}}\tilde{\varphi}_{\lambda}\;.$
(49)
The appropriate boundary conditions are
$\displaystyle\left.\frac{\mathrm{d}\tilde{\varphi}_{\lambda}}{\mathrm{d}r}\right|_{r_{0}}$
$\displaystyle=$ $\displaystyle-1$ $\displaystyle\tilde{\varphi}_{\lambda}(0)$
$\displaystyle=$ $\displaystyle 0\;,$ (50)
and the solution is analytical but very involved (we have found it by using
the Mathematica packpage). The dipole mode is the only exception; if
$\lambda=1$ we have $\tilde{\varphi}_{1}=-r$.
### A.3 Kinetic energy
Once the velocity fields $S(\mathbf{r},t)$ and $\varphi(\mathbf{r},t)$ have
been determined, the collective kinetic energy can be easily calculated to
second order in the collective parameters:
$\displaystyle T$ $\displaystyle=$
$\displaystyle\int\mathrm{d}^{3}\mathbf{r}~{}t[\rho,S,|\Phi|,\varphi]$ (51)
$\displaystyle=$
$\displaystyle\frac{1}{2}m_{He}\int\mathrm{d}^{3}\mathbf{r}\,\rho(\mathbf{r},t)|\nabla
S(\mathbf{r},t)|^{2}+\frac{1}{2}m_{Mg}\int\mathrm{d}^{3}\mathbf{r}\,|\Phi(\mathbf{r},t)|^{2}|\nabla\varphi(\mathbf{r},t)|^{2}$
$\displaystyle\simeq$
$\displaystyle\sum_{\lambda=0}^{\infty}\frac{4\pi}{2\lambda+1}\left\\{\frac{1}{2}m_{He}\int\mathrm{d}r~{}r^{2}~{}\rho_{0}(r)~{}\left[\left(\frac{\mathrm{d}\tilde{S}_{\lambda}}{\mathrm{d}r}\right)^{2}+\frac{\lambda(\lambda+1)}{r^{2}}\tilde{S}_{\lambda}^{2}\right]\right.$
$\displaystyle\left.+\frac{1}{2}m_{Mg}\int\mathrm{d}r~{}r^{2}~{}|\Phi_{0}(r)|^{2}~{}\left[\left(\frac{\mathrm{d}\tilde{\varphi}_{\lambda}}{\mathrm{d}r}\right)^{2}+\frac{\lambda(\lambda+1)}{r^{2}}\tilde{\varphi}_{\lambda}^{2}\right]\right\\}\sum_{\mu=-\lambda}^{\lambda}\dot{\alpha}_{\lambda\mu}^{2}$
$\displaystyle\equiv$
$\displaystyle\frac{1}{2}\sum_{\lambda=0}^{\infty}M_{\lambda}^{*}\sum_{\mu=-\lambda}^{\lambda}\dot{\alpha}_{\lambda\mu}^{2}$
which defines the hydrodynamic mass $M_{\lambda}^{*}$ for each $\lambda$ mode.
Using the continuity equations and the Gauss theorem, one can find an
alternative expression for $M_{\lambda}^{*}$
$M_{\lambda}^{*}=\frac{4\pi}{2\lambda+1}\left\\{m_{He}\int\mathrm{d}r~{}r^{2}~{}\rho^{\prime}_{0}(r)~{}\tilde{S}_{\lambda}(r)+m_{Mg}\int\mathrm{d}r~{}r^{2}~{}2\Phi_{0}(r)\Phi^{\prime}_{0}(r)\tilde{\varphi}_{\lambda}(r)\right\\}\;.$
(52)
We have checked that both expressions yield the same values for
$M_{\lambda}^{*}$, which constitutes a test on the numerical accuracy of the
method. It is easy to see from Eq. (51) that in bulk liquid helium, the
$\lambda=1$ hydrodynamic mass coincides with that given in Ref. Leh02, . Using
that $\tilde{\varphi}_{1}=-r$, it is also easy to check from the above
expressions that the impurity contribution to the $\lambda=1$ hydrodynamic
mass is just the bare mass of the Mg atom.
The sum of Eqs. (39) and (51) represents the Hamiltonian of a set of uncoupled
harmonic oscillators whose frequency only depends on $\lambda$.
## Appendix B
In this Appendix we work out in detail the expressions we have used to
describe the homogeneous broadening of the absorption line. We consider a
spherical ground state defined by a helium density $\rho_{0}(r)$ and impurity
wave function $\Phi_{0}(r)$, both modified by the action of the breathing mode
defined in Eq. (21), namely $\rho(r,\alpha_{0})=\rho_{0}(r+\alpha_{0})\,{\cal
K}^{-1}$, and $\Phi(r,\alpha_{0})$. The computation of the spectra for a given
$\alpha_{0}$ can be carried out starting from Eq. (11)
$I(\omega,\alpha_{0})=4\pi\sum_{i}\int\mathrm{d}r~{}|r~{}\Phi(r,\alpha_{0})|^{2}\delta[\omega+\omega(\alpha_{0})-V^{\mathrm{ex}}_{i}(r,\alpha_{0})/\hbar]\;,$
(53)
where $\hbar\omega(\alpha)=\varepsilon(\alpha_{0})$ is the impurity
eigenenergy and $V^{\mathrm{ex}}_{i}(r,\alpha_{0})$ are the PES defined by Eq.
(13), where $[\alpha]$ reduces to $\alpha_{0}$. Notice that $\alpha_{0}$ is
not introduced perturbatively; it is unnecessary since this mode does not
break the spherical symmetry. Next, we perturbatively introduce the modes with
$\lambda\geq 2$ to first order; Eq. (53) becomes
$\displaystyle I(\omega,[\alpha])$ $\displaystyle\simeq$ $\displaystyle
I(\omega,\alpha_{0})$ (54)
$\displaystyle+\sum_{i}\sum_{\lambda=2}^{\infty}\sum_{\mu=-\lambda}^{\lambda}\left\\{\int\mathrm{d}\mathbf{r}~{}\Phi(r,\alpha_{0})\Phi^{(1)}_{\lambda}(r,\alpha_{0})Z_{\lambda\mu}(\hat{r})\delta[\omega+\omega(\alpha_{0})-V^{\mathrm{ex}}_{i}(r,\alpha_{0})/\hbar]\right.$
$\displaystyle\left.-4\pi\int\mathrm{d}r~{}|r~{}\Phi(r,\alpha_{0})|^{2}\delta^{\prime}[\omega+\omega(\alpha_{0})-V^{\mathrm{ex}}_{i}(r,\alpha_{0})/\hbar]\frac{1}{\hbar}\epsilon_{\lambda\mu}^{i}(r)\right\\}\alpha_{\lambda\mu}\;,$
where $\epsilon_{\lambda\mu}^{i}(r)=\left.\partial
V^{\mathrm{ex}}_{i}(r)/\partial\alpha_{\lambda\mu}\right|_{[\alpha]_{\lambda\geq
2}=0}$. The first integral is zero due to the orthogonality of the spherical
harmonics. To evalue the second integral, we expand
$\epsilon_{\lambda\mu}^{i}(r)$ as a power series of $r$. Taken into account
that the PES have a stationary point at $r=0$ due to the spherical geometry
–the first order term is zero– we can safely stop the expansion at the zeroth
order term, since the wave function $\Phi(r,\alpha_{0})$ is very narrow
$\displaystyle I(\omega,[\alpha])\simeq I(\omega,\alpha_{0})$
$\displaystyle-4\pi\sum_{i}\sum_{\lambda=2}^{\infty}\sum_{\mu=-\lambda}^{\lambda}\int\mathrm{d}r~{}|r~{}\Phi(r,\alpha_{0})|^{2}\delta^{\prime}[\omega-\omega(\alpha_{0})-V^{\mathrm{ex}}_{i}(r,\alpha_{0})/\hbar]\left(\frac{1}{\hbar}\epsilon_{\lambda\mu}^{i}(0)\alpha_{\lambda\mu}\right)\;.$
(55)
This equation may be interpreted as the expansion to first order of a shift in
$\omega$, so it can be written as
$\displaystyle I(\omega,[\alpha])$ $\displaystyle\simeq$ $\displaystyle
4\pi\sum_{i}\int\mathrm{d}r~{}|r~{}\Phi(r,\alpha_{0})|^{2}\delta\left[\omega+\omega(\alpha_{0})-\frac{1}{\hbar}\left(V^{\mathrm{ex}}_{i}(r,\alpha_{0})+\sum_{\lambda=2}^{\infty}\sum_{\mu=-\lambda}^{\lambda}\epsilon_{\lambda\mu}^{i}(0)\alpha_{\lambda\mu}\right)\right]$
(56) $\displaystyle=$
$\displaystyle\sum_{i}I\left[\omega+\sum_{\lambda=2}^{\infty}\sum_{\mu=-\lambda}^{\lambda}\epsilon_{\lambda\mu}^{i}(0)\alpha_{\lambda\mu}/\hbar,\alpha_{0}\right]\;.$
Thus the eigenvalues $\epsilon_{\lambda\mu}^{i}(0)$ are related to the
diagonalization of the expansion of $U_{ij}(\mathbf{r},[\alpha])$ defined in
Eq. (12). Writing this matrix equation as a function of the real spherical
harmonics
$\displaystyle
U(\mathbf{r},[\alpha])=\int\mathrm{d}^{3}\mathbf{r^{\prime}}\rho(\mathbf{r^{\prime}}+\mathbf{r},[\alpha])\left\\{\frac{1}{3}[V_{\Sigma}(r^{\prime})+2V_{\Pi}(r^{\prime})]Z_{00}(\hat{r})\left(\begin{array}[]{ccc}1&0&0\\\
0&1&0\\\ 0&0&1\end{array}\right)\right.$ (60)
$\displaystyle\left.+\frac{1}{\sqrt{3}}[V_{\Sigma}(r^{\prime})-V_{\Pi}(r^{\prime})]\left(\begin{array}[]{ccc}-\frac{1}{\sqrt{3}}Z_{20}(\hat{r})+Z_{22}(\hat{r})&Z_{2-2}(\hat{r})&Z_{21}(\hat{r})\\\
Z_{2-2}(\hat{r})&-\frac{1}{\sqrt{3}}Z_{20}(\hat{r})-Z_{22}(\hat{r})&Z_{2-1}(\hat{r})\\\
Z_{21}(\hat{r})&Z_{2-1}(\hat{r})&\frac{2}{\sqrt{3}}Z_{20}(\hat{r})\end{array}\right)\right\\}\;,$
(64)
expanding $\rho(\mathbf{r^{\prime}}+\mathbf{r},[\alpha])$ to first order in
$[\alpha]_{\lambda\geq 2}$, and evaluating the first order contribution at
$r=0$ we obtain
$\displaystyle U(\mathbf{r},[\alpha])$ $\displaystyle\simeq$ $\displaystyle
U(r,\alpha_{0})+\sum_{\lambda=2}^{\infty}\sum_{\mu=-\lambda}^{\lambda}\frac{4\pi}{2\lambda+1}\int\mathrm{d}r^{\prime}\,r^{\prime
2}\,\rho^{\prime}(r^{\prime},\alpha_{0})\frac{1}{\sqrt{3}}[V_{\Sigma}(r^{\prime})-V_{\Pi}(r^{\prime})]$
(68)
$\displaystyle\times\left(\begin{array}[]{ccc}-\frac{1}{\sqrt{3}}\delta_{2\lambda}\delta_{0\mu}+\delta_{2\lambda}\delta_{2\mu}&\delta_{2\lambda}\delta_{-2\mu}&\delta_{2\lambda}\delta_{1\mu}\\\
\delta_{2\lambda}\delta_{-2\mu}&-\frac{1}{\sqrt{3}}\delta_{2\lambda}\delta_{0\mu}-\delta_{2\lambda}\delta_{2\mu}&\delta_{2\lambda}\delta_{-1\mu}\\\
\delta_{2\lambda}\delta_{1\mu}&\delta_{2\lambda}\delta_{-1\mu}&\frac{2}{\sqrt{3}}\delta_{2\lambda}\delta_{0\mu}\end{array}\right)\alpha_{\lambda\mu}\;,$
This shows that, to first order, only quadrupolar deformations are coupled to
the dipole electronic transition, and that its effect is a shift of the
spectral line, as shown in Eq. (56). At variance with the approach of Ref.
Kin96, , where the above matrix is approximated by its diagonal expression,
implying that only the $\mu=0$ and 2 components of the quadrupole deformation
are considered, our approach incorporates all five components.
The relation between the eigenvalues $\lambda_{i}(r,[\alpha])$ of
$U(r,[\alpha])$ and the coefficients $\epsilon_{\lambda\mu}^{i}(0)$ is
$\lambda_{i}(r,[\alpha])\equiv
V^{\mathrm{ex}}_{i}(r,\alpha_{0})+\sum_{\mu=-2}^{2}\epsilon_{2\mu}^{i}(0)\alpha_{2\mu}\;.$
(69)
Finally, the total spectrum is written as
$\displaystyle I(\omega)$ $\displaystyle\propto$ $\displaystyle
4\pi\int\mathrm{d}\alpha_{0}\,\mathrm{d}^{5}\alpha_{2}|\psi(\alpha_{0},\alpha_{2})|^{2}I(\omega,[\alpha])$
(70) $\displaystyle\simeq$ $\displaystyle
4\pi\int\mathrm{d}\alpha_{0}\,\mathrm{d}^{5}\alpha_{2}|\psi(\alpha_{0},\alpha_{2})|^{2}\sum_{m}\int\mathrm{d}r~{}|r~{}\Phi(r,\alpha_{0})|^{2}$
$\displaystyle\times\delta\left[\omega+\omega^{\mathrm{gs}}_{X}(\alpha_{0})-\frac{1}{\hbar}\lambda_{i}(r,[\alpha])\right]\;,$
## Appendix C
In this Appendix we consider that the doped drop is cylindrically symmetric.
We expand the helium density and ground state wave function into spherical
harmonics with $\mu=0$
$\displaystyle\rho(\mathbf{r})$ $\displaystyle=$
$\displaystyle\sum_{\lambda=0}^{\infty}\beta_{\lambda}\tilde{\rho}_{\lambda}(r)Z_{\lambda
0}(\hat{r})$ $\displaystyle\Phi(\mathbf{r})$ $\displaystyle=$
$\displaystyle\sum_{\lambda=0}^{\infty}\gamma_{\lambda}\tilde{\Phi}_{\lambda}(r)Z_{\lambda
0}(\hat{r})\;,$ (71)
where
$\beta_{\lambda}=\frac{2\lambda+1}{4\pi}\int\mathrm{d}^{3}\mathbf{r}\rho(\mathbf{r})Z_{\lambda
0}(\hat{r})$ and
$\tilde{\rho}_{\lambda}(r)=\frac{2\lambda+1}{4\pi\beta_{\lambda}}\int\mathrm{d}\Omega\rho(\mathbf{r})Z_{\lambda
0}(\hat{r})$, with analogous definitions for $\tilde{\Phi}_{\lambda}$ and
$\gamma_{\lambda}$. If
$\beta_{0}\gg\beta_{\lambda>0}\,(\Rightarrow\gamma_{0}\gg\gamma_{\lambda>0})$,
we can compute the line shape to first order in $\beta_{\lambda>0}$
($\gamma_{\lambda>0}$); in analogy with Eq. (54) we write
$\displaystyle I(\omega,[\beta])\simeq I(\omega,\beta_{0})$
$\displaystyle-4\pi\sum_{i}\sum_{\lambda=1}^{\infty}\int\mathrm{d}r~{}|r~{}\gamma_{0}\tilde{\Phi}_{0}(r)|^{2}\delta^{\prime}[\omega-\omega(\alpha_{0})-V^{\mathrm{ex}}_{i}(r,\beta_{0})/\hbar]\frac{1}{\hbar}\epsilon_{\lambda
0}^{i}(0)\beta_{\lambda}\;,$ (72)
with $\epsilon_{\lambda 0}^{i}$ defined now by the potential matrix
$\displaystyle U(\mathbf{r},[\beta])$ $\displaystyle\simeq$ $\displaystyle
U(r,\beta_{0})+\frac{4\pi}{15}\int\mathrm{d}r^{\prime}\,r^{\prime
2}\,\tilde{\rho}_{2}(r^{\prime})[V_{\Sigma}(r^{\prime})-V_{\Pi}(r^{\prime})]\left(\begin{array}[]{ccc}-1&0&0\\\
0&-1&0\\\ 0&0&2\end{array}\right)\beta_{2}\;.$ (76)
Introducing the shape deformations defined in Eq. (34), to first order this
matrix becomes
$\displaystyle U(\mathbf{r},[\beta],[\alpha])$ $\displaystyle\simeq$
$\displaystyle
U(r,\beta_{0},\alpha_{0})+\frac{4\pi}{5\sqrt{3}}\int\mathrm{d}r^{\prime}\,r^{\prime
2}\,\beta_{0}\,\tilde{\rho}^{\prime}_{0}(r^{\prime},\alpha_{0})[V_{\Sigma}(r^{\prime})-V_{\Pi}(r^{\prime})]$
(80)
$\displaystyle\times\left(\begin{array}[]{ccc}-\frac{1}{\sqrt{3}}\tilde{\alpha}_{20}+\alpha_{22}&\alpha_{2-2}&\alpha_{21}\\\
\alpha_{2-2}&-\frac{1}{\sqrt{3}}\tilde{\alpha}_{20}-\alpha_{22}&\alpha_{2-1}\\\
\alpha_{21}&\alpha_{2-1}&\frac{2}{\sqrt{3}}\tilde{\alpha}_{20}\end{array}\right)\;,$
where we have defined $\tilde{\alpha}_{20}\equiv\alpha_{20}+\beta_{2}C$, with
$C=\frac{\displaystyle\int\mathrm{d}r^{\prime}\,r^{\prime
2}\,\tilde{\rho}_{2}(r^{\prime})[V_{\Sigma}(r^{\prime})-V_{\Pi}(r^{\prime})]}{\displaystyle\int\mathrm{d}r^{\prime}\,r^{\prime
2}\,\beta_{0}\,\tilde{\rho}^{\prime}_{0}(r^{\prime},\alpha_{0})[V_{\Sigma}(r^{\prime})-V_{\Pi}(r^{\prime})]}\;.$
(81)
These equations show that the computation of the line shape for this geometry
is as in the spherical case but with a shift in the $\alpha_{20}$ parameter.
The dipole absorption spectrum is finally obtained as
$\displaystyle I(\omega)$ $\displaystyle\propto$ $\displaystyle
4\pi\int\mathrm{d}\alpha_{0}\,\mathrm{d}^{5}\alpha_{2}|\psi(\alpha_{0},\alpha^{\prime}_{2})|^{2}\sum_{m}\int\mathrm{d}r~{}|r~{}\Phi(r,\alpha_{0})|^{2}$
(82)
$\displaystyle\times\delta\left[\omega+\omega^{\mathrm{gs}}_{X}(\alpha_{0})-\frac{1}{\hbar}\lambda_{i}(r,[\alpha])\right]$
with
${\alpha^{\prime}}_{2}^{2}=({\alpha}_{20}-\beta_{2}C)^{2}+\alpha_{22}^{2}+\alpha_{2-2}^{2}+\alpha_{21}^{2}+\alpha_{2-1}^{2}$.
## References
* (1) F. Stienkemeier and A. F. Vilesov, J. Chem. Phys. 115, 10119 (2001).
* (2) F. Stienkemeier and K. K. Lehmann, J. Phys. B 39, R127 (2006).
* (3) M. Barranco, R. Guardiola, S. Hernández, R. Mayol, and M. Pi, J. Low Temp. Phys. 142, 1 (2006).
* (4) M. Mella, G. Calderoni, and F. Cargnoni, J. Chem. Phys. 123, 054328 (2005).
* (5) M. Elhiyani and M. Lewerenz, contribution to the XXII International Symposium on Molecular Beams, University of Freiburg (2007).
* (6) M. Elhiyani and M. Lewerenz, contribution to the 16th. Symposium on Atomic and Surface Physics and Related Topics, Innsbruck University Press (2008).
* (7) A. Hernando, R. Mayol, M. Pi, M. Barranco, F. Ancilotto, O. Bünermann, and F. Stienkemeier, J. Phys. Chem. A 111, 7303 (2007).
* (8) J. Reho, U. Merker, M. R. Radcliff, K. K. Lehmann, and G. Scoles, J. Chem. Phys. 112, 8409 (2000).
* (9) A. Przystawik, S. Göde, J. Tiggesbäumker, and K-H. Meiwes-Broer, contribution to the XXII International Symposium on Molecular Beams, University of Freiburg (2007); A. Przystawik, S. Göde, T. Döppner, J. Tiggesbäumker, and K-H. Meiwes-Broer, submitted to Phys. Rev. Lett. (2008).
* (10) Y. Moriwaki and N. Morita, Eur. Phys. J. D 5, 53 (1999).
* (11) Y. Moriwaki, K. Inui, K. Kobayashi, F. Matsushima, and N. Morita, J. Mol. Struc. 786, 112 (2006).
* (12) J. P. Toennies and A. F. Vilesov, Angew. Chem. Ind. Ed. 43 2622 (2004).
* (13) Y. Ren and V. V. Kresin, Phys. Rev. A 76, 043204 (2007).
* (14) J. Tiggesbäumker and F. Stienkemeier, Phys. Chem. Chem. Phys. 9, 4748 (2007).
* (15) F. Dalfovo, A. Lastri, L. Pricaupenko, S. Stringari, and J. Treiner, Phys. Rev. B 52, 1193 (1995).
* (16) L. Giacomazzi, F. Toigo, and F. Ancilotto, Phys. Rev. B 67, 104501 (2003).
* (17) L. Lehtovaara, T. Kiljunen, and J. Eloranta, J. of Comp. Phys. 194, 78 (2004).
* (18) F. Ancilotto, M. Barranco, F. Caupin, R. Mayol, and M. Pi, Phys. Rev. B 72, 214522 (2005); F. Ancilotto, M. Pi, R. Mayol, M. Barranco, and K. K. Lehmann, J. Phys. Chem. A 111, 12695 (2007).
* (19) R. J. Hinde, J. Phys. B: At. Mol. Opt. Phys. 36, 3119 (2003).
* (20) H. Partridge, J. R. Stallcop, and E. Levin, J. Chem. Phys. 115, 6471 (2001).
* (21) M. Guilleumas, F. Garcias, M. Barranco, M. Pi, and E. Suraud, Z. Phys. D 26, 385 (1993).
* (22) M. Frigo and S. G. Johnson, Proc. IEEE 93(2), 216 (2005).
* (23) W. H. Press, S. A. Teulosky, W. T. Vetterling, and B. P. Flannery, Numerical Recipes in Fortran 77: The Art of Scientific Computing (Cambridge University Press, Cambridge, 1999).
* (24) F. Ancilotto, D. G. Austing, M. Barranco, R. Mayol, K. Muraki, M. Pi, S. Sasaki, and S. Tarucha, Phys. Rev. B 67, 205311 (2003).
* (25) A. Hernando, M. Barranco, R. Mayol, M. Pi, and M. Krośnicki, Phys. Rev. B 77, 024513 (2008).
* (26) M. Lax, J. Chem. Phys. 20, 1752 (1952).
* (27) F. O. Ellison, J. Am. Chem. Soc. 85, 3540 (1963).
* (28) P. B. Lerner, M. B. Chadwick, and I. M. Sokolov, J. Low Temp. Phys. 90, 319 (1993).
* (29) T. Kinoshita, K. Fukuda, and T. Yabuzaki, Phys. Rev. B 54, 6600 (1996).
* (30) K. K. Lehmann and A. M. Dokter, Phys. Rev. Lett. 92, 173401 (2004).
* (31) We have corrected here an obvious error made in Ref. Her08, .
* (32) F. Stienkemeier, J. Higgins, C. Callegari, S. I. Kanorsky, W. E. Ernst, and G. Scoles, Z. Phys. D 38, 253 (1996).
* (33) O. Bünermann, G. Droppelmann, A. Hernando, R. Mayol, and F. Stienkemeier, J. Phys. Chem. A 111, 12684 (2007).
* (34) The location of the two components of the broad line in liquid helium and in drops is sensibly the same, whereas the relative intensity of the components seems as exchanged.
* (35) F. Stienkemeier, F. Meier, and H. O. Lutz, J. Chem. Phys. 107, 10 816 (1997).
* (36) F. Stienkemeier, F. Meier, and H. O. Lutz, Eur. Phys. J. D 9, 313 (1999).
* (37) Y. Moriwaki and N. Morita, Eur. Phys. J. D 33, 323 (2005).
* (38) W. C. Martin and R. Zalubas, J. Phys. Chem. Ref. Data 9, 1 (1980).
* (39) D. M. Brink, S. Stringari, Z. Phys. D. 15, 257 (1990).
* (40) Since the hydrodynamic mass depends on $R$, we have compute $w(R)$ using several $M^{*}$ values between $M^{*}=40$ a.u., which corresponds to the mass in the bulk liquid, and the free Mg mass $M^{*}=24$ a.u. This yields the band of $w(R)$ values displayed in the figure.
* (41) The mean value $\langle L^{2}\rangle$ is calculated as $Q^{-1}Tr\left(L^{2}e^{-{\cal H}/k_{B}T}\right)=Q^{-1}\sum_{n\ell}\ell(\ell+1)(2\ell+1)e^{-E_{n\ell}/k_{B}T}$ in the quantal approach, and as $Q^{-1}\int_{0}^{\infty}dRR^{2}\sum_{\ell}\ell(\ell+1)(2\ell+1)e^{-V_{\mathrm{eff}}(R)/k_{B}T}$ in the semiclassical approach.
* (42) F. Ancilotto, P. B. Lerner, and M. W. Cole, J. Low Temp. Phys. 101, 5-6 (1995).
* (43) L. Wilets, Theories of Nuclear Fission (Clarendon Press, Oxford, 1964).
* (44) P. Ring and P. Schuck, The Nuclear Many-Body Problem (Springer-Verlag, New York, 1980).
* (45) W. B. Fowler and D. L. Dexter, Phys. Rev. 176, 337 (1968).
* (46) H. Guérin, J. Phys. B: At. Mol. Opt. Phys. 25, 3371 (1992).
* (47) M. J. Renne and B. R. A. Nijboer, J. Phys. C 6, L10 (1973).
* (48) E. Cheng, M. W. Cole, and M. H. Cohen, Phys. Rev. B 50, 1136 (1994); Erratum ibid. 50, 16 134 (1994).
* (49) F. Stienkemeier, O. Bünermann, R. Mayol, F. Ancilotto, M. Barranco, and M. Pi, Phys. Rev. B 70, 214509 (2004).
* (50) M. Lewerenz, B. Schilling, and J.P. Toennies, J. Chem. Phys. 102, 8191 (1995).
* (51) E.B. Gordon, Low Temp. Phys. 30, 756 (2004).
* (52) J. Eloranta, Phys. Rev. B 77, 134301 (2008).
* (53) We note at this point that a quantitative estimation of the energy barrier height temporarily preventing the collapse of two Mg atoms into a dimer is made difficult by the fact that its actual value is determined by a delicate balance between the Mg-He interactions and the long-range part of the Mg-Mg interaction in vacuum, which is affected by some uncertainty (see Refs. Gue92, ; Czu01, ; Tie02, ). Any small difference in the van der Waals tail of the Mg-Mg interaction at distances of $\sim 9-10$ Å would result in a large change of the estimated barrier height. Besides, we want also to stress the difficulty to estimate the attempting frequency [inverse of the prefactor multiplying the exponential in Eq. (28)], in view of the kind of configurations appearing in this problem, see e.g. Fig. 9.
* (54) E. Czuchaj, M. Krośniki, and H. Stoll, Theor. Chem. Acc, 107, 27 (2001).
* (55) E. Tiesinga, S. Kotochigova, and P. S. Julienne, Phys. Rev. A 65, 042722 (2002).
* (56) S. Ogata, J. of Phys. Soc. Jap. 68, 2153 (1999).
* (57) M. Mella, M. C. Colombo, and F. G. Morosi, J. Chem. Phys. 117, 9695 (2002).
* (58) T. Nakatsukasa, K. Yabana, and G. F. Bertsch, Phys. Rev. A 65, 032512 (2002).
* (59) K. K. Lehmann, Phys. Rev. Lett. 88, 145301 (2002).
Figure 1: (Color online) $X^{1}\Sigma$ Mg-He pair potentials used in this
work: squares connected with a solid line, from Ref. Hin03, ; circles
connected with a dotted line, from Ref. Par01, .
Figure 2: Energy of the Mg atom as a function of the number of atoms in the
drop, obtained using the Mg-He potential of Ref. Hin03, (squares). The values
given in Ref. Her07, are also displayed (dots). The lines are drawn to guide
the eye.
Figure 3: (Color online) Bottom panel: total energy (K) of Mg@4He50 as a
function of ${\cal Z}_{0}$ (Å) obtained using the Mg-He potential of Ref.
Hin03, (squares) and of Ref. Par01, (circles). The energies are referred to
their equilibrium values, -157.0 K and -153.8 K, respectively. The vertical
line locates the drop surface at $R_{1/2}=r_{0}N^{1/3}$, with $r_{0}=2.22$ Å.
The horizontal line has been drawn 0.4 K above the equilibrium energy. Top
panel: probability densities for the configurations displayed in the bottom
panel; the single peak distribution corresponds to the Mg-He interaction of
Ref. Par01, .
Figure 4: (Color online) Helium density plots of the Mg@4He50 droplet in the
$y=0$ plane obtained using the Mg-He potential of Ref. Hin03, . From top to
bottom and left to right, the ${\cal Z}_{0}$ values correspond to the
stationary points displayed in Fig. 3, namely 0, 2, 4, and 6 Å, respectively.
The brighter regions are the higher density ones.
Figure 5: (Color online) Helium density profiles of the Mg@4He50 droplet along
the $z$ axis obtained using the Mg-He potential of Ref. Hin03, . The ${\cal
Z}_{0}$ value is indicated in each panel. Dashed lines, pure drops; solid
lines, doped drops.
Figure 6: (Color online) Bottom panel: total energy (K) of Mg@4He1000 as a
function of ${\cal Z}_{0}$ (Å) obtained using the Mg-He potential of Ref.
Hin03, (squares) and of Ref. Par01, (circles). The energies are referred to
their equilibrium values, -5482.0 K and -5476.8 K, respectively. The vertical
lines roughly delimit the drop surface region, conventionally defined as the
radial distance between the points where the density equals 0.1$\rho_{b}$ and
0.9$\rho_{b}$, being $\rho_{b}=0.0218$ Å-3 the bulk liquid density. The
horizontal line has been drawn 0.4 K above the equilibrium energy. Top panel:
total energy (K) of Mg@4HeN with $N=1000$ and 2000 (vertical right scale) as a
function of ${\cal Z}_{0}$ (Å) obtained using the Mg-He potential of Ref.
Hin03, . The energies are referred to their equilibrium values, $-5482.0$ K
and $-11\,629.8$ K, respectively, and the distances (horizontal scale) are
referred to the $R_{1/2}$ radius; also shown are the corresponding probability
densities (vertical left scale): dot(dot-dash) line, $N=1000(2000)$. The solid
line represents the probability density of the $N=10\,000$ drop.
Figure 7: (Color online) Total absorption spectrum of one Mg atom attached to
4He1000 in the vicinity of the $3s3p$ 1P${}_{1}\leftarrow 3s^{2}$ 1S0
transition. The line has been decomposed into its two $\Pi$ and one $\Sigma$
components, the former one is the higher frequency transition. The starred
vertical line represents the gas-phase transition. The experimental curve,
adapted for Ref. Reh00, , has been vertically offset for clarity. Also shown
is the absorption spectrum obtained neglecting homogeneous broadening (hatched
region)
Figure 8: (Color online) From top to bottom and left to right, (Mg+Mg)@4He1000
metastable configurations for Mg-Mg interatomic distances $d=18.5$ Å, 12.9 Å,
and 9.3 Å, and total energies -5567.8 K, -5573.9 K, and -5580.3 K,
respectively. The bottom right panel shows the specularly symmetric
configuration at $d=9.5$ Å with total energy $-5581.4$ K. The brigther regions
are the higher density ones.
Figure 9: (Color online) Helium density profiles of the (Mg+Mg)@4He1000
symmetric configuration ($d=9.5$ Å) along the $z$-axis (solid line) and the
$x$\- or $y$-axis (dashed line).
Figure 10: (Color online) Energy (K) of the Mg+Mg@4He1000 system as a function
of the Mg-Mg distance (Å). The energies have been referred to that of the
metastable equilibrium configuration (local minimum) at $L=0$. The lines have
been obtained by a cubic spline of the actual calculations. From bottom to
top, the curves correspond to $L=0$ to $50\hbar$ in $10\hbar$ steps.
Figure 11: (Color online) Energy (K) of the Mg+Mg@3He1000 system as a function
of the Mg-Mg distance (Å). The line has been drawn to guide the eye.
Figure 12: (Color online) Top panel: helium density profiles of the
(Mg+Mg)@3He1000 complex at $d=8.6$ Å along the $z$-axis (solid line) and the
$x$\- or $y$-axis (dashed line). Bottom panel: equidensity lines corresponding
to the same configuration. The brigther regions are the higher density ones.
Figure 13: (Color online) Total absorption spectrum of one Mg atom attached to
4He1000 in the vicinity of the $3s3p$ 1P${}_{1}\leftarrow 3s^{2}$ 1S0
transition in the distorted environment created by the presence of another Mg
atom. The line has been decomposed into its two $\Pi$ and one $\Sigma$
components, the former one is the higher frequency transition. The starred
vertical line represents the gas-phase transition. The experimental curve,
adapted for Ref. Reh00, , has been vertically offset for clarity. Also shown
is the absorption spectrum obtained neglecting homogeneous broadening (hatched
region)
Table 1: Atomic shift $\Delta\omega$ of Mg@4He1000 ($R_{1/2}=22.2$ Å) as a function of the average distance between the magnesium atom and the center of mass of the 4He1000 moiety. Also indicated is the corresponding wavelength $\lambda$. The value of the transition energy in the gas phase is 35 051 cm-1.Mar80 ${\cal Z}_{0}$ (Å) | $\Delta\omega$ (cm-1) | $\lambda$ (nm)
---|---|---
$0$ | 659.0 | 280.0
$10$ | 642.1 | 280.2
$14$ | 615.8 | 280.4
$18$ | 520.3 | 281.1
$22$ | 492.5 | 281.4
| arxiv-papers | 2008-07-11T14:00:19 | 2024-09-04T02:48:56.713621 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Alberto Hernando, Manuel Barranco, Ricardo Mayol, Marti Pi and\n Francesco Ancilotto",
"submitter": "Alberto Hernando",
"url": "https://arxiv.org/abs/0807.1857"
} |
0807.1894 | # Robust magnetoplasmon spectrum of modulated graphene at finite temperature
M. Tahir1∗ , K. Sabeeh2†, A. MacKinnon1 1Blackett Laboratory, Imperial College
London, South Kensington campus, London SW7 2AZ, United Kingdom. 2Department
of Physics, Quaid-i-Azam University, Islamabad 45320, Pakistan.
([; date; date; date; date)
###### Abstract
In this work, we determine the effects of temperature on the magnetoplasmon
spectrum of an electrically modulated graphene monolayer as well as the two-
dimensional electron gas (2DEG). The intra-Landau-band magnetoplasmon spectrum
within the Self Consistent Field (SCF) approach is investigated for both the
aforementioned systems. Results obtained not only exhibit Shubnikov-de Hass
(SdH) oscillations but also commensurability oscillations (Weiss
oscillations). These oscillations are periodic as a function of inverse
magnetic field. We find that both the magnetic oscillations, SdH and Weiss,
have a greater amplitude and are more robust against temperature in graphene
compared to conventional 2DEG. Furthermore, there is a $\pi$ phase shift
between the magnetoplasmon oscillations in the two systems which can be
attributed to Dirac electrons in graphene acquiring a Berry’s phase as they
traverse a closed path in a magnetic field.
###### pacs:
72.20.My,72.80.Rj,73.50.Dn,73.40.-c
Date text]date
## I Introduction
Remarkable progress made in epitaxial crystal growth techniques has led to the
fabrication of novel semiconductor heterostructures. These modern
microstructuring techniques have made possible the realization of two-
dimensional electron gas (2DEG) system in semiconductor heterostructures. 2DEG
is a condensed matter system where a number of novel phenomena have been
observed over the years. One such phenomena was the observation of
commensurability oscillations in physical properties when the 2DEG in the
presence of a magnetic field is subjected to electric modulation. The electric
modulation introduces an additional length scale in the system and the
occurrence of these oscillations, commonly known as Weiss oscillations, is due
to the commensurability of the electron cyclotron diameter at the Fermi energy
and the period of the electric modulation. These oscillations were found to be
periodic in the inverse magnetic field 1 ; 2 ; 3 . This type of electrical
modulation of the 2D system can be carried out by depositing an array of
parallel metallic strips on the surface or through two interfering laser beams
1 ; 2 ; 3 .
One of the important electronic properties of a system are the collective
excitations (plasmons). Weiss oscillations in the magnetoplasmon spectrum of
2DEG has been the subject of continued interest 4 . Recently, the fabrication
of crytallites of graphene monolayer has generated a lot of interest. The
study of this new material is not only of academic interest but there are
serious efforts underway to investigate whether it can serve as the basic
material for a carbon-based electronics that might replace silicon-based
electronics. Graphene has a honeycomb lattice of carbon atoms. The
quasiparticles in graphene have a band structure in which electron and hole
bands touch at two points in the Brillouin zone. At these Dirac points the
quasiparticles obey the massless Dirac equation. In other words, they behave
as massless Dirac fermions leading to a linear dispersion relation
$\epsilon_{k}=vk$ ( with the characteristic velocity $v\simeq 10^{6}m/s)$.
This difference in the nature of the quasiparticles in graphene from
conventional 2DEG has given rise to a host of new and unusual phenomena such
as anomalous quantum Hall effects and a $\pi$ Berry’s phase5 6 . These
transport experiments have shown results in agreement with the presence of
Dirac fermions. The 2D Dirac-like spectrum was confirmed recently by cyclotron
resonance measurements and also by angle resolved photoelectron spectroscopy
(ARPES) measurements in monolayer graphene7 . Recent theoretical work on
graphene multilayers has also shown the existence of Dirac electrons with a
linear energy spectrum in monolayer graphene8 . Plasmons in graphene were
studied as early as in the eighties9 and more recently 10 . In this work, we
investigate magnetoplasmons in graphene monolayer when it is subjected to
electric modulation at finite temperature. The zero temperature study was
carried out by us in an earlier work 11 where we showed that the intra-Landau
band magnetoplasmons exhibit Weiss oscillations and are present in the system
due to the broadening of the Landau levels as a result of modulation.
Presently, we extend our earlier work by taking into account the effects of
temperature on the magnetoplasmons in graphene. While working on the finite
temperature dispersion relation for magnetoplasmons in graphene we realized
that finite temperature calculation for 2DEG is not available in the
literature so in this work we also determine the finite temperature dispersion
relation for magnetoplasmons in electrically modulated 2DEG. We find that the
magnetoplasmons are more robust against temperature than those occurring in a
conventional 2DEG realized in semiconductor heterostructures. Characteristic
temperatures are determined for the damping of the magnetic oscillations (SdH
and Weiss) in both the graphene and 2DEG systems. The experimental study of
these should be as revealing as the enhanced magnetoconductivity prediction12
and even more interesting: it bears directly on the many-body properties of
the two-dimensional systems, as well as the frequency dependent transport and
optical response.
## II Robust magnetoplasmon spectrum in graphene at finite temperature
We consider two-dimensional Dirac electrons in graphene moving in the
$x-y$-plane. The magnetic field ($B_{z}$) is applied along the $z$-direction
perpendicular to the graphene plane. This system is subjected to weak electric
modulation along the $x$-direction. We employ the Landau gauge and write the
vector potential as $A=(0,Bx,0)$. The two-dimensional Dirac like Hamiltonian
for single electron in the Landau gauge is 1 ; 2 ; 8
$H_{0}=v\sigma.(-i\hslash\nabla+eA).$ (1)
Here $\sigma=\\{\sigma_{x},\sigma_{y}\\}$ are the Pauli matrices and $v$
characterizes the electron velocity. The complete Hamiltonian of our system is
represented as
$H=H_{0}+V(x)$ (2)
where $H_{0}$ is the unmodulated Hamiltonian and $V(x)$ is the one-dimensional
periodic modulation potential along the $x$-direction modelled as
$V(x)=V_{0}\cos(Kx)$ (3)
where $K=2\pi/a,a$ is the period of modulation and $V_{0}$ is the constant
modulation amplitude. Applying standard perturbation theory to determine the
first order correction to the unmodulated energy eigenvalues in the presence
of modulation we obtain
$\varepsilon(n,x_{0})=\hslash\omega_{g}\sqrt{n}+F_{n}\cos(Kx_{0})$ (4)
with $\omega_{g}=v\sqrt{\frac{2eB}{\hslash}}$ is the cyclotron frequency of
Dirac electrons in graphene and $n$ is an integer,
$F_{n}=\frac{1}{2}V_{0}e^{-\chi/2}[L_{n}(\chi)+L_{n-1}(\chi)]$ and
$\chi=K^{2}l^{2}/2,$ $x_{0}=l^{2}k_{y}$, $L_{n}(\chi)$ are Laguerre
polynomials and $l=\sqrt{\hslash/eB}$ is the magnetic length. The Landau level
spectrum for Dirac electrons in graphene is significantly different from the
spectrum for electrons in conventional 2DEG which is given as
$\varepsilon(n)=\hbar\omega_{c}(n+1/2)$, where $\omega_{c}$ is the cyclotron
frequency.
The intra-Landau-band plasmon spectrum is determined within the Ehrenreich-
Cohen self-consistent-field (SCF) approach13 . The plasmon dispersion relation
is given by
$\overset{\sim}{\omega}^{2}=\frac{16e^{2}}{k\bar{q}\pi}\frac{1}{al^{2}}\sin^{2}(\frac{\pi}{a}(x_{0}^{\prime}))\times
A_{n},$ (5)
$A_{n}=\underset{n}{\sum}F_{n}\times\int_{0}^{a/2}dx_{0}f(\varepsilon(n,x_{0}))\cos(Kx_{0})$
(6)
where $x_{0}^{\prime}=l^{2}q_{y},\bar{q}$ is the two dimensional wave number,
$k$ is background dielectric constant and $f(\varepsilon(n,x_{0}))$ is the
Fermi distribution function. This result was obtained previously in 11 where
it was used to obtain the zero temperature dispersion relation. Here we are
interested in temperature effects and we proceed as follows to determine the
finite temperature dispersion relation. In the regime of weak modulation, the
distribution function $f(\varepsilon(n,x_{0}))$ can be expressed as
$f(\varepsilon(n,x_{0}))\simeq
f(\varepsilon_{n})+F_{n}f^{\prime}(\varepsilon_{n})\cos(Kx_{0}),$ (7)
where $f^{\prime}(x)=\frac{d}{dx}f(x)$. Substituting the above expression for
$f(\varepsilon(n,x_{0})),$ the integral over $x_{0}$ can be performed to yield
the intra-Landau band plasmon dispersion relation at finite temperature as
$\overset{\sim}{\omega}^{2}=\frac{4e^{2}}{k\bar{q}\pi
l^{2}}\sin^{2}[\frac{\pi}{a}(x_{0}^{\prime})]\times B_{n},$ (8)
with $B_{n}=\sum F_{n}^{2}\times[-f^{\prime}(\varepsilon)]$. Related physics
can be made more transparent by considering the asymptotic expression of
intra-Landau band magnetoplasmon spectrum, where analytic results in terms of
elementary functions can be obtained.
To obtain the asymptotic expression for the intra-Landau band plasmon spectrum
we employ the following asymptotic expression for the Laguerre polynomials12
$\exp^{-\chi/2}L_{n}(\chi)\rightarrow\frac{1}{\sqrt{\pi\sqrt{n\chi}}}\cos\left(2\sqrt{n\chi}-\frac{\pi}{4}\right)$
(9)
Note that the asymptotic results are valid when many Landau Levels are filled.
We now take the continuum limit:
$n-->\frac{\varepsilon^{2}}{\hslash^{2}\omega_{g}^{2}},\overset{\infty}{\underset{n=0}{{\displaystyle\sum}}}-->{\displaystyle\int\limits_{0}^{\infty}}\frac{2\varepsilon
d\varepsilon}{\hslash^{2}\omega_{g}^{2}}.$ (10)
$B_{n}$ that appears in the expression for $\overset{\sim}{\omega}^{2}$in
equation (8) can be expressed as
$B_{n}=\frac{V_{0}^{2}}{\pi}{\displaystyle\int\limits_{0}^{\infty}}\frac{2\varepsilon
d\varepsilon}{\hslash^{2}\omega_{g}^{2}\frac{\varepsilon}{\hslash\omega_{g}}\sqrt{\chi}}\frac{\beta
g(\varepsilon)}{[g(\varepsilon)+1)]^{2}}\cos^{2}\left(2\sqrt{n\chi}-\frac{\pi}{4}\right)\cos^{2}\left(\frac{1}{2}\sqrt{\frac{\chi}{n}}\right)$
(11)
where $g(\varepsilon)=\exp\beta(\varepsilon-\varepsilon_{F})$,
$\varepsilon_{F}=\hslash vK_{F},K_{F}=\sqrt{2\pi n_{D}}$, $n_{D}$ is the
number density, $\chi=\frac{K^{2}l^{2}}{2}=2\pi^{2}/b$ with
$b=\frac{eBa^{2}}{\hslash}$ and $\beta=\frac{1}{K_{B}T}.$ Now assuming that
the temperature is low such that $\beta^{-1}\ll\varepsilon_{F}$ and
substituting $\varepsilon=\varepsilon_{F}+s\beta^{-1}$, we rewrite the above
integral as
$B_{n}=\frac{2V_{0}^{2}\cos^{2}\left(\frac{\hslash\omega_{g}}{2\varepsilon_{F}}\sqrt{\chi}\right)}{\pi\hslash\omega_{g}\sqrt{\chi}}{\displaystyle\int\limits_{0}^{\infty}}ds\frac{e^{s}}{[e^{s}+1)]^{2}}\cos^{2}\left(\frac{2\varepsilon_{F}\sqrt{\chi}}{\hslash\omega_{g}}-\frac{\pi}{4}+\frac{2\sqrt{\chi}\beta^{-1}}{\hslash\omega_{g}}s\right)$
(12)
with the result
$B_{n}=\frac{V_{0}^{2}\cos^{2}\left(\frac{\hslash\omega_{g}}{2\varepsilon_{F}}\sqrt{\chi}\right)}{2\pi\hslash\omega_{g}\sqrt{\chi}}\left[2-2A\left(\frac{T}{T_{W}}\right)+4A\left(\frac{T}{T_{W}}\right)\cos^{2}\left(\frac{2\varepsilon_{F}\sqrt{\chi}}{\hslash\omega_{g}}-\frac{\pi}{4}\right)\right]$
(13)
where
$A(\frac{T}{T_{W}})=\frac{\frac{T}{T_{W}}}{\sinh(\frac{T}{T_{W}})}-^{(\frac{T}{T_{W}}-->\infty)}->=2\frac{T}{T_{W}}e^{-\frac{T}{T_{W}}}$
, $\frac{T}{T_{W}}=\frac{4\pi\sqrt{\chi}K_{B}T}{\hslash\omega_{g}}$ and
$T_{W}=\frac{\hslash\omega_{g}}{4\pi K_{B}\sqrt{\chi}}$ is the characteristic
damping temperature of the Weiss oscillations. Substituting the expression for
$B_{n}$ in equation (8), the asymptotic expression for intra-Landau spectrum
is obtained
$\displaystyle\overset{\sim}{\omega}^{2}$
$\displaystyle=\frac{2e^{2}}{k\bar{q}\pi
l^{2}}\frac{V_{0}^{2}\cos^{2}\left(\frac{\omega_{g}}{2\varepsilon_{F}}\sqrt{\chi}\right)}{\pi\hslash\omega_{g}\sqrt{\chi}}\sin^{2}\left(\frac{\pi}{a}(x_{0}^{\prime})\right)$
$\displaystyle\qquad\times\left[2-2A\left(\frac{T}{T_{W}}\right)+4A\left(\frac{T}{T_{W}}\right)\cos^{2}\left(\frac{2\varepsilon_{F}\sqrt{\chi}}{\hslash\omega_{g}}-\frac{\pi}{4}\right)\right]$
(14)
The above expression is valid at high temperature
($K_{B}T>>\hslash\omega_{g}/2\pi^{2}$) and is not able to account for the SdH
oscillations occurring in the magnetoplasmon spectrum. To take these into
account we use the following expression for the density of states at low
magnetic fields in the absence of impurity scattering given by
$D(\varepsilon)=\frac{\varepsilon}{\pi
l^{2}\hslash^{2}\omega_{g}^{2}}\left[1-2\cos\left(\frac{2\pi\varepsilon^{2}}{\hslash^{2}\omega_{g}^{2}}\right)\right]$.
This expression for $D(\varepsilon)$ was obtained in 14 in the absence of
scattering and gap $\Delta=0$. The sum appearing in equation (6) can now be
expressed in the continuum approximation as
$\overset{\infty}{\underset{n=0}{{\displaystyle\sum}}}-->2\pi
l^{2}{\displaystyle\int\limits_{0}^{\infty}}D(\varepsilon)d\varepsilon.$
Therefore, the intra-Landau band magnetoplasmon dispersion relation that takes
into account both Weiss and SdH oscillations is given by
$\displaystyle\overset{\sim}{\omega}^{2}$
$\displaystyle=\frac{2e^{2}}{k\bar{q}\pi
l^{2}}\frac{V_{0}^{2}\cos^{2}\left(\frac{\hslash\omega_{g}}{2\varepsilon_{F}}\sqrt{\chi}\right)}{\pi\hslash\omega_{g}\sqrt{\chi}}\sin^{2}\left(\frac{\pi}{a}(x_{0}^{\prime})\right)$
$\displaystyle\qquad\times\left\\{\left[2-2A\left(\frac{T}{T_{W}}\right)+4A\left(\frac{T}{T_{W}}\right)\cos^{2}\left(\frac{2\varepsilon_{F}\sqrt{\chi}}{\hslash\omega_{g}}-\frac{\pi}{4}\right)\right]\right.$
$\displaystyle\qquad\qquad\
\left.-4\frac{2X_{SdH}}{\sinh(X_{SdH})}\cos\left(\frac{2\pi\varepsilon_{F}^{2}}{\hslash^{2}\omega_{g}^{2}}\right)\cos^{2}\left(\frac{2\varepsilon_{F}\sqrt{\chi}}{\hslash\omega_{g}}-\frac{\pi}{4}\right)\right\\}$
(15)
where
$X_{SdH}=\frac{4\pi^{2}\varepsilon_{F}}{\hslash^{2}\omega_{g}^{2}\beta}$,
$\frac{T}{T_{SdH}}=\frac{4\pi^{2}\epsilon_{F}K_{B}T}{\hslash^{2}\omega_{g}^{2}}$
and $T_{SdH}=\frac{\hslash^{2}\omega_{g}^{2}}{4\pi^{2}\epsilon_{F}K_{B}}$ is
the characteristic damping temperature of the SdH oscillations.
Following the same approach as given above for graphene monolayer, we can
obtain the intra-Landau band magnetoplasmon spectrum for conventional 2DEG at
finite temperature with the result
$\overset{\sim}{\omega}^{2}=\frac{4e^{2}m^{\ast}\omega_{c}}{\hslash
k\bar{q}\pi}\sin^{2}[\frac{\pi}{a}(x_{0}^{\prime})]\times B_{n}(C),$ (16)
where $m^{\ast}$ is the standard electron mass, $B_{n}(C)=\sum
F_{n}^{2}(C)\times[-f^{\prime}(\varepsilon)]$, and
$F_{n}(C)=V_{0}e^{-\chi/2}L_{n}(\chi)$ is the modulation width of the
conventional 2DEG. The corresponding asymptotic result is
$\displaystyle\overset{\sim}{\omega}^{2}$
$\displaystyle=\frac{4V_{0}^{2}e^{2}m^{\ast}\omega_{c}}{k\bar{q}2\pi^{2}\hslash\sqrt{\chi\hslash\omega_{c}\varepsilon_{F}}}\sin^{2}\left(\frac{\pi}{a}(x_{0}^{\prime})\right)$
$\displaystyle\qquad\times\left\\{\left[1-A\left(\frac{T}{T_{W}^{p}}\right)+2A\left(\frac{T}{T_{W}^{p}}\right)\cos^{2}\left(2\sqrt{\frac{\chi\varepsilon_{F}}{\hslash\omega_{c}}}-\frac{\pi}{4}\right)\right]\right.$
$\displaystyle\qquad\qquad\left.-4\frac{2X^{p}_{SdH}}{\sinh(X^{p}_{SdH})}\cos\left(\frac{2\pi\varepsilon_{F}}{\hslash\omega_{c}}\right)\cos^{2}\left[2\sqrt{\frac{\chi\varepsilon_{F}}{\hslash\omega_{c}}}-\frac{\pi}{4}\right]\right\\}$
(17)
where $\frac{T}{T_{W}^{p}}=\frac{4\pi^{2}k_{B}T}{\hslash\omega_{c}aK_{F}}$,
$X^{p}_{SdH}=\frac{2\pi^{2}}{\hbar\omega_{c}\beta}$,
$\frac{T}{T_{SdH}^{p}}=\frac{2\pi^{2}k_{B}T}{\hslash\omega_{c}},$
$\chi=\frac{K^{2}l^{2}}{2}=2\pi^{2}/b$ with $b=\frac{eBa^{2}}{\hslash}$.
Now we can compare the exact results, for the temperature dependent
magnetoplasmon in terms of the elementary functions, with those for a
conventional 2DEG. The differences are:
* •
the standard electron energy eigenvalues scale linearly with the magnetic
field whereas those for Dirac electrons in graphene scale as the square root.
* •
the temperature dependence of the Weiss oscillations,
$A\left(\frac{T}{T_{W}}\right)$, is clearly different from that of the
standard 2DEG, $A\left(\frac{T}{T_{W}^{p}}\right)$3 .
* •
the temperature dependence of SdH oscillations in graphene,
$\frac{X_{SdH}}{\sinh(X_{SdH})}$, is different from that of the standard 2DEG,
$\frac{X^{p}_{SdH}}{\sinh(X^{p}_{SdH})}$15 ; 16 .
* •
the density of states term that contains the cosine function responsible for
the SdH oscillations has a different dependence in each of the systems:
$\cos\left(\frac{2\pi\varepsilon_{F}^{2}}{\hslash^{2}\omega_{g}^{2}}\right)$14
and $\cos\left(\frac{2\pi\varepsilon_{F}}{\hslash\omega_{c}}\right)$ 15 ; 16
respectively.
These differences will give different results for the temperature dependent
magnetoplasmon spectrum, as we discuss in the next section.
## III Discussion of Results
The finite temperature intra-Landau band magnetoplasmon dispersion relation
for electrically modulated graphene and 2DEG given by equations (15) and (17)
are the central results of this work. These results allow us to clearly see
the effects of temperature on Weiss and SdH oscillations in the magnetoplasmon
spectrum of the two systems. We find that these two types of oscillations have
different characteristic damping temperatures. This is easily understood if we
realize that the origin of these oscillations is different. The SdH
oscillations arise due to the discreteness of the Landau levels and their
observation requires that the thermal energy $K_{B}T$ acquired by the
electrons at temperature $T$ has to be smaller than the separation between the
levels. Weiss oscillations are related to the commensurability of two lengths:
the size of the cyclotron orbit and the period of the modulation, these
oscillations will be observed if the spread in the cyclotron diameter is
smaller than the modulation period. To elucidate the effects of temperature we
present the magnetoplasmon spectrum in graphene and 2DEG at two different
temperatures in Figs.(1) and (2).
Figure 1: Intra-Landau-band plasma frequency as a function of inverse magnetic
field in a graphene monolayer at $T=1\,\mathrm{K}$ (solid line) and
$T=12\,\mathrm{K}$ (dashed line). Figure 2: Intra-Landau-band plasma frequency
as a function of inverse magnetic field in a conventional 2DEG at
$T=0.3\,\mathrm{K}$ (solid line) and $T=3\,\mathrm{K}$ (dashed line).
In Fig.(1), we show the magnetoplasmon energy as a function of the inverse
magnetic field in graphene monolayer at two different temperatures: $T=1K$ and
$12K.$ The following parameters were employed for graphene10 ; 11 : $k=3$,
$n_{D}=3.16\times 10^{16}$ m-2, $v=10^{6}m/s$, $a=380$ nm and $V_{0}=1.0$ meV.
We also take $q_{x}=0$ and $q_{y}=.01k_{F},$ with $k_{F}=(2\pi n_{D})^{1/2}.$
Weiss oscillations superimposed on the SdH oscillations can be clearly seen in
both the curves. In Eq.(15), the terms containing the characteristic
temperature for Weiss oscillations $T_{w}$ are mainly responsible for Weiss
oscillations whereas terms containing $T_{SdH}$ are responsible for the SdH
oscillations. The intra-Landau-band plasmons have frequencies comparable to
the bandwidth and they occur as a result of broadening of the Landau levels
due to the modulation in our system. These type of intra-Landau-band plasmons
accompanied by regular oscillatory behavior (in $1/B)$ of the SdH type was
first predicted in 17 for tunneling planar superlattice where the overlap of
electron wavefunction in adjacent quantum wells provides the mechanism for
broadening of Landau levels. The SdH oscillations occur as a result of
emptying out of electrons from successive Landau levels when they pass through
the Fermi level as the magnetic field is increased. The amplitude of these
oscillations is a monotonic function of the magnetic field when the Landau
bandwidth is independent of the band index $n.$ In the density modulated case,
the Landau bandwidths oscillate as a function of the band index $n,$ with the
result that in the plasmon spectrum of the intra-Landau band type, there is a
new kind of oscillation called Weiss oscillation which is also periodic in
$1/B$ but with a different period and amplitude from the SdH type oscillation.
At $T=12K,$ we find that the SdH oscillations are washed out while the Weiss
oscillations persist. From Eq.(15), we see that $T_{W}$ and $T_{SdH}$ set the
temperature scale at which these oscillations will be damped. For the 2DEG,
the magnetoplasmon energy as a function of inverse magnetic field is presented
in Fig.(2) at two different temperatures: $T=0.3K$ and $3K$. For conventional
2DEG (a 2DEG at the GaAs-AlGaAs heterojunction) we use the following
parameters1 ; 2 ; 3 ; 4 : $m^{\ast}=.07m_{e}$($m_{e}$ is the electron mass),
$k=12$ and $n_{D}=3.16\times 10^{15}$ m-2 with the modulation strength and
period same as in the graphene system. We find that the SdH oscillations in
the magnetoplasmon spectrum die out at $T=3K$ here while they are present at a
higher temperature ($12K$) in graphene.
Figure 3: Intra-Landau-band plasma frequency as a function of inverse magnetic
field in the region of very strong magnetic field: solid line at
$T=5\,\mathrm{K}$ in a graphene monolayer; dashed ($T=5\,\mathrm{K}$) and
dotted ($1\,\mathrm{K}$) in standard 2DEG.
$\frac{B^{\prime}}{B}=\frac{\hbar}{eBa^{2}}$ where $B$ is the magnetic field
and $B^{\prime}=\frac{\hbar}{ea^{2}}=0.0046\,\mbox{Tesla}$.
To compare the results for the two systems we show in Fig.(3) the
magnetoplasmon spectrum as a function of the inverse magnetic field
($\frac{B^{\prime}}{B}=\frac{\hbar}{Bea^{2}},$ where
$B^{\prime}=\frac{\hbar}{ea^{2}}=0.0046$ Tesla) for both graphene (solid line
at temperature $5K$) and 2DEG (dotted and dashed line at temperature $1K$ and
$5K$ respectively). The oscillations in the conventional 2DEG have been damped
out strongly at $5K$ but are well resolved, significant and have larger
amplitude in graphene at the same temperature. This confirms that the graphene
oscillations are more robust and enhanced against temperature than those in
the conventional 2DEG. The magnetic oscillations, SdH and Weiss, have a higher
amplitude in graphene compared to 2DEG. This can be attributed to the larger
characteristic velocity ($v\sim 10^{6}m/s)$ of electrons in graphene compared
to the Fermi velocity of standard electrons and smaller background dielectric
constant $k$ in graphene in contrast to conventional 2DEG. The temperature at
which we expect the Weiss oscillations to dampen is determined by comparing
the characteristic temperatures for Weiss oscillations in the two systems:
$\frac{T_{W}^{p}}{T_{W}}=\frac{v_{F}}{v}$, the ratio of the characteristic
temperatures is equal to the corresponding velocities at the Fermi surface12 .
For the parameters used in this work $\frac{T_{W}^{p}}{T_{W}}\sim 0.24,$
implying that Weiss oscillations are damped at a higher temperature in
graphene compared to the 2DEG. Similarly, damping of SdH oscillations can also
be compared in the two systems through their corresponding characteristic
temperatures: $\frac{T_{SdH}^{p}}{T_{SdH}}=\frac{\hslash K_{F}}{vm^{\ast}}\sim
0.23$ for $n_{D}=3.16\times 10^{15}$ m-2 whereas it is $\sim$ $0.74$ for
$n_{D}=3.16\times 10^{16}$ m${}^{-2}.$ Therefore, SdH oscillations are damped
at a higher temperature in graphene compared to 2DEG. Furthermore, our results
also show that these oscillations in the magnetoplasmon spectrum differ in
phase by $\pi$ in the two systems which is due to quasiparticles in graphene
acquiring a Berry’s phase of $\pi$ as they move in the magnetic field6 .
## IV Conclusions
We have determined the finite temperature intra-Landau band magnetoplasmon
frequency for electrically modulated graphene as well as the 2DEG in the
presence of a magnetic field employing the SCF approach. We find that the
magnetic oscillations (SdH and Weiss) in the magnetoplasmon spectrum in
graphene monolayer have a higher amplitude compared to the conventional 2DEG
realized in semiconductor heterostructures. Moreover, these oscillations
persist at a higher temperature in graphene compared to the 2DEG. Hence they
are more robust against temperature in graphene. Furthermore, $\pi$ Berry’s
phase acquired by the Dirac electrons leads to $\pi$ phase shift in the
magnetoplasmon spectrum in graphene monolayer compared to 2DEG.
One of us (K. Sabeeh) would like to acknowledge the support of the Pakistan
Science Foundation (PSF) through project No. C-QU/Phys (129). M. Tahir would
like to acknowledge the support of the Pakistan Higher Education Commission
(HEC).
* E-mail: m.tahir06@imperial.ac.uk
† E-mail: ksabeeh@qau.edu.pk, kashifsabeeh@hotmail.com
## References
* (1) D. Weiss, K. v. Klitzing, K. Ploog, and G. Weimann, Europhys. Lett., 8, 179 (1989); R. W. Winkler, J. P. Kotthaus, and K. Ploog, Phys. Rev. Lett. 62, 1177 (1989); R. R. Gerhardts, D. Weiss, and K. v. Klitzing, Phys. Rev. Lett. 62, 1173 (1989).
* (2) P. Vasilopoulos and F. M. Peeters, Phys. Rev. Lett. 63, 2120 (1989).
* (3) F. M. Peeters and P. Vasilopoulos, Phys. Rev. B 46, 4667 (1992).
* (4) H. L. Cui, V. Fessatidis and N. J. M. Horing, Phys. Rev. Lett. 63, 2598 (1989); O. G. Balev, N. Studart, P. Vasilopoulos, Phys. Rev. B 62, 15834 (2000), M.S. Kushwaha, Surface Science Reports 41, 1 (2001) and references therein.
* (5) K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, M. I. Katsnelson, I. V. Grigorieva, S. V. Dubonos, and A. A. Firsov, Nature 438, 197 (2005); Y. Zhang, Y. W. Tan, H. L. Stormer, and P. Kim, Nature 438, 201 (2005).
* (6) Y. Zheng and T. Ando, Phys. Rev. B 65, 245420 (2002); V. P. Gusynin and S. G. Sharapov, Phys. Rev. Lett. 95, 146801 (2005); N. M. R. Perez, F. Guinea, and A. H. Castro Neto, Phys. Rev. B 73, 125411 (2006); M. I. Katsnelson, K. S. Novoselov, and A. K. Geim, Nat. Phys. 2, 620 (2006); K. S. Novoselov, E. McCann, S. V. Morozov, V. I. Fal’ko, M. I. Katsnelson, U. Zeitler, D. Jiang, F. Schedin, and A. K. Geim, Nat. Phys. 2, 177 (2006).
* (7) R. S. Deacon, K-C. Chuang, R. J. Nicholas, K. S. Novoselov, and A. K. Geim, Phys. Rev. B 76, 081406(R) (2007).; S. Y. Zhou, G.-H. Gweon, J. Graf, A. V. Fedorov, C. D. Spataru, R. D. Diehl, Y. Kopelevich, D. H. Lee, S. G. Louie, and A. Lanzara, Nat. Phys. 2, 595 (2006).
* (8) B. Partoens and F. M. Peeters, Phys. Rev. B 75, 193402 (2007).
* (9) K. W. -K. Shung, Phys. Rev. B 34, 979 (1986).
* (10) M. -F. Lin, F. -L. Shyu, J. Phys. Soc. Japan 69, 607 (2000), V. Apalkov, X.-F. Wang and T. Chakraborty, Int. J. Mod. Phys. B 21, 1167 (2007); E. H. Hwang and S. Das Sarma, Phys. Rev. B 75, 205418 (2007); X.-F. Wang and T. Chakraborty, Phys. Rev. B 75, 033408 (2007).
* (11) M. Tahir, K. Sabeeh, Phys. Rev. B 76, 195416 (2007).
* (12) A. Matulis and F. M. Peeters, Phys. Rev. B 75, 125429 (2007).
* (13) H. Ehrenreich and M. H. Cohen, Phys. Rev. 115, 786 (1959).
* (14) V. P. Gusynin and S. G. Sharapov, Phys. Rev. B 71 125124 (2005).; S. G. Sharapov, V. P. Gusynin and H. Beck, Phys. Rev. B 69 075104 (2004).
* (15) R. B. Dingle, Proc. R. Soc. A 211, 517(1952).
* (16) Akira Isihara and Ludvig Smrcka, J. Physics. C: Solid State Phys. 19, 6777 (1986).
* (17) W. M. Que and G. Kirczenow, Phys. Rev. Lett. 62, 1687 (1989); W. M. Que and G. Kirczenow, Phys. Rev. B 36, 6596 (1987).
| arxiv-papers | 2008-07-11T17:11:32 | 2024-09-04T02:48:56.721749 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "M. Tahir, K. Sabeeh, and A. MacKinnon",
"submitter": "Muhammad Tahir",
"url": "https://arxiv.org/abs/0807.1894"
} |
0807.1908 | FTPI-MINN-08/24; UMN-TH-2703/08
Confinement and Localization on Domain Walls
R. Auzzi(1), S. Bolognesi(2), M. Shifman (2,3) and
A. Yung(2,4)
(1)Department of Physics, Swansea University,
Singleton Park, Swansea SA2 8PP, U.K.
(2)William I. Fine Theoretical Physics Institute, University of Minnesota,
116 Church St. S.E., Minneapolis, MN 55455, USA
(3) Laboratoire de Physique Théorique111Unité Mixte de Recherche du CNRS, (UMR
8627). Université de Paris-Sud XI
Bâtiment 210, F-91405 Orsay Cédex, FRANCE
(4) Petersburg Nuclear Physics Institute, Gatchina, St. Petersburg 188300,
RUSSIA
and
Institute of Theoretical and Experimental Physics, Moscow 117259, RUSSIA
We continue the studies of localization of the U(1) gauge fields on domain
walls. Depending on dynamics of the bulk theory the gauge field localized on
the domain wall can be either in the Coulomb phase or squeezed into flux tubes
implying (Abelian) confinement of probe charges on the wall along the wall
surface. First, we consider a simple toy model with one flavor in the bulk at
weak coupling (a minimal model) realizing the latter scenario. We then suggest
a model presenting an extension of the Seiberg–Witten theory which is at
strong coupling, but all theoretical constructions are under full control if
we base our analysis on a dual effective action. Finally, we compare our
findings with the wall in a “nonminimal” theory with two distinct quark
flavors that had been studied previously. In this case the U(1) gauge field
trapped on the wall is exactly massless because it is the Goldstone boson of a
U(1) symmetry in the bulk spontaneously broken on the wall. The theory on the
wall is in the Coulomb phase. We explain why the mechanism of confinement
discussed in the first part of the paper does not work in this case, and
strings are not formed on the walls.
## 1 Introduction
Localization of gauge fields on domain walls which are supported by some four-
dimensional gauge theories is discussed in the literature for a long time [1,
2, 3, 4, 5, 6]. Elementary domain walls localize U(1) fields. As was explained
by Polyakov [7], in 2+1 dimensions the U(1) gauge field is dual to a phase
field $\sigma$ living on $S_{1}$. The U(1) gauge theory in 2+1 dimensions can
exist in distinct regimes: (i) Coulomb, with the long-range interaction $\ln
r$ due to the exchange of the gauge field; (ii) the gauge field is Higgsed,
electric charges are screened, interaction due to the exchange of the gauge
field falls off exponentially; (iii) the gauge field acquires a mass through
the Chern–Simons term, gauge symmetry is unbroken; and (iv) the dual photon
field $\sigma$ gets a mass term. This latter regime is quite peculiar. It
might seem that the mass term of the $\sigma$ field implies short-range
interactions. In fact, it is the opposite! Electric charges (seen as the
$\sigma$ field vortices in the dual language) are connected by a flux line
which plays the role of a confining string. Interaction between the electric
charges grows linearly with the distance $r$. In terms of $\sigma$ the string
is a domain line very similar to the axion domain walls in 3+1 dimensions. The
domain line endpoints are the $\sigma$ field vortices. (For some reviews see
Refs. [8, 9, 10].)
The domain lines of the $\sigma$ field are the essence of the Polyakov
confinement [7]. Polyakov’s model is 2+1 dimensional compact electrodynamics.
It represents the low-energy limit of SU(2) Yang–Mills theory with one adjoint
Higgs field which develops a vacuum expectation value (VEV) breaking SU(2)
down to U(1). The mass term for the dual photon is generated by SU(2) three-
dimensional instantons (in 2+1 dimensions, technically, they are identical to
’t Hooft–Polyakov monopoles [11]). When the U(1) gauge field is dynamically
localized on a wall occurring in 3+1 dimensional theory, which of the four
regimes listed above is in fact implemented depends on details of the bulk
theory.
The first example of a U(1) gauge field localized on a wall, in a fully
controllable theoretical setting, was given in [4]. In this example a global
U(1) symmetry of the bulk theory, spontaneously broken on the wall, guarantees
masslessness of the 2+1 dimensional photon.222The dual photon $\sigma$ is the
Goldstone field. The U(1) theory on the wall is in the Coulomb regime. When
the global U(1) symmetry is explicitly weakly broken in the bulk, the $\sigma$
field becomes quasi-Goldstone, a $\sigma$ mass term is generated implying
confinement of the electric charges on the wall [5].
In a recent paper [6] a mechanism (developing a concept put forward in [1])
has been suggested that leads to confinement on domain walls. Unlike the
models discussed in [4, 5], consideration of Ref. [6] was carried out in
nonsupersymmetric setting, although the mechanism per se is general and can be
implemented in a wide class of bulk theories, both supersymmetric or
nonsupersymmetric. The only requirements to these theories are: they should
support both domain walls and Abrikosov–Nielsen–Olesen (ANO) flux tubes [12]
and be minimal (in which sense minimal will be explained later). Far away from
the wall the charged field condensate responsible for the ANO flux tubes is
“large” and squeezes the flux tube from all directions in the perpendicular
plane. Now, if we place such tube inside the wall, where the above condensate
vanishes with an exponential accuracy, in the first approximation the
confining regime gives place to the Coulomb regime on the wall. The flux tube
is still squeezed inside the wall in the direction perpendicular to the wall;
however, it swells in the directions parallel to the wall.
In the next approximation one should take into account the fact that there is
a residual charged field condensate inside the wall. Although it is
exponentially small, it still limits the swelling of the flux tube placed
inside the wall in the directions parallel to the wall. The thickness of the
flux tube in these directions is exponentially large, but finite. If we go to
still larger distances along the wall, (magnetic) charges attached to the
endpoints of such a tube experience linear confinement.
The above description is phrased in terms of the charged field condensate and
magnetic flux tubes. Needless to say, in actuality we keep in mind a dual
picture, presented in Fig. 1: the monopole condensation leading to electric
flux tubes. In what follows the dualization will be tacitly assumed. Thus,
when we speak of matter fields that condense, we will keep in mind that these
local fields present an effective description of monopoles, much in the same
way as in the Seiberg–Witten construction [13].
The suggestion put forward in [6] is inspirational. At the same time,
operational mode of this mechanism remained unclear, as well as its relation
to other regimes implementable in the models with the U(1) gauge field
localized on the walls. Moreover, particular models considered in [6] suffer
from the wall-antiwall instability. With these instabilities, working out
quantitative details does not seem possible.
The purpose of the present paper is to address these issues. We focus on
investigation of how this mechanism actually works, and what deformations or
modifications lead to deconfinement. We suggest two stable model examples: one
at weak coupling and another using the Seiberg–Witten solution [13] at strong
coupling. These models are demonstrated to be working examples of confinement
on the domain wall. En route, we will also clarify some aspect regarding
localization of the gauge fields on the domain walls.
To ensure stability of the model it is necessary to require that two vacua in
which the matter fields condense are two distinct vacua. Let us call them
Confining 1 and Confining 2 (Fig. 1).333In the dual language Confining 1 and
Confining 2 read Higgs 1 and Higgs 2. If the monopole mass in the Coulomb
phase is $m$, the condensate in the center of the wall is roughly $ve^{-md/2}$
where $d$ is the wall thickness. If $md$ is large enough, the condensate
inside the wall almost vanishes, and the gauge theory exists inside the wall
in the (almost) Coulomb regime. Deviations of this almost Coulomb regime from
the perfect Coulomb regime determine the thickness of the flux tube in the
directions parallel to the wall.
Figure 1: The condensate profile for the wall of the type Confining
1–Coulomb–Confining 2.
The exponentially small deviation from the perfect Coulomb regime inside the
wall has a clear-cut interpretation in terms of the field $\sigma$. Instead of
being the modulus field, as in [4], it becomes a quasimodulus. We work out a
method which allows one to calculate its mass directly from the bulk theory.
Moreover, introduction of the second matter field, as in [4], can restore the
exact modulus status of $\sigma$.
The paper is organized as follows. In Sect. 2 we discuss some conceptual
aspects. Section 3 introduces a “minimal” model, the simplest example which
realizes the confinement mechanism discussed by Dvali et al. Special attention
is given here to discussion of the localization mechanism for the gauge field
and formation of the flux tubes in the wall. This example is quasiclassical
(the model is weakly coupled) and is phrased in dual terms. The quark field
condenses while the monopoles are confined by the Abrikosov–Nielsen–Olesen
string. Section 4 provides an example at strong coupling in which condensation
that occurs is that of the monopole field. This example is an extension of
${\cal N}=2$ super-Yang–Mills theory (slightly broken to ${\cal N}=1$)
considered by Seiberg and Witten. The Seiberg–Witten solution is an essential
ingredient which allows us to treat the theory at strong coupling. In Sect. 5
we consider a “nonminimal” theory with two flavors and explain why, in
contradistinction with the minimal model of Sect. 2, the gauge field on the
wall remain massless despite the presence of a residual condensate in the
middle of the wall. The theory on the wall is in the Coulomb regime. The
masslessness is backed up by the Goldstone theorem: in the bulk we have an
exact global U(1) symmetry which is spontaneously Section 7 summarizes our
findings.
## 2 Conceptual aspects
The basic idea of the gauge field localization suggested in [1] assumes that
the bulk four-dimensional theory is in the confining regime while inside the
1+2-dimensional wall we have “less confinement” (or no confinement at all).
Then the chromoelectric flux coming from the bulk through a tube spreads out
inside the wall. The flux tube-wall junction plays the role of the color
source inside the wall. In the dual formulation the bulk theory is Higgsed
while inside the wall it is “less Higgsed” (not Higgsed at all in the case of
U(1)).
The technical implementation of this idea is not quite straightforward.
Indeed, say, in the U(1) case which has just been mentioned the magnetic
charges are confined in the bulk. Thus, the magnetic flux from a distant
magnetic monopole is squeezed into a tube, and when this tube hits the wall,
it spreads out inside the wall. To describe this phenomenon in terms of the
standard 1+2-dimensional U(1) gauge theory on the wall surface we have to use
a duality transformation which converts the magnetic field inside the wall
into a dual electric field of the effective theory on the wall surface. The
flux tube-wall junction acts as an electric charge source in 1+2-dimensional
electrodynamics on the wall. This duality transformation is a crucial element
of the construction. The relation between the gauge potential in the bulk and
that in the effective low-energy theory on the wall surface is nonlocal.
In 1+2 dimensions the gauge field $A_{\mu}$ is dual to a phase field $\sigma$
[7]. If a U(1) symmetry is an exact symmetry of the bulk theory, spontaneously
broken on the wall, occurrence of the Goldstone field $\sigma$ localized on
the wall is inevitable. Dualizing the above Goldstone field we get massless
electrodynamics on the wall. The electric flux from a charge source in the
effective theory on the wall surface is spread according to the Coulomb law.
If, on the other hand, a global U(1) symmetry is only an approximate symmetry
of the bulk theory [5], or even just an approximate symmetry of the domain
wall solution [1], we should expect a pseudo-Goldstone mode localized on the
wall, with a small mass term. With this pseudo-Goldstone mode we get
electrodynamics with confinement on the wall, à la Polyakov. The electric flux
of the world-volume theory is, in its turn, squeezed on the wall, forming a
band of thickness inversely proportional to the pseudo-Goldstone mass. This
thickness is exponentially larger than that of the bulk magnetic flux tubes.
These two scenarios are realized in two-flavor and one-flavor models,
respectively.
## 3 The Simplest Example at Weak Coupling
### 3.1 Theoretical Setting
To introduce the reader to the subject we will start with a toy model that
contains all relevant features of the physical phenomenon we want to describe.
Consider a U$(1)$ gauge theory with a charged scalar field $Q$. Our task is to
study a domain wall interpolating between Higgs–Coulomb–Higgs vacua. The model
is nonsupersymmetric. The simplest choice of the potential seems to be as
follows:
$|Q|^{2}(|Q|^{2}-v^{2})^{2}\,,$ (1)
as suggested in [6]. However, there is a problem with (1), namely the
Higgs–Coulomb–Higgs interpolation is a wall-antiwall configuration in this
model, which is unstable. Of course, the instability can be made exponentially
small, but so are the effects we try to trace.
To create a stable configuration we need at least an extra real neutral field.
Consider a system in the Higgs phase, with the U$(1)$ gauge group, a scalar
field $Q$ with charge $+1$, and an uncharged scalar $a$, with the following
Lagrangian:
$L=-\frac{1}{4e^{2}}(F^{\mu\nu})^{2}+\frac{1}{2e^{2}}(\partial_{\mu}a)^{2}+|\nabla_{\mu}Q|^{2}-V\,,$
(2)
with the potential
$V=\frac{(a-m)^{2}(a+m)^{2}|Q|^{2}}{m^{2}}+\frac{e^{2}}{2}(v^{2}-|Q|^{2})^{2}\,.$
(3)
This model is non-renormalizable, but we can still consider it as an effective
theory in the infrared.444This example can be embedded in supersymmetric QED
(SQED) with the Fayet–Iliopolous $D$-term, the superpotential
$W=Q\tilde{Q}\,\,\frac{(a-m)(a+m)}{m}\,,$ and in a vacuum with $\tilde{Q}=0$.
It is obvious then that we have two distinct vacua $a=\pm m$, $|Q|=v$. If we
keep $m\gg ev$, there is a large intermediate region inside the domain wall
where the VEV of $Q$ is almost zero (see Fig. 2). Even if the theory has no
Coulomb vacuum, if we look at the domain wall profile we immediately see that
this inside region of the wall is almost in the Coulomb phase. In the limit
$m\rightarrow\infty$ the wall becomes infinitely thick, and the Higgs VEV in
the wall center tends to zero.
Figure 2: Domain wall in the toy model (2), (3). The profile of the $a(z)$
field is presented by the solid line while that of $Q(z)$ by dashed. The
horizontal axis presents the direction $z$ perpendicular to the wall.
To see this we write down the equations of motion for our model. Since we are
looking for the domain wall solution we assume that all fields are static and
depend only on a single spatial coordinate $z$ (the wall is perpendicular to
the $z$-axis), and drop the gauge field. We have
$\displaystyle\partial_{z}^{2}a=4a\,\frac{a^{2}-m^{2}}{m^{2}}\,Q^{2}\,,$
$\displaystyle\partial_{z}^{2}Q-Q\,\frac{(a^{2}-m^{2})^{2}}{m^{2}}-e^{2}(|Q|^{2}-v^{2})\,Q\,=0\,.$
(4)
In the leading approximation the quark field $Q$ vanishes inside the wall.
Then the first equation above gives for the neutral scalar $a$
$a\approx 2m\,\frac{z-z_{0}}{d},$ (5)
where $z_{0}$ is the center of the wall, while $d$ is its thickness, to be
determined below. In Eq. (5) we take into account the $a$-field boundary
conditions, $a\to\pm m$ at $z\to\pm\infty$. Substituting (5) in the second
equation in (4) we can write the following equation for the profile field $Q$,
neglecting the non-linear terms:
$\partial_{z}^{2}Q=Q\,\left\\{\frac{16m^{2}}{d^{4}}\left[(z-z_{0})^{2}-\frac{d^{2}}{4}\right]^{2}-e^{2}v^{2}\right\\}\,.$
(6)
Below we will show that $d\sim m/e^{2}v^{2}$. Taking this into account, inside
the wall, near its left edge,
$1/ev\ll z-z_{0}+\frac{d}{2}\ll d\,,$ (7)
we can simplify the above equation by dropping the second term in the curly
brackets and replacing $z-z_{0}-d/2$ by $-d$. Then we get
$\partial_{z}^{2}Q=Q\,\frac{16m^{2}}{d^{2}}\left(z-z_{0}+\frac{d}{2}\right)^{2}.$
(8)
The solution of this equation obviously has the form
$Q\approx v\,e^{-\frac{2m}{d}(z-z_{0}+\frac{d}{2})^{2}},$ (9)
where we use Eq. (7).
Much in the same way, near the right edge of the wall at
$1/ev\ll-\left(z-z_{0}-\frac{d}{2}\right)\ll d$ (10)
we derive from (4)
$Q\approx v\,e^{-\frac{2m}{d}(z-z_{0}-\frac{d}{2})^{2}}\,.$ (11)
Note that these quark profiles are similar to those obtained in [4] in
supersymmetric QED with two quark flavors, see Sect. 4.
Let us estimate $d$ in the limit $m\gg ev$. In the inside region the VEV of
$Q$ almost vanishes while $a$ is linear in $z$. In order to estimate the
thickness of this region (i.e. the wall thickness), let us first estimate the
wall tension as a function of $d$ and then minimize it with respect to $d$,
$T_{\rm wall}\sim\frac{e^{2}v^{4}}{2}d+\frac{(2m)^{2}}{de^{2}}\,.$ (12)
The assumptions for this estimation are that for $m\gg ev$ the dominating
contributions to the energy come from the potential and kinetic terms of the
$a$ field. The minimum is achieved at
$d=\frac{2\sqrt{2}m}{e^{2}v^{2}}\,,$ (13)
where $ev$ is the photon mass in the bulk. The tension of the wall is of the
order of
$T_{\rm wall}\sim 2\sqrt{2}\,m\,v^{2}.$ (14)
In what follows, we shall be interested in trapping gauge fields, inside this
domain wall. Localization of gauge fields on lower-dimensional topological
objects is, generally speaking, a nontrivial task. Massless scalars can be
localized as Goldstone bosons of continuous symmetries spontaneously broken on
the given topological defects. Massless fermions can be localized via
Jackiw–Rebbi’s and other index theorems [14] (see [15, 16]). We shall discuss
this issue in more detail in Section 6.
As discussed in Ref. [4], we can consider the following gauge invariant order
parameter:
$e^{i\sigma}=v^{-2}\bar{Q}(-\infty)e^{i\int A_{z}dz}Q(+\infty)\,.$ (15)
The difference with the $2$-flavor model considered in [4], and discussed in
Section 5, is that the two scalar fields at the edges of the Wilson line are
now the same. This implies that we do not have a strictly massless gauge field
localized on the wall. This is because the expectation value of the matter
field never exactly vanishes inside the wall.
Let us denote the expectation value of the condensate $|Q|$ in the wall center
$(z=z_{0})$ by $v_{0}$,
$v_{0}=\left|Q(z=z_{0})\right|\,.$ (16)
In the limit in which $m\gg ev$, the domain wall is thick and $v_{0}$ is very
small. A numerical fit in the range $0.3\,v<m<0.6\,v$ and $0.15<e<0.2$ shows
that to a very good approximation 555This formula, strictly speaking, is valid
in the range of parameters indicated above. However, it is likely that it
works also for larger $d$, although it is difficult to check this assumption
because accurate numerical calculations are more difficult for larger values
of $d$.
$v_{0}\approx
v\exp\left(-0.88\,\frac{m^{2}}{e^{2}\,v^{2}}\right)=v\exp\left(-0.31\,d\,m\right)\,.$
(17)
The mass of the gauge field in the bulk is $ev$; Higgsing inside the wall is
exponentially weaker, so that the gauge field mass is $\sim e\,v_{0}$. To what
extent can we speak of localization? The answer depends on the range of the
parameters.
Since the domain wall is an object with thickness $d$ a low-energy effective
action makes sense only up to energy scales $\sim 1/d$. At higher energies
excitations of the wall internal structure become important. Fluctuations of
the wall as a whole in the transverse direction (Goldstone modes of the
translational symmetry) are massless. They always belong to the low-energy
effective action. Other — massive — excitations can be considered a part of
the $2+1$ dimensional effective action as long as their mass is much smaller
than $1/d$. Physics changes in passing from one of the following regimes to
another:
$(i)\qquad 1/d\ll e\,v_{0}\ll e\,v\,.$
In this regime the mass of the gauge field inside the wall is much larger than
$1/d$. We can not speak about localization, physics of the wall is essentially
four-dimensional.
$(ii)\qquad e\,v_{0}\ll 1/d\ll e\,v\,.$
This is the localization limit. Up to energies $\sim 1/d$ the gauge field can
be considered as a field localized on the $2+1$ dimensional world volume.
$(iii)\qquad e\,v_{0}\ll e\,v\ll 1/d\,.$
In this case the gauge field is localized only up to energies $e\,v$. Due to
leakage in the bulk no localization occurs at energies from $\sim ev$ to
$1/d$.
Focusing on the regimes (ii) and (iii) we ask ourselves whether or not a
quasimoduli field lives on the wall in these cases.
### 3.2 Modulus or quasimodulus on the wall
First, we need to explain why we expect a U$(1)$ quasimodulus on the wall
world volume. We begin by presenting the simplest solutions that describe
localization of the gauge field: a constant magnetic field and a constant
electric field. From now on we will always work in the gauge $A_{z}=0$.
A constant magnetic field inside the wall is parallel to the wall surface. Let
us assume the magnetic field to be aligned along the $x$ axis,
$\vec{B}=B\hat{x}$ where $\hat{x}$ is the unit vector along $x$. We can
construct it in the following way. At negative $z$ we take the field
$Q=ve^{iky}$ and the gauge field $A_{2}=k$ (or, which is the same, $A_{y}=-k$.
At positive $z$ we take $Q=v$ and $A_{y}=0$. In this way in two vacua, to the
left and to the right of the wall, the field configuration is pure gauge.
Inside the wall, $A_{y}$ linearly interpolates between $-k$ and $0$ on the
interval of size $d$. The magnetic field is $B_{x}=-\partial_{z}A_{y}=k/d$.
The magnetic flux per unit length in the $y$ direction is $\int dzB_{x}=k$.
The magnetic field inside the wall is a vector on the wall since it can be
oriented either along $\hat{x}$ or along $\hat{y}$.
The electric field inside the wall can only be perpendicular to the wall
surface, aligned along $\hat{z}$. Thus, on the wall it must be interpreted as
a pseudoscalar. To obtain such an electric field inside the wall consider
$Q=ve^{i\omega t}$ at negative $z$ and $A_{t}=\omega$. At positive $z$ we have
$Q=v$ and $A_{t}=0$. Inside the wall $A_{t}$ linearly interpolates between
$\omega$ and $0$. The electric field inside the wall is
$E_{z}=-\partial_{z}A_{t}=\omega/d$.
Of course, from the $2+1$ dimensional point of view the picture must be
dualized, since in 2+1 dimensions it is the electric field $F_{0i}$ which is a
vector while the magnetic field $F_{12}$ is a scalar. For example,
$E^{(2+1)}_{x}=B^{(2+1)}_{x}=-\partial_{z}A_{x}$ and
$B^{(2+1)}=B^{(2+1)}_{z}=-\partial_{z}A_{t}$. In other words,
$F^{(2+1)}_{\mu\nu}=\frac{1}{2}\epsilon^{\mu\nu\rho z}F_{\rho
z}\,,\qquad\mu,\nu,\rho=0,1,2\,.$ (18)
Now we can further dualize $F^{(2+1)}_{\mu\nu}$ à la Polyakov, expressing
$F^{(2+1)}_{\mu\nu}$ in terms of a phase fields $\sigma$,
$\frac{1}{2}\epsilon_{\mu\nu\rho}F^{(2+1)}_{\nu\rho}=\frac{1}{d}\partial_{\mu}\sigma$
(19)
Assembling everything together we have, with our gauge choice
$\displaystyle B^{(3+1)}_{x}=-\partial_{z}A_{y}\qquad$ $\displaystyle=$
$\displaystyle\qquad E^{(2+1)}_{x}=\frac{1}{d}\,\partial_{y}\sigma\,,$ (20)
$\displaystyle B^{(3+1)}_{y}=\partial_{z}A_{x}\qquad$ $\displaystyle=$
$\displaystyle\qquad E^{(2+1)}_{y}=-\frac{1}{d}\,\partial_{x}\sigma\,,$ (21)
$\displaystyle E^{(3+1)}_{z}=-\partial_{z}A_{t}\qquad$ $\displaystyle=$
$\displaystyle\qquad B^{(2+1)}=\frac{1}{d}\,\partial_{t}\sigma\,.$ (22)
Note that the angle field $\sigma$ exactly corresponds to the phase of $Q$ at
negative $z$ relative to that at positive $z$.
We can use unitary gauge which ensures that $Q=v$ in both vacua at
$z\to\pm\infty$. In this gauge the constant magnetic field inside the wall
looks as follows. The gauge field $A_{\mu}=0$ in both vacua, while inside the
wall
$A_{z}=\frac{1}{d}\,\sigma(x,y),\qquad A_{x}=A_{y}=0.$ (23)
This gauge potential gives the magnetic field strength shown in equations
above.
As discussed before, one can consider the gauge invariant order parameter
(15). The kinetic term for the gauge field, expressed as a function of the
$\sigma$ modulus, takes the form (cf. [4])
${\cal
L}_{2+1}=\frac{1}{2\,e^{2}_{3+1}\,d}\,\,\partial_{\mu}\sigma\partial_{\mu}\sigma\,.$
(24)
### 3.3 Potential for the quasimodulus
As was mentioned, in the model at hand, the phase field $\sigma$ is not an
exact modulus. A potential $V(\sigma)$ is generated forcing $\sigma=0$ in the
true vacuum which is unique. In the localization limit of large $d$ we can
nevertheless speak of a $2+1$ dimensional effective theory for $\sigma$, since
the $\sigma$ field mass is much smaller than the excitation energy of the
domain wall $\sim 1/d$. In the leading approximation the $\sigma$ field
Lagrangian will be of the sine-Gordon type,
${\cal
L}_{2+1}=\frac{1}{2e^{2}\,d}\,\,\partial_{\mu}\sigma\partial_{\mu}\sigma-\Delta\cdot\left(\sin{\frac{\sigma}{2}}\right)^{2}$
(25)
where $\Delta$ is the difference between two tensions: the tension of the
$\sigma=0$ wall and and that of the $\sigma=\pi$ wall. The mass of $\sigma$ is
$m_{\sigma}\approx\sqrt{\frac{\Delta\,e^{2}\,d}{2}}\,.$ (26)
Explicit calculations of the quasimodulus potential $V(\sigma)$ are not easy.
The minimal tension domain wall corresponds to the lowest energy $\sigma=0$.
Walls with $\sigma\neq 0$ are not static solutions of the equation of motion
since they decay. To study such solutions of the equation of motion we have to
deal with more than one parameter.
To explain what the last sentence means we consider the solution with a
constant magnetic field in the $\hat{y}$ direction (see Fig. 3).
Figure 3: An (almost) constant magnetic field inside the domain wall. From the
point of view of the quasi modulus $\sigma$ it correspond to an “almost”
linear solution $\sigma=kx+\dots$. The dots stand for small modulations. When
$\sigma\approx 0$, the wall is thinner and the magnetic field larger. At
$\sigma\approx\pi$ the magnetic field is smaller and the wall thicker. The
amplitude of the oscillations in $|\vec{B}|$ and thickness depends on the mass
of the quasimodulus. In the limit of vanishing mass we recover the constant
magnetic field solution.
As a first approximation, we can choose $Q=ve^{ikx}$ and $A_{x}=k$ at $z<0$,
and $Q=v$ and $A_{x}=0$ at $z>0$, similar to the discussion above. This gives
a constant magnetic field $B_{y}=k/d$ inside the wall. But it is clear that,
due to a nonvanishing (albeit small) $Q$ condensate inside the wall, this is
not the exact solution. The latter requires modulations of $k$. One can
understand this circumstance both from the bulk and from the brane point of
view. The bulk explanation (see Fig. 3) is as follows. Due to topological
reasons, the field $Q$ must exactly vanish at some $x$ and $z\approx 0$ each
time the relative phase $\sigma$ rotates by $2\pi$. The lines (in the $y$
direction) on which $Q$ vanishes are the lines where the magnetic field
reaches its maximum. We also expect the thickness of the wall to be a little
bit larger around these lines.
From the point of view of the $2+1$ dimensional effective action (25) it is
also clear that $\sigma(y)=kx$ with $k$ constant is not a solution once the
sine term is switched on. The derivative $d\sigma/dx$ will be larger at the
top of the potential ($\sigma=\pi$) and smaller at the bottom ($\sigma=0$).
Figure 4: Left: the stable wall ($\sigma=0$). Right: the unstable wall
($\sigma=\pi$). In this specific example, even though the Higgs VEV is not so
small in the middle of the stable wall, there is just a $2\%$ t difference
between the tensions of the unstable and stable walls.
Figure 5: Left: the energy density inside the stable wall (solid line) and the
unstable wall (dashed line). In the center of the wall, the unstable wall has
a lower energy density. On the other hand, the unstable wall is thicker and
this make its total tension larger than that of the stable wall. Right: both
domain walls in the $(a,Q)$ plane.
There is a trick one can use to detect the quasimodulus. At $\sigma=\pi$ the
wall solution corresponds to the maximum of the potential $V(\sigma)$. It is
stationary but unstable. In this solution the $Q$ field vanishes exactly in
the middle of the wall, at $z=0$. We can easily get the solution if we impose
two conditions: (i) the field $Q$ is real; (ii) $Q$ changes sign in passing
from one side of the wall to the other, see Figs. 4 and 5. The difference
between the tensions of the unstable and stable walls will determine $\Delta$
in Eq. (25).
The mass term for the $\sigma$ field, is what induces confinement inside the
domain wall. In Figure 3 we discussed how a constant magnetic field solution
is deformed by the mass term. There are modulations of the thickness of the
wall, and magnitude of the magnetic field. These modulations are nothing but
the flux tubes, all lined together. A single winding in the phase $\sigma$, a
kink ot the sine-Gordon model (25), is a domain line from the wall point of
view. The mass term is what stabilizes the thickness of this kink, that
otherwise would spread to $\infty$. These, from the point of view of the bulk
theory, are the confining vortices bounded inside the domain wall.
### 3.4 A direct estimate of the $\sigma$ mass term
The Lagrangian of the model under consideration has no exact or approximate
global U(1) symmetry. However, the domain wall solution does have such an
approximate symmetry. Indeed, let us start from the limit $m/(ev)\to\infty$.
In this limit the wall edges become infinitely sharp while the absolute value
of the field $Q$ inside the wall vanishes. Then the field configuration under
consideration has two U(1) symmetries: $Q(z>z_{0}+d/2)\to
e^{i\alpha}Q(z>z_{0}+d/2)$ and $Q(z<z_{0}-d/2)\to e^{i\beta}Q(z<z_{0}-d/2)$,
with independent phases $\alpha$ and $\beta$. The common U(1) symmetry, with
$\alpha=\beta$, is gauged. The U(1) phase rotation of $Q(z>z_{0}+d/2)$
relative to $Q(z<z_{0}-d/2)$ remains as an exact global U(1) symmetry.
If we keep $m/(ev)$ finite, then $Q(z_{0})$ does not vanish and, because of
continuity of the field $Q$, the above symmetry becomes approximate. This is
due to the fact that now one must continuously evolve the phase of $Q$ from
zero at $z=z_{0}-d/2$ up to a given $\sigma$ at $z=z_{0}+d/2$. With the $z$
dependent phase inside the wall, the wall configuration no longer presents the
exact solution. The question is how to deal with such a situation.
This is quite similar to the treatment of quasizero modes in the instanton-
anti-instanton background field. To deal with them one routinely uses the
valley method [17]: in the functional space one finds a direction
corresponding to the bottom of the valley and fixes the coordinate along this
direction “by hand.” In the problem at hand the coordinate to be fixed can be
defined as the phase $\sigma$ in the expression
$Q(z=z_{0}+d/2)=e^{i\sigma}Q(z=z_{0}-d/2)$ (in the $A_{z}=0$ gauge). For
simplicity we will assume $\sigma(z=z_{0}+d/2)$ to be small. Then, given the
above boundary condition, minimization of the energy functional gives that
inside the wall
$\sigma(z)\approx\frac{z-z_{0}+(\delta/2)}{\delta}\,\sigma\,,$ (27)
where $\delta$ is the size of the region inside the wall where phase $\sigma$
is changing, $\delta<d$. It is a free parameter and we estimate it below by
minimization procedure.
As for the ansatz for the scalar field inside the wall we choose the sum of
the profile functions (9) and (11) at the left and right edges of the wall. We
have
$Q\approx
v\,\left[e^{-\frac{2m}{d}(z-z_{0}+\frac{d}{2})^{2}}+e^{-\frac{2m}{d}(z-z_{0}-\frac{d}{2})^{2}}\right]\,\exp{\left(i\,\frac{z-z_{0}+(\delta/2)}{\delta}\,\sigma\right)}.$
(28)
It is quite obvious that the only extra term in the energy functional (2), (3)
comes from $|\partial_{z}Q|^{2}$ with the derivative acting on the phase, i.e.
$(Q\partial_{z}\sigma)^{2}$. Thus, the potential energy associated with
$\sigma\neq 0$ (at small $\sigma$) is
$V(\sigma)=\int
dz\,\left(Q\frac{\sigma}{\delta}\right)^{2}\sim\sigma^{2}\,\frac{v^{2}}{\delta^{2}}\,\int_{-\delta/2}^{\infty}dz\,e^{-\frac{2m}{d}(z+\frac{d}{2})^{2}}\sim\sigma^{2}\,\frac{v^{2}}{\delta^{2}ev}\,e^{-\frac{m}{2d}(d-\delta)^{2}}\,.$
(29)
Minimizing this with respect to $\delta$ we get
$e^{2}v^{2}(d-\delta)\sim\frac{1}{\delta}.$ (30)
Assuming that $\delta\ll d$ we have
$\delta\sim\frac{1}{e^{2}v^{2}\,d}\sim\frac{1}{m},$ (31)
which confirms our assumption, since $\delta/d\sim\frac{e^{2}v^{2}}{m^{2}}\ll
1$.
Substituting $\delta$ from (31) to the potential (29) we get
$V(\sigma)\sim\sigma^{2}\,v^{2}\,\frac{m^{2}}{ev}\,e^{-\frac{md}{2}},$ (32)
which shows that the potential for $\sigma$ is exponentially small and is
determined by the value of the scalar field in the middle of the wall. Using
more accurate numerical estimate (16) for the value $v_{0}$ we finally get
$V(\sigma)\sim\frac{v_{0}^{2}\,m^{2}}{ev}\,\sigma^{2}\,.$ (33)
This estimate implies, in turn, that
$\tilde{m}_{\sigma}\sim ev_{0}\,\left(\frac{m}{ev}\right)^{3/2}\,.$ (34)
### 3.5 A numerical check
We have now two different estimations of $m_{\sigma}$. One of them (Eq. (26))
is given in term of $\Delta$, which is the difference in tension between the
$\sigma=\pi$ and the $\sigma=0$ wall. The other one (Eq. (34)) is given in
terms of the residual condensate $v_{0}$ at the center of the domain wall. It
is interesting to compare them for a cross check.
| m
---|---
| $0.30$ | $0.35$ | $0.40$ | $0.45$ | $0.50$
e | $0.2$ | $2.5\times 10^{-2}$ | $6.0\times 10^{-3}$ | $1.1\times 10^{-3}$ | $1.6\times 10^{-4}$ | $1.8\times 10^{-5}$
$0.19$ | $1.6\times 10^{-2}$ | $3.2\times 10^{-3}$ | $4.8\times 10^{-4}$ | $5.7\times 10^{-5}$ | $5.1\times 10^{-6}$
$0.18$ | $9.0\times 10^{-3}$ | $1.5\times 10^{-3}$ | $1.8\times 10^{-4}$ | $1.7\times 10^{-5}$ | $1.1\times 10^{-6}$
$0.17$ | $4.7\times 10^{-3}$ | $6.2\times 10^{-4}$ | $5.8\times 10^{-5}$ | $3.9\times 10^{-6}$ | $1.9\times 10^{-7}$
$0.16$ | $2.1\times 10^{-3}$ | $2.1\times 10^{-4}$ | $1.5\times 10^{-5}$ | $7.0\times 10^{-7}$ | $2.2\times 10^{-8}$
Table 1: Values of $\Delta$ for $v=1$ and different values of $(m,e)$. | m
---|---
| $0.30$ | $0.35$ | $0.40$ | $0.45$ | $0.50$
e | $0.2$ | $2.6\times 10^{-1}$ | $1.1\times 10^{-1}$ | $4.1\times 10^{-2}$ | $1.4\times 10^{-2}$ | $4.5\times 10^{-3}$
$0.19$ | $2.0\times 10^{-1}$ | $7.7\times 10^{-2}$ | $2.7\times 10^{-2}$ | $8.4\times 10^{-3}$ | $2.3\times 10^{-3}$
$0.18$ | $1.5\times 10^{-1}$ | $5.2\times 10^{-2}$ | $1.6\times 10^{-2}$ | $4.5\times 10^{-3}$ | $1.1\times 10^{-3}$
$0.17$ | $1.0\times 10^{-1}$ | $3.2\times 10^{-2}$ | $9.0\times 10^{-3}$ | $2.2\times 10^{-3}$ | $4.5\times 10^{-4}$
$0.16$ | $6.8\times 10^{-2}$ | $1.9\times 10^{-2}$ | $4.5\times 10^{-3}$ | $9.1\times 10^{-4}$ | $1.5\times 10^{-4}$
Table 2: Values of $v_{0}$ for $v=1$ and different values of $(m,e)$.
In Table 1 numerical results for $\Delta$ are shown for some values of the
parameters. In Table 2 the corresponding values of $v_{0}$ are presented. We
can then use these values for a comparison the two independent estimations
$m_{\sigma}$, $\tilde{m}_{\sigma}$. The results are shown in Table 3. The
proximity of the ratio $m_{\sigma}/\tilde{m}_{\sigma}$ to unity is obvious.
The agreement is indeed quite good. We can also verify that the localization
condition $e\,v_{0}\ll 1/d$ is satisfied very well for $m\geq 0.4\,v$ and
$e<0.2$.
| m
---|---
| $0.30$ | $0.35$ | $0.40$ | $0.45$ | $0.50$
e | $0.2$ | $1.08$ | $1.08$ | $1.08$ | $1.05$ | $1.01$
$0.19$ | $1.07$ | $1.08$ | $1.07$ | $1.03$ | $0.99$
$0.18$ | $1.08$ | $1.08$ | $1.05$ | $1.01$ | $0.97$
$0.17$ | $1.08$ | $1.07$ | $1.03$ | $0.99$ | $0.95$
$0.16$ | $1.08$ | $1.06$ | $1.01$ | $0.97$ | $0.92$
Table 3: Values of $m_{\sigma}/\tilde{m}_{\sigma}$ for $v=1$ and different
values of $(m,e)$. $m_{\sigma}$ is given in Eq. (26) and $\tilde{m}_{\sigma}$
given by Eq. (34).
## 4 A Modified Seiberg–Witten framework
### 4.1 Theoretical Setting
We now turn to a strong coupling example of the confinement phenomenon inside
domain walls. The model we will dwell on below is a rather straightforward
modification of the Seiberg–Witten (SW) model [13], which supports both domain
walls and strings.
The theory of interest is $\mathcal{N}=2$ gauge theory, with the gauge group
${\rm U}(2)={\rm SU}(2)\times{\rm U}(1)/Z_{2}$, with no matter
hypermultiplets. The following superpotential which breaks the extended
supersymmetry down to $\mathcal{N}=1$ is then added:
$W=\alpha\,{\rm Tr}\,\left(\frac{\Phi^{3}}{3}-\xi\Phi\right).$ (35)
Classically, we have three vacua, with $\phi$ equal to
$\left(\begin{array}[]{cc}\phantom{.}\sqrt{\xi}&0\\\\[5.69054pt]
0&\phantom{-}\sqrt{\xi}\end{array}\right)\
,\qquad\left(\begin{array}[]{cc}\phantom{.}\sqrt{\xi}&0\\\\[5.69054pt]
0&-\sqrt{\xi}\end{array}\right)\
,\qquad\left(\begin{array}[]{cc}-\sqrt{\xi}&0\\\\[5.69054pt]
0&-\sqrt{\xi}\end{array}\right)\ .$ (36)
The first and the last vacua preserve the non-Abelian SU(2) gauge symmetry.
Strong coupling effects à la Seiberg and Witten will then split each of them
into two vacua (the monopole and dyon vacua). The vacuum in the middle
preserves only the U$(1)\times{\rm U}(1)$ gauge symmetry, and is not split.
We, thus, expect in total five vacua, for generic values of $\xi$.
If we set $\alpha$ at zero, the U$(1)$ and SU$(2)$ sectors get completely
decoupled. Since there are no matter hypermultiplets, only a nonvanishing
superpotential can make the two sectors communicate with each other. Dynamics
of the U$(1)$ sector is trivial, while the SU$(2)$ sector is described by the
Seiberg–Witten solution [13]. We parametrize the moduli space by
$\Phi=\left(\begin{array}[]{cc}a_{0}+a_{3}&0\\\ 0&a_{0}-a_{3}\\\
\end{array}\right).$ (37)
The conventional SW solution is written in terms of the invariant
$u=2a_{3}^{2}\,.$ (38)
In our case
${\rm Tr}\,\Phi=2a_{0},\qquad{\rm
Tr}\,\Phi^{2}=2a_{0}^{2}+2a_{3}^{2},\qquad{\rm
Tr}\,\Phi^{3}=2a_{0}^{3}+6a_{3}^{2}a_{0}\,,$ (39)
and there is an invariant way to parameterize the moduli space [18, 19] by
$u_{1}={\rm Tr}\,\Phi\,,\qquad u_{2}=\frac{1}{2}{\rm Tr}\,\Phi^{2}\,.$ (40)
The trace of of $\Phi^{3}$ can be expressed in terms of $u_{1,2}$,
${\rm Tr}\,\Phi^{3}=\frac{3}{2}({\rm Tr}\,\Phi)({\rm
Tr}\,\Phi^{2})-\frac{1}{2}({\rm Tr}\,\Phi)^{3}\,.$ (41)
This relation is not modified by quantum corrections.
After the superpotential (35) is switched on, the moduli space is lifted. Five
discrete vacua described above survive. This is a special case of the set-up
considered in Refs. [18, 19]. The positions of the vacua are the following
(see Appendix Appendix A. Vacuum structure for more details). The value of
$u_{2}$ is $\xi$ for all five vacua. It is not modified by quantum
corrections. The Coulomb vacuum in the middle is not modified by quantum
correction either. It lies at
$u_{1}=0\,,\quad u_{2}=\xi\,.$ (42)
The monopole-$1$ and dyon-$1$ vacua $\phi={\rm diag}\,(\xi,\xi)$ are at
$u_{1}=2\sqrt{\xi-\Lambda^{2}},\,\,\,u_{2}=\xi\quad{\rm and}\quad
u_{1}=2\sqrt{\xi+\Lambda^{2}},\,\,\,u_{2}=\xi\,,$ (43)
respectively. The dyon-$2$ and monopole-$2$ vacua from $\phi={\rm
diag}\,(-\xi,-\xi)$ are at
$u_{1}=-2\sqrt{\xi-\Lambda^{2}},\,\,\,u_{2}=\xi\quad{\rm and}\quad
u_{1}=-2\sqrt{\xi+\Lambda^{2}},\,\,\,u_{2}=\xi\,.$ (44)
Figure 6: Five vacua of the model in the limit $\sqrt{\xi}\gg\Lambda$. (In the
plot we set $\Lambda=1$, $\xi=4$). The dashed line connects the monopole-$1$,
Coulomb, and monopole-$2$ vacua. This is the composite domain wall we will
analyze in what follows. Note that all five vacua are aligned.
Figure 7: As $\xi$ becomes smaller, we reach a critical point where the
Coulomb vacua lies exactly on the monopole singularity. Around this point we
can use the magnetic effective action to describe the wall. It remains weakly
coupled in all three vacua of interest ($\Lambda=1$, $\xi=1.45$ in this plot).
In the limit $\sqrt{\xi}\gg\Lambda$ the Coulomb vacuum is such that the
electric coupling is small,
$e_{3}^{2}=\frac{1}{\ln\,(\sqrt{\xi}/\Lambda)}\,.$ (45)
The five vacua in this limit are depicted in Fig. 6.
As $\xi$ decreases and becomes of order $\Lambda$, the Coulomb vacuum enters a
strong coupling regime. At the critical value $\xi=\Lambda^{2}$ the Coulomb
vacuum lies exactly in the monopole singularity and coalesces with two
monopole vacua. Around this critical value, the Coulomb vacuum is such that
the magnetic coupling is small, so we can use the same set of low-energy
effective variables to describe both the Coulomb and the confining vacua. In
this section we will study (in the limit $|\xi-\Lambda^{2}|\ll\Lambda^{2}$)
the domain walls connecting the Coulomb and the confining vacua. It is
convenient to introduce a dimensionless parameter
$\epsilon^{2}=\frac{\xi-\Lambda^{2}}{\Lambda^{2}}\,.$ (46)
We will work in the limit $|\epsilon|\ll 1$, but keeping three vacua separate
(i.e. $\epsilon\neq 0$).
The effect of the superpotential $W$ in the infrared theory, near the monopole
vacuum, is described by the following effective superpotential:
$\tilde{W}=\frac{A_{3D}M\tilde{M}}{\sqrt{2}}+\frac{\alpha(2A_{0}^{3}+6u(A_{3D})A_{0})}{3}-2\alpha\xi
A_{0}\,,$ (47)
where we can use
$u(A_{3D})\approx\Lambda^{2}-2i\Lambda A_{3D}-\frac{1}{4}A_{3D}^{2}+\ldots\,.$
(48)
The $F$-term part of the scalar potential is
$V_{F}=e_{3D}^{2}|\frac{M\tilde{M}}{\sqrt{2}}+2\alpha
u^{\prime}A_{0}|^{2}+4e_{0}^{2}\alpha^{2}|A_{0}^{2}+u-\xi|^{2}+\frac{|MA_{3D}|^{2}+|\tilde{M}A_{3D}|^{2}}{2}\,,$
(49)
while the $D$-term part is
$V_{D}=\frac{e_{3D}^{2}}{2}(MM^{\dagger}-\tilde{M}\tilde{M}^{\dagger})^{2}\,,$
(50)
where $M$, $\tilde{M}$ are the monopole superfields. The vacuum expectation
values in the confining vacua are (in the limit $\epsilon\ll 1$)
$A_{3D}=0,\qquad A_{0}=\pm\Lambda\epsilon,\qquad M\tilde{M}=\pm
4\sqrt{2}\,i\,\Lambda^{2}\alpha\epsilon\,.$ (51)
The dual description is valid in these vacua provided that the monopole
condensate is $\ll\Lambda^{2}$ implying $|\alpha\epsilon|\ll 1$.
The Coulomb vacuum is defined by the following constraints:
$M=\tilde{M}=0\,,\qquad A_{0}=0\,,\qquad u(A_{3D})=\xi\,.$ (52)
The VEV of $A_{3D}$ in the limit $\epsilon\ll 1$ is
$A_{3D}=\frac{\epsilon^{2}\Lambda i}{2}\,.$ (53)
A nice feature of the limit $\epsilon\ll 1$ is that we can use the same weakly
coupled effective description in all three vacua of interest (Coulomb,
monopole-1 and monopole-2). This unified description is also valid for the
domain wall interpolating between them. In this sense the situation
drastically differs from the case considered in Ref. [20], where no unified
description was possible for the domain wall interpolating between the
monopole and the dyon vacua considered in [20].
The kinetic terms are
$\displaystyle\mathcal{L}_{\rm kin}$ $\displaystyle=$
$\displaystyle\frac{1}{4e_{3D}^{2}}\left(F_{3D}^{\mu\nu}\right)^{2}+\frac{1}{4e_{0}^{2}}\left(F_{0}^{\mu\nu}\right)^{2}$
(54) $\displaystyle+$
$\displaystyle\frac{1}{2e_{3D}^{2}}(\partial_{\mu}{a_{3D}})^{2}+\frac{1}{2e_{0}^{2}}(\partial_{\mu}{a_{0}})^{2}+|\nabla_{\mu}M|^{2}+|\nabla_{\mu}\tilde{M}|^{2}\,.$
In the confining vacua the value of the coupling $e_{3D}$ is determined by the
monopole condensate,
$e_{3D}^{2}=\frac{1}{\ln\,(\Lambda/\sqrt{|M\tilde{M}|})}\approx\frac{1}{\ln\,(1/\sqrt{|\alpha\epsilon|})}\,.$
(55)
In the Coulomb vacuum the expression for $e_{3D}^{2}$ can be found from the
Seiberg–Witten expression for $\tau=\theta/(2\pi)+4\pi i/g^{2}$,
$e_{3D}^{2}\approx\frac{1}{\ln\,(1/|\epsilon|)}\,.$ (56)
An effective coupling $e_{3D}(z)$ along the domain wall profile is a function
of the field condensates inside the wall. It interpolates between the two
values above, (55) and (56). If we choose $\alpha\sim\epsilon$, then
$e_{3D}^{2}$ is approximately constant.
Replacing the leading terms in $u(A_{3D})$ in the scalar potential, we get
$\displaystyle V$ $\displaystyle=$ $\displaystyle
e_{3D}^{2}\left|\frac{M\tilde{M}}{\sqrt{2}}-4i\alpha\Lambda
A_{0}\right|^{2}+4e_{0}^{2}\alpha^{2}\left|A_{0}^{2}-2i\Lambda
A_{3D}+\Lambda^{2}-\xi\right|^{2}$ (57) $\displaystyle+$
$\displaystyle\frac{|MA_{3D}|^{2}+|\tilde{M}A_{3D}|^{2}}{2}+\frac{e_{3D}^{2}}{2}(MM^{\dagger}-\tilde{M}\tilde{M}^{\dagger})^{2}\,.$
### 4.2 Elementary and composite walls
We look for the wall solution in the ansatz:
$|\tilde{M}|=|M|\,,$ (58)
which automatically guarantees vanishing of the $D$ term. We will assume the
phases of $\tilde{M}$ and $M$ to be constant constant in our wall solution.
This assumption is to be checked a posteriori. Once the phases are constant,
they can be chosen at will by virtue of a global gauge rotation. It is
convenient to choose
$\tilde{M}=iM\,.$ (59)
The value of the superpotential in the two confining vacua, monopole-1 and
monopole-2, is
$W=\pm\,\frac{4}{3}\,\Lambda^{3}\,\alpha\epsilon^{3}\,.$ (60)
In the Coulomb vacuum the superpotential vanishes. Hence, the tension of the
BPS wall interpolating between monopole-1 and monopole-2 vacua, if it existed,
would be twice the tension of the BPS wall interpolating between the Coulomb
and confining vacua. The latter walls will be referred to as elementary. The
former wall can be called composite.
For the elementary wall, we can take $M$ and $A_{0}$ real and $A_{3D}$ pure
imaginary. Then we can write the following BPS equations: 666 Note that
$e_{3D}$ is a function of the fields $M\tilde{M},\,\,a_{3D}$; this does not
change the form of the BPS equations. Our choice $\alpha\approx\epsilon$
guarantees constancy of $e_{3D}$ to a very good approximation. For simplicity,
in the numerical calculations we keep $e_{3D}$ constant.
$\displaystyle\frac{\partial M}{\partial z}-i\frac{M\,A_{3D}}{\sqrt{2}}=0\,,$
$\displaystyle\frac{\partial A_{0}}{\partial z}-2\alpha
e_{0}^{2}(A_{0}^{2}-2i\Lambda A_{3D}-\xi+\Lambda^{2})=0\,,$
$\displaystyle\frac{\partial A_{3D}}{\partial
z}+ie_{3D}^{2}\left(\frac{M^{2}}{\sqrt{2}}-4\alpha\Lambda A_{0}\right)=0\,.$
(61)
The profile functions for the elementary wall interpolating between the
confining and the Coulomb vacua, are shown if Fig. 8.
Figure 8: Elementary wall at $\mu=0$ (see Eq. (62)), $A_{0}$ (solid), $A_{3}$
(long dashes) and $M$ (short dashes).
In this theoretical set-up, no BPS wall interpolating between two confining
vacua, monopole-1 and monopole-2, exists. In other words, a composite wall
built of two elementary walls at a finite distance from each other, does not
exist. Supersymmetric solutions correspond to viscous flows in the first-order
equations, starting from monopole-1, following the profile $\tilde{W}$ and
ending in monopole-2. It is not difficult to see that such flow cannot be
realized in the case at hand. A field configuration interpolating between
monopole-1 and monopole-2 is always time-dependent; it represents two
elementary walls moving under the influence of a repulsive force between them
(see Ref. [21] for a discussion in supersymmetric sigma models). This force
falls off exponentially with the wall separation.
### 4.3 Composite wall stabilization
In order to avoid the problem discussed in Sect. 4.2 and stabilize the
composite domain wall, an extra term is introduced in the superpotential,
$W=\alpha\left({\rm
Tr}\left(\frac{\Phi^{3}}{3}-\xi\Phi\right)+\frac{i\mu}{2}({\rm
Tr}\,\Phi)^{2}\right),$ (62)
where $\xi$ is chosen as a real parameter, with $\xi>\Lambda^{2}$, and $\mu$
is a real mass parameter, $\mu<\epsilon\,\Lambda$.
In the effective low-energy superpotential we get
$\tilde{W}=\frac{A_{3D}M\tilde{M}}{\sqrt{2}}+\frac{\alpha(2A_{0}^{3}+6u(A_{3D})A_{0})}{3}-2\alpha\xi
A_{0}+2i\alpha\mu A_{0}^{2}\,\,,$ (63)
Then, the $F$ term takes the form
$\displaystyle V_{F}$ $\displaystyle=$ $\displaystyle
e_{3D}^{2}\left|\frac{M\tilde{M}}{\sqrt{2}}+2\alpha
u^{\prime}A_{0}\right|^{2}+4e_{0}^{2}\alpha^{2}\left|A_{0}^{2}+2i\mu
A_{0}+u-\xi\right|^{2}$ (64) $\displaystyle+$
$\displaystyle\frac{|MA_{3D}|^{2}\,|\tilde{M}A_{3D}|^{2}}{2}\,.$
Figure 9: Superpotential values in three vacua, at nonvanishing $\mu$ (in the
complex $\tilde{W}$ plane).
The Coulomb vacuum remains intact when $\mu\neq 0$ is switched on, and so is
the value of the superpotential in the Coulomb vacuum. Both confining vacua
have $A_{3D}=0$. The values of $A_{0}$ and the monopole field condensate are
$A_{0}=-\mu i\pm\sqrt{\xi-\Lambda^{2}-\mu^{2}},\quad
M\tilde{M}=i4\sqrt{2}\alpha\Lambda\left(-i\mu\pm\sqrt{\xi-\Lambda^{2}-\mu^{2}}\right)\,.$
(65)
The values of the superpotential in both confining vacua change (see Fig. 9),
being shifted upwards in the complex plane,
$\tilde{W}_{1,2}=\frac{2}{3}\alpha\left[-2i\mu^{3}+3i(\xi-\Lambda^{2})\mu\mp
2(\xi-\Lambda^{2}-\mu^{2})^{3/2}\right].$ (66)
The tension of the BPS domain wall is given by the absolute value of the
difference of the superpotentials at two vacua between which the given wall
interpolates. For this reason, if the composite BPS walls exist at $\mu\neq
0$, the composite wall will be stable, see Fig. 9.
In order to write the BPS equations for the elementary wall, we need complex
profile functions for each field, $M$, $A_{3D}$ and $A_{0}$. The ansatz
$\tilde{M}=iM$ can still be used. Let us introduce a phase (see Ref. [22] for
a detailed discussion)
$\omega={\rm
Arg}\,\,\frac{3\,(\xi-\Lambda^{2})\,\mu-2\mu^{3}}{2\left(\xi-\Lambda^{2}-\mu^{2}\right)^{3/2}}\,.$
(67)
The BPS equations generalizing those in Eq. (61) to the case $\mu\neq 0$ are
$\displaystyle\frac{\partial M^{\dagger}}{\partial
z}-ie^{i\omega}\frac{M\,A_{3D}}{\sqrt{2}}=0\,,$ $\displaystyle\frac{\partial
A_{0}^{\dagger}}{\partial z}-e^{i\omega}2\alpha e_{0}^{2}(A_{0}^{2}-2i\Lambda
A_{3D}+2i\mu\Lambda A_{0}-\xi+\Lambda^{2})=0\,,$ $\displaystyle\frac{\partial
A_{3D}^{\dagger}}{\partial
z}+ie^{i\omega}e_{3D}^{2}\left(\frac{M^{2}}{\sqrt{2}}-4\alpha\Lambda
A_{0}\right)=0\,.$ (68)
Numerical solution for the profile functions of the elementary walls at
$\mu\neq 0$ is displayed in Fig. 10.
Figure 10: Elementary wall profile functions at nonvanishing $\mu$. Left:
${\rm Re}\,A_{0}$ (solid), ${\rm Im}\,A_{0}$ (long dashes), ${\rm Im}\,A_{3}$
(short dashes), ${\rm Re}\,A_{3}$ (dots). Right: ${\rm Re}\,M$ (solid) and
${\rm Im}\,M$ (dashed).
For the composite wall we again can use the ansatz $\tilde{M}=iM$. The BPS
equations are very similar to the ones for the elementary walls. The only
difference is that for the composite wall $\omega=0$ and the profiles of ${\rm
Im}\,A_{0}$ and ${\rm Re}\,A_{3D}$ are constant,
${\rm Im}\,A_{0}=-\mu\,,\qquad{\rm Re}\,A_{3D}=0\,.$ (69)
The BPS equations for the non-constant profile functions are
$\displaystyle\frac{\partial({\rm Re}\,M)}{\partial z}=-\frac{({\rm
Re}\,M)\,({\rm Im}A_{3D})}{\sqrt{2}}\,,$ $\displaystyle\frac{\partial({\rm
Im}\,M)}{\partial z}=\frac{({\rm Im}\,M)\,({\rm Im}A_{3D})}{\sqrt{2}}\,,$
$\displaystyle\frac{\partial({\rm Re}\,A_{0})}{\partial z}=2\alpha
e_{0}^{2}(({\rm Re}\,A_{0})^{2}+2({\rm
Im}\,A_{3D})\Lambda-\epsilon^{2}\Lambda^{2}+\mu^{2})\,,$
$\displaystyle\frac{\partial({\rm Im}\,A_{3D})}{\partial
z}=-e_{3}^{2}\left(\frac{({\rm Re}\,M)^{2}-({\rm
Im}\,M)^{2}}{\sqrt{2}}-4\alpha\Lambda({\rm Re}\,A_{0})\right)\,.$ (70)
From these equations we deduce that the following quantity remains constant
($z$-independent):
$({\rm Re}\,M)(\,{\rm Im}\,M)=-2^{3/2}\alpha\mu\Lambda\,.$ (71)
The corresponding profiles are shown in Fig. 11. The boundary conditions are
such that $M(\infty)=-iM(-\infty)$. Note that if we try to use the boundary
conditions $M(\infty)=iM(-\infty)$, we get an unstable wall, whose tension is
larger than twice the tension of the elementary wall.
Figure 11: The profile functions of the stable composite wall profiles at
nonvanishing $\mu$ Left: ${\rm Re}\,A_{0}$ (solid), ${\rm Im}\,A_{0}$ (long
dashes), ${\rm Im}\,A_{3}$ (short dashes). Right: ${\rm Re}\,M$ (solid) and
${\rm Im}\,M$ (dashed).
The tensions of the BPS walls are given by the central charges,
$\displaystyle T_{12}$ $\displaystyle=$ $\displaystyle
T_{23}=\frac{4\alpha}{3}\sqrt{4(\epsilon^{2}\Lambda^{2}-\mu^{2})^{3}+(3\mu\epsilon^{2}\Lambda^{2}-2\mu^{3})^{2}}\,,$
$\displaystyle T_{13}$ $\displaystyle=$
$\displaystyle\frac{16\alpha}{3}(\epsilon^{2}\Lambda^{2}-\mu^{2})^{3/2}\,,$
(72)
implying that the composite wall is stable. Note also that the parameter
$\omega$ is different for the composite and elementary walls. Two out of four
supercharges will annihilate each domain wall; but they will be different for
the composite and elementary walls.
On symmetry grounds one can state that the real and imaginary parts of the $M$
condensate in the wall center (i.e. at the point $z=0$) are equal in absolute
value. Using the fact that $({\rm Re}\,M(z))(\,{\rm Im}\,M(z))$ is constant,
it is straightforward to analytically calculate the monopole condensate at the
center of the composite domain wall. The expression for the condensate in the
wall center is very concise,
$|{\rm Re}\,M(z=0)|=|{\rm Im}\,M(z=0)|\approx
2^{3/4}\sqrt{\mu\alpha\Lambda}\,.$ (73)
In the limit $\mu\ll\epsilon\Lambda$ the condensate outside the wall is
$|{\rm Re}\,M(z=\pm\infty)|\approx
2^{5/4}\Lambda\sqrt{\alpha\epsilon}\,,\quad|{\rm Im}\,M(z=\pm\infty)|\approx
2^{1/4}\mu\sqrt{\frac{\alpha}{\epsilon}}\,.$ (74)
The ratio of the absolute values of the monopole condensates inside and
outside the wall is proportional to $\sqrt{{\mu}/{(\epsilon\Lambda)}}$.
### 4.4 Confinement on the composite domain wall
As in the toy model discussed in Sect. 3, we would like to understand the
localization of the (massive) gauge field on the wall as a quasimodulus
$\sigma$ localized on the wall world volume. Previous consideration suggests
us to look for an opposite direction rotation of the U$(1)$ phase of the
monopole field at $z<0$ and $z>0$, respectively. Our target is an exited
domain wall (corresponding to $\sigma=\pi$) in which the charged field profile
vanishes in the center of the wall (see Fig. 4 pertinent to the toy model of
Sect. 3). There is an important difference between the toy model and that of
Sect. 4. In the former case the Coulomb phase was not a true vacuum of the
theory, while in the latter it is. This implies that imposing the condition
$\sigma=\pi$ we get, in fact, two elementary walls, with no binding energy,
separated by infinite distance. Needless to say, the condition that typical
energies in the low-energy theory must be $\ll 1/d$ cannot be met then. In
this formulation it makes no sense to speak of localization and reduction to
$2+1$ dimensions.
Figure 12: A sketch of an electric flux tube trapped in the middle of the
composite wall.
To discuss confinement and localization on the composite wall we must change
the setting. Consider the field configuration shown in Fig. 12, which displays
a flux tube trapped in the middle of the composite wall. Let us call $R$ the
radius of the region where the electric field is localized.
The two-component wall is stabilized at a distance $d_{0}$ between the
elementary walls. To separate them will cost a finite amount of energy per
unit surface of the wall, to be referred to as $\delta T_{\rm w}$. In the
model discussed in this section, at the first nonvanishing order in $\mu$, we
have
$\delta T_{\rm w}=6\,\alpha\,\epsilon\,\Lambda\,\mu^{2}\,.$ (75)
Denote the electric flux of the vortex by $f$. An estimate for the tension of
the flux tube is
$T_{\rm ft}\approx\frac{1}{2e_{3D}^{2}}\,\,\frac{f^{2}}{\pi R^{2}}+2R\,\delta
T_{\rm w}\,.$ (76)
Minimizing with respect to $R$ we arrive at
$R_{*}=\left(\frac{f^{2}}{2\pi\,e_{3D}^{2}\,\delta T_{w}}\right)^{1/3}\,.$
(77)
This shows that $R_{*}$ is a finite quantity, and the flux is indeed squeezed
into flux tubes inside the composite domain wall. This is, of course,
applicable provided that $\delta T_{\rm w}$ is positive (i.e. the force
between two elementary domain walls is attractive, which is the case in the
model at hand).
Now, let us consider the case of a constant electric flux per unit length of
the wall in the perpendicular direction (analogous to the picture in Fig. 3)
We will denote this quantity — flux per unit length — by the same letter $f$.
This flux is trapped in the middle Coulomb vacuum. As a consequence, there is
a repulsive contribution to the potential between the two components of the
composite wall,
$T_{\rm wf}=\frac{1}{e_{3D}^{2}}\,\,\frac{f^{2}}{d}\,.$ (78)
In order to find the minimum of the overall potential, we have to add this to
the domain wall tension $T_{\rm w}$ considered as a function of $d$. At large
distance $T_{\rm w}$ has just an exponential tail in $d$; therefore, the-long
range force is repulsive, as follows from Eq. (78). If the flux $f$ is small,
however, the global minimum of the inter-wall potential is still at finite
$d$. There is a critical value of $f$ — let us call it $f_{0}$ — for which the
elementary domain walls at the equilibrium will be separated by a infinite
distance. A simple energy estimate can be used to evaluate the value of
$f_{0}$,
$\frac{1}{e_{3D}^{2}}\,\,\frac{f^{2}_{0}}{d_{0}}>\delta T_{\rm w}\,,\quad
f_{0}=e_{3D}\sqrt{d_{0}\,\delta T_{\rm w}}\,,$ (79)
where $d_{0}$ is the domain wall separation at zero $f$.
Thus, although the low-energy description in the Seiberg–Witten motivated
model at hand is not of the sine-Gordon type (cf. (25)), the quasimodulus-
based low-energy description is still valid at $|\sigma|\ll\pi$: a mass term
$m\sigma^{2}$ is generated.
## 5 Two-Flavor Model
### 5.1 $\mathcal{N}=2$ SQED with two flavors
At this point one is tempted to say that the fact that the gauge field
localized on the wall weakly confines is due to residual exponentially small
VEV’s of the Higgs field inside the wall. In this case, the absence of the
Coulomb regime on the wall would be a universal phenomenon. In fact, we know
that this is not the case [4] (see also the review paper [8]). In Ref. [4]
$\mathcal{N}=2$ SQED with two flavors was considered. This model has a domain
wall with a phase field $\sigma$ localized on it. This field is a Goldstone of
a spontaneously broken U(1) and, thus, remains exactly massless. Upon
Polyakov’s dualization it represents a U(1) gauge field on the wall in the
Coulomb regime. Moreover, in the same paper [4] it was shown that a magnetic
flux tube coming from the bulk in the perpendicular to the wall direction ends
on the wall creating a vortex of the $\sigma$ field. Upon Polyakov’s
dualization this vortex is interpreted as an electric charge, a source of the
electric field on the wall. Two such sources interact through the Coulomb
potential at large distances (logarithmic in 2+1 dimensions).
Residual exponentially small VEV’s of the Higgs field inside the wall are
certainly important. For instance, in the toy model considered in Sect. 3 the
$\sigma$ field turns out to be a quasimodulus only because the wall at hand is
very thick implying exponential suppression of the Higgs field inside the
wall. If the wall was thin, $\sigma$ would be heavy, and the only light field
on its world volume would be the translational modulus. In the two-flavor
model we can introduce $\sigma$ as a relative phase between $q_{1}$ on one
side of the wall and $q_{2}$ on the other side. Because these fields are
different, there is no need in a $z$ dependent interpolation of $\sigma$
inside the wall, as was the case in the one-flavor model. This is the
technical reason for its masslessness.
Below we will illustrate the emergence of the exact moduli field $\sigma$ in
$\mathcal{N}=2$ SQED with $2$ flavors analyzing it in a regime somewhat
different from that of Ref. [4]. This moduli field remains massless even in
the limit of thin wall. Thus, the mode of implementation of 2+1 dimensional
electrodynamics on the wall world volume — Coulomb vs. confinement — is a
dynamical issue.
Consider $\mathcal{N}=2$ SQED with two flavors and the matter mass terms
chosen as follows:
$m_{1}=m\,,\qquad m_{2}=-m\,,\quad\Delta m\equiv m_{1}-m_{2}\equiv 2m\,.$ (80)
We will introduce the Fayet–Iliopoulos term $\xi$ through the superpotential.
Then the bosonic part of the action can be written as
$S=\int
d^{4}x\left\\{\frac{1}{4e^{2}}F_{\mu\nu}^{2}+\frac{1}{e^{2}}|\partial_{\mu}a|^{2}+\bar{\nabla}_{\mu}\bar{q}_{A}\nabla_{\mu}q^{A}+\bar{\nabla}_{\mu}\tilde{q}_{A}\nabla_{\mu}\bar{\tilde{q}}^{A}+V_{D}+V_{F}\right\\},\\\
\mbox{ }$
where the scalar potential is the sum of $D$ and $F$ terms,
$\displaystyle V_{D}$ $\displaystyle=$
$\displaystyle\frac{e^{2}}{8}\left(|q^{B}|^{2}-|\tilde{q}_{B}|^{2}\right)^{2}\,,$
$\displaystyle V_{F}$ $\displaystyle=$
$\displaystyle\frac{1}{2}\left|q^{B}(a+\sqrt{2}m_{B})\right|^{2}+\frac{1}{2}\,\left|\tilde{q}_{B}(a+\sqrt{2}m_{B})\right|^{2}$
(81) $\displaystyle+$
$\displaystyle\frac{e^{2}}{2}\left|\tilde{q}_{A}q^{A}-\frac{\xi}{2}\right|^{2}\,.$
This theory has two vacua, and a domain wall interpolating between them. It
was fully analyzed in [4] in the limit of thick wall
$m\gg e\sqrt{\xi}\,,$ (82)
when the overlap between two edges of the wall is exponentially small.
Now we will discuss the same problem in the opposite limit of a thin wall,
with a strong overlap,
$m\ll e\sqrt{\xi}\,.$ (83)
This limit is usually referred to as the sigma-model limit. If $m\ll
e\sqrt{\xi}$, the “photonic” supermultiplet becomes heavy, since the photon
mass in the bulk $\sim e\sqrt{\xi}$. Therefore, it can be integrated out,
leaving us with the theory of fields from the matter supermultiplets, nearly
massless in the scale $e\sqrt{\xi}$, which interact through a nonlinear sigma
model with the Kähler term corresponding to the Eguchi–Hanson metric. The
manifold parametrized by these (nearly) massless fields is four-dimensional.
The two vacua of the model vacua lie at the base of this manifold. Therefore,
in considering the domain wall solutions in the sigma-model limit $m\to 0$
[23, 24, 25] one can use the ansatz $q=\tilde{q}^{\dagger}$ and limit oneself
to the base manifold, which is, in fact, a two-dimensional sphere. In this way
we arrive at the problem of the domain wall in the CP(1) model deformed by a
twisted mass term (related to a nonvanishing $\Delta m=2m$). In this
formulation the problem was first addressed in [24].
In the sigma-model limit one can readily find explicitly the wall profiles,
$\displaystyle q^{1}$ $\displaystyle=$
$\displaystyle\bar{\tilde{q}}^{1}=\sqrt{\frac{\xi}{2}}\left(\cos\frac{\eta(z)}{2}\right)\,,$
$\displaystyle q^{2}$ $\displaystyle=$
$\displaystyle\bar{\tilde{q}}^{2}=\sqrt{\frac{\xi}{2}}\left(\sin\frac{\eta(z)}{2}\right)e^{i\sigma}\,,$
$\displaystyle a$ $\displaystyle=$ $\displaystyle
m\sqrt{2}\left(\sin^{2}\frac{\eta}{2}-\cos^{2}\frac{\eta}{2}\right)=-m\sqrt{2}\,\cos\eta\,,$
(84)
where
$\eta(z)=2\arctan\,(\exp(2mz))\,.$ (85)
Note that
$|q^{1}|^{2}+|q^{2}|^{2}={\xi/2}$ (86)
for all $z$. The modulus $\sigma$ in Eq. (84) reflects the fact that the
target space of the CP(1) model with the twisted mass has U(1) symmetry. It is
spontaneously broken on each given wall solution. More details on kinks in the
CP(1) model with the twisted mass, which appear as domain walls in the problem
at hand, can be found in [8].
Since $e^{2}\to\infty$ in the sigma-model limit, in the bulk action we can
neglect the contribution due to the gauge field strength tensor. The gauge
field then becomes non dynamical; it is expressible in terms of the matter
fields,
$A_{\mu}=\frac{i\left(\bar{q}\stackrel{{\scriptstyle\leftrightarrow}}{{\partial_{\mu}}}q-\bar{\tilde{q}}\stackrel{{\scriptstyle\leftrightarrow}}{{\partial_{\mu}}}\tilde{q}\right)}{\bar{q}q+\bar{\tilde{q}}\tilde{q}}\,.$
(87)
The scalar field kinetic term tends to zero too, implying
$a=\sqrt{2}\,m\,\,\frac{|q_{2}|^{2}+|\tilde{q}_{2}|^{2}-|q_{1}|^{2}-|\tilde{q}_{1}|^{2}}{\bar{q}q+\bar{\tilde{q}}\tilde{q}}\,.$
(88)
As usual, we promote $\sigma$ to a $(x,y,t)$-dependent field on the wall
world-volume. In our gauge the field $A_{z}$ vanishes while the nonvanishing
component are
$A_{k}=-2\sin^{2}\frac{\eta(z)}{2}(\partial_{k}\sigma)\,,\qquad k=1,2\,.$ (89)
The field strength then takes the form
$F_{kz}=\frac{2m}{\cosh^{2}2mz}\,(\partial_{k}\sigma)\,,\qquad k=1,2\,.$ (90)
If we substitute these expressions back in the action, we get
$S=\int
dz\,dt\,d^{2}x(\partial_{k}\sigma)^{2}\left(\frac{\xi}{2\cosh^{2}2mz}+\frac{2m^{2}}{e^{2}\,\cosh^{4}2mz}\right)\,,$
(91)
the the first term is due to the covariant derivative and the second term is
due to the field strength tensor. The second term is negligible and can be
omitted. This shows that our approximation is self-consistent. Keeping only
the first term and integrating over $z$ gives us the normalization of the
world-volume effective action,
$\displaystyle S$ $\displaystyle=$ $\displaystyle\int
d^{3}x\,\beta\,(\partial_{k}\sigma)^{2}\,,$ $\displaystyle\beta$
$\displaystyle=$ $\displaystyle\int
dz\,\frac{\xi}{2\cosh^{2}2mz}=\frac{\xi}{2m}\,.$ (92)
The $\sigma$ field is strictly massless, the gauge field that dualizes
$\sigma$ is in the Coulomb phase on the wall world volume.
Quantitatively, the thin wall (sigma model) approximation is not
parametrically supported. However, it reveals the existence of a massless
modulus on the wall in the most straightforward way. Moreover, it can be
conveniently used to discuss a “boojum” configuration, with a flux tube ending
on the wall. The approximation is analytic at sufficiently large distances
from the point of the wall-tube junction.
The wall-tube junction solution in the sigma-model limit was found (in a
different notation) in Refs. [25, 26]. For illustrative purposes we will
reproduce it in our notation. The BPS equations for a vortex ending on the
wall are (see Ref. [4])
$\displaystyle
B_{3}-\frac{g^{2}}{2}\left(2|q_{k}|^{2}-\xi\right)-\sqrt{2}\partial_{3}a=0\,,\quad
B_{1}-iB_{2}-\sqrt{2}(\partial_{1}-i\partial_{2})a=0\,,$
$\displaystyle\nabla_{3}q_{k}=-\frac{1}{\sqrt{2}}q_{k}(a+\sqrt{2}m_{k})\,,\quad(\nabla_{1}-i\nabla_{2})q_{k}=0\,.$
(93)
In the sigma-model limit only $q_{k}$’s are dynamical variables. Thus, we need
to solve two equations in the second line in Eq. (93). This can be done using
the following ansatz in the cylindrical coordinates $z,r,\phi$:
$q_{1}=\sqrt{\frac{\xi}{2}}\,e^{i\phi}\,\cos\frac{\theta(z,r)}{2}\,,\qquad
q_{2}=\sqrt{\frac{\xi}{2}}\,\sin\frac{\theta(z,r)}{2}\,.$ (94)
Then the BPS equations can be written as
$\frac{\partial\theta}{\partial
r}=\frac{1}{r}\sin\theta\,,\qquad\frac{\partial\theta}{\partial
z}=2m\,\sin\theta\,.$ (95)
The solution can be readily found, namely,
$\theta=2\arctan\,\left[\frac{\,\exp\left(2m(z-z_{0})\right)}{r}\right]\,,$
(96)
where $z_{0}$ is an integration constant which parametrizes the $z$ position
of the object. Now we can use Eqs. (87) and (88) to determine $A_{\mu}$ and
$a$,
$a=-\sqrt{2}m\cos\theta\,,\qquad
A_{\hat{\phi}}=-\frac{\cos^{2}(\theta/2)}{r}\,.$ (97)
This solution is valid in the sigma-model limit ($m\ll e\sqrt{\xi}$) and at
distances $r\gg{1}/({e\sqrt{\xi}})$ from the point of the wall-tube junction.
We see that the gauge field is localized inside the wall and has the Coulomb
$1/r$ behavior at large $r$. This shows that we do have the Coulomb phase on
the wall – even in the sigma model limit – and confirms that $\sigma$ is a
strictly massless modulus.
To conclude this section let us return to the gauge theory limit (82) of the
problem at hand studied in [4]. One might naively suspect that there are two
different moduli fields in the two-flavor model. One is the modulus $\sigma$
which is exactly massless because it is related to the global U(1) symmetry
broken by the wall solution. One might guess that this modulus has nothing to
do with the bulk gauge field. Allegedly, the bulk gauge field localized on the
wall through the same mechanism as was discussed in the one-flavor model is a
quasimodulus $\tilde{\sigma}$ which is different from the modulus $\sigma$,
and $\tilde{\sigma}$ acquires a small mass due to exponentially small quark
fields inside the wall, as in Sect. 3.
The solution for the wall-string junction found in [4] shows that this naive
picture is incorrect. In [4] it was shown that the string orthogonal to the
domain wall can end on the wall, and the magnetic flux it carries penetrates
in the wall. The endpoint of the string plays the role of a vortex for the
$\sigma$ field localized on the wall. In fact, the solution for the wall (far
away from the string endpoint) is approximately given by the unperturbed
domain wall solution with the collective coordinate $\sigma$ determined by [4]
$\sigma\,=\,\alpha,$ (98)
where $\alpha$ is the polar angle on the wall plane. The endpoint of the
string creates a vortex of the field $\sigma$. This shows that in fact there
is no extra field $\tilde{\sigma}$. The gauge field localized on the wall is
dual to $\sigma$ which is strictly massless in the two-flavor model.
### 5.2 Further discussion of moduli vs. quasimoduli
The example of Sect. 5.1 shows that the two-flavor model has a massless phase
field $\sigma$ localized on the domain wall. This is not related to
supersymmetry. Any model with a global U(1) symmetry, two distinct vacua in
which the global symmetry is unbroken, and a domain wall that spontaneously
breaks this symmetry, automatically has a massless Goldstone boson localized
on the domain wall. In particular, the wall solution with $\sigma=\pi$ has the
same tension as that with $\sigma=0$, in an obvious contradistinction with the
solution of the toy model discussed in Sect. 3. Note that in the latter case,
in the $\sigma=\pi$ wall the matter field vanishes on the plane lying in the
middle of the wall. No such zero-plane occurs in the former case.
To better understand the relation between the one- and two-flavor models, we
should dwell on the following question: why there is no confinement on the
wall in the two-flavor models, although the matter field condensates do not
exactly vanish inside the wall? Why the $\sigma$ field remains massless in
this case? A (partial) answer to this puzzling question is as follows. 3+1
dimensional physics inside the wall, between its edges, is not the only thing
to consider. Existence vs. nonexistence of a massless modulus is a global
effect. The boundary conditions at $z=\pm\infty$ play a crucial role. In the
two-flavor model we deal with four phases:
$q_{1}(z=-\infty),\,\,\,q_{2}(z=-\infty)\,\,\,{\rm
and}\,\,\,q_{1}(z=+\infty),\,\,\,q_{2}(z=+\infty)\,.$ (99)
Moreover,
$q_{1}(z=-\infty)\to\sqrt{\frac{\xi}{2}}\,,\quad{\rm
and}\,\,\,q_{2}(z=+\infty)\to\sqrt{\frac{\xi}{2}}\,,$ (100)
while
$|q_{2}(z=-\infty)|\to 0\,,\quad|q_{1}(z=+\infty)|\to 0\,.$ (101)
The massless modulus $\sigma$ corresponds to a rotation of $q_{1}(z=-\infty)$
and $q_{1}(z=+\infty)$ in the same direction while $q_{2}(z=-\infty)$ and
$q_{2}(z=+\infty)$ remain fixed.
The would-be quasimodulus of Dvali et al. corresponds to a rotation of the
phases of $q_{1}(z=+\infty)$ and $q_{2}(z=+\infty)$ in the same direction
while $q_{1}(z=-\infty)$ and $q_{2}(z=-\infty)$ remain fixed. The would-be
quasimodulus has a tachyonic direction and classically decays into the modulus
$\sigma$.
Let us try to understand mechanisms responsible for this phenomenon in more
detail. Consider a string parallel to the domain wall (a grid of such strings
is depicted in Fig. 13).
Figure 13: A grid of flux tubes parallel to the domain wall. At $z>0$ they are
made out of the $q_{2}$ field condensate. The line of zeroes of the $q_{2}$
field is surrounded by the magnetic field. At $z=0$ the vortices lies in the
middle of the wall. Proceeding further the zeroes of $q_{2}$ go to $z$
negative while the magnetic field remains on the domain wall surface. In the
limit where the zeroes are at $-\infty$ we recover the solution describing a
constant magnetic field on the wall.
If the distance between the wall and the string is very large we have an
ordinary ANO string with thickness $\sim{1}/{(e\sqrt{\xi})}$. As we move the
string toward the wall, the quark condensate decreases and, in the center of
the wall, the thickness of the string (in the directions parallel to the wall)
becomes
$\sim R=\frac{1}{e\sqrt{\xi}e^{-dm/2}}\,,$ (102)
where $d$ is the thickness of the wall. The thickness of the wall depends on
the regime in which we find ourselves. In the limit ${m}/{(e\sqrt{\xi})}\gg 1$
the thickness is entirely determined by the matter field and is
$d\sim{m}/{(e^{2}\xi)}$. In the opposite limit ${m}/{(e\sqrt{\xi})}\ll 1$ (the
sigma-model limit) the thickness is $d\sim{1}/{m}$.
If we want to compare the thickness of the domain wall with that of the string
in the middle of the wall we should compare
${\rm Max}\left[\frac{1}{m},\frac{m}{e^{2}\xi}\right]$
with
$\frac{1}{e\sqrt{\xi}e^{-dm/2}}\,.$
Multiplying both sides by $e\sqrt{\xi}$ we can express everything in terms of
a dimensionless parameter
$x=\frac{m}{e\sqrt{\xi}}\,.$ (103)
Thus, we should compare
${\rm Max}\left[\frac{1}{x},x\right]\,\,\,{\rm
with}\,\,\,e^{\frac{x^{2}}{2}}\,.$
These two functions are plotted in Fig. 14.
Figure 14: Comparison between the thickness of the domain wall, and the
transverse size of the flux tube located inside the domain wall. In the region
$x\ll 1$ corresponding to the sigma-model limit, the transverse size of the
flux tube is much smaller than the wall thickness.
In the $x\gg 1$ limit the thickness of the string is much larger than the
thickness of the domain wall. This is qualitatively consistent with the
existence of a Coulomb phase in the middle of the wall. The opposite limit
$x\ll 1$ is problematic. The thickness of the string is much smaller than the
thickness of the wall. We are thus tempted to conclude that in this regime the
wall lives in a confining phase, which is clearly in contradiction with the
rigorous proof above of the existence of a massless modulus and, hence, the
Coulomb phase.
The way out of this paradox is as follows. In the previous discussion we have
compared the thickness of the wall with that of the string in the middle of
the wall. So we (erroneously) assumed that dynamics on the wall world volume
is directly deducible from consideration of 3+1 dimensional dynamics inside
the wall, between its two edges. In the $x\gg 0$ limit this creates little
problem. In this limit we have a thick region inside the wall in which the
matter field condensate essentially vanishes. But in the $x\ll 1$ limit this
way of thinking creates a paradox and, thus, reveal its inconsistency.
In fact, the wall world-volume dynamics reflects not only what happens at
$z\sim 0$, but also what happens at $z\to\pm\infty$.
Consider a grid of flux tubes at $z>0$ parallel to the domain wall (Fig. 13).
We should remember that $z>0$ is the region where $q_{2}$ condenses while
$z<0$ the region where $q_{1}$ condenses. Each flux tube has a line of zeroes
for the $q_{2}$ field and a magnetic field surrounding this line of zeroes. We
then move the grid of flux tubes toward the domain wall. We want to understand
what happens to these flux tubes as we pass through the wall and then move
toward negative infinity. At $z=0$ the flux tube grid is in the middle of the
wall; the thickness of each flux tube is larger. If we move on, something new
happens. The lines of zeroes go to $z$ negative and get separated from their
magnetic field. The magnetic field remains trapped inside the wall. In the
limit where the zeroes are at $z\to-\infty$, we recover the solution of
constant magnetic field inside the domain wall. So, the $q_{2}$ flux tube,
passing through the wall to the other side of the wall, does not become the
$q_{1}$ flux tube.
This is not in contradiction with the symmetries of the theory. The domain
wall is symmetric under the $Z_{2}$ transformation $z\leftrightarrow-z$
combined with $q_{2}\leftrightarrow q_{1}$. The one-flavor case is different.
The domain wall is symmetric under the parity transformation
$z\leftrightarrow-z$ and the $q$ flux tube passing through the wall to the
other side of the wall preserves its “identity” remaining the $q$ flux tube.
Of course, physically the line of the $q_{2}$ zeroes at $z=-\infty$ (more
exactly, the plane of the $q_{2}$ zeroes 777 A flux tube has line of zeroes of
a charged scalar inside it, while in order to have a Coulomb phase on the wall
we need the whole plane of the $q$ zeroes (parallel to $x,y$-plane).) is in
fact not so “far away.” The $q_{2}$ quark has the exponential profile
$\sim\exp{(-e^{2}\xi\,z^{2})}$ inside the wall and essentially becomes zero at
distances of the order of $1/e\sqrt{\xi}$. Thus, the plane of the $q_{2}$
zeroes is shifted from the region where the magnetic field is concentrated by
separation of the order of thickness of the flux tube in the vacuum. This is
in accord with our physical intuition.
Here we arrive at a crucial distinction of the two-flavor model from the one-
flavor model of Sect. 2. In the one-flavor model the “empty” domain wall (i.e.
without magnetic field) has a nonvanishing $Q$ field everywhere. Inside the
wall it becomes small, but still does not vanish. In order for a magnetic
field to penetrate the wall we need to have zeroes of the $Q$-field. Clearly,
it costs less energy to create a line of $Q$-zeroes in the $(x,y)$-plane than
the whole $(x,y)$-plane of the $Q$-zeroes. This qualitatively explains why we
have confinement on the wall in the one-flavor model ($\sigma$ is a
quasimodulus) and the Coulomb phase on the wall in the two-flavor model
($\sigma$ is strictly massless).
Now, let us discuss energetics of this process. First of all, consider the
string grid when it is far away from the wall in the positive-$z$ region.
Assume we deal with a homogeneous grid with density $f$, so that the flux per
unit of length is $4\pi f$. The tension of this configuration is the sum of
the wall tension plus the string tensions
$T_{z\gg 0}=\xi\Delta m+2\pi\xi f\,.$ (104)
In the opposite position, when the zeroes are at negative $z$, far away from
the wall, we can also easily compute the tension. It is just that of a domain
wall with a constant magnetic flux inside it. An easy way to get the result is
the thin-edge approximation (similar to that adopted in [27] for the
$Q$-wall). The wall tension is now given by a sum of three terms,
$T\left(d\right)=\frac{1}{d}\left(\frac{2(\Delta
m)^{2}}{g^{2}}+\frac{8\pi^{2}f^{2}}{g}\right)+\frac{g^{2}\xi^{2}}{8}d\,.$
(105)
Minimizing with respect to $d$ we obtain
$d=\frac{4}{g^{2}\,\xi}\left((\Delta m)^{2}+4\pi^{2}f^{2}\right)^{1/2},$ (106)
and
$T_{z\ll 0}=\xi\sqrt{(\Delta m)^{2}+4\pi^{2}f^{2}}\,.$ (107)
The expression in Eq. (107) can also be obtained from a more rigorous
derivation using the Bogomol’nyi completion method (see Ref. [28]). The
physical situation is very similar to the $Q$-kinks discussed in detail in
Refs. [29] for $1+1$ dimensional sigma models. Following [4] we denote
$q^{A}=\bar{\widetilde{q}}_{A}\equiv\frac{1}{\sqrt{2}}\,\varphi^{A}\,,$ (108)
where we introduce a new complex field $\varphi^{A}$. The action then reduces
to
$\displaystyle S_{\rm red}$ $\displaystyle=$ $\displaystyle\int
d^{4}x\left\\{\frac{1}{4g^{2}}F_{\mu\nu}^{2}+\frac{1}{g^{2}}|\partial_{\mu}a|^{2}+\bar{\nabla}_{\mu}\bar{\varphi}_{A}\nabla_{\mu}\varphi^{A}\right.$
(109) $\displaystyle+$
$\displaystyle\left.\frac{g^{2}}{8}\left(|\varphi^{A}|^{2}-\xi\right)^{2}+\frac{1}{2}\left|\varphi^{A}\right|^{2}\,\left|a+\sqrt{2}m_{A}\right|^{2}\right\\}\,.$
The Bogomol’nyi completion of the wall$+$tube energy functional can be written
as
$\displaystyle T_{\rm w}=\int dz$
$\displaystyle\left\\{\,\,\left|\cos{\alpha\,}\nabla_{z}\varphi^{A}\pm\frac{1}{\sqrt{2}}\varphi^{A}(a+\sqrt{2}m_{A})\right|^{2}\right.$
(110) $\displaystyle+\left|\sin{\alpha\,}\nabla_{z}\varphi^{A}\mp
i\,\nabla_{x}\varphi^{A}\right|^{2}$
$\displaystyle+\left|\frac{1}{g}\partial_{z}a\pm\cos{\alpha\,}\frac{g}{2\sqrt{2}}(|\varphi^{A}|^{2}-\xi)\right|^{2}$
$\displaystyle+\left|\frac{1}{\sqrt{2}g}\partial_{z}A_{x}\pm\sin{\alpha\,}\frac{g}{2\sqrt{2}}(|\varphi^{A}|^{2}-\xi)\right|^{2}$
$\displaystyle\pm\left.\frac{1}{\sqrt{2}}\cos{\alpha\,}\xi\partial_{z}a\pm\frac{1}{2}\sin{\alpha\,}\xi\partial_{z}A_{x}\right\\}.$
The BPS equations are obtained by putting to zero each of the first four lines
of Eq. (110). In order to find an explicit solution, let us choose a gauge
where $A_{z}=0$; the following ansatz can then be used:
$\varphi_{k}=\eta_{k}(z)\exp(i\lambda_{k}x)\,,\qquad a=a(z)\,,\qquad
A_{x}=-f(z)\,.$ (111)
From a particular linear combination of the BPS equations we can find the
value of $\lambda_{k}$ and also that the profile for $a$ and for $f$ are
proportional:
$f(z)=\sqrt{2}\tan\alpha\,a(z)\,,\qquad\lambda_{k}=m_{k}\tan\alpha\,.$ (112)
The equations for the other profiles give the following first order system:
$\sqrt{2}\,\partial_{z}a+\frac{g^{2}}{2}\cos\alpha(\eta_{1}^{2}+\eta_{2}^{2}-\xi)=0\,,$
(113)
$\partial_{z}\eta_{k}+\frac{1}{\sqrt{2}\cos\alpha}\eta_{k}(a+\sqrt{2}m_{k})=0\,.$
A numerical solution is shown in Fig. 15.
Figure 15: Left: Domain wall without magnetic flux ($\alpha=0$). Right: Domain
wall with magnetic flux ($\alpha=\pi/3$). Due to the magnetic field, the
domain wall thickness increases. The following numerical values were used:
$m=0.3$, $e=0.2$, $\xi=1$.
For the moment $\alpha$ is an arbitrary angle (this is a usual trick used in
analyzing dyons or $Q$-kinks). The wall$+$tube boundary conditions are as
follows. At $z\to+\infty$
$\varphi^{1}=0\ ,\quad\varphi^{2}=\sqrt{\xi}e^{i2\pi f}\ ,\quad A_{x}=4\pi
f\,.$ (114)
At $z\to-\infty$
$\varphi^{1}=\sqrt{\xi}\ ,\quad\varphi^{2}=0\ ,\quad A_{x}=0\,.$ (115)
The decomposition (110) give the upper bound
$T_{\rm w}\geq\cos{\alpha\,}\xi\Delta m+\sin{\alpha\,}2\pi\xi f\,.$ (116)
Maximizing with respect to $\alpha$ we get exactly the expression in Eq.
(107).]
This presents a more quantitative proof of our assertion. The $q_{2}$ string
parallel to the wall reaches the minimum of the energy when the line of zeroes
of $q_{2}$ is at $z\to-\infty$, and the tension is given by (107). If,
instead, we consider a set of $q_{1}$ strings parallel to the wall, the
minimum is reached when the line of zeroes of $q_{1}$ is at $z\to+\infty$. In
both cases the energy is always localized inside the domain wall in the form
of a constant magnetic field. The constant magnetic field corresponds to the
Coulomb phase on the wall world volume, with the massless modulus $\sigma=ky$.
Figure 16: A grid of strings perpendicular to the domain wall. The strings
form a linear lattice with distance $1/(2f)$ between them. The domain wall is
deformed into the shape shown in this figure in order to balance the tension
of the strings. Half flux of the flux tubes goes into each of two semi-planes
into which the wall is divided. The angle $\theta$ is determined in the text,
see Eq. (117). It tends to $0$ when the string density $2f$ is very small. It
tends to $\pi/2$ when the magnetic flux is very large.
We can also discuss a different system that will help us to further elucidate
these issues. Consider a grid of flux tubes perpendicular to the domain wall
(Fig. 16). The flux tubes are aligned along the $z$ axis and the wall,
initially, is aligned in the plane $x,y$. We assume the flux tubes to be
equidistant separated by intervals $1/(2f)$. (We consider twice the density of
the previous system for later convenience). Hence, $2f$ is the linear density.
The magnetic flux density is $8\pi f$. This simplification will allow us to
make computations very quickly. Simultaneously, this set-up still provides us
with essential information on physics of the “flux tubes perpendicular to the
wall” system.
This flux tube grid, for sufficiently large density $2f$, can be considered as
a surface with tension $4\pi\xi f$ and linear flux $8\pi f$. When this surface
intersects with the domain wall we have a three-surface junction: the flux
tube grid and two semi-planes into which the domain wall is divided. The angle
$\theta$ shown in Fig. 16 defines geometry of the junction. The magnetic flux
carried by the flux tubes is divided exactly into two equal parts, so that
each semi-plane carries linear flux $4\pi f$. The wall tension is given by
$\xi\sqrt{(\Delta m)^{2}+4\pi^{2}f^{2}}$. The angle $\theta$ is determined by
a simple balance of tensions,
$\cos{\theta}=\frac{2\pi f}{\sqrt{(\Delta m)^{2}+4\pi^{2}f^{2}}}\,.$ (117)
Note that $\theta=\pi/2-\alpha$ where $\alpha$ is the angle that maximizes the
BPS bound (116).
From this simple example we can learn an important lesson. First of all the
flux tube grid is a source of a constant magnetic field inside the wall. This
means, as was already discussed, that the $2+1$ dimensional theory on the
domain wall is in the Coulomb phase. Second, we can conclude that the wall
with a constant magnetic field is $1/4$ BPS saturated. This follows from the
fact that the flux tube perpendicular to the wall is known to be $1/4$ BPS
(See [4] and Appendix B.) It is straightforward to verify that, if we take the
BPS equations for a system of strings ending on the wall (which are in Eq.
(93)) and rotate them by an angle $\alpha$ around the $x$ axis, we recover the
same equations as those that we get from the Bogomol’nyi completion in Eq.
(110).
## 6 Peculiarities of the Gauge Field Localization. Can we uplift the problem
to five dimensions?
As was mentioned in Sect. 2, localization of gauge fields on domain walls is
not similar to that of, say, spinor field since dualization is important. This
makes the procedure non-local with respect to the gauge potentials in the bulk
and on the brane. The latter is not just a mode reduction of the former.
Let us explain it in more detail. To begin with, consider the conventional
localization mechanism on topological defects. The bulk theory is defined in
space-time $X^{M}$ where $M=0,\dots,d-1$. The soliton is a topological object
extended in $x^{\mu}$ where $\mu=0,\dots,p$. The transverse coordinates are
$s^{a}$ with $a=p+1,\dots,d-1$. The theory has some bosonic fields
$\phi^{(j)}(x,s)$. The soliton is a topologically stable solution made of the
bosonic fields $\phi^{(a)}$ which are independent of $x$. The soliton is a
$p$-brane spanned on the coordinates $x^{\mu}$.
Now if we want to find the spectrum of a particular scalar field $\varphi$ in
the soliton background we can separate variables
$\varphi(x,s)=\sum_{n}\varphi_{\parallel}^{(n)}(x)\varphi_{\perp}^{(n)}(s)\,.$
(118)
In the quadratic in $\varphi$ approximation the Lagrangian takes the form
$L=\int d^{p}x\,\int
d^{d-p}s\left(\partial_{M}\varphi\partial^{M}\varphi-f(\phi)\varphi^{2}\right)$
(119)
implying the following (linear) equations of motion for $\varphi$
$(\partial_{M}\partial^{M}+f(\phi))\varphi(X)=0\,.$ (120)
Inserting here Eq. (118) we get two equations
$\displaystyle(\partial_{a}\partial_{a}+f(\phi))\varphi_{\perp}^{(n)}(s)$
$\displaystyle=$ $\displaystyle m_{(n)}^{2}\varphi_{\perp}^{(n)}(s)\,,$ (121)
$\displaystyle(\partial_{\mu}\partial^{\mu}+m_{(n)}^{2})\varphi_{\parallel}^{(n)}(x)$
$\displaystyle=$ $\displaystyle 0\,,$ (122)
where we used the fact that
$\partial_{M}\partial^{M}=\partial_{\mu}\partial^{\mu}+\partial_{a}\partial_{a}$.
The physical meaning of this formula is as follows. First we solve the
“transverse” field equation (121) and we find the mass eigenvalues. The
corresponding longitudinal field $\varphi_{\parallel}(x)$ is (generally
speaking) a massive field in the longitudinal variables $x$. As long as
$\varphi_{\perp}^{(n)}$ is normalizable on $s^{a}$ and $m_{(n)}$ is much
smaller than the inverse soliton thickness, we can keep this soliton-localized
field in the low-energy Lagrangian emerging on the soliton world volume. For
this particle to be massless, the corresponding transverse mode must be a zero
mode of the equation (121).
To write the effective Lagrangian for this localized field we use the bulk
Lagrangian (119), expand in longitudinal and transverse fields, and integrate
over the the transverse variables,
$\int ds\varphi_{\perp}(s)^{2}\int
d^{p}x\left(\partial_{\mu}\varphi_{\parallel}(x)\partial^{\mu}\varphi_{\parallel}(x)-m^{2}\varphi_{\parallel}(x)^{2}\right).$
(123)
The norm of the perpendicular field factorizes out providing a normalization
for the parallel field. Only normalizable modes of (121) lead to soliton-
localized fluctuations.
Now let us return to gauge fields on 1+2-dimensional walls [4], try to follow
the way outlined above and see that this is not the right procedure to
localize the gauge field.
The bulk fields we start from are the gauge field $A_{M}$, and two complex
scalar fields $q_{1}$ and $q_{2}$. The bulk theory is four-dimensional, and
$q_{2}$. If we want to parallelize the procedure outlined above we “nominate”
$A_{\mu}$ ($\mu=0,1,2)$ as “our” fields and apply the standard decomposition
$A_{\mu}(X)=\varphi_{\perp}(z)A_{\mu}(x)$ (124)
(with normalizable $\varphi_{\perp}$ implying that at $z\to\pm\infty$). Then
we note that the magnetic field inside the wall (parallel to the wall) always
involves a derivative with over $z$. The total magnetic flux inside the wall
thus contains $\int dz\partial_{z}\varphi_{\perp}$ and obviously vanishes.
The mode decomposition (124) is suitable for massive vector fields on the
wall. But there is no index theorem that can protect any zero energy solutions
of equation (121).
What is the actual procedure leading to the gauge field localization? We must
use a global U(1) symmetry (exact in the two-flavor case or approximate in the
one flavor case). Spontaneous breaking of this symmetry localizes, through the
Goldstone theorem, a phase field on the wall. Dualization of this field gives
rise to QED on the wall world volume, with electric field directed along the
wall. At the same time, the flux inside the wall (parallel to the wall) is
that of the magnetic field of the original bulk theory.
The necessity of dualization explains why in five- (and higher dimensions) it
is so hard to localize gauge fields on the 1+3-dimensional wall within a
field-theoretic framework. (By field theory we mean here something without
gravity, or at least where gravity does not play essential role in the
mechanism of localization.) Such mechanisms could be of enormous
phenomenological interest. Ideas as to how one could address this problem can
be found in the literature, see e.g. [3, 30, 31].
The construction discussed in the present paper cannot be uplifted to higher
dimensions. Let us discuss in more detail what happens if we just lift our
four-dimensional models to five dimensions. In four-dimensional theory, what
is localized on the 1+2-dimensional wall is a gauge boson
${F}_{ij}=\epsilon_{ijk}\partial_{k}\sigma\,.$ (125)
where ${F}_{ij}=\partial_{[i}\widetilde{A}_{j]}$, and $\widetilde{A}_{j}$ is
the localized gauge potential.
As we uplift the model to, say, five dimensions, we observe that what is
localized here is not a gauge field but a $2$-form, or the Kalb–Ramond filed
$\widetilde{H}_{\mu\nu}$. Indeed, now
${C}_{\mu\nu\rho}=\epsilon_{\mu\nu\rho\sigma}\partial_{\sigma}\sigma\,,\qquad{C}_{\mu\nu\rho}=\partial_{[\mu}\widetilde{H}_{\nu\rho]}\,.$
(126)
To interpret this result let us ask ourselves what are the sources charged
under this Kalb–Ramond field. We know that $2$-forms couple naturally to
strings, or $1$-branes, much in the same way $1$-forms couple to particles. We
have, in fact, a natural candidate for this string. This is the world sheet
spanned by the uplifted ANO vortices ending on the domain wall. The ANO
vortices in five dimensions are $2$-branes. This is because their codimension
is fixed to be $2$ from the homotopy $\pi_{1}({\rm U}(1))$. These ANO vortices
($2$-branes) end on the domain walls ($3$-branes), and the intersection
between them is exactly the string world sheet, the source of the localized
Kalb–Ramond field.
## 7 Conclusions
Summarizing, we conclude that in certain models a string (flux tube) inside
the domain wall is possible. These are the models which we called minimal: in
particular, those in which the domain wall interpolates between two vacua
where one and the same field condenses. The condensate is exponentially
suppressed in the center of the domain wall. A string in the bulk parallel to
the domain wall is attracted to the wall. The lowest-energy configuration is
achieved when the zeroes of the field are at $z=0$. Figure 17 illustrates this
example.
Figure 17: The energy of a string parallel to the wall as a function of the
distance in the one-flavor model of Sect. 2 (the minimal model). The string on
one side of the wall is the same as the string on the other side. The matter
field condensate becomes smaller in the center of the wall; that’s why the
tension of the string also reaches a minimum at this point. One finds a bound
state of string inside the wall.
As a result, the U(1) filed trapped on the wall is in the confinement regime.
We compared this regime with that of the two-flavor model. The lowest-energy
configuration here is achieved when the zeroes are at $z=-\infty$ (the mirror
reflected solution $z\leftrightarrow-z$ is also possible). Now, the Coulomb
phase on the wall is energetically preferred to the confining phase (see Fig.
18).
Figure 18: In the two-flavor model discussed in Sect. 4, we consider a domain
wall and a set of parallel flux tubes with density $f$ placed at a certain
distance $z$ form the wall. At large positive $z$ the tension of this
configuration is given by the sum of the tensions of the wall and the grid of
the flux tubes. The energy minimum is reached when the distance (defined as
the position of the zeroes) is $-\infty$. Then the configuration looks as the
domain wall with a constant magnetic field on it.
Underlying dynamics is not so transparent as in the minimal model, with the
“confinement on the wall” regime. Technically the distinction is due to the
fact that in the two-flavor model $q_{1}$ vanishes on one side of the wall and
$q_{2}$ on the other, so there is no relative phase between, say, $q_{1}$ on
the left of the wall and $q_{1}$ on the right of the wall. There is only a
phase between $q_{1}$ on the left and $q_{2}$ on the right, and it is not $z$
dependent.
This technical argument is backed up by an iron-clad symmetry argument. The
two-flavor model has a global U(1) in the bulk, which is spontaneously broken
on the wall. The Goldstone boson of this breaking is the $\sigma$ field. It is
strictly massless. On the other hand, one can trace its connection to the
photon of the bulk theory which is Higgsed in the bulk [4].
The mechanisms we discussed in this paper are quite general. We presented two
working examples: a toy minimal model in Sect. 3 and a strong coupling example
in Sect. 4. In both examples there is a condensate which does not vanish in
two distinct vacua separated by the domain wall, and a residual (suppressed)
condensate in the middle of the wall. The quanta of the fields which condense
in both vacua are very heavy inside the wall; hence, one can view the residual
condensate as a “tunneling effect” [1].
## Acknowledgments
We want to thank G. Dvali for correspondence and useful discussions. R.A. is
grateful to FTPI for their hospitality in February 2008 when a part of this
work was done. S.B. wants to thank Ki-Myeong Lee and people at KIAS for
hospitality extended to him in June 2008 and for interesting discussions. This
paper was presented by S.B. at the conference CAQCD-08 in Minneapolis in May
2008.
This work is supported by DOE grant DE-FG02-94ER40823. The work of M.S. was
supported in part by Chaire Internationalle de Recherche Blaise Pascal de
l’Etat et de la Régoin d’Ile-de-France, gérée par la Fondation de l’Ecole
Normale Supérieure. The work of A.Y. was supported by FTPI, University of
Minnesota, by RFBR Grant No. 09-02-00475a, and by Russian State Grant for
Scientific Schools RSGSS-11242003.2.
## Appendix A. Vacuum structure
This appendix is devoted to exact computation of the vacuum structure of the
theory discussed in Sect. 4 using the technique developed in [19]. The U(2)
$\mathcal{N}=2$ gauge theory, without hypermultiplets, has the following
Seiberg–Witten curve
$\displaystyle{y}^{2}$ $\displaystyle=$
$\displaystyle(z-\phi_{1})^{2}(z-\phi_{2})^{2}-\Lambda^{4}$ (A.1)
$\displaystyle=$
$\displaystyle(z^{2}-u_{1}z+\frac{u_{1}^{2}}{2}-u_{2})^{2}-\Lambda^{4}\,.$
The moduli space consists of the Coulomb branch parametrized by $u_{1}={\rm
Tr}\phi$ and $u_{2}=\frac{1}{2}{\rm Tr}\phi^{2}$. The four roots of the curve
are
$\displaystyle z^{-}_{1,2}$ $\displaystyle=$
$\displaystyle\frac{u_{1}}{2}\pm\sqrt{-\frac{u_{1}^{2}}{4}+u_{2}+\Lambda^{2}}\
,$ $\displaystyle z^{+}_{1,2}$ $\displaystyle=$
$\displaystyle\frac{u_{1}}{2}\pm\sqrt{-\frac{u_{1}^{2}}{4}+u_{2}-\Lambda^{2})}\,.$
(A.2)
Since we have no hypermultiplets the singularity structure is very simple and
consists just of the monopole and dyon singularities (co-dimension $2$
surfaces) without intersections
$u_{2}=\frac{u_{1}^{2}}{4}\pm\Lambda^{2}=0$ (A.3)
Now, let us break extended supersymmetry down to ${\cal N}=1$ by virtue of the
superpotential
$W=\alpha\,{\rm Tr}\,\left(\frac{\Phi^{3}}{3}-\xi\Phi\right).$ (A.4)
Classically, we have the three vacua (36). The quantum solution goes as
follows. The Seiberg–Witten curve factorizes,
$y^{2}=P_{N}^{2}-\Lambda^{2N}=F_{2n}H_{N-n}^{2}\,,$ (A.5)
where $n$ is the number of the unbroken U(1) factors in the low-energy theory.
This factorization is then related to the superpotential parameters,
$y_{m}^{2}=W_{k}^{\prime 2}+f_{k-1}=F_{2n}\widetilde{Q}_{k-n}^{2}$ (A.6)
with
$\widetilde{Q}_{k-n}=V_{k-N}H_{N-n}+Q_{N-n-1}\,.$
The unknown parameters are the coefficients of $f_{k-1}$, the coefficients of
$V_{k-N}$ and $Q_{N-n-1}$, and, finally $n$ parameters of the U$(1)^{n}$
Coulomb moduli space. In total $k+n+(k-N)+(N-n-1)=2k-1$, exactly the number of
equations from the polynomial equality (A.6).
In the case of interest $N=2$, $k=2$ and $n=1,2$ depending on whether we deal
with the confining vacua or the Coulomb one. The solutions are as follows.
(i) Monopole
We must lie in the monopole singularity
$u_{2}=\frac{u_{1}^{2}}{4}+\Lambda^{2}=0\,.$
On this surface the factorization equation is
$y_{m}^{2}=(z^{2}-\xi)^{2}+f_{1}=\left(z^{2}+-u_{1}z+\frac{u_{1}^{2}}{4}-\Lambda^{2}\right)\left(z-\frac{u_{1}}{2}-a\right)^{2}\,.$
(A.7)
The solution is
$f_{1}=\mp
4\Lambda^{2}\sqrt{\xi-\Lambda^{2}}z+4\Lambda^{2}\xi+3\Lambda^{4}\,\,\,{\rm
and}\,\,\,a=u_{1}=\pm 2\sqrt{\xi-\Lambda^{2}}\,,$ (A.8)
where $\pm$ correspond to two classical non-Abelian vacua in Eq. (36).
(ii) Coulomb
The factorization in this case is very simple
$y_{m}^{2}=(z^{2}-\xi)^{2}+f_{1}=P_{2}^{2}-\Lambda^{4}\,.$ (A.9)
The solution is $u_{1}=0$, $u_{2}=\xi$ and $f_{1}=-\Lambda^{4}$. There is no
change from the classical formula.
(iii) Dyon
We proceed in the same as above in the monopole case, but now we lie in the
singularity
$u_{2}=\frac{u_{1}^{2}}{4}-\Lambda^{2}=0\,.$
The factorization is
$y_{m}^{2}=(z^{2}-\xi)^{2}+f_{1}=(z^{2}+-u_{1}z+\frac{u_{1}^{2}}{4}+2\Lambda^{2})(z-\frac{u_{1}}{2}-a)^{2}\,.$
(A.10)
The solution is
$f_{1}=\pm
4\Lambda^{2}\sqrt{\xi+\Lambda^{2}}z-4\Lambda^{2}\xi+3\Lambda^{4}\,\,\,{\rm
and}\,\,\,a=u_{1}=\pm 2\sqrt{\xi+\Lambda^{2}}\,.$ (A.11)
Summarizing, we have five vacua, in total. One is the Coulomb vacuum, whose
position in the moduli space is not modified by quantum corrections. Four
others are two monopole and two dyon vacua. Their $u_{2}$ coordinate is not
modified by quantum correction, but the $u_{1}$ coordinate is changed. This is
why they are aligned in Fig. 6. At the critical value $\xi=\Lambda^{2}$ the
two monopole vacua and the Coulomb vacuum coalesce together. In Sect. 4 we
performed our analysis near this critical value, in order to have a low-energy
effective action which is weakly coupled on the domain wall profile. Another
critical value is at $\xi=-\Lambda^{2}$ were the dyon vacua coalesce with the
Coulomb one.
## Appendix B. Comments on supercharges
In this appendix we will show, basing on the central charges, that in
$\mathcal{N}=2$ SQED configurations with flux tubes parallel to the domain
wall are not BPS saturated. We will follow the formalism of Ref. [4].
The supersymmetry transformations in $\mathcal{N}=2$ SQED are given by the
following expressions (where $f,p=1,2$ are SU$(2)_{R}$ indices):
$\displaystyle\delta\lambda^{f\alpha}=\frac{1}{2}(\sigma_{\mu}\bar{\sigma}_{\nu}\varepsilon^{f})^{\alpha}F_{\mu\nu}+\varepsilon^{\alpha
p}D^{a}(\tau^{a})^{f}_{p}+i\sqrt{2}\partial/^{\,\,\alpha\dot{\alpha}}\,a\,\bar{\varepsilon}^{f}_{\dot{\alpha}}\
,$ $\displaystyle\delta\psi^{\alpha A}\ =\ i\sqrt{2}\
\nabla/^{\,\,\alpha\dot{\alpha}}q^{fA}\bar{\varepsilon}_{f\dot{\alpha}}+\sqrt{2}\varepsilon^{\alpha
f}F_{f}^{A}\ ,$ $\displaystyle\delta\tilde{\psi}^{\alpha}_{A}\ =\ i\sqrt{2}\
\nabla/^{\,\,\alpha\dot{\alpha}}\bar{q}^{f}_{A}\bar{\varepsilon}_{f\dot{\alpha}}+\sqrt{2}\varepsilon^{\alpha
f}\bar{F}_{fA}\ ,$ (B.1)
$D^{a}$ is the SU(2)R triplet of $D$ terms:
$D^{1}=i\frac{g^{2}}{2}\left(|\varphi^{A}|^{2}-\xi\right),\;D^{2}=D^{3}=0\,,$
(B.2)
while $F^{f}$ and $\bar{F}_{f}$ are the matter $F$ terms,
$F^{fA}=i\frac{1}{\sqrt{2}}\left(a+\sqrt{2}m_{A}\right)q^{fA},\qquad\bar{F}_{Af}=i\frac{1}{\sqrt{2}}\left(\bar{a}+\sqrt{2}m_{A}\right)\bar{q}_{Af}.$
(B.3)
Let us consider a BPS domain wall, oriented in the $(x,y)$ plane. The
following supersymmetry transformations are left unbroken by the domain wall:
$\displaystyle\bar{\varepsilon}^{2}_{\dot{2}}$ $\displaystyle=$
$\displaystyle-i\varepsilon^{21},\quad\bar{\varepsilon}^{1}_{\dot{2}}=-i\varepsilon^{22},$
$\displaystyle\bar{\varepsilon}^{1}_{\dot{1}}$ $\displaystyle=$ $\displaystyle
i\varepsilon^{12},\qquad\bar{\varepsilon}^{2}_{\dot{1}}=i\varepsilon^{11}.$
(B.4)
Let us consider a vortex parallel to the $z$ axis. The following
supertransformations are left unbroken:
$\displaystyle\varepsilon^{12}=-\varepsilon^{11}\,,\quad\bar{\varepsilon}_{\dot{1}}^{2}=-\bar{\varepsilon}_{\dot{1}}^{1}\,,$
$\displaystyle\varepsilon^{21}=\varepsilon^{22}\,,\qquad\bar{\varepsilon}_{\dot{2}}^{1}=\bar{\varepsilon}_{\dot{2}}^{2}\,.$
(B.5)
It is possible then to find a $1/4$-BPS soliton corresponding to a vortex
perpendicular to the wall, because Eqs. (B.4,B.5) are compatible in the sense
that we can solve both constraints taking
$\displaystyle\varepsilon^{12}=-\varepsilon^{11}\ ,$
$\displaystyle\varepsilon^{21}=\varepsilon^{22},$ (B.6)
and the $\bar{\epsilon}$ given by Eq. (B.4).
Let us then consider a vortex parallel to the $y$ axis. From a simple spinor
rotation, we can find supertransformations left unbroken,
$\displaystyle\varepsilon^{12}=-i\varepsilon^{21}\,,\quad\bar{\varepsilon}_{\dot{1}}^{2}=-i\bar{\varepsilon}_{\dot{2}}^{1}\,,$
$\displaystyle\varepsilon^{11}=-i\varepsilon^{22}\,,\qquad\bar{\varepsilon}_{\dot{1}}^{1}=-i\bar{\varepsilon}_{\dot{2}}^{2}\,.$
(B.7)
It is easy to check that it is not possible to find a non-trivial solution to
the constraints in Eqs. (B.4) and (B.7). This shows that the configuration
with the flux tube parallel to the domain wall breaks all supersymmetries of
the theory.
Let us then consider a domain wall in the $(x,y)$ plane with some constant
magnetic field along the $x$ axis. As discussed in Figure 16, the unbroken
supercharges are the same of the system of a vortex ending on a wall (modulo a
rotation by the angle in Eq. (117) along the $y$ axis).
## References
* [1] G. R. Dvali and M. A. Shifman, Phys. Lett. B 396, 64 (1997) [Erratum-ibid. B 407, 452 (1997)] [arXiv:hep-th/9612128].
* [2] B. S. Acharya and C. Vafa, arXiv:hep-th/0103011.
* [3] S. L. Dubovsky and V. A. Rubakov, Int. J. Mod. Phys. A 16, 4331 (2001) [arXiv:hep-th/0105243].
* [4] M. Shifman and A. Yung, Phys. Rev. D 67, 125007 (2003) [arXiv:hep-th/0212293].
* [5] R. Auzzi, M. Shifman and A. Yung, Phys. Rev. D 74, 045007 (2006) [arXiv:hep-th/0606060].
* [6] G. Dvali, H. B. Nielsen and N. Tetradis, Phys. Rev. D 77, 085005 (2008) [arXiv:0710.5051 [hep-th]].
* [7] A. M. Polyakov, Nucl. Phys. B 120, 429 (1977).
* [8] M. Shifman and A. Yung, Rev. Mod. Phys. 79, 1139 (2007) [arXiv:hep-th/0703267].
* [9] D. Tong, arXiv:hep-th/0509216.
* [10] M. Eto, Y. Isozumi, M. Nitta, K. Ohashi and N. Sakai, J. Phys. A 39 (2006) R315 [arXiv:hep-th/0602170].
* [11] G. ’t Hooft, Nucl. Phys. B 79, 276 (1974); A. M. Polyakov, JETP Lett. 20, 194 (1974) [reprinted in Solitons and Particles, Eds. C. Rebbi and G. Soliani, (World Scientific, Singapore, 1985), p. 522].
* [12] A. Abrikosov, Sov. Phys. JETP 32 1442 (1957) [Reprinted in Solitons and Particles, Eds. C. Rebbi and G. Soliani (World Scientific, Singapore, 1984), p. 356]; H. Nielsen and P. Olesen, Nucl. Phys. B61 45 (1973) [Reprinted in Solitons and Particles, Eds. C. Rebbi and G. Soliani (World Scientific, Singapore, 1984), p. 365].
* [13] N. Seiberg and E. Witten, Nucl. Phys. B 426, 19 (1994) [Erratum-ibid. B 430, 485 (1994)] [arXiv:hep-th/9407087].
* [14] R. Jackiw and C. Rebbi, Phys. Rev. D 13, 3398 (1976).
* [15] J. E. Kiskis, Phys. Rev. D 15, 2329 (1977); M. M. Ansourian, Phys. Lett. B 70, 301 (1977); N. K. Nielsen and B. Schroer, Nucl. Phys. B 120, 62 (1977); R. Jackiw and P. Rossi, Nucl. Phys. B 190, 681 (1981); E. J. Weinberg, Phys. Rev. D 24, 2669 (1981).
* [16] A. Gorsky, M. Shifman and A. Yung, Phys. Rev. D 75, 065032 (2007) [arXiv:hep-th/0701040].
* [17] I. I. Balitsky and A. V. Yung, Phys. Lett. B 168, 113 (1986).
* [18] F. Cachazo, M. R. Douglas, N. Seiberg and E. Witten, JHEP 0212 (2002) 071 [arXiv:hep-th/0211170].
* [19] F. Cachazo, N. Seiberg and E. Witten, JHEP 0302, 042 (2003) [arXiv:hep-th/0301006];
* [20] V. S. Kaplunovsky, J. Sonnenschein and S. Yankielowicz, Nucl. Phys. B 552 (1999) 209 [arXiv:hep-th/9811195].
* [21] R. Portugues and P. K. Townsend, Phys. Lett. B 530, 227 (2002) [arXiv:hep-th/0112077].
* [22] B. Chibisov and M. A. Shifman, Phys. Rev. D 56 (1997) 7990 [Erratum-ibid. D 58 (1998) 109901] [arXiv:hep-th/9706141].
* [23] J. P. Gauntlett, D. Tong and P. K. Townsend, Phys. Rev. D 64, 025010 (2001) [hep-th/0012178].
* [24] D. Tong, Phys. Rev. D 66, 025013 (2002) [hep-th/0202012].
* [25] J. P. Gauntlett, R. Portugues, D. Tong and P. K. Townsend, Phys. Rev. D 63, 085002 (2001) [arXiv:hep-th/0008221].
* [26] Y. Isozumi, M. Nitta, K. Ohashi and N. Sakai, Phys. Rev. D 71, 065018 (2005) [arXiv:hep-th/0405129].
* [27] S. Bolognesi and M. Shifman, Phys. Rev. D 76, 125024 (2007) [arXiv:0705.0379 [hep-th]].
* [28] E. Bogomol’nyi, The stability of classical solutions, Sov. J. Nucl. Phys. 24, 449 (1976) [Reprinted in Solitons and Particles, Eds. C. Rebbi and G. Soliani (World Scientific, Singapore, 1984), p. 389].
* [29] E. R. C. Abraham and P. K. Townsend, Phys. Lett. B 291 (1992) 85; E. R. C. Abraham and P. K. Townsend, Phys. Lett. B 295 (1992) 225.
* [30] S. L. Dubovsky, V. A. Rubakov and S. M. Sibiryakov, JHEP 0201, 037 (2002) [arXiv:hep-th/0201025].
* [31] Y. Isozumi, K. Ohashi and N. Sakai, JHEP 0311, 061 (2003) [arXiv:hep-th/0310130]; JHEP 0311, 060 (2003) [arXiv:hep-th/0310189].
| arxiv-papers | 2008-07-11T19:02:16 | 2024-09-04T02:48:56.727159 | {
"license": "Public Domain",
"authors": "R. Auzzi, S. Bolognesi, M. Shifman and A. Yung",
"submitter": "Stefano Bolognesi",
"url": "https://arxiv.org/abs/0807.1908"
} |
0807.1951 | # Phenomenological models of baryon
Do Quoc Tuan
Dept. of Theoretical Physics, College of Science, VNU, Hanoi, Vietnam
Abstract
In this paper, I present almost my works performed during my time at VNU. I
was interested in the composite Skyrme model proposed by H. Y. Cheung and F.
Gursey. The expanding for this model based on results obtained from the
original Skyrme model given by T. H. R. Skyrme in 1961 was general and
interesting.
Acknowledgments
I want to show deep gratitude to my supervisor Dr. Pham Thuc Tuyen (Department
of Theoretical Physics, College of Science (another name is Hanoi University
of Science), Vietnam National University) because of everything that he has
been teaching and supporting me during a student’s time! I also thank
Department of Theoretical Physics and Faculty of Physics for giving me the
best conditions to completing the thesis! Finally, and most of all, I want to
give my thesis to my parent, my beloved as a special gift for the supports
over a long period!
Ha noi, 6/2007
Do Quoc Tuan111email: do.tocxoan@gmail.com
###### Contents
1. 1 Why is the Skyrme model
1. 1.1 The weak point of QCD
2. 1.2 The idea of Skyrme
2. 2 The massless composite Skyrme model
1. 2.1 The formalism
2. 2.2 Calculations for static quantities of baryon
3. 3 The composite Skyrme model with pion’s mass
1. 3.1 The term of pion’mass
2. 3.2 The field equation and numerical results
4. 4 The supersymmetric composite Skyrme model
5. 5 Appendix
1. 5.1 Some notations for Chapter 2 and 3
2. 5.2 Some notations for Chapter 4
## Chapter 1 Why is the Skyrme model
### 1.1 The weak point of QCD
In physically, we can divide the nature in two types of particles: matter
particles and field particles. Matter particles interact together by fields
concluding transfer interaction particles. There are four fields corresponding
four interactions (four forces) and their particles. The graviton is a
transfer interaction particle of gravitational force, the photon is a transfer
interaction particle of electromagnetic force, $W^{\pm}$ and $Z^{o}$ are
transfer interaction particle of weak force, the gluon is transfer interaction
particle of strong force. It is known well, quarks are interact together by
the strong force. In other word, the interaction of quarks is the exchange of
gluons. And the theory that describes quarks and their interaction is $QCD$
(Quantum Chromo Dynamics). In mathematically, Gell-Mann and Neuman proposed
that quarks make up a multiple which represents the gauge $SU(3)$ being the
group of $3\times 3$ unitary matrices with determinant 1. Gell-Mann firstly
gives generators of $SU(3)$, therefore, they are called Gell-Mann matrices
$\lambda$ . They satisfy the Lie algebra
$\left[{\lambda_{i},\lambda_{j}}\right]=2if_{ijk}\lambda_{k},$ (1.1)
where $f_{ijk}$ is the structure constant. It is totally antisymmetric
$f_{ijk}=-f_{jik}.$ (1.2)
To describe the interaction between quarks, we need using the gluon field
$A_{\mu}^{\alpha}$ . And now, the Lagrangian density is the Yang-Mills
Lagrangian density
${\cal
L}=-\frac{1}{4}F_{\mu\nu}^{\alpha}F^{\mu\nu\alpha}+i\bar{\psi}\gamma^{\mu}(\partial_{\mu}+igA_{\mu})\psi+M\bar{\psi}\psi,$
(1.3) $A_{\mu}=A_{\mu}^{\alpha}\lambda^{\alpha},$ (1.4)
$F_{\mu\nu}^{\alpha}=\partial_{\mu}A_{\nu}^{\alpha}-\partial_{\nu}A_{\mu}^{\alpha}-gf^{\alpha\beta\gamma}A_{\mu}^{\beta}A_{\nu}^{\gamma},$
(1.5)
where g is a constant of interaction. At the high energy scale ($>TeV$), $QCD$
is very good to use. But, when the energy lower (in $GeV$), it meets with some
problems in which describe quantities of quarks and gluons. The main cause is
the high value of g, it obstructs an expansion of interaction terms of hadron.
To repair this problem, t’Hooft proposed that $1/N_{c}$ plays as g($N_{c}$ is
color number in a gauge theory). And, he gave that in large-$N_{c}$ limit,
$QCD$ becomes equivalent to an effective field theory of meson.
### 1.2 The idea of Skyrme
In 1961, Skyrme gave the new model in which baryons may be obtained from the
theory of meson [1]. He began with the Lagrangian density of pion
${\cal
L}=\frac{1}{2}\partial_{\mu}\vec{\pi}\partial^{\mu}\vec{\pi}-\frac{{m^{2}}}{2}\vec{\pi}\vec{\pi},$
(1.6)
in which $\pi$’s make up a three dimensional isotopic space. After, he
expanded the isotopic $SU\left(2\right)$ becomes a new group with the algebra
$\left[{V_{i},V_{j}}\right]=2i\varepsilon_{ijk}V_{k},$ (1.7)
$\left[{A_{i},A_{j}}\right]=2i\varepsilon_{ijk}V_{k},$ (1.8)
$\left[{V_{i},A_{j}}\right]=2i\varepsilon_{ijk}A_{k},$ (1.9)
where generators $V_{i}$ are vector charges, generators $A_{i}$ are axial
charges. And now, an isotopic space have four components
$\left\\{{\vec{\pi}}\right\\}\to\left\\{{\vec{\varphi},\varphi_{4}}\right\\}$
transforming in a rule
$\left[{V_{i},\varphi_{j}}\right]=2i\varepsilon_{ijk}\varphi_{k},$ (1.10)
$\left[{V_{i},\varphi_{4}}\right]=0,$ (1.11)
$\left[{A_{i},\varphi_{j}}\right]=-2i\varepsilon_{ij}\varphi_{4},$ (1.12)
$\left[{A_{i},\varphi_{4}}\right]=2i\varphi_{i}.$ (1.13)
Particles described by $\varphi_{i}\left({i=1,2,3}\right)$ are pseudo- scalars
and making isotopic vectors. A particle described by $\varphi_{4}$ has a
positive parity, a spin and an isotopic spin equals zero, and it is a vacuum.
If replace $V_{i}$ and $A_{i}$ by $L_{i}$ and $R_{i}$ defined
$L_{i}=\frac{1}{2}\left({V_{i}-A_{i}}\right),$ (1.14)
$R_{i}=\frac{1}{2}\left({V_{i}+A_{i}}\right),$ (1.15)
then obtaining a new algebra
$\left[{L_{i},L_{j}}\right]=\left[{\frac{1}{2}\left({V_{i}-A_{i}}\right),\frac{1}{2}\left({V_{j}-A_{j}}\right)}\right]$
,
$=\frac{1}{4}\left[{V_{i},V_{j}}\right]-\frac{1}{4}\left[{V_{i},A_{j}}\right]-\frac{1}{4}\left[{A_{i},V_{j}}\right]+\frac{1}{4}\left[{A_{i},A_{j}}\right]$
,
$=\frac{1}{2}i\varepsilon_{ijk}V_{k}-\frac{1}{2}i\varepsilon_{ijk}A_{k}+\frac{1}{2}i\varepsilon_{jik}A_{k}+\frac{1}{2}i\varepsilon_{ijk}V_{k}$
,
$=i\varepsilon_{ijk}V_{k}-i\varepsilon_{jik}A_{k}$ ,
$\bf{\Rightarrow\left[{L_{i},L_{j}}\right]=2i\varepsilon_{ijk}L_{k}}$ (1.16)
and
$\left[{R_{i},R_{j}}\right]=\left[{\frac{1}{2}\left({V_{i}+A_{i}}\right),\frac{1}{2}\left({V_{j}+A_{j}}\right)}\right]$
,
$=\frac{1}{4}\left[{V_{i},V_{j}}\right]+\frac{1}{4}\left[{V_{i},A_{j}}\right]+\frac{1}{4}\left[{A_{i},V_{j}}\right]+\frac{1}{4}\left[{A_{i},A_{j}}\right]$
,
$=\frac{1}{2}i\varepsilon_{ijk}V_{k}+\frac{1}{2}i\varepsilon_{ijk}A_{k}-\frac{1}{2}i\varepsilon_{jik}A_{k}+\frac{1}{2}i\varepsilon_{ijk}V_{k}$
,
$=i\varepsilon_{ijk}V_{k}+i\varepsilon_{jik}A_{k}$ ,
$\bf{\Rightarrow\left[{R_{i},R_{j}}\right]=2i\varepsilon_{ijk}R_{k}}$ (1.17)
and
$\left[{L_{i},R_{j}}\right]=\left[{\frac{1}{2}\left({V_{i}-A_{i}}\right),\frac{1}{2}\left({V_{j}+A_{j}}\right)}\right]$
,
$=\frac{1}{4}\left[{V_{i},V_{j}}\right]+\frac{1}{4}\left[{V_{i},A_{j}}\right]-\frac{1}{4}\left[{A_{i},V_{j}}\right]-\frac{1}{4}\left[{A_{i},A_{j}}\right]$
,
$=\frac{1}{2}i\varepsilon_{ijk}V_{k}+\frac{1}{2}i\varepsilon_{ijk}A_{k}+\frac{1}{2}i\varepsilon_{jik}A_{k}-\frac{1}{2}i\varepsilon_{ijk}V_{k}$
,
$\bf{\Rightarrow\left[{L_{i},R_{j}}\right]=0}.$ (1.18)
This is an algebra of the chiral group $SU\left(2\right)\times
SU\left(2\right)$. Thus, the expanding group is the chiral group
$SU\left(2\right)\times SU\left(2\right)$. Continuously, he proposed that
$\left\\{{\vec{\varphi},\varphi_{4}}\right\\}$ form a three dimensional sphere
$\vec{\varphi}^{2}+\varphi_{4}^{2}=1.$ (1.19)
Recently, $\left\\{{\vec{\varphi},\varphi_{4}}\right\\}$ are associated with
$2\times 2$ unitary $U\left({x,t}\right)$ matrix
$U\left({x,t}\right)=\exp i\vec{\tau}\vec{\pi}\left({x,t}\right)$ ,
$=\exp i\vec{\tau}\frac{{\vec{\pi}}}{\pi}\pi$ ,
$=\cos\pi+i\vec{\tau}\frac{{\vec{\pi}}}{\pi}\sin\pi,$ (1.20)
where $\pi=\sqrt{\vec{\pi}^{2}}$, $\tau$’s are Pauli matrices. Pion fields now
are associated with $\vec{\pi}$ fields
$\varphi_{4}\Leftrightarrow\cos\pi,$ (1.21)
$\vec{\varphi}\Leftrightarrow\frac{{\vec{\pi}}}{\pi}\sin\pi.$ (1.22)
Really
$\vec{\varphi}^{2}+\varphi_{4}^{2}=\frac{{\vec{\pi}^{2}}}{{\pi^{2}}}\sin^{2}\pi+\cos^{2}\pi=1.$
(1.23)
The Lagrangian density proposed by Skyrme [1] is
${\cal
L}=\frac{{F_{\pi}^{2}}}{{16}}Tr\left({\partial_{\mu}U\partial_{\mu}U^{\dagger}}\right)+\frac{1}{{32e^{2}}}Tr\left[{\left({\partial_{\mu}U}\right)U^{\dagger},\left({\partial_{\nu}U}\right)U^{\dagger}}\right]^{2},$
(1.24)
here $U$ is an $SU\left(2\right)$ matrix transforming as $U\to AUB^{-1}$ under
an chiral $SU\left(2\right)\times SU\left(2\right)$. He proved that the model
has a soliton solution (called hedgehog111firstly called by Polyakov, see at
L. Ryder, Quantum Field Theory (second edition), Cambridge University Press
solution)
$U_{o}=\exp\left({i\vec{\tau}\frac{{\vec{r}}}{r}F\left(r\right)}\right),$
(1.25)
where $F\left(r\right)$ is the chiral angle, baryons are obtained from the
quantization of soliton solution(see chapter 2).
It is well known that in the large-$N$ limit, $QCD$ may be regarded as the
theory of effective meson fields. E. Witten argued that baryons may be
regarded as solitons of this effective meson theory without an further
reference to their quark content [2]. We know that solitons have not a spin.
But, by quantizing around solitons, we obtain the special topological
construct of these solitons which makes a spin. On the other hand, soliton
solutions have a limited spatial distribution, therefore, they are correlative
to particles that have a limited spatial dimension [2,15].
Figure 1.1: The configuration of hedgehog, arrows on it are isotopic vectors
of $\vec{c}$
## Chapter 2 The massless composite Skyrme model
### 2.1 The formalism
In 1983, G. S. Adkins, C. R. Nappi and E. Witten used the Skyrme’s idea to
computed the static properties of baryons on the chiral group
$SU\left(2\right)\times SU\left(2\right)$ and found that the model gives
predictions that agree well with experiments [2]. However, this model has a
problem in which the situation of particles at low energy are not interested
but it regards that particles in free states, and this point can reduce the
exactitude of theoretical datum. In 1989, this problem is treated by P. T.
Tuyen and N. A. Viet [3]. They considered that at low energy particles are not
free but they would shrink together into the cluster. Therefore, the group
that describes them is $SU\left(2\right)\times
SU\left(2\right)/SU\left(2\right)_{diag}$. This model gives few datum better
than the sigma model. In 1990, H. Y. Cheung, F. Gursey [4] gave the general
model in which the sigma model and the V-T model are two special cases
corresponding with $n=1$ and $n=2$. They introduced a composite _SU(2)_
configuration $V_{n}=UPUP...PU$, where _U(x,t)_ is the _SU(2)_ matrix and
$P=p_{o}+i\vec{\tau}.\vec{p}$ is the constant _SU(2)_ matrix, and they
proposed the composite Skyrme lagrangian density
${\cal
L}_{n}=-\frac{{F_{\pi}^{2}}}{{16n^{2}}}Tr\left({\partial_{\mu}V^{-n}\partial^{\mu}V^{n}}\right)+\frac{1}{{32e^{2}n^{4}}}Tr\left({\left[{V^{-n}\partial_{\mu}V^{n},V^{-n}\partial_{\nu}V^{n}}\right]^{2}}\right),$
(2.1)
with $F_{\pi}$ is the pion decay constant, _e_ is a dimensionless parameter.
Under the chiral group $SU\left(2\right)\times SU\left(2\right)$, _U_
transforms as $U\to AUB^{\dagger}$, _P_ transforms as $P\to BPA^{\dagger}$,
and so $V_{n}$ transforms as $V_{n}\to AV_{n}B^{\dagger}$. In the limit $P\to
I$, $V_{n}\to U_{n}$, the lagrangian (2.1) becomes approximated by the
lagrangian density
${\cal
L}_{n}=-\frac{{F_{\pi}^{2}}}{{16n^{2}}}Tr\left({\partial_{\mu}U^{-n}\partial^{\mu}U^{n}}\right)+\frac{1}{{32e^{2}n^{4}}}Tr\left({\left[{U^{-n}\partial_{\mu}U^{n},U^{-n}\partial_{\nu}U^{n}}\right]^{2}}\right).$
(2.2)
Now, using the hedgehog soliton solution gave by Skyrme
$U_{0}\left(r\right)=\exp\left[{i\tau.\hat{r}F\left(r\right)}\right]$, where
$\tau$’s are Pauli’s matrices and $\hat{r}=\vec{r}/r$, the lagrangian density
(2.2) is
${\cal
L}_{n}=\frac{{e^{2}F_{\pi}^{4}}}{8}\left[{\left({\frac{{dF}}{{d\tilde{r}}}}\right)^{2}+2\frac{{\sin^{2}nF}}{{n^{2}\tilde{r}^{2}}}}\right]+\frac{{e^{2}F_{\pi}^{4}\sin^{2}nF}}{{2n^{2}\tilde{r}^{2}}}\left[{2\left({\frac{{dF}}{{d\tilde{r}}}}\right)^{2}+\frac{{\sin^{2}nF}}{{n^{2}\tilde{r}^{2}}}}\right],$
(2.3)
where $\tilde{r}=eF_{\pi}r$ is the dimensionless variable. And the static
energy of hedgehog is
${\cal E}_{n}=\frac{{F_{\pi}}}{e}A_{n}=M_{n},$ (2.4)
$A_{n}=\int\limits_{0}^{\infty}{4\pi\tilde{r}^{2}\left\\{{\frac{1}{8}\left[{\left({\frac{{dF}}{{d\tilde{r}}}}\right)^{2}+2\frac{{\sin^{2}nF}}{{n^{2}\tilde{r}^{2}}}}\right]+\frac{{e^{2}F_{\pi}^{4}\sin^{2}nF}}{{2n^{2}\tilde{r}^{2}}}\left[{2\left({\frac{{dF}}{{d\tilde{r}}}}\right)^{2}+\frac{{\sin^{2}nF}}{{n^{2}\tilde{r}^{2}}}}\right]}\right\\}}d\tilde{r}.$
(2.5)
We obtain the non-linear differential equation of _F(r)_ from minimum
condition of the hedgehog’s energy
$\delta_{F}{\cal E}_{n}=0,$ (2.6)
or
$\frac{d}{{dr}}\frac{{\delta{\cal E}_{n}}}{{\delta
F^{\prime}}}-\frac{{\delta{\cal E}_{n}}}{{\delta F}}=0,(Euler-Lagrange-
equation)$ (2.7)
$\left({\frac{{\tilde{r}^{2}}}{4}+\frac{2}{{n^{2}}}\sin^{2}nF}\right)F^{\prime\prime}+\frac{{\tilde{r}}}{2}F^{\prime}+\frac{1}{n}\sin
2nF\left({F^{\prime
2}-\frac{{\sin^{2}nF}}{{n^{2}\tilde{r}^{2}}}-\frac{1}{4}}\right)=0.$ (2.8)
To solve this equation we need boundary conditions for the _F(r)_. The first
condition of _F(r)_ that relates to the limit of hedgehog’s energy when
$r\to\infty$ is $F\left(\infty\right)=0$. The second condition relating to the
exist of baryonic number B is $F\left(0\right)=\pi$. This equation can not
give us the analytic solution. So we will solve it by the numerical method.
Figure 2.1: The numerical solution of (2.8) for cases of n=2 and n=4 [22, 25].
Now we will quantize the soliton by collective coordinates
$A\left(t\right)=a_{o}\left(t\right)+i\vec{a}\left(t\right)\vec{\tau}$ with
$a_{\mu}\left(t\right)$ $\left({\mu=0,1,2,3}\right)$ are collective
coordinates, $A\left(t\right)$ is $SU(2)$ matrix. From properties of $SU(2)$
matrix, we have
$A\left(t\right)A^{\dagger}\left(t\right)=1$ ,
$\left\\{{a_{o}\left(t\right)+i\vec{a}\left(t\right)\vec{\tau}}\right\\}\left\\{{a_{o}\left(t\right)-i\vec{a}\left(t\right)\vec{\tau}}\right\\}=1$
,
$\bf{a_{\mu}a_{\mu}=1}.$ (2.9)
By time derivative, resulting
$\bf{a_{\mu}\dot{a}_{\mu}=0}$ (2.10)
By this way, solitons to be particles. Different quantization take different
particles. Thus, it takes us to the new idea about the origin of particles in
nature in which they are made by a basic particle (called soliton). With the
_A(t)_ matrix, the $U^{n}\left({x,t}\right)$ will transform as
$U^{n}\left({x,t}\right)=A\left(t\right)U_{0}^{n}\left(x\right)A^{-1}\left(t\right)$
, and we put it into (2.2), the lagrangian now is
$L_{n}=-{\cal
E}_{n}+\Gamma_{n}Tr\left[{\partial_{0}A\left(t\right)\partial_{0}A^{-1}\left(t\right)}\right]$
,
$=-{\cal E}_{n}+2\Gamma_{n}\sum\limits_{\mu=0}^{3}{\dot{a}_{\mu}^{2}},$ (2.11)
with
$\Gamma_{n}=\frac{1}{{e^{3}F_{\pi}}}\Phi_{n},$ (2.12)
$\Phi_{n}=\frac{{2\pi}}{{3n^{2}}}\int\limits_{0}^{\infty}{\tilde{r}^{2}\sin^{2}nF\left\\{{1+4\left[{\left({\frac{{dF}}{{d\tilde{r}}}}\right)^{2}+\frac{{\sin^{2}nF}}{{n^{2}\tilde{r}^{2}}}}\right]}\right\\}}d\tilde{r}.$
(2.13)
The conjugate momenta are obtained from the derivative of lagrangian following
$\dot{a}_{\mu}(t)$
$\pi_{\mu}=\frac{{\partial L_{n}}}{{\partial\dot{a}_{\mu}}}$ ,
$=\frac{{\partial\left[{-M_{n}+2\Gamma_{n}\sum\limits_{\mu=0}^{3}{\dot{a}_{\mu}^{2}}}\right]}}{{\partial\dot{a}_{\mu}}}$
,
$=4\Gamma_{n}\dot{a}_{\mu}.$ (2.14)
We go to the hamiltonian
$H_{n}=\pi_{\mu}\dot{a}_{\mu}-L_{n}$ ,
$=4\Gamma_{n}\dot{a}_{\mu}\dot{a}_{\mu}-L_{n}$ ,
$=M_{n}+2\Gamma_{n}\dot{a}_{\mu}\dot{a}_{\mu}$ ,
$=M_{n}+\frac{{\left({\sum\limits_{\mu=0}^{3}{\pi_{\mu}^{2}}}\right)}}{{8\Gamma_{n}}}.$
(2.15)
Canonical quantizing momenta $\pi_{\mu}=-i\frac{\partial}{{\partial a_{\mu}}}$
let the hamiltonian as
$H_{n}=M_{n}+\frac{1}{{8\Gamma_{n}}}\sum\limits_{\mu=0}^{3}{\left({\frac{{-\partial^{2}}}{{\partial
a_{\mu}^{2}}}}\right)}.$ (2.16)
Now, we will define the isotopic vector $\vec{c}$ following the way
$\dot{A}^{\dagger}A=\left({\dot{a}_{0}-i\dot{a}_{i}\tau_{i}}\right)\left({a_{0}+ia_{i}\tau_{i}}\right)$
,
$=\dot{a}_{0}a_{0}+\dot{\vec{a}}\vec{a}-i\left({a_{0}\dot{\vec{a}}-\dot{a}_{0}\vec{a}+\vec{a}\times\dot{\vec{a}}}\right)\vec{\tau}$
,
note (2.10), we have
$\dot{A}^{\dagger}A=-i\vec{c}\vec{\tau},$ (2.17)
with
$\vec{c}=\left({a_{0}\dot{\vec{a}}-\dot{a}_{0}\vec{a}+\vec{a}\times\dot{\vec{a}}}\right).$
(2.18)
The square isotopic vector $\vec{c}$ is
$\vec{c}^{2}=\dot{a}_{0}^{2}\vec{a}^{2}+\dot{\vec{a}}^{2}a_{0}^{2}+\vec{a}^{2}\dot{\vec{a}}^{2}-(\vec{a}\dot{\vec{a}})^{2}-2\dot{a}_{0}a_{0}\dot{\vec{a}}\vec{a}$
,
$=\dot{a}_{0}^{2}\vec{a}^{2}+\dot{\vec{a}}^{2}a_{0}^{2}+\vec{a}^{2}\dot{\vec{a}}^{2}-(\vec{a}\dot{\vec{a}})^{2}+2(\vec{a}\dot{\vec{a}})^{2}$
,
$=\dot{a}_{\mu}\dot{a}_{\mu}.$ (2.19)
In the configuration of hedgehog, since the isotopic vector and the radius
vector have the same direction, rotation in the co-ordinate space can make an
isospin of particle, and we assume that a spin equals to an isospin. The
canonical momenta that correspond with collective coordinates are isospins. We
can homogenize $\vec{c}^{2}$ with the square of a spin and an isospin
$\vec{c}^{2}=\frac{1}{{16\Gamma_{n}^{2}}}\vec{J}^{2}=\frac{1}{{16\Gamma_{n}^{2}}}\vec{T}^{2}.$
(2.20)
Thus, we have
$H_{n}=M_{n}+\frac{1}{{8\Gamma_{n}}}\vec{J}^{2}=M_{n}+\frac{1}{{8\Gamma_{n}}}\vec{T}^{2},$
(2.21)
with $\vec{J}$ is a spin and $\vec{T}$ is an isospin. Eigenvalues of
hamiltonian (2.21) are
$E_{n}=M_{n}+\frac{{l\left({l+2}\right)}}{{8\Gamma_{n}}},$ (2.22)
with _l=2J_ and _J_ is the quantum number of spin. So, the nucleon
$\left({J=1/2}\right)$ and the delta $\left({J=3/2}\right)$ masses are given
by
$M_{N}=M_{n}+\frac{3}{{8\Gamma_{n}}},$ (2.23)
$M_{\Delta}=M_{n}+\frac{{15}}{{8\Gamma_{n}}}.$ (2.24)
Note (2.12), (2.13), (2.23), (2.24) we have
$F_{\pi}=\sqrt[4]{{\frac{{\left({5M_{N}-M_{\Delta}}\right)^{3}\left({M_{\Delta}-M_{N}}\right)\Phi_{n}}}{{384A_{n}^{3}}}}},$
(2.25) $e=\frac{{4A_{n}F_{\pi}}}{{\left({5M_{N}-M_{\Delta}}\right)}}.$ (2.26)
### 2.2 Calculations for static quantities of baryon
After constructing the formalism of massless composite model, we apply to the
baryon (nucleon). Since this model is constructed depend on some ideas of
theory of topology then we can start at the topological current to
calculations for static quantities of baryon [2, 3, 4, 15]. The topological
current (a baryonic current) is defined
$B^{\mu}=\frac{{\varepsilon^{\mu\nu\alpha\beta}}}{{24\pi^{2}n}}Tr\left[{\left({U^{{\dagger}n}\partial_{\nu}U^{n}}\right)\left({U^{{\dagger}n}\partial_{\alpha}U^{n}}\right)\left({U^{{\dagger}n}\partial_{\beta}U^{n}}\right)}\right],$
(2.27)
with $\varepsilon_{0123}=-\varepsilon^{0123}=1$ (totally antisymmetry). If we
put
$U^{n}\left({x,t}\right)=A\left(t\right)U_{0}^{n}\left(x\right)A^{-1}\left(t\right)$
into (2.27), we will have
$B^{0}=-\frac{1}{{2\pi^{2}}}\frac{{\sin^{2}nF}}{{r^{2}}}F^{\prime},$ (2.28)
$B^{i}=\frac{{i\varepsilon^{{\rm{ijk}}}}}{{2\pi^{2}}}\frac{{\sin^{2}nF}}{{r^{2}}}F^{\prime}x_{k}Tr\left[{\dot{A}^{\dagger}A\tau_{j}}\right].$
(2.29)
Note (2.17) and (2.20), then (2.29) is now
$B^{i}=\frac{{\varepsilon^{{\rm{ijk}}}}}{{2\pi^{2}}}\frac{{\sin^{2}nF}}{{r^{2}}}F^{\prime}x_{k}\frac{{J_{j}}}{{2\Gamma_{n}}}.$
(2.30)
The baryonic number is
$B=\int\limits_{0}^{\infty}{4\pi r^{2}}B^{0}\left(r\right)dr=1.$ (2.31)
The iso-scalar electric square radius is
$\left\langle{r^{2}}\right\rangle_{E,I=0}=\int\limits_{0}^{\infty}{4\pi
r^{2}}B^{0}\left(r\right)r^{2}dr.$ (2.32)
The iso-scalar magnetic square radius is
$\left\langle{r^{2}}\right\rangle_{M,I=0}=\frac{3}{5}\frac{{\int\limits_{0}^{\infty}{4\pi
r^{2}B^{0}\left(r\right)r^{4}}}}{{\left\langle{r^{2}}\right\rangle_{E,I=0}}}.$
(2.33)
We have the time’s component of vector current $\vec{V}$
$V_{i}^{0}=\frac{i}{3}\frac{{\sin^{2}nF}}{{n^{2}r^{2}}}\left({F_{\pi}^{2}+\frac{4}{{e^{2}}}\left[{\left({\frac{{dF}}{{dr}}}\right)^{2}+\frac{{\sin^{2}nF}}{{n^{2}r^{2}}}}\right]}\right)Tr\left({\dot{A}A^{\dagger}\tau_{i}}\right),$
(2.34)
note (2.17)
$V_{i}^{0}=\frac{{\sin^{2}nF}}{{3n^{2}r^{2}}}\left({F_{\pi}^{2}+\frac{4}{{e^{2}}}\left[{\left({\frac{{dF}}{{dr}}}\right)^{2}+\frac{{\sin^{2}nF}}{{n^{2}r^{2}}}}\right]}\right)\frac{{T_{i}}}{{2\Gamma_{n}}}.$
(2.35)
The density of charge of nucleon is defined by Gell-Mann - Nishijima
$Q=\int\limits_{0}^{\infty}{\rho_{nucleon}}\left(r\right)dr=I_{3}+\frac{1}{2}B,$
(2.36)
and
$I_{3}=\int\limits_{0}^{\infty}{r^{2}}V_{3}^{0}d\Omega
dr=\int\limits_{0}^{\infty}{\frac{{4\pi\sin^{2}nF}}{{3n^{2}}}\left({F_{\pi}^{2}+\frac{4}{{e^{2}}}\left[{\left({\frac{{dF}}{{dr}}}\right)^{2}+\frac{{\sin^{2}nF}}{{n^{2}r^{2}}}}\right]}\right)\frac{{T_{3}}}{{2\Gamma_{n}}}}dr.$
(2.37)
So, we have
$\rho_{nu}\left(r\right)=\frac{{4\pi\sin^{2}nF}}{{3n^{2}}}\left({F_{\pi}^{2}+\frac{4}{{e^{2}}}\left[{\left({\frac{{dF}}{{dr}}}\right)^{2}+\frac{{\sin^{2}nF}}{{n^{2}r^{2}}}}\right]}\right)\frac{{T_{3}}}{{2\Gamma_{n}}}-\frac{{\sin^{2}nF}}{\pi}F^{\prime}.$
(2.38)
With $\bf{T_{3}=\frac{1}{2}}$ and $\bf{T_{3}=-\frac{1}{2}}$, we have the
charge’s densities of proton and neutron.
After integrating, we obtain charges of a proton and a neutron
$Q_{pro}=\int\limits_{0}^{\infty}{\rho_{pro}}\left(r\right)dr=1,$ (2.39)
$Q_{neu}=\int\limits_{0}^{\infty}{\rho_{neu}}\left(r\right)dr=0.$ (2.40)
The iso-vector electric square radius and the iso-vector magnetic square
radius are
$\left\langle{r^{2}}\right\rangle_{M,I=1}=\left\langle{r^{2}}\right\rangle_{E,I=1}=\int\limits_{0}^{\infty}{r^{2}V_{3}^{0}}dr=\infty.$
(2.41)
However, in the experiment the iso-vector electric radius is 0.88, the iso-
vector magnetic radius is 0.8. This is the weak point of the massless
composite Skyrme model! The iso-scalar magnetic moment is defined
$\vec{\mu}_{I=0}=\frac{1}{2}\int{\left({\vec{r}\times\vec{B}}\right)d^{3}x},$
(2.42)
and the result is
$\vec{\mu}_{I=0}=\frac{1}{{3\Gamma_{n}}}\left\langle{r^{2}}\right\rangle_{I=0}\vec{J}.$
(2.43)
The connection between the spin and the magnetic moment is
$\vec{\mu}=\frac{g}{{2M_{N}}}\vec{J}.$ (2.44)
Note (2.23, 2.24), we have
$g_{I=0}=g_{proton}+g_{neutron}=\frac{4}{9}M_{N}\left({M_{\Delta}-M_{N}}\right)\left\langle{r^{2}}\right\rangle_{I=0}.$
(2.45)
The iso-vector magnetic moment is defined
$\vec{\mu}_{I=1}=\frac{1}{2}\int{\left({\vec{r}\times\vec{V}_{3}}\right)d^{3}}x,$
(2.46)
where $\vec{V}_{3}$ is the vector current. After few calculations, the result
is
$\vec{\mu}_{I=1}=\frac{1}{3}\Gamma_{n}\vec{J}.$ (2.47)
And we also have
$g_{I=1}=g_{proton}-g_{neutron}=\frac{{2M_{N}}}{{M_{\Delta}-M_{N}}}.$ (2.48)
Using (2.44) and (2.47), we can find
$\mu_{pro}=\frac{{g_{pro}}}{2}=\frac{1}{4}\left[{\frac{4}{9}M_{N}\left({M_{\Delta}-M_{N}}\right)\left\langle{r^{2}}\right\rangle_{I=0}+\frac{{2M_{N}}}{{M_{\Delta}-M_{N}}}}\right],$
(2.49)
$\mu_{neu}=\frac{{g_{neu}}}{2}=\frac{1}{4}\left[{\frac{4}{9}M_{N}\left({M_{\Delta}-M_{N}}\right)\left\langle{r^{2}}\right\rangle_{I=0}-\frac{{2M_{N}}}{{M_{\Delta}-M_{N}}}}\right].$
(2.50)
The axial coupling is defined
$g_{A}=-\frac{\pi}{{3e^{2}}}D_{n},$ (2.51)
with
$D_{n}=\int\limits_{0}^{\infty}{\tilde{r}^{2}}\left\\{{\frac{{\sin
2nF}}{{n\tilde{r}}}\left[{1+4\left({F^{\prime
2}+\frac{{\sin^{2}nF}}{{n^{2}\tilde{r}^{2}}}}\right)}\right]+F^{\prime}\left({1+\frac{{8\sin^{2}nF}}{{n^{2}\tilde{r}^{2}}}}\right)}\right\\}d\tilde{r}.$
(2.52)
Figure 2.2: Plots of density of neutron charge and proton charge for a case of
$n=2$ [25]. Figure 2.3: Plots of density of neutron charge and proton charge
for a case of $n=4$ [25].
NUMERICAL RESULTS TABLE FOR THE CASE OF N=4 [25]
Quantity Prediction Experiment $e$ 1.10 - $F_{\pi}$(MeV) 168 186
$\left\langle{r^{2}}\right\rangle_{E,I=0}^{1/2}$(fm) 1.10 0.72
$\left\langle{r^{2}}\right\rangle_{M,I=0}^{1/2}$(fm) 1.93 0.81
$\left\langle{r^{2}}\right\rangle_{E,I=1}^{1/2}$(fm) $\infty$ 0.88
$\left\langle{r^{2}}\right\rangle_{M,I=1}^{1/2}$(fm) $\infty$ 0.8
$\mu_{pro}$(mag) 2.55 2.79 $\mu_{neu}$(mag) -0.66 -1.91
$\left|{\frac{{\mu_{pro}}}{{\mu_{neu}}}}\right|$ 3.86 1.46 $g_{A}$ 3.52 1.23
## Chapter 3 The composite Skyrme model with pion’s mass
### 3.1 The term of pion’mass
When Skyrme constructed a model from the theory of pi-meson, he did not let
the term of mass of pion because of its spontaneous symmetry breaking.
However, with $n=1,2,3,4$, datum of prediction are not really exact. One of
causes that we do not let the term of pion’s mass into the lagrangian [2, 3].
Therefore, we will propose the composite lagrangian density with the pion’s
mass [22, 24]. The lagrangian density now is
${\cal L}_{n}^{pion}={\cal
L}_{n}+\frac{1}{{8n}}m_{\pi}^{2}F_{\pi}^{2}\left[{Tr\left({U^{n}}\right)-2}\right].$
(3.1)
Now using the hedgehog form gave by Skyrme
$U_{o}\left(r\right)=\exp\left[{i\tau.\hat{r}F\left(r\right)}\right]$, where
$\tau$’s are Pauli’s matrices and
$\hat{r}=\frac{{\vec{r}}}{{\left|{\vec{r}}\right|}}$, the lagrangian density
is
${\cal L}_{n}^{pion}={\cal
L}_{n}+\frac{1}{{4n}}m_{\pi}^{2}F_{\pi}^{2}\left({\cos nF-1}\right),$ (3.2)
${\cal
L}_{n}=\frac{{F_{\pi}^{2}}}{8}\left[{\left({\frac{{dF}}{{dr}}}\right)^{2}+2\frac{{\sin^{2}nF}}{{n^{2}r^{2}}}}\right]+\frac{{\sin^{2}nF}}{{2e^{2}n^{2}r^{2}}}\left[{2\left({\frac{{dF}}{{dr}}}\right)^{2}+\frac{{\sin^{2}nF}}{{n^{2}r^{2}}}}\right].$
(3.3)
The energy of hedgehog is
${\cal
E}_{n}^{pion}=\frac{{F_{\pi}}}{e}A_{n}+\frac{1}{{e^{3}F_{\pi}}}\Theta_{n},$
(3.4)
with
$A_{n}=\int\limits_{0}^{\infty}{4\pi\tilde{r}^{2}\left\\{{\left[{\frac{{F^{\prime
2}}}{8}+\frac{{\sin^{2}nF}}{{4n^{2}\tilde{r}^{2}}}}\right]+\frac{{e^{2}F_{\pi}^{4}\sin^{2}nF}}{{n^{2}\tilde{r}^{2}}}\left[{F^{\prime
2}+\frac{{\sin^{2}nF}}{{2n^{2}\tilde{r}^{2}}}}\right]}\right\\}}d\tilde{r},$
(3.5)
and
$\Theta_{n}=\int\limits_{0}^{\infty}{\frac{{\pi\tilde{r}^{2}m_{\pi}^{2}}}{n}}\left({\cos
nF-1}\right)d\tilde{r},$ (3.6)
$\tilde{r}=eF_{\pi}r$ is a dimensionless variable.
After some steps of calculation as chapter 2, we go to the hamiltonian
$H_{n}={\cal E}_{n}^{pion}+\frac{1}{{8\Gamma_{n}}}\vec{J}^{2}={\cal
E}_{n}^{pion}+\frac{1}{{8\Gamma_{n}}}\vec{T}^{2},$ (3.7)
with $\vec{J}$ is a spin and $\vec{T}$ is an isospin. Eigenvalues of
hamiltonian (3.7) are
$M_{n}=E_{n}=\frac{{F_{\pi}}}{e}A_{n}+\frac{1}{{e^{3}F_{\pi}}}\Theta_{n}+\frac{{l\left({l+2}\right)}}{{8\Gamma_{n}}},$
(3.8)
$l=2J$ and $J$ is the quantum number of spin. So, the nucleon
$\left({J=1/2}\right)$ and delta $\left({J=3/2}\right)$ masses are given by
$M_{N}=\frac{{F_{\pi}}}{e}A_{n}+\frac{1}{{e^{3}F_{\pi}}}\Theta_{n}+\frac{{3e^{3}F_{\pi}}}{{8\Phi_{n}}},$
(3.9)
$M_{\Delta}=\frac{{F_{\pi}}}{e}A_{n}+\frac{1}{{e^{3}F_{\pi}}}\Theta_{n}+\frac{{15e^{3}F_{\pi}}}{{8\Phi_{n}}}.$
(3.10)
From (3.9) and (3.10), we obtain
$F_{\pi}=\frac{{2\left({M_{\Delta}-M_{N}}\right)\Phi_{n}}}{{3e^{3}}},$ (3.11)
$e=\sqrt[4]{{\frac{{4\left({M_{\Delta}-M_{N}}\right)^{3}\Phi_{n}^{2}A_{n}}}{{6M_{N}\left({M_{\Delta}-M_{N}}\right)^{2}\Phi_{n}-9\Theta_{n}-3\left({M_{\Delta}-M_{N}}\right)^{2}\Phi_{n}}}}}.$
(3.12)
### 3.2 The field equation and numerical results
We obtain the nonlinear differential equation of $F\left(r\right)$ from
minimum condition of the hedgehog’s energy
$\delta_{F}{\cal E}_{n}^{pion}=0,$ (3.13)
or
$\frac{d}{{dr}}\frac{{\delta{\cal E}_{n}^{pion}}}{{\delta
F^{\prime}}}-\frac{{\delta{\cal E}_{n}^{pion}}}{{\delta F}}=0(Euler-Lagrange-
equation),$ (3.14)
$\left({\frac{{\tilde{r}^{2}}}{4}+\frac{2}{{n^{2}}}\sin^{2}nF}\right)F^{\prime\prime}+\frac{{\tilde{r}}}{2}F^{\prime}+\frac{{\sin
2nF}}{n}\left({F^{\prime
2}-\frac{{\sin^{2}nF}}{{n^{2}\tilde{r}^{2}}}-\frac{1}{4}}\right)-\beta\tilde{r}^{2}\sin
nF=0,$ (3.15)
where
$\beta=\frac{{m_{\pi}^{2}}}{{4e^{2}F_{\pi}^{2}}}.$ (3.16)
We see that (3.16) contains the constant of $e$ and $F_{\pi}$ but they are not
known because they are must found after solving (3.15). Thus, we will solve
(3.15) by the way is taking the value of $\beta$ in spite of $e$ and $F_{\pi}$
and raising it until it equals approximately the $\beta$ that defined by
(3.11, 3.12, 3.16). In this case $n=4$, we need the value of $\beta$ is 0.0025
[24].
Figure 3.1: Plots of numerical solution for a case of $n=4$ [24]. Figure 3.2:
Plots of density of neutron charge and proton charge for a case of $n=4$ [24].
Following results that are obtained above by us (chapter 2), quantities of
baryon (proton and neutron) are described below [24]
NUMERICAL RESULTS TABLE FOR THE CASE OF N=4
Quantity Prediction Experiment $e$ 2.87 - $F_{\pi}$(MeV) 446.56 186
$\left\langle{r^{2}}\right\rangle_{E,I=0}^{1/2}$(fm) 0.21 0.71
$\left\langle{r^{2}}\right\rangle_{E,I=1}^{1/2}$(fm) 2.26 0.88
$\left\langle{r^{2}}\right\rangle_{M,I=0}^{1/2}$(fm) 1.0 0.81
$\left\langle{r^{2}}\right\rangle_{M,I=1}^{1/2}$(fm) 2.26 0.8 $\mu_{pro}$(mag)
1.64 2.79 $\mu_{neu}$(mag) -1.57 -1.91
$\left|{\frac{{\mu_{pro}}}{{\mu_{neu}}}}\right|$ 1.04 1.46 $g_{A}$ 1.22 1.23
## Chapter 4 The supersymmetric composite Skyrme model
The fundamental theories of particles consider that elementary particles are
points. They all have the spin half-one and combining them together will make
particles which have the structural space. But, the model gave by Skyrme is
quite otherwise. He consider that there are solitons. In these models, one
consider that soliton solutions have the structural space and the spin zero.
By quantizing these solitons, we obtain real particles that have the spin non-
zero. Since working with solitons, these models are all nonlinear. There is an
important point that these models work at the ideal case (a pion’s mass is
zero). That makes datum of prediction are not really exact. Thus, we need to
let the term of pion’s mass in these models. But, with only pion, the models
would not be exact, because we also have other mesons. Must let all of mesons
into the model. Thus, we extend the model onto the _SU(3)_ group. But then
will be an another cause that makes the model is not really exact. In general
we must consider the contributions of the particle partners of the mesons. On
the other hand, each meson has a ”superpartner” (make up from quark
superpartners). When all of these contributions are counted in the model, it
may be exact [23].
In 1984, E. A. Bergshoeff, R. I. Nepomachie, H. J. Schnitzer proposed Skyrmion
of four-dimensional supersymmetric non-linear sigma model [5]. The composite
Skyrme model proposed by H. Y. Cheung and F. Gursey in 1990. It’s the general
model in which the sigma model and $V-T$ model are two special cases
corresponding $n=1,2$. Thus, we will calculate the composite model in the
supersymmetry by the way proposed by E. A. Bergshoeff et al [26].
We recall the Lagrangian density of composite Skyrme model
${\cal
L}_{n}=-\frac{{f_{\pi}^{2}}}{{16n^{2}}}Tr\left({\partial_{\mu}U^{-n}\partial^{\mu}U^{n}}\right)+\frac{1}{{32e^{2}n^{4}}}Tr\left({\left[{U^{-n}\partial_{\mu}U^{n},U^{-n}\partial_{\nu}U^{n}}\right]^{2}}\right),$
(4.1)
with $f_{\pi}$ is the pion decay constant, _e_ is a dimensionless parameter.
Replacing ordinary derivatives in the Lagrangian density into covariant
derivatives
$D_{\mu}U^{n}=\partial_{\mu}U^{n}-iV_{\mu}U^{n}\tau_{3},$ (4.2)
the Lagrangian density (4.1) is now
${\cal
L}_{n}=-\frac{{f_{\pi}^{2}}}{{16n^{2}}}Tr\left({D_{\mu}U^{-n}D^{\mu}U^{n}}\right)+\frac{1}{{32e^{2}n^{4}}}Tr\left({\left[{U^{-n}D_{\mu}U^{n},U^{-n}D_{\nu}U^{n}}\right]^{2}}\right).$
(4.3)
and is invariant under local $U\left(1\right)_{R}$ and global
$SU\left(2\right)_{L}$ transformations
$U^{n}\left(r\right)\to
AU^{n}\left(x\right)e^{i\lambda\left(r\right)\tau_{3}},A\in
SU\left(2\right)_{L},$ (4.4) $V_{\mu}\left(r\right)\to
V_{\mu}\left(r\right)+\partial_{\mu}\lambda\left(r\right).$ (4.5)
Really, we have
$\left({D_{\mu}U^{-n}}\right)^{\prime}=\left[{\partial_{\mu}+i\left({V_{\mu}+\partial_{\mu}\lambda}\right)\tau_{3}}\right]\left({U^{-n}e^{-i\lambda\left(x\right)\tau_{3}}}\right)A^{\dagger}$
,
$=\left({\partial_{\mu}U^{-n}}\right)e^{-i\lambda\tau_{3}}A^{\dagger}-i\left({\partial_{\mu}\lambda}\right)\tau_{3}U^{-n}e^{-i\lambda\tau_{3}}A^{\dagger}+$
$+iV_{\mu}\tau_{3}U^{-n}e^{-i\lambda\tau_{3}}A^{\dagger}+i\left({\partial_{\mu}\lambda}\right)\tau_{3}U^{-n}e^{-i\lambda\tau_{3}}A^{\dagger}$
,
$=\left[{\left({\partial_{\mu}+iV_{\mu}\tau_{3}}\right)U^{-n}}\right]e^{-\lambda\left(x\right)\tau_{3}}A^{\dagger}$
,
$\bf{=\left({D_{\mu}U^{-n}}\right)e^{-i\lambda\left(x\right)\tau_{3}}A^{\dagger}};$
(4.6)
$\left({D^{\mu}U^{n}}\right)^{\prime}=A\left[{\partial^{\mu}-i\left({V^{\mu}+\partial^{\mu}\lambda}\right)\tau_{3}}\right]\left({U^{n}e^{i\lambda\left(x\right)\tau_{3}}}\right)$,
$=A\left({\partial^{\mu}U^{n}}\right)e^{i\lambda\tau_{3}}+iA\left({\partial^{\mu}\lambda}\right)\tau_{3}U^{n}e^{i\lambda\tau_{3}}-iAV^{\mu}\tau_{3}U^{n}e^{i\lambda\tau_{3}}-iA\left({\partial^{\mu}\lambda}\right)\tau_{3}U^{n}e^{i\lambda\tau_{3}}$
,
$=A\left({\partial^{\mu}U^{n}}\right)e^{i\lambda\tau_{3}}-iAV^{\mu}\tau_{3}U^{n}e^{i\lambda\tau_{3}}=A\left[{\left({\partial^{\mu}-iV^{\mu}\tau_{3}}\right)U^{n}}\right]e^{i\lambda\tau_{3}}$
,
$\bf{=A\left({D^{\mu}U^{n}}\right)e^{i\lambda\tau_{3}}};$ (4.7)
$\Rightarrow\left({D_{\mu}U^{-n}}\right)^{\prime}\left({D^{\mu}U^{n}}\right)^{\prime}=\left({D_{\mu}U^{-n}}\right)e^{-i\lambda\left(x\right)\tau_{3}}A^{\dagger}Ae^{i\lambda\tau_{3}}\left({D^{\mu}U^{n}}\right)$
,
$=\left({D_{\mu}U^{-n}}\right)\left({D^{\mu}U^{n}}\right).$ (4.8) $\Rightarrow
Tr\left[{\left({D_{\mu}U^{-n}}\right)^{\prime}\left({D^{\mu}U^{n}}\right)^{\prime}}\right]=Tr\left[{\left({D_{\mu}U^{-n}}\right)\left({D^{\mu}U^{n}}\right)}\right].$
(4.9)
The second term of Lagrangian is proved similarly.
The gauge field $V_{\mu}\left(r\right)$ is defined
$V_{\mu}=-\frac{i}{{2n}}Tr\left({U^{-n}\partial_{\mu}U^{n}\tau_{3}}\right).$
(4.10)
We parameterize the $SU\left(2\right)$ matrix $U^{n}\left(r\right)$ in terms
of the complex scalars $A_{i}$
$U^{n}\left(r\right)=\left({\begin{array}[]{*{20}c}{A_{1}}&{-A_{2}^{*}}\\\
{A_{2}}&{A_{1}^{*}}\\\ \end{array}}\right),$ (4.11)
with the unitary constraint
$U^{\dagger}U=1\Rightarrow\bar{A}^{i}A_{i}=A_{1}^{*}A_{1}+A_{2}^{*}A_{2}=1.$
(4.12)
Now, (4.2) is
$D_{\mu}\left({\begin{array}[]{*{20}c}{A_{1}}&{-A_{2}^{*}}\\\
{A_{2}}&{A_{1}^{*}}\\\
\end{array}}\right)=\partial_{\mu}\left({\begin{array}[]{*{20}c}{A_{1}}&{-A_{2}^{*}}\\\
{A_{2}}&{A_{1}^{*}}\\\
\end{array}}\right)-iV_{\mu}\left({\begin{array}[]{*{20}c}{A_{1}}&{A_{2}^{*}}\\\
{A_{2}}&{-A_{1}^{*}}\\\ \end{array}}\right)$ (4.13)
or
$D_{\mu}A_{i}=\left({\partial_{\mu}-iV_{\mu}}\right)A_{i},$ (4.14)
$D_{\mu}\bar{A}_{i}=\left({\partial_{\mu}+iV_{\mu}}\right)\bar{A}_{i}.$ (4.15)
The form of gauge field is then
$V_{\mu}=-\frac{i}{{2n}}Tr\left({U^{-n}\partial_{\mu}U^{n}\tau_{3}}\right)$ ,
$=-\frac{i}{{2n}}Tr\left[{\left({\begin{array}[]{*{20}c}{A_{1}^{*}}&{A_{2}^{*}}\\\
{-A_{2}}&{A_{1}}\\\
\end{array}}\right)\partial_{\mu}\left({\begin{array}[]{*{20}c}{A_{1}}&{-A_{2}^{*}}\\\
{A_{2}}&{A_{1}^{*}}\\\ \end{array}}\right)\left({\begin{array}[]{*{20}c}1&0\\\
0&{-1}\\\ \end{array}}\right)}\right]$,
$=-\frac{i}{{2n}}Tr\left[{\left({\begin{array}[]{*{20}c}{A_{1}^{*}}&{A_{2}^{*}}\\\
{-A_{2}}&{A_{1}}\\\
\end{array}}\right)\left({\begin{array}[]{*{20}c}{\partial_{\mu}A_{1}}&{\partial_{\mu}A_{2}^{*}}\\\
{\partial_{\mu}A_{2}}&{-\partial_{\mu}A_{1}^{*}}\\\
\end{array}}\right)}\right]$,
$=-\frac{i}{{2n}}Tr\left[{\begin{array}[]{*{20}c}{A_{1}^{*}\partial_{\mu}A_{1}+A_{2}^{*}\partial_{\mu}A_{2}}&{}\hfil\\\
{}\hfil&{-A_{2}\partial_{\mu}A_{2}^{*}-A_{1}\partial_{\mu}A_{1}^{*}}\\\
\end{array}}\right]$ ,
$=-\frac{i}{{2n}}\left[{\bar{A}^{i}\partial_{\mu}A_{i}-\left({\partial_{\mu}\bar{A}^{i}}\right)A_{i}}\right]$
,
$\bf\Rightarrow{V_{\mu}\left(r\right)=-\frac{i}{{2n}}\bar{A}^{i}\mathord{\buildrel{\lower
3.0pt\hbox{$\scriptscriptstyle\leftrightarrow$}}\over{\partial}}A_{i}}.$
(4.16)
With matrix $U^{n}\left(r\right)$ is defined in (4.11) and the gauge field is
defined in (4.16), the supersymmetric Lagrangian density (4.3) is then
${\cal
L}_{n}=-\frac{{f_{\pi}^{2}}}{{8n^{2}}}\bar{D}_{\mu}\bar{A}D^{\mu}A-\frac{1}{{16e^{2}n^{2}}}F_{\mu\nu}^{2}\left(V\right),$
(4.17)
with
$F_{\mu\nu}\left(V\right)=\partial_{\mu}V_{\nu}-\partial_{\nu}V_{\mu}.$ (4.18)
Really, we have \+ The first term
${\cal
L}_{n}^{1}=-\frac{1}{{16n^{2}}}f_{\pi}^{2}Tr\left[{D^{\mu}\left({\begin{array}[]{*{20}c}{A_{1}^{*}}&{A_{2}^{*}}\\\
{-A_{2}}&{A_{1}}\\\
\end{array}}\right)D_{\mu}\left({\begin{array}[]{*{20}c}{A_{1}}&{-A_{2}^{*}}\\\
{A_{2}}&{A_{1}^{*}}\\\ \end{array}}\right)}\right]$,
$=-\frac{1}{{16n^{2}}}f_{\pi}^{2}Tr\left({\begin{array}[]{*{20}c}{D^{\mu}A_{1}^{*}D_{\mu}A_{1}+D^{\mu}A_{2}^{*}D_{\mu}A_{2}^{*}}&{}\hfil\\\
{}\hfil&{D^{\mu}A_{2}D_{\mu}A_{2}^{*}+D^{\mu}A_{1}D_{\mu}A_{1}^{*}}\\\
\end{array}}\right)$ ,
$\Rightarrow{\bf{\cal
L}_{n}^{1}=-\frac{1}{{8n^{2}}}f_{\pi}^{2}D^{\mu}\bar{A}^{i}D_{\mu}A_{i}}.$
(4.19)
\+ The second term
$U^{-n}D_{\mu}U^{n}U^{-n}D_{\nu}U^{n}-U^{-n}D_{\nu}U^{n}U^{-n}D_{\mu}U^{n}$,
$=U^{-n}\left[{\left({\partial_{\mu}-iV_{\mu}\tau_{3}}\right)U^{n}}\right]U^{-n}\left[{\left({\partial_{\nu}-iV_{\nu}\tau_{3}}\right)U^{n}}\right]-$
$-U^{-n}\left[{\left({\partial_{\nu}-iV_{\nu}\tau_{3}}\right)U^{n}}\right]U^{-n}\left[{\left({\partial_{\mu}-iV_{\mu}\tau_{3}}\right)U^{n}}\right]$,
$=\left({U^{-n}\partial_{\mu}U^{n}-iU^{-n}V_{\mu}\tau_{3}U^{n}}\right)\left({U^{-n}\partial_{\nu}U^{n}-iU^{-n}V_{\nu}\tau_{3}U^{n}}\right)-$
$-\left({U^{-n}\partial_{\nu}U^{n}-iU^{-n}V_{\nu}\tau_{3}U^{n}}\right)\left({U^{-n}\partial_{\mu}U^{n}-iU^{-n}V_{\mu}\tau_{3}U^{n}}\right)$,
$=U^{-n}\partial_{\mu}U^{n}U^{-n}\partial_{\nu}U^{n}-U^{-n}\partial_{\nu}U^{n}U^{-n}\partial_{\mu}U^{n}.$
(4.20)
Using
$U^{-n}U^{n}=1\Rightarrow\partial_{\mu}U^{-n}U^{n}+U^{-n}\partial_{\mu}U^{n}=0,$
(4.21)
(4.20) is now
$\left({4.20}\right)=-\partial_{\mu}U^{-n}\partial_{\nu}U^{n}+\partial_{\nu}U^{-n}\partial_{\mu}U^{n}.$
(4.22)
The second term is now
$\Rightarrow{\cal
L}_{n}^{2}=\frac{1}{{32e^{2}n^{4}}}Tr\left[{\left({\partial_{\mu}U^{-n}\partial_{\nu}U^{n}-\partial_{\nu}U^{-n}\partial_{\mu}U^{n}}\right)^{2}}\right].$
(4.23)
We define
$F_{\mu\nu}\left(V\right)\equiv\partial_{\mu}V_{\nu}-\partial_{\nu}V_{\mu},$
(4.24)
$=-\frac{i}{{2n}}Tr\left[{\partial_{\mu}\left({U^{-n}\partial_{\nu}U^{n}\tau_{3}}\right)-\partial_{\nu}\left({U^{-n}\partial_{\mu}U^{n}\tau_{3}}\right)}\right],$
(4.25)
$=-\frac{i}{{2n}}Tr\left[{\partial_{\mu}U^{-n}\partial_{\nu}U^{n}\tau_{3}-\partial_{\nu}U^{-n}\partial_{\mu}U^{n}\tau_{3}}\right].$
(4.26) $\Rightarrow
F_{\mu\nu}^{2}=\frac{1}{{2n^{2}}}Tr\left[{\left({\partial_{\mu}U^{-n}\partial_{\nu}U^{n}-\partial_{\nu}U^{-n}\partial_{\mu}U^{n}}\right)^{2}}\right].$
(4.27)
From (4.23) and (4.27), we have
${\bf{\cal L}_{n}^{2}=\frac{1}{{16e^{2}n^{2}}}F_{\mu\nu}^{2}}.$ (4.28)
To supersymmetrise this model, we extend $A_{i}$ to chiral scalar multiples
$\left({A_{i},\psi_{\alpha i},F_{i}}\right)$ $\left({i,\alpha=1,2}\right)$ and
the vector $V_{\mu}\left(r\right)$ to real vector multiples
$\left({V_{\mu},\lambda_{\alpha},D}\right)$. Here, the fields $F_{i}$ are
complex scalars, $D$ is real scalar, $\psi_{\alpha i}$, $\lambda_{\alpha}$ are
Majorana two-component spinors. $\psi_{\alpha i}$ corresponds to a left-handed
chiral spinor, $\bar{\psi}^{\alpha i}=\left({\psi_{i}^{\alpha}}\right)^{*}$
corresponds to a right-handed one. The supersymmetric Lagrangian density is
given by
${\cal
L}_{susy}=\frac{{f_{\pi}^{2}}}{{8n^{2}}}\left[{-D^{\mu}\bar{A}^{i}D_{\mu}A_{i}-\frac{1}{2}i\bar{\psi}^{\dot{\alpha}i}\left({\sigma_{\mu}}\right)_{\alpha\dot{\alpha}}\mathord{\buildrel{\lower
3.0pt\hbox{$\scriptscriptstyle\leftrightarrow$}}\over{D}}^{\mu}\psi_{i}^{\alpha}+\bar{F}^{i}F_{i}-}\right.$
(4.29) $\left.{-i\bar{A}^{i}\lambda^{\alpha}\psi_{\alpha
i}+iA_{i}\bar{\lambda}^{\dot{\alpha}}\bar{\psi}_{\dot{\alpha}}^{i}+D\left({\bar{A}^{i}A_{i}-1}\right)}\right]+$
(4.30)
$+\frac{1}{{8e^{2}n^{2}}}\left[{-\frac{1}{2}F_{\mu\nu}^{2}-i\bar{\lambda}^{\dot{\alpha}}\left({\sigma^{\mu}}\right)_{\dot{\alpha}}^{\alpha}\partial_{\mu}\lambda_{\alpha}+D^{2}}\right],$
(4.31)
is invariant under supersymmetric transformations
$\delta A_{i}=-\varepsilon^{\alpha}\psi_{\alpha i},$ (4.32)
$\delta\psi_{\alpha
i}=-i\bar{\varepsilon}^{\dot{\alpha}}\left({\sigma^{\mu}}\right)_{\alpha\dot{\alpha}}D_{\mu}A_{i}+\varepsilon_{\alpha}F_{i},$
(4.33) $\delta
F_{i}=-i\bar{\varepsilon}^{\dot{\alpha}}\left({\sigma^{\mu}}\right)_{\dot{\alpha}}^{\alpha}D_{\mu}\psi_{\alpha
i}-i\bar{\varepsilon}^{\dot{\alpha}}A_{i}\bar{\lambda}_{\dot{\alpha}},$ (4.34)
$\delta
V_{\mu}=-\frac{1}{2}i\left({\sigma_{\mu}}\right)^{\alpha\dot{\alpha}}\left({\bar{\varepsilon}_{\dot{\alpha}}\lambda_{\alpha}+\varepsilon_{\alpha}\bar{\lambda}_{\dot{\alpha}}}\right),$
(4.35)
$\delta\lambda_{\alpha}=\varepsilon^{\beta}\left({\sigma^{\mu\nu}}\right)_{\beta\alpha}F_{\mu\nu}+i\varepsilon_{\alpha}D,$
(4.36) $\delta
D=\frac{1}{2}\left({\sigma^{\mu}}\right)_{\alpha\dot{\alpha}}\partial_{\mu}\left({\bar{\varepsilon}^{\dot{\alpha}}\lambda^{\alpha}-\varepsilon^{\alpha}\bar{\lambda}^{\dot{\alpha}}}\right).$
(4.37)
The field equation and their supersymmetric transformations lead to the
following constraints with
$\bar{A}^{i}A_{i}=0,$ (4.38) $\bar{A}^{i}\psi_{\alpha i}=0,$ (4.39)
$\bar{A}^{i}F_{i}=0$ (4.40)
and following algebraic expressions for
$V_{\mu}=-\frac{1}{2}\left\\{{i\bar{A}^{i}\mathord{\buildrel{\lower
3.0pt\hbox{$\scriptscriptstyle\leftrightarrow$}}\over{\partial}}_{\mu}A_{i}+\left({\sigma_{\mu}}\right)^{\alpha\dot{\alpha}}\bar{\psi}_{\dot{\alpha}}^{i}\psi_{\alpha
i}}\right\\},$ (4.41) $\lambda_{\alpha}=-i\left\\{{\bar{F}^{i}\psi_{\alpha
i}+i\left({\sigma^{\mu}}\right)_{\alpha\dot{\alpha}}\left({D_{\mu}A_{i}}\right)\bar{\psi}^{\dot{\alpha}i}}\right\\},$
(4.42)
$D=D^{\mu}\bar{A}^{i}D_{\mu}A_{i}+\frac{1}{2}i\bar{\psi}^{\dot{\alpha}i}\left({\sigma^{\mu}}\right)_{\alpha\dot{\alpha}}\left({\mathord{\buildrel{\lower
3.0pt\hbox{$\scriptscriptstyle\leftrightarrow$}}\over{D}}_{\mu}\psi_{i}^{\alpha}}\right)-\bar{F}^{i}F_{i}.$
(4.43)
Setting $\psi_{\alpha}=F_{i}=0$, then
${\cal
L}_{susy}=-\frac{{f_{\pi}^{2}}}{{8n^{2}}}\bar{D}^{\mu}\bar{A}D_{\mu}A+\frac{1}{{8e^{2}n^{2}}}\left[{-\frac{1}{2}F_{\mu\nu}^{2}+\left({\bar{D}^{\mu}\bar{A}D_{\mu}A}\right)^{2}}\right].$
(4.44)
The second term is fourth-order in derivatives. However, there is other
possible fourth-order term
$\square\bar{A}\square A-\left({\bar{D}^{\mu}\bar{A}D_{\mu}A}\right)^{2},$
(4.45)
where $\square\equiv D^{\mu}D_{\mu}$ is the gauge covariant d’Alembertian.
Thus, we may let it to the Lagrangian density
${\cal
L}_{susy}=-\frac{{f_{\pi}^{2}}}{{8n^{2}}}\bar{D}^{\mu}\bar{A}D_{\mu}A+\frac{1}{{8e^{2}n^{2}}}\left[{\alpha\left\\{{-\frac{1}{2}F_{\mu\nu}^{2}+\left({\bar{D}^{\mu}\bar{A}D_{\mu}A}\right)^{2}}\right\\}+}\right.$
(4.46) $\left.{+\beta\left\\{{\square\bar{A}\square
A-\left({\bar{D}^{\mu}\bar{A}D_{\mu}A}\right)^{2}}\right\\}}\right].$ (4.47)
Now, we recall the hedgehog ansatz
$U^{n}\left(r\right)=\cos nf\left(r\right)+i\vec{\tau}\frac{{\vec{r}}}{r}\sin
nf\left(r\right).$ (4.48)
From (4.11), we have
$A_{1}=\cos nf\left(r\right)+i\cos\theta\sin nf\left(r\right),$ (4.49)
$A_{2}=ie^{i\varphi}\sin\theta\sin nf\left(r\right).$ (4.50)
Inserting the hedgehog ansatz into the supersymmetric Lagrangian density, we
obtain the supersymmetric static energy
$E=4\pi\frac{{f_{\pi}}}{e}\int{dxx^{2}}\left\\{{\frac{1}{{12}}\left({f^{\prime
2}+\frac{{2\sin^{2}nf}}{{n^{2}x^{2}}}}\right)+\left({\frac{{\alpha+\beta}}{{15}}}\right)\left({f^{\prime
2}-\frac{{\sin^{2}nf}}{{n^{2}x^{2}}}}\right)+}\right.$ (4.51)
$\left.{+\frac{\beta}{{12}}\left({f^{\prime\prime}+\frac{{2f^{\prime}}}{{nx}}-\frac{{\sin
2nf}}{{n^{2}x^{2}}}}\right)}\right\\}.$ (4.52)
The Euler-Lagrange equation takes the field equation
$-x^{2}f^{\prime\prime}-2xf^{\prime}+\frac{{\sin
2nf}}{n}+\frac{{4\left({\alpha+\beta}\right)}}{5}\left[{\frac{{2f^{\prime\prime}\sin^{2}nf}}{{n^{2}}}-6x^{2}f^{\prime
2}f^{\prime\prime}-4xf^{\prime 3}+}\right.$ (4.53) $\left.{+\frac{{f^{\prime
2}\sin 2nf}}{n}+\frac{{\sin^{2}nf\sin
2nf}}{{n^{3}x^{2}}}}\right]+\beta\left[{x^{2}f^{\left(4\right)}+\frac{{4xf^{\left(3\right)}}}{n}-\frac{{4f^{\prime\prime}\cos
2nf}}{n}+}\right.$ (4.54) $\left.{+\frac{{4f^{\prime 2}\sin
2nf}}{{n^{2}}}-\frac{{4\sin^{2}nf\sin 2nf}}{{n^{3}x^{2}}}}\right]=0,$ (4.55)
with $x=ef_{\pi}r$ is the dimensionless variable and boundary conditions
$f\left(0\right)=\pi,$ (4.56) $f\left(\infty\right)=0.$ (4.57)
Figure 4.1: Plots of supersymmetric numerical solution of $f\left(x\right)$
for a case of $\alpha=1,\beta=0,n=2,3$ [26].
## Chapter 5 Appendix
### 5.1 Some notations for Chapter 2 and 3
There are some notations to calculations in chapter 2 and 3
$\frac{\partial}{{\partial
x^{\mu}}}\equiv\partial_{\mu}=\left({\partial_{0},\partial_{i}}\right)=\left({\frac{\partial}{{\partial
t}},\vec{\nabla}}\right);$ (5.1) $\frac{\partial}{{\partial
x_{\mu}}}\equiv\partial^{\mu}=\left({\partial^{0},\partial^{i}}\right)=g^{\mu\nu}\partial_{\nu}=diag\left({1,-1,-1,-1}\right)\partial_{\nu}=\left({\frac{\partial}{{\partial
t}},-\vec{\nabla}}\right);$ (5.2) $\partial_{i}F\left(r\right)=\frac{{\partial
F\left(r\right)}}{{\partial x^{i}}}=\frac{{\partial F}}{{\partial
r}}\frac{{\partial r}}{{\partial x^{i}}}=F^{\prime}\frac{{x^{i}}}{r}.$ (5.3)
$U_{o}^{n}\left(r\right)=\exp\left[{i\tau.\hat{r}\left\\{{nF\left(r\right)}\right\\}}\right]=\cos\left\\{{nF\left(r\right)}\right\\}+i\vec{\tau}.\hat{r}\sin\left\\{{nF\left(r\right)}\right\\};$
(5.4)
$U_{o}^{-n}\left(r\right)=\exp\left[{-i\tau.\hat{r}\left\\{{nF\left(r\right)}\right\\}}\right]=\cos\left\\{{nF\left(r\right)}\right\\}-i\vec{\tau}.\hat{r}\sin\left\\{{nF\left(r\right)}\right\\};$
(5.5)
Some identities of Pauli matrices are
$\tau_{i}\tau_{j}=i\varepsilon_{ijk}\tau_{k};$ (5.6)
$\left({\tau_{i}}\right)^{2}=1;$ (5.7) $\tau_{i}\tau_{j}=-\tau_{j}\tau_{i};$
(5.8) $Tr\left({\tau_{i}^{2}}\right)=2;$ (5.9) $Tr\left({\tau_{i}}\right)=0;$
(5.10)
We will define
$\partial_{\mu}U_{o}^{n}\left(r\right)=\partial_{\mu}\left({\cos
nF}\right)+i\vec{\tau}\partial_{\mu}\left({\hat{r}\sin nF}\right)$ ,
$=-n\partial_{i}F\sin nF+\frac{{i\vec{\tau}}}{r}\partial_{i}\vec{r}\sin
nF+i\vec{\tau}\vec{r}\partial_{i}\left({\frac{1}{r}}\right)\sin
nF+in\vec{\tau}\frac{{\vec{r}}}{r}\partial_{i}F\cos nF$ ,
$=-n\frac{{F^{\prime}x^{i}}}{r}\sin nF+\frac{{i\tau_{i}\vec{e}^{i}}}{r}\sin
nF-\frac{{i\vec{\tau}\vec{r}x^{i}}}{{r^{3}}}\sin
nF+in\frac{{\vec{\tau}\vec{r}F^{\prime}x^{i}}}{{r^{2}}}\cos nF;$ (5.11)
$\partial_{\mu}U_{o}^{-n}\left(r\right)=\partial_{\mu}\left({\cos
nF}\right)-i\vec{\tau}\partial_{\mu}\left({\hat{r}\sin nF}\right)$ ,
$=-n\frac{{F^{\prime}x^{i}}}{r}\sin nF-\frac{{i\tau_{i}\vec{e}^{i}}}{r}\sin
nF+\frac{{i\vec{\tau}\vec{r}x^{i}}}{{r^{3}}}\sin nF-
in\frac{{\vec{\tau}\vec{r}F^{\prime}x^{i}}}{{r^{2}}}\cos nF;$ (5.12)
$\partial^{\mu}U_{o}^{n}\left(r\right)=\partial^{\mu}\left({\cos
nF}\right)+i\vec{\tau}\partial^{\mu}\left({\hat{r}\sin nF}\right)$ ,
$=n\frac{{F^{\prime}x_{i}}}{r}\sin nF-\frac{{i\tau_{i}\vec{e}_{i}}}{r}\sin
nF+\frac{{i\vec{\tau}\vec{r}x_{i}}}{{r^{3}}}\sin nF-
in\frac{{\vec{\tau}\vec{r}F^{\prime}x_{i}}}{{r^{2}}}\cos nF;$ (5.13)
The first term of the Lagrangian density (2.2) is defined
$\Rightarrow{\cal
L}_{n}^{1}=Tr\left({\partial_{\mu}U_{o}^{-n}\partial^{\mu}U_{o}^{n}}\right)$
$=Tr\left\\{{\left({-n\frac{{F^{\prime}x^{i}}}{r}\sin
nF-\frac{{i\tau_{i}\vec{e}^{i}}}{r}\sin
nF+\frac{{i\vec{\tau}\vec{r}x^{i}}}{{r^{3}}}\sin nF-
in\frac{{\vec{\tau}\vec{r}F^{\prime}x_{i}}}{{r^{2}}}\cos
nF}\right)\times}\right.$
$\times\left.{\left({n\frac{{F^{\prime}x_{i}}}{r}\sin
nF-\frac{{i\tau_{i}\vec{e}_{i}}}{r}\sin
nF+\frac{{i\vec{\tau}\vec{r}x_{i}}}{{r^{3}}}\sin nF-
in\frac{{\vec{\tau}\vec{r}F^{\prime}x_{i}}}{{r^{2}}}\cos nF}\right)}\right\\}$
$=Tr\left\\{{-n^{2}F^{\prime
2}\sin^{2}nF+inF^{\prime}\frac{{\vec{\tau}\vec{r}}}{{r^{2}}}\sin^{2}nF-
inF^{\prime}\frac{{\vec{\tau}\vec{r}}}{{r^{2}}}\sin^{2}nF+}\right.$
$+in^{2}F^{\prime 2}\frac{{\vec{\tau}\vec{r}}}{{2r}}\sin 2nF-
inF^{\prime}\frac{{\vec{\tau}\vec{r}}}{{r^{2}}}\sin^{2}nF-3\frac{{\sin^{2}nF}}{{r^{2}}}+\frac{{\sin^{2}nF}}{{r^{2}}}-$
$-n\frac{{F^{\prime}}}{{2r}}\sin
2nF+inF^{\prime}\frac{{\vec{\tau}\vec{r}}}{{r^{2}}}\sin^{2}nF+\frac{{\sin^{2}nF}}{{r^{2}}}-\frac{{\sin^{2}nF}}{{r^{2}}}+n\frac{{F^{\prime}}}{{2r}}\sin
2nF-$
$\left.{-in^{2}F^{\prime 2}\frac{{\vec{\tau}\vec{r}}}{{2r}}\sin
2nF-n\frac{{F^{\prime}}}{{2r}}\sin 2nF+n\frac{{F^{\prime}}}{{2r}}\sin
2nF-n^{2}F^{\prime 2}\cos^{2}nF}\right\\}$
$=Tr\left({-n^{2}F^{\prime 2}-2\frac{{\sin^{2}nF}}{{r^{2}}}}\right)$
$\Rightarrow{\cal
L}_{n}^{1}=-2n^{2}\left({\frac{{dF}}{{dr}}}\right)^{2}-4\frac{{\sin^{2}nF}}{{r^{2}}}.$
(5.14)
To defined the second term of the Lagrangian density (2.2), we will go to the
intermediary step using the unitary condition of $U_{o}^{n}$
$U_{o}^{-n}U_{o}^{n}=1\Rightarrow\partial_{\mu}U_{o}^{-n}U_{o}^{n}+U_{o}^{-n}\partial_{\mu}U_{o}^{n}=0.$
(5.15)
The second term of Lagrangian density (2.2) is now
$\Rightarrow{\cal
L}_{n}^{2}=Tr\left[{\left({\partial_{\mu}U_{o}^{-n}\partial_{\nu}U_{o}^{n}-\partial_{\nu}U_{o}^{-n}\partial_{\mu}U_{o}^{n}}\right)^{2}}\right].$
(5.16)
Note (5.11), (5.12) we can define
${\cal
L}_{n}^{2}=32n^{2}\frac{{\sin^{2}nF}}{{r^{2}}}\left({\frac{{dF}}{{dr}}}\right)^{2}+16\frac{{\sin^{4}nF}}{{r^{4}}}.$
(5.17)
### 5.2 Some notations for Chapter 4
Dotted and undotted spinor indices run from 1 to 2 and are denoted by the
early letters of the English alphabet. Spinor indices are raised and lowered
by the $\varepsilon$ tensors
$\varepsilon^{ab}=\left({\begin{array}[]{*{20}c}0&1\\\ {-1}&0\\\
\end{array}}\right),$ (5.18)
$\varepsilon_{ab}=\left({\begin{array}[]{*{20}c}0&{-1}\\\ 1&0\\\
\end{array}}\right).$ (5.19)
For example
$\psi^{a}=\varepsilon^{ab}\psi_{b},$ (5.20)
$\chi^{a}=\varepsilon_{ab}\chi^{b}.$ (5.21)
The product $\psi\chi$ is
$\psi\chi=\psi^{a}\chi_{a}=-\psi_{a}\chi^{a}=\chi^{a}\psi_{a}=\chi\psi.$
(5.22)
In the special case
$\psi^{a}\psi^{b}=-\frac{1}{2}\psi\psi\varepsilon^{ab}.$ (5.23)
The hermitian conjugates of the spinors $\psi$ and $\chi$ are $\bar{\psi}$ and
$\bar{\chi}$
$\bar{\psi}^{\dot{a}}=\left({\psi^{a}}\right)^{\dagger},$ (5.24)
$\bar{\chi}=\left({\chi_{a}}\right)^{\dagger}.$ (5.25)
Their dotted indices are raised and lowered by the tensors
$\varepsilon^{\dot{a}\dot{b}}$ and $\varepsilon_{\dot{a}\dot{b}}$ which are
equal to their undotted counterparts $\varepsilon^{ab}$ and
$\varepsilon_{ab}$. The product of them is
$\bar{\psi}\bar{\chi}=\bar{\psi}_{\dot{a}}\bar{\chi}^{\dot{a}}=-\bar{\psi}^{\dot{a}}\bar{\chi}_{\dot{a}}=\bar{\chi}_{\dot{a}}\bar{\psi}^{\dot{a}}=\bar{\chi}\bar{\psi}.$
(5.26)
In the special case
$\bar{\psi}^{\dot{a}}\bar{\psi}^{\dot{b}}=\frac{1}{2}\bar{\psi}\bar{\psi}\varepsilon^{\dot{a}\dot{b}}.$
(5.27)
The hermitian conjugates of the product $\chi\psi$ is
$\left({\chi\psi}\right)^{\dagger}=\left({\chi^{a}\psi_{a}}\right)^{\dagger}=\bar{\psi}_{\dot{a}}\bar{\chi}^{\dot{a}}=\bar{\psi}\bar{\chi}=\bar{\chi}\bar{\psi}.$
(5.28)
The Pauli matrices $\sigma_{a\dot{b}}^{m}$ are
$\sigma^{0}=\left({\begin{array}[]{*{20}c}{1}&0\\\ 0&{1}\\\
\end{array}}\right),\sigma^{1}=\left({\begin{array}[]{*{20}c}0&1\\\ 1&0\\\
\end{array}}\right)$ ,
$\sigma^{2}=\left({\begin{array}[]{*{20}c}0&{-i}\\\ i&0\\\
\end{array}}\right),\sigma^{3}=\left({\begin{array}[]{*{20}c}1&0\\\ 0&{-1}\\\
\end{array}}\right).$ (5.29)
Other useful identities are
$\chi\sigma^{n}\bar{\psi}=-\bar{\psi}\bar{\sigma}^{n}\chi,$ (5.30)
$\chi\sigma^{m}\bar{\sigma}^{n}\psi=\psi\sigma^{n}\bar{\sigma}^{m}\chi,$
(5.31)
$\left({\chi\sigma^{m}\bar{\psi}}\right)^{\dagger}=\psi\sigma^{m}\bar{\chi},$
(5.32)
$\left({\chi\sigma^{m}\bar{\sigma}^{n}\psi}\right)^{\dagger}=\bar{\psi}\bar{\sigma}^{n}\sigma^{m}\bar{\chi}.$
(5.33)
Using these above notations, we can define (from…)
$\delta\bar{A}_{i}=-\bar{\psi}_{\dot{\alpha}i}\bar{\varepsilon}^{\dot{\alpha}},$
(5.34)
$\delta\bar{\psi}_{\dot{\alpha}i}=i\left({\sigma^{\mu}}\right)_{\alpha\dot{\alpha}}\varepsilon^{\alpha}D_{\mu}\bar{A}_{i}+\bar{\varepsilon}_{\dot{\alpha}}\bar{F}_{i},$
(5.35)
$\delta\bar{F}_{i}=iD_{\mu}\bar{\psi}_{\dot{\alpha}i}\left({\sigma^{\mu}}\right)_{\dot{\alpha}}^{\alpha}\varepsilon^{\alpha}+i\bar{A}_{i}\lambda_{\alpha}\varepsilon^{\alpha},$
(5.36)
$\delta\bar{V}_{\mu}=\frac{1}{2}i\left({\bar{\lambda}_{\dot{\alpha}}\varepsilon_{\alpha}+\lambda_{\alpha}\bar{\varepsilon}_{\dot{\alpha}}}\right)\left({\sigma_{\mu}}\right)^{\alpha\dot{\alpha}},$
(5.37)
$\delta\bar{\lambda}_{\dot{\alpha}}=\left({\sigma^{\mu\nu}}\right)_{\beta\alpha}\bar{\varepsilon}^{\dot{\beta}}F_{\mu\nu}-i\bar{\varepsilon}_{\dot{\alpha}}\bar{D},$
(5.38)
$\delta\bar{D}=\frac{1}{2}\partial_{\mu}\left({\bar{\lambda}^{\dot{\alpha}}\varepsilon^{\alpha}-\lambda^{\alpha}\bar{\varepsilon}^{\dot{\alpha}}}\right)\left({\sigma^{\mu}}\right)_{\alpha\dot{\alpha}}.$
(5.39)
Since $\delta$ and $D_{\mu}$ can permute together, $\delta{\cal L}_{susy}$ is
defined
$\delta{\cal
L}_{susy}=\frac{{f_{\pi}^{2}}}{{8n^{2}}}\left[{-D^{\mu}\left({\delta\bar{A}^{i}}\right)D_{\mu}A_{i}-D^{\mu}\bar{A}^{i}D_{\mu}\left({\delta
A_{i}}\right)-\frac{1}{2}i\left({\delta\bar{\psi}^{\dot{\alpha}i}}\right)\left({\sigma_{\mu}}\right)_{\alpha\dot{\alpha}}\mathord{\buildrel{\lower
3.0pt\hbox{$\scriptscriptstyle\leftrightarrow$}}\over{D}}^{\mu}\psi_{i}^{\alpha}-}\right.$
$-\frac{1}{2}i\bar{\psi}^{\dot{\alpha}i}\left({\sigma_{\mu}}\right)_{\alpha\dot{\alpha}}\mathord{\buildrel{\lower
3.0pt\hbox{$\scriptscriptstyle\leftrightarrow$}}\over{D}}^{\mu}\left({\delta\psi_{i}^{\alpha}}\right)+\delta\bar{F}^{i}F_{i}+\bar{F}^{i}\delta
F_{i}-i\left({\delta\bar{A}^{i}}\right)\lambda^{\alpha}\psi_{\alpha
i}-i\bar{A}^{i}\left({\delta\lambda^{\alpha}}\right)\psi_{\alpha
i}-i\bar{A}^{i}\lambda^{\alpha}\delta\psi_{\alpha i}+$
$\left.{+i\left({\delta
A_{i}}\right)\bar{\lambda}^{\dot{\alpha}}\bar{\psi}_{\dot{\alpha}}^{i}+iA_{i}\left({\delta\bar{\lambda}^{\dot{\alpha}}}\right)\bar{\psi}_{\dot{\alpha}}^{i}+iA_{i}\bar{\lambda}^{\dot{\alpha}}\left({\delta\bar{\psi}_{\dot{\alpha}}^{i}}\right)+\left({\delta
D}\right)\left({\bar{A}^{i}A_{i}-1}\right)+D\left({\delta\bar{A}^{i}}\right)A_{i}+D\bar{A}^{i}\left({\delta
A_{i}}\right)}\right]+$
$+\frac{1}{{8e^{2}n^{4}}}\left[{-F_{\mu\nu}\delta
F_{\mu\nu}-i\left({\delta\bar{\lambda}^{\dot{\alpha}}}\right)\left({\sigma^{\mu}}\right)_{\dot{\alpha}}^{\alpha}\partial_{\mu}\lambda_{\alpha}-i\bar{\lambda}^{\dot{\alpha}}\left({\sigma^{\mu}}\right)_{\dot{\alpha}}^{\alpha}\partial_{\mu}\left({\delta\lambda_{\alpha}}\right)+2D\delta
D}\right].$ (5.40)
## Bibliography
* [1] T. H. R. Skyrme, Nucl. Phys., 31 (1962) 566.
* [2] G. S. Adkins, C. R. Nappi and E. Witten, Nucl. Phys., B228 (1983) 552. G. S. Adkins, Nucl. Phys., B249 (1985) 507. G.S. Adkins, C. R. Nappi, Nucl. Phys., B233 (1984) 109. E. Witten, Nucl. Phys., B223 (1983) 422. E. Witten, Nucl. Phys., B223 (1983) 433.
* [3] N. A. Viet, P. T. Tuyen, J. Phys., G15 (July 1989) 937.
* [4] H. Y. Cheung, Feza Gursey, Modern Physics Letters A, Vol.5 No.21 (1990) 1685
* [5] E. A. Bergshoeff, R. I. Nepomachie, H. J. Schnitzer, Nucl. Phys., B249 (1985) 93
* [6] N. Shiiki, N. Sawado, S.Oryu, hep-th/0603069 (2006)
* [7] A. Acus, hep-ph/9901240 (1999)
* [8] C. Houghton, S. Magee, hep-th/0602227 (2006)
* [9] J. Wess, B. Zumino, Nucl. Phys., B70 (1974) 39. J. Wess, B. Zumino, Nucl. Phys., B78 (1974) 1.
* [10] A. Salam, J. Strathdee, Nucl. Phys., B76 (1974) 477. A. Salam, J. Strathdee, Nucl. Phys., B80 (1974) 499.
* [11] M. F. Sohnius, Physics Report, 128, Nos 2 and 3, (1985), 39 - 204
* [12] P. T. Tuyen, Quantum Mechanics, in Vietnamese, College of Science, VNU, (2004)
* [13] P. T. Tuyen, Theory of Fundamental Particles, in Vietnamese, College of Science, VNU, (2006)
* [14] P. T. Tuyen, Theory of Group, in Vietnamese, College of Science, VNU, (2006)
* [15] L. Ryder, Quantum Field Theory, second edition, Cambridge University Press, (1996)
* [16] S. Weinberg, The Quantum Theory of Fields, vol I, II, III, Cambridge University Press, (1995)
* [17] M. E. Peskin, D. V. Shroeder, An Introduction to Quantum Field Theory, Westview Press, (1995)
* [18] H. Georgi, Lie Algebras in Particle Physics, second edition, Perseus Books, (1999)
* [19] W. Greiner, J. Reinhardt, Field Quantization, Springer, (1996)
* [20] A. Bilal, hep-th/0101055 (2001)
* [21] K. Cahill, hep-ph/9907295 (1999)
* [22] D. Q. Tuan, P. T. Tuyen, J. Science. (Math - Phys), VNU, T.XXII, No 2AP (2006) 194
* [23] D. Q. Tuan, P. T. Tuyen, Phenomenal models of Baryon, Talk in National Conference on Theoretical Physics 31, Cua Lo, Nghe An, 2006 (no-published)
* [24] D. Q. Tuan, P. T. Tuyen, The composite Skyrme model with the mass of pion, (no-published) 2006
* [25] D. Q. Tuan, P. T. Tuyen, The composite Skyrme model with N=4, (no-published) 2006
* [26] D. Q. Tuan, P. T. Tuyen, The supersymmetric composite skyrmion, (no-published) 2007
* [27] Nikolaev V.A. , Tkachev O.G. , hep-ph/0109192 (2001)
* [28] O.V.Pavlovskii, hep-th/0312343 (2003)
* [29] R. A. Battye, P. M. Sutcliffe, hep-th/0602220 (2006)
* [30] D. Q. Tuan, P. T. Tuyen, The formalism of gravitational composite Skyrme model, Talk in National Conference on Theoretical Physics 32, Nha Trang, Khanh Hoa, 2007 (no-published)
| arxiv-papers | 2008-07-12T03:10:05 | 2024-09-04T02:48:56.738086 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Do Quoc Tuan",
"submitter": "Tuan Do quoc",
"url": "https://arxiv.org/abs/0807.1951"
} |
0807.2046 | # Balanced split sets and Hamilton-Jacobi equations
Pablo Angulo Ardoy
Department of Mathematics
Universidad Autónoma de Madrid Luis Guijarro
Department of Mathematics
Universidad Autónoma de Madrid.
ICMAT CSIC-UAM-UCM-UC3M
###### Abstract
We study the singular set of solutions to Hamilton-Jacobi equations with a
Hamiltonian independent of $u$. In a previous paper, we proved that the
singular set is what we called a balanced split locus. In this paper, we find
and classify all balanced split loci, identifying the cases where the only
balanced split locus is the singular locus, and the cases where this does not
hold. This clarifies the relationship between viscosity solutions and the
classical approach of characteristics, providing equations for the singular
set. Along the way, we prove more structure results about the singular sets.
## 1 Introduction
In this paper we consider the following boundary value problem:
$\displaystyle H(p,du(p))=1$ $\displaystyle p\in\Omega$ (1.1) $\displaystyle
u(p)=g(p)$ $\displaystyle p\in\partial\Omega$ (1.2)
for a smooth compact manifold $\Omega$ of dimension $n$ with boundary, $H$
smooth, $H^{-1}(1)\cap T^{\ast}_{p}\Omega$ strictly convex for every $p$, and
$g$ smooth and satisfying the compatibility condition:
$|g(y)-g(z)|<d(y,z)\qquad\forall y,z\in\partial\Omega$ (1.3)
where $d$ is the distance induced by the Finsler metric:
$\varphi_{p}(v)=\sup\left\\{\left\langle
v,\alpha\right\rangle_{p}\,:\,\alpha\in
T^{\ast}_{p}\Omega,\,H(p,\alpha)=1\right\\}$ (1.4)
This definition gives a norm in every tangent space $T_{p}\Omega$. Indeed, $H$
is a norm at every tangent space if we make the harmless assumption that $H$
is positively homogeneous of order $1$: $H(p,\lambda\alpha)=\lambda
H(p,\alpha)$ for $\lambda>0$.
A unique _viscosity solution_ is given by the Lax-Oleinik formula:
$u(p)=\inf_{q\in\partial\Omega}\left\\{d(p,q)+g(q)\right\\}$ (1.5)
A local _classical_ solution can be computed near $\partial\Omega$ following
_characteristic_ curves, which are geodesics of the metric $\varphi$ starting
from a point in $\partial\Omega$ with initial speed given by a vector field on
$\partial\Omega$ that is determined by $H$ and $g$ (see 3.1): if
$\gamma:[0,t)\rightarrow\Omega$ is the unique (projected) characteristic from
a point $q\in\partial\Omega$ to $p=\gamma(t)$ that does not intersect $Sing$,
then $u(p)=g(q)+t$. The viscosity solution can be thought of as a way to
extend the classical solution to the whole $\Omega$.
Let $Sing$ be the closure of the singular set of the viscosity solution $u$ to
the above problem. $Sing$ has a key property: any point in $\Omega\setminus
Sing$ can be joined to $\partial\Omega$ by a unique characteristic curve that
does not intersect $Sing$. A set with this property is said to _split_
$\Omega$. Once characteristic curves are known, if we replace $Sing$ by any
set $S$ that splits $\Omega$, we can still apply the formula in the last
paragraph to obtain another function with some resemblance to the viscosity
solution (see definition 2.4).
The set $Sing$ has an extra property: it is a _balanced split locus_. This
notion, introduced in [AG] and inspired originally by the paper [IT], is
related to the notion of _semiconcave_ functions that is now common in the
study of Hamilton-Jacobi equations (see section 2.1). Our goal in this paper
is to determine whether there is a unique balanced split locus. In the cases
when this is not true, we also give an interpretation of the multiple balanced
split loci.
Finally, we recall that the distance function to the boundary in Riemannian
and Finsler geometry is the viscosity solution of a Hamilton-Jacobi equation
([MM]), and the _cut locus_ is the closure of the singular set of the distance
function to the boundary ([LN]). Thus, our results also apply to cut loci in
Finsler geometry.
### 1.1 Outline
In section 2 we state our results, give examples, and comment on possible
extensions. Section 3 gathers some of the results from the literature we will
need, and includes a few new lemmas that we use later. Section 4 contains our
proof that the _distance to a balanced split locus_ and _distance to the
$k$-th conjugate point_ are Lipschitz. Section 5 contains the proof of the
main theorems, modulo a result that is proved in section 6. This last section
also features detailed descriptions of a balanced split set at each of the
points in the classification introduced in [AG].
### 1.2 Acknowledgements.
The authors express their gratitude to Ireneo Peral, Yanyan Li, Luc Nguyen,
Marco Fontelos and Juan Carlos Álvarez for interesting conversations about
this problem. The paper also benefited greatly from the referee’s input, and
the authors want to thank him for it. Both authors were partially supported
during the preparation of this work by grants MTM2007-61982 and MTM2008-02686
of the MEC and the MCINN respectively.
## 2 Statement of results.
### 2.1 Setting
We study a Hamilton-Jacobi equation given by (1.1) and (1.2) in a $C^{\infty}$
compact manifold with boundary $\Omega$. $H$ is smooth and strictly convex in
the second argument and the boundary data $g$ is smooth and satisfies (1.3).
The solution by characteristics gives the _characteristic vector field_ on
$\partial\Omega$, which we write as a map $\Gamma:\partial\Omega\rightarrow
T\Omega$ that is a section of the projection map
$\pi:T\Omega\rightarrow\Omega$ of the tangent to $\Omega$. The characteristic
curves are the integral curves of the geodesic vector field in $T\Omega$ with
initial point $\Gamma(z)$ for $z\in\partial\Omega$. The projected
characteristics are the projection to $\Omega$ of the characteristics. The
characteristic vector field is smooth and points inwards (see 3.1).
Let $\Phi$ be the geodesic flow in $T\Omega$, and $D(\Phi)$ its domain. We
introduce the set $V\subset\mathbb{R}\times\partial\Omega$:
$V=\left\\{x=(t,z),z\in\partial\Omega,t\geq 0,(t,\Gamma(z))\in
D(\Phi)\right\\}$ (2.1)
$V$ has coordinates given by $z\in\partial\Omega$ and $t\in\mathbb{R}$. We set
$F:V\rightarrow\Omega$ to be the map given by following the projected
characteristic with initial value $\Gamma(z)$ a time $t$:
$F(t,z)=\pi(\Phi(t,\Gamma(z)))$ (2.2)
The vector $r$ given as $\frac{\partial}{\partial t}$ in the above coordinates
maps under $F$ to the tangent to the projected characteristic.
###### Definition 2.1.
For a set $S\subset\Omega$, let $A(S)\subset V$ be the set of all $x=(t,z)\in
V$ such that $F(s,z)\notin S,\,\forall\,0\leq s<t$. We say that a set
$S\subset\Omega$ _splits_ $\Omega$ iff $F$ restricts to a bijection between
$A(S)$ and $\Omega\setminus S$.
Whenever $S$ splits $\Omega$, we can define a vector field $R_{p}$ in
$\Omega\setminus S$ to be $dF_{x}(r_{x})$ for the unique $x$ in $A(S)$ such
that $F(x)=p$.
###### Definition 2.2.
For a point $a\in S$, we define the _limit set_ $R_{a}$ as the set of vectors
in $T_{a}\Omega$ that are limits of sequences of the vectors $R_{p}$ defined
above at points $p\in\Omega\setminus S$.
###### Definition 2.3.
If $S$ splits $\Omega$, we also define a set $Q_{p}\subset V$ for $p\in\Omega$
by
$Q_{p}=\left(F|_{\overline{A(S)}}\right)^{-1}(p)$
The following relation holds between the sets $R_{p}$ and $Q_{p}$:
$R_{p}=\left\\{dF_{x}(r_{x}):x\in Q_{p}\right\\}$
###### Definition 2.4.
If $S$ splits $\Omega$, we can define a real-valued function $h$ in
$\Omega\setminus S$ by setting:
$h(p)=g(z)+t$
where $(t,z)$ is the unique point in $A(S)$ with $F(t,z)=p$.
If we start with the viscosity solution $u$ to the Hamilton-Jacobi equations,
and let $S=Sing$ be the closure of the set where $u$ is not $C^{1}$, then $S$
splits $\Omega$. If we follow the above definition involving $A(S)$ to get a
new function $h$, then we find $h=u$.
###### Definition 2.5.
A set $S$ that splits $\Omega$ is a split locus iff
$S=\overline{\left\\{p\in S:\quad\sharp R_{p}\geq 2\right\\}}$
The role of this condition is to restrict $S$ to its essential part. A set
that merely splits $\Omega$ could be too big: actually $\Omega$ itself splits
$\Omega$. The following lemma may clarify this condition.
###### Lemma 2.6.
A set $S$ that splits $\Omega$ is a split locus if and only if $S$ is closed
and it has no proper closed subsets that split $\Omega$.
###### Proof.
The only if part is trivial, so we will only prove the other implication.
Assume $S$ is a split locus and let $S^{\prime}\subset S$ be a closed set
splitting $\Omega$. Let $q\in S\setminus S^{\prime}$ a point with $\sharp
R_{q}\geq 2$. Since $S^{\prime}$ is closed, there is a neighborhood of $q$
away from $S^{\prime}$; so, if $\gamma_{1}$ is a segment of a geodesic in
$\Omega\setminus S^{\prime}$ joining $\partial\Omega$ with $q$, there is a
point $q_{1}$ in $\gamma_{1}$ lying beyond $q$. Furthermore, we can choose the
point $q_{1}$ not lying in $S$, so there is a second geodesic $\gamma_{2}$
contained in $\Omega\setminus S\subset\Omega\setminus S^{\prime}$ from
$\partial\Omega$ to $q_{1}$. As $q\in S$, we see $\gamma_{2}$ is necessarily
different from $\gamma_{1}$, which is a contradiction if $S^{\prime}$ is
split. Therefore we learn $S^{\prime}\supset\left\\{p\in S:\quad\sharp
R_{p}\geq 2\right\\}$, so $S=\overline{\left\\{p\in S:\quad\sharp R_{p}\geq
2\right\\}}\subset S^{\prime}$.
∎
Finally, we introduce the following more restrictive condition (see 3.3 for
the definition of $v_{p}(q)$, the _vector from $p$ to $q$_, and 3.2 for the
Finsler dual of a vector).
###### Definition 2.7.
We say a split locus $S\subset\Omega$ is _balanced_ for given $\Omega$, $H$
and $g$ (or simply that it is balanced if there is no risk of confusion) iff
for all $p\in S$, all sequences $p_{i}\to p$ with $v_{p_{i}}(p)\to v\in
T_{p}\Omega$, and any sequence of vectors $X_{i}\in R_{p_{i}}\to X_{\infty}\in
R_{p}$, then
$w_{\infty}(v)=\max\left\\{w(v),\text{ $w$ is dual to some $R\in
R_{p}$}\right\\}$
where $w_{\infty}$ is the dual of $X_{\infty}$.
#### Remark.
We proved in [AG] that the cut locus of a submanifold in a Finsler manifold
and the closure of the singular locus of the solution to (1.1) and (1.2) are
always balanced split loci. The proof (and the definition of balanced itself)
was inspired by the paper [IT], and consists basically of an application of
the first variation formula.
We give now another proof that relates the balanced condition to the notion of
semiconcave functions, which is now common in the study of Hamilton-Jacobi
equations. More precisely, we simply translate theorem 3.3.15 in the book [CS]
to our language to get the following lemma:
###### Lemma 2.8.
The closure of the singular set of the viscosity solution to (1.1) and (1.2)
is a balanced split locus.
###### Proof.
Let $u$ be the viscosity solution to (1.1) and (1.2), and let $Sing$ be the
closure of its singular set. We leave to the reader the proof that $Sing$ is a
split locus (otherwise, see [AG]).
It is well known that $u$ is semiconcave (see for example [CS, 5.3.7]). The
superdifferential $D^{+}u(p)$ of $u$ at $p$ is the convex hull of the set of
limits of differentials of $u$ at points where $u$ is $C^{1}$ (see [CS,
3.3.6]). At a point where $u$ is $C^{1}$, the dual of the speed vector of a
characteristic is the differential of $u$. Thus, the superdifferential at $p$
is the convex hull of the duals to the vectors in $R_{p}$. We deduce:
$\max\left\\{w(v),\text{ $w$ is dual to some $R\in
R_{p}$}\right\\}=\max\left\\{w(v),w\in D^{+}u(p)\right\\}$
Given $p\in\Omega$, and $v\in T_{p}\Omega$, the _exposed face_ of $D^{+}u(p)$
in the direction $v$ is given by:
$D^{+}(p,v)=\\{\tilde{w}\in D^{+}u(p):\tilde{w}(v)\leq w(v)\;\forall w\in
D^{+}u(p)\\}$
The balanced condition can be rephrased in these terms as:
> Let $p_{i}\rightarrow p\in S$ be a sequence with $v_{p_{i}}(p)\to v\in
> T_{p}\Omega$, and let $w_{i}\in D^{+}u(p_{i})$ be a sequence converging to
> $w\in D^{+}u(p)$.
>
> Then $w\in D^{+}u(p,-v)$
which is exactly the statement of theorem [CS, 3.3.15], with two minor
remarks:
1. 1.
The condition is restricted to points $p\in S$. At points in $\Omega\setminus
S$, the balanced condition is trivial.
2. 2.
In the balanced condition, we use the vectors $v_{p_{i}}(p)$ from $p_{i}$ to
$p$, contrary to the reference [CS]. Thus the minus sign in the statement.
∎
In the light of this new proof, we can regard the balanced condition as a
differential version of the semiconcavity condition. A semiconcave function
that is a solution to (1.1) is also a viscosity solution (see [CS, 5.3.1]).
This papers tries to recover the same result under the balanced condition.
### 2.2 Results
For _fixed_ $\Omega$, $H$ and $g$ satisfying the conditions stated earlier,
there is always at least one balanced split locus, namely the singular set of
the solution of (1.1) and (1.2). In general, there might be more than one
balanced split loci, depending on the topology of $\Omega$.
Our first theorem covers a situation where there is uniqueness.
###### Theorem 2.9.
Assume $\Omega$ is simply connected and $\partial\Omega$ is connected.
Then there is a unique balanced split locus, which is the singular locus of
the solution of (1.1) and (1.2).
The next theorem removes the assumption that $\partial\Omega$ is connected, at
the price of losing uniqueness:
###### Theorem 2.10.
Assume $\Omega$ is simply connected and $\partial\Omega$ has several connected
components. Let $S\subset\Omega$ be a balanced split locus.
Then $S$ is the singular locus of the solution of (1.1) and (1.2) with
boundary data $g+a$ where the function $a$ is constant at each connected
component of $\partial\Omega$.
The above theorem describes precisely all the balanced split loci in a
situation where there is non-uniqueness. If $\Omega$ is not simply connected,
the balanced split loci are more complicated to describe. We provide a
somewhat involved procedure using the universal cover of the manifold.
However, the final answer is very natural in light of the examples.
###### Theorem 2.11.
There exists a bijection between balanced split loci for given $\Omega$, $H$
and $g$ and an open subset of the homology space
$H^{1}(\Omega,\partial\Omega)$ containing zero.
In fact, this theorem follows immediately from the next, where we construct
such bijection:
###### Theorem 2.12.
Let $\widetilde{\Omega}$ be the universal cover of $\Omega$, and lift both $H$
and $g$ to $\widetilde{\Omega}$.
Let $a:[\partial\widetilde{\Omega}]\rightarrow\mathbb{R}$ be an assignment of
a constant to each connected component of $\partial\widetilde{\Omega}$ that is
equivariant for the action of the automorphism group of the covering and such
that $\widetilde{g}(z)+a(z)$ satisfies the compatibility condition (1.3) in
$\widetilde{\Omega}$. Then the singular locus $\widetilde{S}$ of the solution
$\widetilde{u}$ to:
$\widetilde{H}(x,d\widetilde{u}(x))=1\quad x\in\widetilde{\Omega}$
$\widetilde{u}(x)=\widetilde{g}(x)+a(z)\quad x\in\partial\widetilde{\Omega}$
is invariant by the automorphism group of the covering, and its quotient is a
set $S$ that is a balanced split locus for $\Omega$, $H$ and $g$. Furthermore:
1. 1.
The procedure above yields a bijection between balanced split loci for given
$\Omega$, $H$ and $g$ and _equivariant compatible_ functions
$a:[\partial\widetilde{\Omega}]\rightarrow\mathbb{R}$.
2. 2.
Among the set of equivariant functions
$a:[\partial\widetilde{\Omega}]\rightarrow\mathbb{R}$ (that can be identified
naturally with $H^{1}(\Omega,\partial\Omega)$), those compatible correspond to
an open subset of $H^{1}(\Omega,\partial\Omega)$ that contains $0$.
#### Remark.
The space $H^{1}(\Omega,\partial\Omega)$ is dual to $H_{n-1}(\Omega)$ by
Lefschetz theorem. The proof of the above theorems rely on the construction
from $S$ of a $(n-1)$-dimensional current $T_{S}$ that is shown to be closed
and thus represents a cohomology class in $H_{n-1}(\Omega)$. The proof of the
above theorem also shows that the map sending $S$ to the homology class of
$T_{S}$ is a bijection from the set of balanced split loci onto a subset of
$H_{n-1}(\Omega)$.
In order to prove the above we will make heavy use of some structure results
for balanced split loci. To begin with, we use the results of [AG], which were
stated for a balanced split locus with this paper in mind. In the last
section, we improve the description of the cut locus near each of these types
of points.
We also study some very important functions for the study of the cut locus.
Recall the global coordinates in $V$ given by $z\in\partial\Omega$ and
$t\in\mathbb{R}$. Let $\lambda_{j}(z)$ be the value of $t$ at which the
geodesic $s\rightarrow\Phi(s,z)$ has its $j$-th conjugate point (counting
multiplicities), or $\infty$ if there is no such point. Let
$\rho_{S}:\partial\Omega\rightarrow\mathbb{R}$ be the minimum $t$ such that
$F(t,z)\in S$.
###### Lemma 2.13.
All functions $\lambda_{j}:\partial\Omega\rightarrow\mathbb{R}$ are Lipschitz
continuous.
###### Lemma 2.14.
The function $\rho_{S}:\partial\Omega\rightarrow\mathbb{R}$ is Lipschitz
continuous if $S$ is balanced.
Both results were proven in [IT] for Riemannian manifolds, and the second one
was given in [LN]. Thus, our results are not new for a cut locus, but the
proof is different from the previous ones and may be of interest. We have
recently known of another proof that $\rho$ and $\lambda_{1}$ are Lipschitz
([CR]).
### 2.3 Examples
Take as $\Omega$ any ring in a euclidean $n$-space bounded by two concentric
spheres. Solve the Hamilton-Jacobi equations with $H(x,p)=|p|$ and $g=0$. The
solution is the distance to the spheres, and the cut locus is the sphere
concentric to the other two and equidistant from each of them. However, any
sphere concentric to the other two and lying between them is a balanced split
set, so there is a one parameter family of split balanced sets. When $n>2$,
this situation is a typical application of 2.10. In the $n=2$ case, there is
also only one free parameter, which is in accord with 2.12, as the rank of the
$H_{1}$ homology space of the ring is one.
Figure 1: Balanced split set in a torus
For a more interesting example, we study balanced split sets with respect to a
point in a euclidean torus. We take as a model the unit square in the
euclidean plane with its borders identified. It is equivalent to study the
distance with respect to a point in this euclidean torus, or the solution to
Hamilton-Jacobi equations with respect to a small distance sphere centered at
the point with the Hamiltonian $H(p)=|p|$ and $g=0$.
A branch of cleave points (see 3.5) must keep constant the difference of the
distances from either sides (recall prop 7.2 in [AG], or read the beginning of
section 5). Moving to the covering plane of the torus, we see they must be
segments of hyperbolas. A balanced split locus is the union of the cleave
segments and a few triple or quadruple points. The set of all balanced split
loci is a $2$-parameter family, as predicted by our theorem 2.12.
### 2.4 Extensions
The techniques in this paper could be applied to other first order PDEs, or
systems of PDEs. In particular, we can mention the Cauchy problem with a
Hamiltonian dependent on $t$, and both Cauchy and Dirichlet problems with a
Hamiltonian dependent on $u$. Characteristic curves are well behaved on those
cases (though some extra hypothesis are needed for a Hamiltonian dependent on
$u$).
In particular, we believe our proofs of 2.13 and 2.14 are more easily
extensible to other settings than the previous ones in the literature. This
may simplify the task of proving that the singular locus for other PDEs have
locally finite $n-1$ Hausdorff measure.
In this paper and its predecessor [AG] strong regularity assumptions were
assumed. There are powerful reasons to weaken the regularity assumptions when
studying Hamilton-Jacobi equations. The definition of a balanced split locus
itself does not require strong regularity. Less regular data, though, could
produce qualitatively different behavior. In the structure result 3.5, the
dimensions of the sets of points of each type may become higher, as a
consequence of the general Morse-Sard-Federer theorem (see [F]). Also, if $g$
is not $C^{1}$, we can expect non-trivial intersections between the singular
set and $\partial\Omega$, or _rarefaction waves_.
## 3 Preliminaries
### 3.1 Definitions
###### Definition 3.1.
Let $v\in T_{p}\Omega$ be a tangent vector at $p$ in a Finsler manifold
$(\Omega,\varphi)$. The _Riemannian metric_ at $(p,v)$ is given by:
$g_{(p,v)}(X,Y)=\frac{\partial^{2}\varphi}{\partial
v^{i}v^{j}}(p,v)X^{i}Y^{j}$
###### Definition 3.2.
The _dual one form_ to a vector $X\in T_{p}\Omega$ with respect to a Finsler
metric $\varphi$ is the unique one form $w\in T^{\ast}_{p}\Omega$ such that
$w(X)=\varphi(X)^{2}$ and $w|_{H}=0$, where $H$ is the hyperplane tangent to
the level set
$\left\\{Y\in T_{p}\Omega,\varphi(Y)=\varphi(X)\right\\}$
at $X$. It coincides with the usual definition of dual one form in Riemannian
geometry.
For a vector field, the dual differential one-form is obtained by applying the
above construction at every point.
We will often use the notation $\widehat{X}$ for the dual one-form to the
vector $X$.
In coordinates, the dual one form $w$ to the vector $X$ is given by:
$w_{j}=\frac{\partial\varphi}{\partial v^{j}}(p,X)$
and also, in terms of the Riemannian metric at $(p,X)$:
$w(\cdot)=g_{(p,X)}(X,\cdot)$
With this notion of dual form, we can restate the usual equations for the
characteristic vector field at points $p\in\partial\Omega$:
$\displaystyle\varphi_{p}(X_{p})=1$
$\displaystyle\widehat{X_{p}}|_{T(\partial\Omega)}=dg$ $\displaystyle
X_{p}\text{ points inwards}$ (3.1)
###### Definition 3.3.
Whenever there is a unique unit speed minimizing geodesic $\gamma$ joining the
points $p$ and $q$ in $\Omega$, we define, following [IT],
$v_{p}(q)=\dot{\gamma}(0)$ (3.2)
For fixed $p\in\Omega$, then any $q$ sufficiently close to $p$ is joined to
$p$ by a unique unit speed minimizing geodesic, so $v_{p}(q)$ is well defined.
###### Definition 3.4.
Let $z\in\partial\Omega$ and $x=(t,z)\in V$.
We say $x$ is _conjugate_ iff $F$ is not a local diffeomorphism at $x$. The
_order_ of conjugacy is the dimension of the kernel of $dF$.
We say $x$ is _a first conjugate vector_ iff no point $(s,x)$ for $s<t$ is
conjugate.
We recall from [AG] a result on the structure of balanced split loci (in that
paper, conjugate points are called focal points):
###### Theorem 3.5.
A balanced split locus consists of the following types of points:
* •
Cleave points: Points at which $R_{p}$ consists of two non-conjugate vectors.
The set of cleave points is a smooth hypersurface;
* •
Edge points: Points at which $R_{p}$ consists of exactly one conjugate vector
of order 1. This is a set of Hausdorff dimension at most $n-2$;
* •
Degenerate cleave points: Points at which $R_{p}$ consists of two vectors,
such that one of them is conjugate of order 1, and the other may be non-
conjugate or conjugate of order 1. This is a set of Hausdorff dimension at
most $n-2$;
* •
Crossing points: Points at which $R_{p}$ consists of non-conjugate and
conjugate vectors of order 1, and $R^{\ast}_{p}$ spans an affine subspace of
dimension $2$ ($R^{\ast}_{p}$ is the set of duals to vectors in $R_{p}$). This
is a rectifiable set of dimension $n-2$;
* •
Remainder: A set of Hausdorff dimension at most $n-3$;
### 3.2 Special coordinates
In [AG], we used only a few properties of the exponential map essentially
introduced in [W]. Those properties, stated in proposition 8.3 of [AG], were
shown enough to guarantee the existence of special coordinates for $F$ near a
conjugate point of order $k$ (see the paragraph on special coordinates before
theorem 6.3 of [AG]). Near a point $x^{0}\in V$ and its image
$F(x^{0})\in\Omega$, we can find coordinates such that $x^{0}$ has coordinates
$0$, and $F$ is written as
$F(x_{1},\dots,x_{n})=(x_{1},\dots,x_{n-k},F_{n-k+1}(x),\dots,F_{n}(x))$ (3.3)
where
* •
$\frac{\partial}{\partial x_{i}}F_{j}(x^{0})$ is $0$ for any $i$ and
$n-k+1\leq j\leq n$,
* •
$\frac{\partial}{\partial x_{i}}\frac{\partial}{\partial x_{1}}F_{j}(x^{0})$
is $\delta^{i}_{j}$, for $n-k+1\leq i\leq n$ and $n-k+1\leq j\leq n$.
* •
$\frac{\partial}{\partial x_{1}}(x^{0})=r_{x^{0}}$
### 3.3 Lagrangian submanifolds of $T^{\ast}\Omega$
Let $D$ be the homeomorphism between $T\Omega$ and $T^{\ast}\Omega$ induced by
the Finsler metric as in definition 3.2 ($D$ is actually a $C^{\infty}$
diffeomorphism away from the zero section). We define a map:
$\Delta(t,z)=D(\Phi(t,\Gamma(z)))$ (3.4)
and a subset of $T^{\ast}\Omega$:
$\Theta=\Delta(V)$ (3.5)
where $\Phi$ is the geodesic flow in $T\Omega$. This is a smooth
$n$-submanifold of $T^{\ast}\Omega$ with boundary.
It is a standard fact that, for a function $u:\Omega\rightarrow\mathbb{R}$,
the graph of its differential $du$ is a Lagrangian submanifold of
$T^{\ast}\Omega$. The subset of $\Theta$ corresponding to small $t$ is the
graph of the differential of the solution $u$ to the HJ equations by
characteristics. Indeed, all of $\Theta$ is a lagrangian submanifold of
$T^{\ast}\Omega$ (see [D]).
We can also carry over the geodesic vector field from $T\Omega$ into
$T^{\ast}\Omega$ (outside the zero sections). This vector field in
$T^{\ast}\Omega$ is tangent to $\Theta$. Then, as we follow an integral curve
$\gamma(t)$ within $\Theta$, the tangent space to $\Theta$ describes a curve
$\lambda(t)$ in the bundle $G$ of lagrangian subspaces of $T^{\ast}\Omega$. It
is a standard fact that the vector subspace $\lambda(t)\subset
T^{\ast}_{\gamma(t)}\Omega$ intersects the vertical subspace of
$T^{\ast}_{\gamma(t)}\Omega$ in a non-trivial subspace for a discrete set of
times. We will review this fact, in elementary terms, and prove a lemma that
will be important for the proof of lemma 2.14.
Let $\eta(t)$ be an integral curve of $r$ with $x_{0}=\eta(0)$ a conjugate
point of order $k$. In special coordinates near $x_{0}$, for $t$ close to $0$,
the differential of $F$ along $\eta$ has the form:
$dF(\eta(t))=\begin{pmatrix}I_{n-k}&0\\\
\ast&\ast\end{pmatrix}=\begin{pmatrix}I_{n-k}&0\\\
0&0\end{pmatrix}+t\begin{pmatrix}0&0\\\
0&I_{k}\end{pmatrix}+\begin{pmatrix}0&0\\\
\ast&R(t)\end{pmatrix}+\begin{pmatrix}0&0\\\ \ast&E(t)\end{pmatrix}$
where $|R^{\prime}(t)|<\varepsilon$ and $|E|<\varepsilon$, with $E=0$ if
$\gamma(0)=x_{0}$.
Let $w\in\ker dF(\eta(t_{1}))$ and $v\in\ker dF(\eta(t_{2}))$ be unit vectors
in the kernel of $dF$ for $t_{1}<t_{2}$ close to $0$. It follows that both $v$
and $w$ are spanned by the last $k$ coordinates. We then find:
$0=w\cdot dF(\eta(t_{2}))\cdot v-v\cdot dF(\eta(t_{1}))\cdot
w=(t_{2}-t_{1})w\cdot v+w(R(t_{2})-R(t_{1}))v+w(E(t_{2})-E(t_{1}))v$
and it follows (for some $t_{1}<t^{\ast}<t_{2}$):
$(t_{2}-t_{1})w\cdot
v<|w||v|(|R^{\prime}(t^{\ast})|+2\varepsilon)(t_{2}-t_{1})<3\varepsilon(t_{2}-t_{1})$
or
$w\cdot v<3\varepsilon$ (3.6)
This also shows that the set of $t$’s such that $dF(\eta(t))$ is singular is
discrete.
Say the point $x_{0}=(z_{0},t_{0})$ is the $j$-th conjugate point along the
integral curve of $r$ through $x_{0}$ from $z_{0}$, and recall that it is of
order $k$ as conjugate point. As $z$ moves towards $z_{0}$, all functions
$\lambda_{j}(z),\dots,\lambda_{j+k}(z)$ converge to $t_{0}$. Let $z_{i}$ be a
sequence of points converging to $z_{0}$ such that the integral curve through
$z_{i}$ meets its $k$ conjugate points near $z_{0}$ at $M$ linear subspaces
(e.g. $\lambda_{j}(z_{i})=\dots=\lambda_{j+k_{1}}(z_{i})$;
$\lambda_{j+k_{1}+1}(z_{i})=\dots=\lambda_{j+k_{2}}(z_{i})$; …;
$\lambda_{j+k_{M-1}+1}(z_{i})=\dots=\lambda_{j+k_{M}}(z_{i})$). we get the
following theorem (see also lemma 1.1 in [IT]):
###### Lemma 3.6.
The subspaces $\ker d_{(\lambda_{j+k_{l}}(z_{i}),z_{i})}F$ for $l=1,\dots,M$
converge to orthogonal subspaces of $\ker d_{(\lambda_{j}(z_{0}),z_{0})}F$,
for the standard inner product in the special coordinates at the point
$(\lambda_{j}(z_{0}),z_{0})$.
### 3.4 A useful lemma
###### Lemma 3.7.
Let $U$ be an open set in $\mathbb{R}^{n}$, $A\subset U$ a proper open set,
$C^{+}$ an open cone, $V\subset U$ an arbitrary open set and $\varepsilon>0$
such that at any point $q\in\partial A\cap V$, we have
$(q+C^{+})\cap(q+B_{\varepsilon})\subset A$.
Then $\partial A\cap V$ is a Lipschitz hypersurface. Moreover, for any vector
$X\in C^{+}$, take coordinates so that $X=\frac{\partial}{\partial x_{1}}$.
Then $\partial A\cap V$ is a graph
$S=\\{(h(x_{2},..,x_{n}),x_{2},..,x_{n})\\}$ for a Lipschitz function $h$.
###### Proof.
Choose the vector $X\in C^{+}$ and coordinate system in the statement. Assume
$X$ has norm $1$, so that $q+tX\in q+B_{t}$ for small positive $t$. Take any
point $p\in\partial A\cap V$. We notice that all points
$p+t\frac{\partial}{\partial x_{1}}$ for $0<t<\varepsilon$ belong to $A$, and
all points $p+t\frac{\partial}{\partial x_{1}}$ for $-\varepsilon<t<0$ belongs
to $U\setminus A$. Indeed, there cannot be a point
$p+t\frac{\partial}{\partial x_{1}}\in A$ for $-\varepsilon<t<0$ because the
set $(p+t\frac{\partial}{\partial x_{1}})+(C^{+}\cap B_{\varepsilon})$
contains an open neighborhood of $p$, which contains points not in $A$. In
particular, there is at most one point of $\partial A\cap V$ in each line with
direction vector $\frac{\partial}{\partial x_{1}}$.
Take two points $q_{1},q_{2}\in\mathbb{R}^{n-1}$ sufficiently close and
consider the lines $L_{1}=\\{(t,q_{1}),t\in\mathbb{R}\\}$ and
$L_{2}=\\{(t,q_{2}),t\in\mathbb{R}\\}$. Assume there is a $t_{1}$ such that
$(t_{1},q_{1})$ belongs to $\partial A$. If there is no point of $\partial A$
in $L_{2}$ then either all points of $L_{2}$ belong to $A$ or they belong to
$U\setminus A$. Both of these options lead to a contradiction if
$((t_{1},q_{1})+C^{+})\cap((t_{1},q_{1})+B_{\varepsilon})\cap
L_{2}\neq\emptyset$ (this condition is equivalent to
$K|q_{1}-q_{2}|<\varepsilon$ for a constant $K$ that depends on $C^{+}$ and
the choice of $X\in C^{+}$ and the coordinate system).
Thus there is a point $(t_{2},q_{2})\in\partial A$. For the constant $K$ above
and $t\geq t_{1}+K|q_{1}-q_{2}|$, the point $(t,q_{2})$ lies in the set
$(t_{1},q_{1})+C^{+}$, so we have
$t_{2}<t_{1}+K|q_{1}-q_{2}|$
The points $q_{1}$ and $q_{2}$ are arbitrary, and the lemma follows. ∎
### 3.5 Some generalities on HJ equations.
###### Lemma 3.8.
For fixed $\Omega$ and $H$, two functions
$g,g^{\prime}:\partial\Omega\rightarrow\mathbb{R}$ have the same
characteristic vector field in $\partial\Omega$ iff $g^{\prime}$ can be
obtained from $g$ by addition of a constant at each connected component of
$\partial\Omega$.
###### Proof.
It follows from (3.1) that $g$ and $g^{\prime}$ have the same characteristic
vector field at all points if and only $dg=dg^{\prime}$ at all points. ∎
For our next definition, observe that given $\Omega$, $H$ and $g$, we can
define a map $\tilde{u}:V\rightarrow\mathbb{R}$ by $\tilde{u}(t,z)=t+g(z)$.
###### Definition 3.9.
We say that a function $u:\Omega\rightarrow\mathbb{R}$ is _made from
characteristics_ iff $u|_{\partial\Omega}=g$ and $u$ can be written as
$u(p)=\tilde{u}\circ s$ for a (not necessarily discontinuous) section $s$ of
$F:V\rightarrow\Omega$.
#### Remark.
In the paper [Me], the same idea is expressed in different terms: all
characteristics are used to build a multi-valued solution, and then some
criterion is used to select a one-valued solution. The criterion used there is
to select the characteristic with the minimum value of $\tilde{u}$.
###### Lemma 3.10.
The viscosity solution to (1.1) and (1.2) is the unique continuous function
that is made from characteristics.
###### Proof.
Let $h$ be a function made from characteristics, and $u$ be the function given
by formula (1.5). Let $Sing$ be the closure of the singular set of $u$.
Take a point $z\in\partial\Omega$. Define:
$t^{\ast}_{z}=\sup\left\\{t\geq 0:\;h(F(\tau,z))=u(F(\tau,z))\;\forall
0\leq\tau<t\right\\}$
Claim: $t^{\ast}_{z}<\rho_{Sing}(z)$ implies $h$ is discontinuous at
$F(t^{\ast}_{z},z)$.
Proof of the claim: Assume that $t^{\ast}_{z}<\rho_{Sing}(z)$ and $h$ is
continuous at $F(t^{\ast}_{z},z)$ for some $z\in\partial\Omega$.
As $t^{\ast}_{z}<\rho_{Sing}(z)<\lambda_{1}(z)$, there is an open neighborhood
$O$ of $(t^{\ast}_{z},z)$ such that $F|_{O}$ is a diffeomorphism onto a
neighborhood of $p=F(t^{\ast}_{z},z)$.
By hypothesis, there is a sequence $t_{n}\rightarrow t^{\ast}_{z}$ and
$p_{n}=F(t_{n},z)$ such that $h(p_{n})\neq u(p_{n})$. As $h$ is built from
characteristics using a section $s$, we have
$h(p_{n})=\tilde{u}(s(p_{n}))=\tilde{u}((s_{n},y_{n}))=s_{n}+g(y_{n})$, for
$(s_{n},y_{n})\neq(t_{n},z)$.
For $n$ big enough, the point $(s_{n},y_{n})$ does not belong to $O$, as
$(t_{n},z)$ is the only preimage of $p_{n}$ in $O$. As $h(p_{n})\rightarrow
h(p)$, and $\partial\Omega$ is compact, we deduce the $s_{n}$ are bounded. We
can take a subsequence of $(s_{n},y_{n})$ converging to
$(s_{\infty},y_{\infty})\not\in O$. So we have
$p=F(t^{\ast}_{z},z)=F(s_{\infty},y_{\infty})$. If $p\not\in Sing$, we deduce
that
$\lim_{n\rightarrow\infty}h(p_{n})=\tilde{u}(s_{\infty},y_{\infty})>h(p)=u(p)=\tilde{u}(t^{\ast}_{z},z)$,
so $h$ is discontinuous at $p$.
Using the claim, we conclude the proof: if $h$ is continuous, then
$\rho_{Sing}(z)\leq t^{\ast}_{z}$ for all $z\in\partial\Omega$, and $u=h$, as
any point in $\Omega$ can be expressed as $F(t,z)$ for some $z$, and some
$t\leq\rho_{Sing}(z)$.
∎
We will need later the following version of the same principle:
###### Lemma 3.11.
Let $S$ be a split locus, and $h$ be the function associated to $S$ as in
definition 2.4. If $\rho_{S}$ is continuous, and $h$ can be extended to
$\Omega$ so that it is continuous except for a set of null $\mathcal{H}^{n-1}$
measure, then $S=Sing$.
###### Proof.
Define
$Y_{0}=\left\\{z\in\partial\Omega:\;h(F(t,z))\neq u(F(t,z))\;\text{ for some
}t\in[0,\rho_{Sing}(z))\right\\}$
By the claim in the previous lemma, $Y_{0}$ is contained in:
$Y=\left\\{z\in\partial\Omega:\;h\text{ discontinuous at }F(t,z)\;\text{ for
some }t\in[0,\rho_{Sing}(z))\right\\}$
Let $A=A(Sing)$ be the set in definition 2.1. The map $F$ restricts to a
diffeomorphism from $A$ onto $\Omega\setminus Sing$. The set $Y$ can be
expressed as:
$Y=\pi_{2}\circ(F|_{A})^{-1}\left(\\{p\in\Omega\setminus Sing:\;h\text{
discontinuous at }p\\})\right)$
and thus by the hypothesis has null $\mathcal{H}^{n-1}$ measure. Therefore,
$\partial\Omega\setminus Y_{0}$ is dense in $\partial\Omega$.
We claim now that $S\subset Sing$. To see this, let $p\in S\setminus Sing$.
Then $p=F(t^{\ast},z^{\ast})$ for a unique $(t^{\ast},z^{\ast})\in A$. It
follows $\rho_{S}(z^{\ast})\leq t^{\ast}<\rho_{Sing}(z^{\ast})$. As $\rho_{S}$
is continuous, $\rho_{S}(z)<\rho_{Sing}(z)$ holds for all $z$ in a
neighborhood of $z^{\ast}$ in $\partial\Omega$ and, in particular, for some
$z\in\partial\Omega\setminus Y_{0}$. This is a contradiction because, for
$\rho_{S}(z)<t<\rho_{Sing}(z)$, $h(F(t,z))=\tilde{u}(t^{\prime},z^{\prime})$
for $(t^{\prime},z^{\prime})\neq(t,z)$, and $t<\rho_{Sing}(z)$ implies
$h(F(t,z))=\tilde{u}(t^{\prime},z^{\prime})>\tilde{u}(t,z)=u(F(t,z))$, forcing
$z\in Y_{0}$.
We deduce $S=Sing$ using lemma 2.6 and the fact that $Sing$ is a split locus.
∎
## 4 $\rho_{S}$ is Lipschitz
In this section we study the functions $\rho_{S}$ and $\lambda_{j}$ defined
earlier. The fact that $\rho_{S}$ is Lipschitz will be of great importance
later. The definitions and the general approach in this section follow [IT],
but our proofs are shorter, provide no precise quantitative bounds, use no
constructions from Riemannian or Finsler geometry, and work for Finsler
manifolds, thus providing a new and shorter proof for the main result in [LN].
The proof that $\lambda_{j}$ are Lipschitz functions was new for Finsler
manifolds when we published the first version of the preprint of this paper.
Since then, another preprint has appeared which shows that $\lambda_{1}$ is
actually semi-concave.
###### Proof of 2.13.
It is immediate to see that the functions $\lambda_{j}$ are continuous, since
this is property (R3) of Warner (see [W, pp. 577-578 and Theorem 4.5 ]).
Near a conjugate point $x^{0}$ of order $k$, we can take special coordinates
as in 3.2:
$F(x_{1},\dots,x_{n})=(x_{1},\dots,x_{n-k},F_{n-k+1},\dots,F_{n})$
Conjugate points near $x$ are the solutions of
$d(x_{1},\dots,x_{n})=det(dF)=\sum_{\sigma}(-1)^{\sigma}\frac{\partial
F_{\sigma(n-k+1)}}{\partial x_{n-k+1}}\dots\frac{\partial
F_{\sigma(n)}}{\partial x_{n}}=0$
>From the properties of the special coordinates, we deduce that:
$D^{\alpha}d(0)=0\qquad\forall|\alpha|<k$ (4.1)
and
$\frac{\partial^{k}}{\partial x_{1}^{k}}d=1$
We can use the preparation theorem of Malgrange (see [GG]) to find real valued
functions $q$ and $l_{i}$ in an open neighborhood $U$ of $x$ such that
$q(x)\neq 0$ and:
$q(x_{1},\dots,x_{n})d(x_{1},\dots,x_{n})=x_{1}^{k}+x_{1}^{k-1}l_{1}(x_{2},\dots,x_{n})+\dots+l_{k}(x_{2},\dots,x_{n})$
and we deduce from (4.1) that
$D^{\alpha}l_{i}(0)=0\qquad\forall|\alpha|<i$ (4.2)
which implies
$|l_{i}(x_{2},\dots,x_{n})|<\bar{C}\max\\{|x_{2}|,\dots,|x_{n}|\\}^{i}$ (4.3)
At any conjugate point $(x_{1},\dots,x_{n})$, we have $q(x)=0$, so:
$-x_{1}^{k}=x_{1}^{k-1}l_{1}(x_{2},\dots,x_{n})+\dots+l_{k}(x_{2},\dots,x_{n})\\\
$
and therefore
$|x_{1}|^{k}<|x_{1}|^{k-1}|l_{1}|+\dots+|l_{k}|$
Combining this and (4.3), we get an inequality for $|x_{1}|$ at any conjugate
point $(x_{1},\dots,x_{n})$, where the constant $C$ ultimately depends on
bounds for the first few derivatives of $F$:
$|x_{1}|^{k}<C\max\\{|x_{1}|,\dots,|x_{n}|\\}^{k-1}\max\\{|x_{2}|,\dots,|x_{n}|\\}$
(4.4)
We notice that $|x_{1}|>\max\\{|x_{2}|,\dots,|x_{n}|\\}$ implies
$|x_{1}|^{k}<C|x_{1}|^{k-1}\max\\{|x_{2}|,\dots,|x_{n}|\\}$. In other words:
$|x_{1}|<\max\\{C,1\\}\max\\{|x_{2}|,\dots,|x_{n}|\\}$
This is the statement that all conjugate points near $x$ lie in a cone of
fixed width containing the hyperplane $x_{1}=0$. Thus all functions
$\lambda_{j}$ to $\lambda_{j+k}$ are Lipschitz at $(x_{2}\dots,x_{n})$ with a
constant independent of $x$. ∎
#### Remark.
A proof of lemma 2.13 in the lines of section 3.3 seems possible: let
$\Lambda(\Omega)$ be the bundle of Lagrangian submanifolds of the symplectic
linear spaces $T^{\ast}_{p}\Omega$ and let $\Sigma(\Omega)$ be the union of
the Maslov cycles within each $\Lambda_{p}(\Omega)$. Define
$\lambda:V\rightarrow\Lambda(\Omega)$ where $\lambda(x)$ is the tangent to
$\Theta$ at $D(\Phi(x))$ (recall 3.5). The graphs of the functions
$\lambda_{k}$ are the preimage of the Maslov cycle $\Sigma(\Omega)$. The
geodesic vector field (transported to $T^{\ast}\Omega$), is transversal to the
Maslov cycle. With some effort, the angle (in an arbitrary metric) between
this vector field and the Maslov cycle at points of intersection can be
bounded from below. This is sufficient to show that the $\lambda_{k}$ are
Lipschitz.
###### Lemma 4.1.
For any split locus $S$ and point $y\in\partial\Omega$, there are no conjugate
points in the curve $t\rightarrow\exp(ty)$ for $t<\rho_{S}(y)$. In other
words, $\rho_{S}\leq\lambda_{1}$.
###### Proof.
Assume there is $x$ with $\rho_{S}(x)-\varepsilon>\lambda_{1}(x)$. By [W,
3.4], the map $F$ is not injective in any neighborhood of $(x,t)$. There are
points $(x_{n},t_{n})$ of $S$ with $x_{n}\rightarrow x$ and
$t_{n}<\rho_{S}(x)-\varepsilon$ (otherwise $S$ does not split $\Omega$).
Taking limits, we see $F(x,t)$ is in $S$ for some $t<\rho_{S}(x)-\varepsilon$,
which contradicts the definition of $\rho_{S}(x)$. ∎
>From now on and for the rest of the paper, $S$ will always be a balanced
split locus:
###### Lemma 4.2.
Let $E\subset\partial\Omega$ be an open subset whose closure is compact and
has a neighborhood where $\rho<\lambda_{1}$. Then $\rho_{S}$ is Lipschitz in
$E$.
###### Proof.
The map $x\rightarrow(F(x),dF_{x}(r))$ is an embedding of $V$ into $TM$. There
is a constant $c$ such that for $x,y\in V$:
$|F(x)-F(y)|+|dF_{x}(r)-dF_{y}(r)|\geq c\min\\{|x-y|,1\\}$ (4.5)
Recall the exponential map is a local diffeomorphism before the first
conjugate point. Points $p=F((z,\rho(z)))$ for $z\in E$ have a set $R_{p}$
consisting of the vector $dF_{(z,\rho(z))}(r)$, and vectors coming from
$V\setminus E$. Choose one such point $p$, and a neighborhood $U$ of $p$. The
above inequality shows that there is a constant $m$ such that:
$|dF_{x}(r)-dF_{y}(r)|\geq m$
for $x=(z,\rho(z))$ with $z\in E$ and $y=(w,\rho(w))\in Q_{p}$ with $w\in
V\setminus E$. By the balanced condition 2.7, any unit vector $v$ tangent to
$S$ satisfies $\widehat{dF_{x}(r)}(v)=\widehat{dF_{y}(r)}(v)$ for some such
$y$ and so:
$\widehat{dF_{x}(r)}(v)<1-\varepsilon$
Thus for any vector $w$ tangent to $E$ both vectors $(w,d\rho_{-}(w))$ and
$(w,d\rho_{+}(w))$ lie in a cone of fixed amplitude around the kernel of
$\widehat{dF_{x}(r)}$ (the hyperplane tangent to the indicatrix at $x$).
Application of lemma 3.7 shows that $\rho$ is Lipschitz.
∎
###### Lemma 4.3.
Let $z_{0}\in\partial\Omega$ be a point such that
$\rho(z_{0})=\lambda_{1}(z_{0})$. Then there is a neighborhood $E$ of $z_{0}$
and a constant $C$ such that for all $z$ in $E$ with $\rho(z)<\lambda_{1}(z)$,
$\rho$ is Lipschitz near $z$ with Lipschitz constant $C$.
###### Proof.
Let $O$ be a compact neighborhood of $(z_{0},\lambda_{1}(z_{0}))$ where
special coordinates apply. Let $x=(z,\rho(z))\in O$ be such that
$\rho(z)<\lambda_{1}(z)$. We can apply the previous lemma and find $\rho$ is
Lipschitz near $z$. We just need to estimate the Lipschitz constant uniformly.
Vectors in $R_{F(x)}$ that are of the form $dF_{y}(r)$ for $y\in V\setminus
O$, are separated from $dF_{x}(r)$ as in the previous lemma and pose no
trouble, but now there might be other vectors $dF_{y}(r)$ for $y\in O$.
Fix the metric $\langle\cdot\rangle$ in $O$ whose matrix in special
coordinates is the identity. Any tangent vector to $S$ satisfies
$\widehat{dF_{x}(r)}(v)=\widehat{dF_{y}(r)}(v)$, for some $y\in O\cap
Q_{F(x)}$. A uniform Lipschitz constant for $\rho$ is found if we bound from
below the angle in the metric $\langle\cdot\rangle$ between $r$ and
$d_{x}F^{-1}(v)$ for any vector $v$ with this property. Fix a point $y\in O$
with $F(x)=F(y)$, let $X=dF_{x}(r)$, $Y=dF_{y}(r)$ and
$\alpha=\widehat{X}-\widehat{Y}$. We need to bound from below the angle
between $r$ and the hyperplane $\ker\alpha$.
Our goal is equivalent to proving that there is $\varepsilon_{1}>0$
independent of $x$ such that:
$\dfrac{F_{x}^{\ast}\alpha(r)}{\|F_{x}^{\ast}\alpha\|}>\varepsilon_{1}$
which is equivalent to:
$\widehat{Y}(X)<1-\varepsilon_{1}\|F_{x}^{\ast}\alpha\|$
in the norm $\|\cdot\|$ associated to $\langle\cdot\rangle$.
Notice first that $X$ and $Y$ belong to the indicatrix at $F(x)=F(y)$, which
is strictly convex. By this and (4.5), we see that for some
$\varepsilon_{2}>0$:
$\widehat{Y}(X)<1-\varepsilon_{2}\|X-Y\|^{2}<1-c\varepsilon_{2}\|x-y\|^{2}$
So it is sufficient to show that for some $C_{1}$ independent of $x$:
$\|F_{x}^{\ast}\alpha\|<C_{1}\|x-y\|^{2}$
Using a Taylor expansion of $\frac{\partial\varphi}{\partial x_{j}}$ in the
second entry, we see the form $F_{x}^{\ast}\alpha$ can be written in
coordinates:
$\begin{array}[]{rcl}F_{x}^{\ast}\alpha&=&\left(\frac{\partial\varphi}{\partial
x_{j}}(p,X)-\frac{\partial\varphi}{\partial x_{j}}(p,Y)\right)\frac{\partial
F_{j}}{\partial x_{l}}\\\ &=&\frac{\partial^{2}\varphi}{\partial
x_{i}x_{j}}(p,X)\left(X_{i}-Y_{i}\right)\frac{\partial F_{j}}{\partial
x_{l}}+O(\|X-Y\|)^{2}\\\ &=&\frac{\partial^{2}\varphi}{\partial
x_{i}x_{j}}(p,X)\left(X_{i}-Y_{i}\right)\frac{\partial F_{j}}{\partial
x_{l}}+O(\|x-y\|)^{2}\end{array}$
The argument goes as follows: we need the inequality
$\|F_{x}^{\ast}\alpha(v)\|<C_{1}\|v\|\|x-y\|^{2}$, so we want to bound the
bilinear map $g_{(p,X)}$ evaluated at $X-Y$ and the vector $dF(v)$. The bound
on the norm is achieved when $dF(v)$ is proportional to $X-Y$. The map
$d_{x}F$ is invertible, so for the vector
$v=\frac{dF^{-1}(X-Y)}{\|dF^{-1}(X-Y)\|}$, we have:
$\|F_{x}^{\ast}\alpha\|=\|F_{x}^{\ast}\alpha(v)\|$
Thus we have:
$\|F_{(z,\rho(z))}^{\ast}\alpha\|<C_{2}\frac{\|X-Y\|^{2}}{\|dF^{-1}(X-Y)\|}+O(\|x-y\|)^{2}<C_{3}\frac{\|x-y\|^{2}}{\|dF^{-1}(X-Y)\|}+O(\|x-y\|)^{2}$
for constants $C_{2}$ and $C_{3}$, and it is enough to show there is
$\varepsilon_{3}$ independent of $x$ and $y$ such that:
$\|dF^{-1}(X-Y)\|>\varepsilon_{3}$ (4.6)
Let $G(x)=d_{x}F(r)$. We have:
$X-Y=G(x)-G(y)=dG_{x}(x-y)+O(\|x-y\|)$
so it is equivalent to show the following:
$\|dF^{-1}dG_{x}(x-y)\|>\varepsilon_{4}$
for $\varepsilon_{4}$ independent of $x$ and $y$.
Assume that $(\rho(z_{0}),z_{0})$ is conjugate of order $k$, so that
$\rho(z_{0})=\lambda_{1}(z_{0})=\dots=\lambda_{k}(z_{0})$. Thanks to Lemma
2.13 and reducing to a smaller $O$, we can assume that
$a_{1}=(\lambda_{1}(z),z)$ to $a_{k}=(\lambda_{k}(z),z)$ all lie within $O$
(some of them may coincide). Let $d_{i}=\lambda_{i}(z)-\rho(z)$ be the
distance from $x$ to the $a_{i}$. At each of the $a_{i}$ there is a vector
$w_{i}\in\ker d_{a_{i}}F$ such that all the $w_{i}$ span a $k$-dimensional
subspace. Recall from section 3.3 that we can choose $w_{i}$ forming an almost
orthonormal subset for the above metric, in the sense that $\left\langle
w_{i},w_{j}\right\rangle=\delta_{i,j}+\varepsilon_{i,j}$.
The kernel of $d_{y}F$ is contained in $K=\langle\dfrac{\partial}{\partial
x_{n-k+1}},\dots,\dfrac{\partial}{\partial x_{n}}\rangle$ for all $y\in O$,
and thus $K=\langle w_{1},\dots,w_{k}\rangle$. Write $w_{i}=\sum_{j\geq
n-k+1}w_{i}^{j}\dfrac{\partial}{\partial x^{j}}$. Then we have
$\dfrac{\partial}{\partial x^{1}}\dfrac{\partial}{\partial
w_{i}}F(a)=z_{i}+R_{i}(a)$, for $z_{i}=\sum w_{i}^{k}\dfrac{\partial}{\partial
y^{k}}$, $\|R_{i}(a)\|<\varepsilon$ and $a\in O$. We deduce
$\dfrac{\partial}{\partial w_{i}}F(x)=\dfrac{\partial}{\partial
w_{i}}F(a_{i})+d_{i}(z_{i}+v_{i})=d_{i}(z_{i}+v_{i})$ for
$\|v_{i}\|<\varepsilon$.
By the form of the special coordinates, $x-y\in K$. Let $x-y=\sum b_{i}w_{i}$.
Since $|w_{i}|$ is almost $1$, there is an index $i_{0}$ such that
$|b_{i_{0}}|>\frac{1}{2n}\|x-y\|$. We have the identity:
$0=F(y)-F(x)=d_{x}F(y-x)+O(\|x-y\|^{2})=\sum
b_{i}d_{i}(z_{i}+v_{i})+O(\|x-y\|^{2})$
Multiplying the above by $\pm z_{j}$, we deduce
$d_{j}|b_{j}|=-\sum|b_{i}|d_{j}(\varepsilon_{i,j}+v_{i}z_{j})+O(\|x-y\|^{2})$,
which leads to
$\sum|b_{i}|d_{i}<C_{4}\|x-y\|^{2}$ (4.7)
At the point $x$, the image by $d_{x}F$ of the unit ball $B_{x}V$ in $T_{x}V$
is contained in a neighborhood of $Im(d_{a_{i}}F)$ of radius $2d_{i}$. We use
the identity
$\|dF^{-1}dG_{x}(\frac{x-y}{\|x-y\|})\|^{-1}=\sup\\{t:tdG_{x}(\frac{x-y}{\|x-y\|})\in
d_{x}F(B_{x}V)\\}$
We can assume the distance between the vectors $dG_{x}(\frac{x-y}{\|x-y\|})$
and $\sum\frac{b_{i}}{\|x-y\|}z_{i}$ is smaller than $\frac{1}{4n}$. In
particular, looking at the $i_{0}$ coordinate chosen above, we see that the
vector $dG_{x}(\frac{x-y}{\|x-y\|})$ needs to be rescaled at least by the
amount $8nd_{i}$ in order to fit within the image of the unit ball.
$\|dF^{-1}dG_{x}(\frac{x-y}{\|x-y\|})\|>\frac{1}{8nd_{i_{0}}}>\frac{|b_{i_{0}}|}{8nC_{4}\|x-y\|^{2}}>\frac{\varepsilon_{4}}{\|x-y\|}$
for $\varepsilon_{4}=\frac{1}{16n^{2}C_{4}}>0$, which is the desired
inequality. ∎
###### Proof of Lemma 2.14.
We prove that $\rho$ is Lipschitz close to a point $z^{0}$. Let $E$ be a
neighborhood of $z^{0}$ such that $\lambda_{1}$ has Lipschitz constant $L$,
and $\rho$ has Lipschitz constant $K$ for all $z\in E$ such that
$\rho(z)<\lambda(z)$. Let $z^{1},z^{2}\in E$ be such that
$\rho(z^{1})<\rho(z^{2})$.
If $\rho(z^{1})=\lambda_{1}(z^{1})$ we can compute
$|\rho(z^{2})-\rho(z^{1})|=\rho(z^{2})-\rho(z^{1})<\lambda(z^{2})-\lambda(z^{1})<L|z^{2}-z^{1}|$
where $L$ is a Lipschitz constant $L$ for $\lambda$ in $U$.
Otherwise take a linear path with unit speed
$\xi:[0,t]\rightarrow\partial\Omega$ from $z^{1}$ to $z^{2}$ and let $a$ be
the supremum of all $s$ such that $\rho(\xi(s))<\lambda(\xi(s))$. Then
$|\rho(z^{2})-\rho(z^{1})|<|\rho(z^{2})-\rho(\xi(a))|+|\rho(\xi(a))-\rho(z^{1})|$
The second term can be bound:
$|\rho(\xi(a))-\rho(z^{1})|<Ka$
If $\rho(z^{2})\geq\rho(\xi(a))$, we can bound the first term as
$|\rho(z^{2})-\rho(\xi(a))|=\rho(z^{2})-\rho(\xi(a))<\lambda(z_{2})-\lambda(\xi(a))<L|t-a|$
while if $\rho(z^{2})<\rho(\xi(a))$, we have
$|\rho(z^{2})-\rho(z^{1})|<|\rho(\xi(a))-\rho(z^{1})|$
so in all cases, the following holds:
$|\rho(z^{2})-\rho(z^{1})|<\max\\{L,K\\}t<\max\\{L,K\\}|z^{2}-z^{1}|$
∎
## 5 Proof of the main theorems.
Take the function $h$ associated to $S$ as in definition 2.4. At a cleave
point $x$ there are two geodesics arriving from $\partial\Omega$; each one
yields a value of $h$ by evaluation of $\tilde{u}$. The _balanced_ condition
implies that $\widehat{X}_{1}(v)=\widehat{X}_{2}(v)$ for the speed vectors
$X_{1}$ and $X_{2}$ of the characteristics reaching $x$ and any vector $v$
tangent to $S$. Furthermore, $\widehat{X}$ is exactly $dh$, so the difference
of the values of $h$ from either side is constant in every connected component
of the cleave locus.
We define an $(n-1)$-current $T$ in this way: Fix an orientation $\mathcal{O}$
in $\Omega$. For every smooth $(n-1)$ differential form $\phi$, restrict it to
the set of cleave points $\mathcal{C}$ (including degenerate cleave points).
In every component $\mathcal{C}_{j}$ of $\mathcal{C}$ compute the following
integrals
$\int_{\mathcal{C}_{j,i}}h_{i}\phi\qquad i=1,2$ (5.1)
where $\mathcal{C}_{j,i}$ is the component $\mathcal{C}_{j}$ with the
orientation induced by $\mathcal{O}$ and the incoming vector $V_{i}$, and
$h_{i}$ for $i=1,2$ are the limit values of $h$ from each side of
$\mathcal{C}_{j}$.
We define the current $T(\phi)$ to be the sum:
$T(\phi)=\sum_{j}\int_{\mathcal{C}_{j,1}}h_{1}\phi+\int_{\mathcal{C}_{j,2}}h_{2}\phi=\sum_{j}\int_{\mathcal{C}_{j,1}}(h_{1}-h_{2})\phi$
(5.2)
The function $h$ is bounded and the $\mathcal{H}^{n-1}$ measure of
$\mathcal{C}$ is finite (thanks to lemma 2.14) so that $T$ is a real flat
current that represents integrals of test functions against the difference
between the values of $h$ from both sides.
If $T=0$, we can apply lemma 3.11 and find $u=h$.
We will prove later that the boundary of $T$ as a current is zero. Assume for
the moment that $\partial T=0$. It defines an element of the homology space
$H_{n-1}(\Omega)$ of dimension $n-1$ with real coefficients. We can study this
space using the long exact sequence of homology with real coefficients for the
pair $(\Omega,\partial\Omega)$:
$\displaystyle 0\rightarrow H_{n}(\Omega)\rightarrow
H_{n}(\Omega,\partial\Omega)\rightarrow$ $\displaystyle
H_{n-1}(\partial\Omega)\rightarrow H_{n-1}(\Omega)\rightarrow
H_{n-1}(\Omega,\partial\Omega)\rightarrow\dots$ (5.3)
### 5.1 Proof of Theorem 2.9.
We prove that under the hypothesis of 2.9, the space $H_{n-1}(\Omega)$ is
zero, and then we deduce that $T=0$.
As $\Omega$ is open, $H_{n}(\Omega)\approx 0$. As $\Omega$ is simply
connected, it is orientable, so we can apply Lefschetz duality with real
coefficients ([Ha, 3.43]) which implies:
$H_{n}(\Omega,\partial\Omega)\approx H^{0}(\Omega)$
and
$H_{n-1}(\Omega,\partial\Omega)\approx H^{1}(\Omega)=0$
As $\partial\Omega$ is connected, we deduce $H_{n-1}(\Omega)$ has rank $0$,
and $T=\partial P$ for some $n$-dimensional flat current $P$. The flat top-
dimensional current $P$ can be represented by a density $f\in L^{n}(\Omega)$
(see [F, p 376, 4.1.18]):
$P(\omega)=\int_{\Omega}f\omega,,\qquad\omega\in\Lambda^{n}(\Omega)$ (5.4)
We deduce from (5.2) that the restriction of $P$ to any open set disjoint with
$S$ is closed, so $f$ is a constant in such open set. As $\Omega\setminus S$
is open and connected, $f$ is constant $a.e.$, and $T=0$.
### 5.2 Proof of Theorem 2.10.
Assume now that $\partial\Omega$ has $k$ connected components $\Gamma_{i}$. We
look at (5), and recall the map $H_{n-1}(\partial\Omega)\rightarrow
H_{n-1}(\Omega)$ is induced by inclusion. We know by Poincaré duality that
$H_{n-1}(\partial\Omega)$ is isomorphic to the linear combinations of the
fundamental classes of the connected components of $\partial\Omega$ with real
coefficients. We deduce that $H_{n-1}(\Omega)$ is generated by the fundamental
classes of the connected components of $\partial\Omega$, and that it is
isomorphic to the quotient of all linear combinations by the subspace of those
linear combinations with equal coefficients. Let
$R=\sum a_{i}\left[\Gamma_{i}\right]$
be the cycle to which $T$ is homologous (the orientation of $\Gamma_{i}$ is
such that, together with the inwards pointing vector, yields the ambient
orientation).
If we define $a(x)=a_{i}$, $\forall x\in\Gamma_{i}$, solve the HJ equations
with boundary data $g-a$ and compute the corresponding current $\widehat{T}$,
we see that $\widehat{T}=T-j_{\sharp}R$, where $j$ is the retraction $j$ of
$\Omega$ onto $S$ that fixes points of $S$ and follows characteristics
otherwise. Then the homology class of $\widehat{T}$ is zero, and we can prove
$\widehat{T}=0$ as before. It follows that $S$ is the singular set to the
solution of the Hamilton-Jacobi equations with boundary data $g-a$.
### 5.3 Proof of Theorem 2.12.
For this result we cannot simply use the sequence (5). We first give a
procedure for obtaining balanced split loci in $\Omega$ other than the cut
locus.
A function $a:[\partial\widetilde{\Omega}]\rightarrow\mathbb{R}$ that assigns
a real number to each connected component of $\partial\widetilde{\Omega}$ is
_equivariant_ iff for any automorphism of the cover $\varphi$ there is a real
number $c(\varphi)$ such that $a\circ\varphi=a+c(\varphi)$.
A function $a:[\partial\widetilde{\Omega}]\rightarrow\mathbb{R}$ is
_compatible_ iff $\widetilde{g}-a$ satisfies the compatibility condition
(1.3).
An equivariant function $a$ yields a group homomorphism from
$\pi_{1}(\Omega,\partial\Omega)$ into $\mathbb{R}$ in this way:
$\sigma\rightarrow a(\widetilde{\sigma}(1))-a(\widetilde{\sigma}(0))$ (5.5)
where $\sigma:[0,1]\rightarrow\Omega$ is a path with endpoints in
$\partial\Omega$ and $\widetilde{\sigma}$ is any lift to $\widetilde{\Omega}$
. The result is independent of the lift because $a$ is equivariant. On the
other hand, choosing an arbitrary component $[\Gamma_{0}]$ of $\partial\Omega$
and a constant $a_{0}=a([\Gamma])$, the formula:
$[\Gamma]\rightarrow a([\Gamma_{0}])+l(\pi\circ\tilde{\sigma}),\text{ for any
path }\tilde{\sigma}\text{ with
}\tilde{\sigma}(0)\in\Gamma_{0},\sigma(1)\in\Gamma$ (5.6)
assigns an equivariant function $a$ to an element $l$ of
$Hom(\pi_{1}(\Omega,\partial\Omega),\mathbb{R})\sim
H^{1}(\Omega,\partial\Omega)$.
Up to addition of a global constant, these two maps are inverse of one
another, so there is a one-to-one correspondence between elements of
$H^{1}(\Omega,\partial\Omega)$ and equivariant functions $a$ (with $a+c$
identified with $a$ for any constant $c$). The compatible equivariant
functions up to addition of a global constant can be identified with an open
subset of $H^{1}(\Omega,\partial\Omega)$ that contains the zero cohomology
class.
Let $\widetilde{\Omega}$ be the universal cover of $\Omega$. We can lift the
Hamiltonian $H$ to a function $\widetilde{H}$ defined on
$T^{\ast}\widetilde{\Omega}$ and the function $g$ to a function $\tilde{g}$
defined on $\partial\widetilde{\Omega}$. The preimage of a balanced split
locus for $\Omega$, $H$ and $g$ is a balanced split locus for
$\widetilde{\Omega}$, $\widetilde{H}$ and $\tilde{g}$ that is invariant by the
automorphism group of the cover, and conversely, a balanced split locus
$\widetilde{S}$ in $\widetilde{\Omega}$ that is invariant by the automorphism
group of the cover descends to a balanced split locus on $\Omega$.
Any function $a$ that is both equivariant and compatible can be used to solve
the Hamilton-Jacobi problem $\widetilde{H}(p,du(p))=1$ in $\widetilde{\Omega}$
and $u(p)=\widetilde{g}(p)-a(p)$. If $\pi_{1}(\Omega)$ is not finite,
$\widetilde{\Omega}$ will not be compact, but this is not a problem (see
remark 5.5 in page 125 of [L]). The singular set is a balanced split locus
that is invariant under the action of $\pi_{1}(\Omega)$ and hence it yields a
balanced split locus in $\Omega$. We write $S[a]$ for this set. It is not hard
to see that the map $a\rightarrow S[a]$ is injective.
Conversely, a balanced split locus in $\Omega$ lifts to a balanced split locus
$\widetilde{S}$ in $\widetilde{\Omega}$. The reader may check that the current
$T_{\widetilde{S}}$ is the lift of $T_{S}$, and in particular it is closed. As
in the proof of Theorem 2.10, we have $H^{1}(\widetilde{\Omega})=0$, and we
deduce
$T_{\widetilde{S}}=\sum_{j}a_{j}[\Lambda_{j}]+\partial P$
where $\Lambda_{j}$ are the connected components of
$\partial\widetilde{\Omega}$.
This class is the lift of the class of $T\in H_{n-1}(\Omega)$ and thus it is
invariant under the action of the group of automorphisms of the cover.
Equivalently, the map defined in (5.5) is a homomorphism. Thus $a$ is
equivariant. Similar arguments as before show that $S=S[a]$.
Thus the map $a\rightarrow S[a]$ is also surjective, which completes the proof
that there is a bijection between equivariant compatible functions
$a:[\partial\widetilde{\Omega}]\rightarrow\mathbb{R}$ and balanced split loci.
## 6 Proof that $\partial T=0$
It is enough to show that $\partial T=0$ at all points of $\Omega$ except for
a set of zero $(n-2)$-dimensional Hausdorff measure. This is clear for points
not in $S$. Due to the structure result 3.5, we need to show the same at
cleave points (including degenerate ones), edge points and crossing points.
Along the proof, we will learn more about the structure of $S$ near those
kinds of points.
Throughout this section, we assume $n=dim(\Omega)>2$. This is only to simplify
notation, but the case $n=2$ is covered too. We shall comment on the necessary
changes to cover the case $n=2$, but do not bother with the simple case $n=1$.
### 6.1 Conjugate points of order $1$.
We now take a closer look at points of $A(S)$ that are also conjugate points
of order $1$. Fortunately, because of 3.5 we do not need to deal with higher
order conjugate points. In a neighborhood $O$ of a point $x^{0}$ of order $1$,
in the special coordinates of section 3.2, we have $x^{0}=0$ and $F$ looks
like:
$\displaystyle
F(x_{1},x_{2},\dots,x_{n})=(x_{1},x_{2},\dots,F_{n}(x_{1},\dots,x_{n}))$ (6.1)
Let $\widetilde{S}$ be the boundary of $A(S)$, but without the points $(0,z)$
for $z\in\partial\Omega$. It follows from 2.14 that $\widetilde{S}$ is a
Lipschitz graph on coordinates given by the vector field $r$ and $n-1$
transversal coordinates. It is not hard to see that it is also a Lipschitz
graph $x_{1}=\tilde{t}(x_{2},\dots,x_{n})$ in the above coordinates $x_{i}$,
possibly after restricting to a smaller open set.
Because of Lemma 4.1, we know $x^{0}$ is a first conjugate point, so we can
assume that $O$ is a coordinate cube
$\prod(-\varepsilon_{i},\varepsilon_{i})$, and that $F$ is a diffeomorphism
when restricted to $\\{x_{1}=s\\}$ for $s<-\varepsilon_{1}/2$.
###### Definition 6.1.
A set $O\subset V$ is _univocal_ iff for any $p\in\Omega$ and $x^{1},x^{2}\in
Q_{p}\cap O$ we have $\tilde{u}(x^{1})=\tilde{u}(x^{2})$.
#### Remark.
The most common case of univocal set is a set $O$ such that $F|O$ is
injective.
###### Lemma 6.2.
Let $x^{0}\in V$ be a conjugate point of order 1. Then $x^{0}$ has an univocal
neighborhood.
###### Proof.
Let $O_{1}$ and $U_{1}$ be neighborhoods of $x^{0}$ and $F(x^{0})$ where the
special coordinates (6.1) hold; let $x_{i}$ be the coordinates in $O_{1}$ and
$y_{i}$ be those in $U_{1}$.
Choose smaller $U\subset U_{1}$ and $O\subset F^{-1}(U)\cap O_{1}$ so that we
can assume that if a point $x^{\prime}\in V\setminus O_{1}$ maps to a point in
$U$, then for the vector $Z=dF_{x^{\prime}}(r)$ we have
$\hat{Z}(\frac{\partial}{\partial y_{1}})<\hat{X}(\frac{\partial}{\partial
y_{1}})$ (6.2)
for any $X=dF_{x}(r)$ with $x\in O$ and also
$\hat{Y}(\frac{\partial}{\partial y_{1}})>1-k$ (6.3)
for some $k>0$ sufficiently small and all $Y=dF_{y^{\prime}}(r)$ for
$y^{\prime}\in O_{1}$.
Take $x^{1},x^{2}\in Q_{q}\cap O$ for $q\in U$. The hypothesis $x^{1},x^{2}\in
Q_{q}$ implies $q=F(x^{1})=F(x^{2})$, and so $x^{1}_{j}=x^{2}_{j}$ follows for
all $j<n$. Let us write $a_{j}=x^{1}_{j}=x^{2}_{j}$ for $j<n$,
$s^{1}=x^{1}_{n}$ and $s^{2}=x^{2}_{n}$. Fix $a_{2},\dots,a_{n-1}$ and
consider the set
$H_{a}=\left\\{x\in O:x_{i}=a_{i};i=2,\dots,n-1\right\\}$
Its image by $F$ is a subset of a plane in the $y_{i}$ coordinates:
$L_{a}=\left\\{y\in U:y_{i}=a_{i},i=2,\dots,n-1\right\\}$
Points of $O_{1}$ not in $H_{a}$ map to other planes. If $n=2$, we keep the
same notation, but the meaning is that $H_{a}=O$ and $L_{a}=V$.
There is $\varepsilon>0$ such that for $t<-\varepsilon/2$, the line
$\\{x_{1}=t\\}\cap H_{a}$ maps diffeomorphically to $\\{y_{1}=t\\}\cap L_{a}$.
Due to the comments at the beginning of this section, $\widetilde{S}$ is given
as a Lipschitz graph $x_{1}=\tilde{t}(x_{2},\dots,x_{n})$. The identity
$a_{1}=\tilde{t}(a_{2},\dots,a_{n-1},s^{i})$ holds for $i=1,2$ because
$x^{1},x^{2}\in Q_{q}$. We define a curve
$\sigma:[s^{1},s^{2}]\rightarrow\widetilde{S}$ by
$\sigma(s)=(\tilde{t}(a_{2},\dots,a_{n-1},s),a_{2},\dots,a_{n-1},s)$. The
image of $\sigma$ by $F$ stays in $S$, describing a closed loop based at $q$;
we will establish the lemma by examining the variation of $\tilde{u}$ along
$\sigma$.
For $i=1,2$, let $\eta^{i}:(-\varepsilon_{i},a_{1}]\rightarrow H_{a}$ given by
$\eta^{i}(t)=(t,a_{2},\dots,a_{n-1},s^{i})$ be the segments parallel to the
$x_{1}$ direction that end at $x^{i}$, defined from the first point in the
segment that is in $O$. We can assume that the intersection of $O$ with any
line parallel to $\frac{\partial}{\partial x_{1}}$ is connected, and that the
intersection of $U$ with any line parallel to $\frac{\partial}{\partial
y_{1}}$ is connected too. We can also assume $\varepsilon_{i}<\varepsilon$.
Let $D$ be the closed subset of $H_{a}$ delimited by the Lipschitz curves
$\eta^{1}$, $\eta^{2}$ and $\sigma$, and let $E$ be the closed subset of
$L_{a}$ delimited by the image of $\eta^{1}$ and $\eta^{2}$.
We claim $D$ is mapped onto $E$. First, no point in $int(D)$ can map to the
image of the two lines, cause this contradicts either $\rho\leq\lambda_{1}$,
or the fact that $\rho(a_{2},\dots,a_{n-1},s^{i})$ is the first time that the
line parallel to the $x_{1}$ direction hits $\widetilde{S}$, for either $i=1$
or $i=2$. We deduce $D$ is mapped into $E$.
Now assume $G=E\setminus F(D)$ is nonempty, and contains a point
$p=(p_{1},\dots,p_{n})$. If $Q_{p}$ contains a point $x\in O_{1}\setminus
F(D)$, following the curve $t\rightarrow(t,x_{2},\dots,x_{n})$ backwards from
$x=(x_{1},x_{2},\dots,x_{n})$, we must hit either a point in the image of
$\eta^{i}|_{(-\varepsilon_{1},a_{1})}$ (which is a contradiction with the fact
that both $(t,\dots,x_{n})$ for $t<x_{1}$ and $(t,a_{2},\dots,a_{n-1},s^{i})$
for $t<a_{1}$ are in $A(S)$; see definition 2.1), or the point $q$ (which
contradicts (6.3)). Thus for any point $p\in G$, we have $Q_{p}\subset
V\setminus O_{1}$.
Now take a point $p\in\partial G$, and pick up a sequence approaching it from
within $G$ and contained in a line with speed vector $\frac{\partial}{\partial
y_{1}}$. By the above, the set $Q$ for points in this sequence is contained in
$V\setminus O_{1}$. We can take a subsequence carrying a convergent sequence
of vectors, and thus $R_{p}$ has a vector of the form $dF_{x^{\ast}}(r)$ for
$x^{\ast}\in D\subset O$. This violates the balanced condition, because of
(6.2). This implies $\partial G=\emptyset$, thus $G=\emptyset$ because $E$ is
connected and $F(D)\neq\emptyset$.
Finally, we claim there are no vectors coming from $V\setminus O_{1}$ in
$R_{p}$ for $p\in int(E)$. The argument is as above, but we now approach a
point with a vector from $V\setminus O_{1}$ in $R_{p}$ within $E=F(D)$ and
with speed $-\frac{\partial}{\partial y_{1}}$. The approaching sequence may be
chosen so that it carries a convergent sequence of vectors from $F(D)$, and
again (6.2) gives a contradiction with the balanced condition.
We now compute:
$\tilde{u}(x^{1})-\tilde{u}(x^{2})=\int_{s_{1}}^{s_{2}}\frac{d(\tilde{u}\circ\sigma)}{ds}=\int_{\sigma}d\tilde{u}$
(6.4)
The curve $F\circ\sigma$ runs through points of $S$. If $F(\sigma(s))$ is a
cleave point, then $F\circ\sigma$ is a smooth curve near $s$. We show that
cleave points are the only contributors to the above integral. If a point is
not cleave, either it is the image of a conjugate vector, or has more than $2$
incoming geodesics. As $F\circ\sigma$ maps into $int(E)$, all vectors in
$R_{F(\sigma(s))}$ come from $O$.
Let $N$ be the set of $s$ such that $\sigma(s)$ is conjugate. We notice that
$\sigma(s)$ is not an A2 point for $s\in N$. This is proposition 6.2 in [AG],
and is a standard result for cut loci in Riemannian manifolds. This means that
at those points the kernel of $dF$ is contained in the tangent to
$\widetilde{S}$. The intersection of $\widetilde{S}$ with the plane $H_{a}$ is
the image of the curve $\sigma$. Thus, for $s\in N$ the tangent to the curve
$\lambda_{1}$ is the kernel of $d_{\sigma(s)}F$. If $\sigma$ is differentiable
at a point $s$ we deduce, thanks to 4.1, that the tangent to the curve
$\lambda_{1}$ is the kernel of $d_{\sigma(s)}F$.
We now use a _variation of length_ argument to get a variant of the Finsler
Gauss lemma. Let $c=(l,w)$ be a tangent vector to
$V\subset\mathbb{R}\times\partial\Omega$ at the point $x=(t,z)$, and assume
$d_{x}F(c)=0$. We show that this implies $d\tilde{u}(c)=0$. Let $\gamma_{s}$
be a variation through geodesics with initial point in $z(s)\in\partial\Omega$
and the characteristic vector field at $z(s)$ as the initial speed vector,
such that $\frac{\partial}{\partial s}z(s)=w$, and with total length $t+sl$ .
By the first variation formula and the equation for the characteristic vector
field at $\partial\Omega$, the variation of the length of the curve
$\gamma_{s}$ is $\frac{\partial\varphi}{\partial v}(p,d_{x}F(r_{x}))\cdot
d_{x}F(c)-\frac{\partial\varphi}{\partial v}(p,d_{z}F(r))\cdot w=-dg(w)$, and
by the definition of $\gamma_{s}$, it is also $l$. We deduce $l=-dg(w)$, and
thus $d\tilde{u}(c)=l+dg(w)=0$.
It follows that $d_{\sigma(s)}F(\sigma^{\prime}(s))=0$ at points $s\in N$
where $\sigma$ is differentiable. As $\sigma$ is Lipschitz, the set of $s$
where it is not differentiable has measure $0$, and we deduce:
$\int_{N}d\tilde{u}(\sigma^{\prime})=0$
$N$ is contained in the set of points where $d(F\circ\sigma)$ vanishes. Thus,
by the Sard-Federer theorem, the image of $N$ has Hausdorff dimension $0$.
Let $\Sigma_{2}$ be the set of points in $L_{a}$ with more than $2$ incoming
geodesics. From the proof of [AG, 7.3], we see that the tangent to
$\Sigma_{2}$ has dimension $0$ and thus $\Sigma_{2}$ has Hausdorff dimension
$0$.
As $F$ is non-singular at points in $[s_{1},s_{2}]\setminus N$, the set of $s$
in $[s_{1},s_{2}]\setminus N$ mapping to a point in $\Sigma_{2}\cup N$ has
measure zero.
Altogether, we see that the integral (6.4) can be restricted to the set $C$ of
$s$ mapping to a cleave point. $C$ is an open set and thus can be expressed as
the disjoint union of a countable amount of intervals. Let $A_{1}$ be one of
those intervals. It is mapped by $F\circ\sigma$ diffeomorphically onto a
smooth curve $c_{0}$ of cleave points contained in $L_{a}$. Points of the form
$(t,a_{2},\dots,a_{n-1},s)$ for $t<\tilde{t}(a_{2},\dots,a_{n-1},s)$ map
through $F$ to a half open ball in $E$. There must be points of $D$ mapping to
the other side of $c_{0}$. Because of all the above, $c_{0}$ is also the image
of other points in $[s_{1},s_{2}]$. As $c$ is made of cleave points, it must
be the image of another component of $C$, which we call $B_{1}$, also mapping
diffeomorphically onto $c_{0}$. Choose a new component $A_{2}$, which is
matched to another component $B_{2}$, different from the above, and so on,
till the $A_{i}$ and $B_{i}$ are all the components of $C$.
We can write the integral on $B_{i}$ as an integral on $A_{i}$ (we add a minus
sign, because the curve is traversed in opposite directions):
$\int_{A_{i}}d\tilde{u}(\sigma^{\prime})+\int_{B_{i}}d\tilde{u}(\sigma^{\prime})=\int_{A_{i}}d\tilde{u}_{l}(\sigma^{\prime})-d\tilde{u}_{r}(\sigma^{\prime})$
where $d\tilde{u}_{l}$ and $d\tilde{u}_{r}$ are the values of $d\tilde{u}$
computed from both sides. The balanced condition implies
$\sigma^{\prime}\in\ker(d\tilde{u}_{l}-d\tilde{u}_{r})$, and thus the above
integral vanishes. The integral (6.4) is absolutely convergent by Lemma 2.14,
and the proof follows.
∎
#### Remark.
The above proof took some inspiration from [H, 5.2]. The reader may be
interested in James Hebda’s _tree-like curves_.
### 6.2 Structure of S near cleave and crossing points
In this section we prove some more results about the structure of a balanced
split locus near degenerate cleave and crossing points. Besides their
importance for proving that $\partial T=0$, we believe they are interesting in
their own sake.
###### Lemma 6.3.
Let $p\in S$ be a (possibly degenerate) _cleave_ point, and let
$Q_{p}=\\{x^{1},x^{2}\\}$.
There are disjoint univocal neighborhoods $O_{1}$ and $O_{2}$ of $x^{1}$ and
$x^{2}$, and a neighborhood $U$ of $p$ such that for any $q\in U$, $Q_{q}$ is
contained in $O_{1}\cup O_{2}$.
Furthermore, if we define:
$A_{i}=\left\\{q\in U\text{ such that }Q_{q}\cap O_{i}\neq\emptyset\right\\}$
for $i=1,2$, then $A_{1}\cap A_{2}$ is the graph of a Lipschitz function, for
adequate coordinates in $U$.
###### Proof.
The points $x^{1}$ and $x^{2}$ are at most of first order, so we can take
univocal neighborhoods $O_{1}$ and $O_{2}$ of $x^{1}$ and $x^{2}$. By
definition of $Q_{p}$ and the compactness of $\Omega$, we can achieve the
first property, reducing $U$ if necessary.
We know $\widehat{d_{x^{1}}F(r)}$ is different from $\widehat{d_{x^{2}}F(r)}$.
For fixed arbitrary coordinates in $U$, we can assume that
$\\{\widehat{d_{x}F(r)}\text{ for }x\in O_{1}\\}$ can be separated by a
hyperplane from $\\{\widehat{d_{x}F(r)}\text{ for }x\in O_{2}\\}$, after
reducing $U$, $O_{1}$, $O_{2}$ if necessary. Therefore, there is a vector
$Z_{0}\in T_{p}\Omega$ and a number $\delta>0$ such that
$\widehat{d_{x}F(r)}(Z)<\widehat{d_{x^{\prime}}F(r)}(Z)+\delta\qquad\forall\,x\in
O_{1},\,x^{\prime}\in O_{2}$ (6.5)
for any unit vector $Z$ in a neighborhood $G$ of $Z_{0}$. Let
$C^{+}=\\{tZ:t>0,Z\in G\\}$ be a one-sided cone containing $Z$. We write
$q+C^{+}$ for the cone displaced to have a vertex in $q$.
Choose $q\in A_{1}\cap A_{2}$, and $Z\in G$. Let $\mathcal{R}=\\{q^{\prime}\in
U:q^{\prime}=q+tZ,t>0\\}$ be a ray contained in $(q+C^{+})\cap U$. We claim
$\mathcal{R}\subset A_{1}\setminus A_{2}$.
For two points $q_{1}=q+t_{1}Z,q_{2}=q+t_{2}Z\in\mathcal{R}$, we say
$q_{1}<q_{2}$ if and only $t_{1}<t_{2}$. If $\mathcal{R}\cap
A_{2}\neq\emptyset$, let $q_{0}$ be the infimum of all points $p>0$ in
$\mathcal{R}\cap A_{2}$, for the above order in $\mathcal{R}$. If $q_{0}\in
A_{1}$ (whether $q_{0}=q$ or not), we can approach $q_{0}$ with a sequence of
points $q_{n}=F(x_{n})>q_{0}$ carrying vectors $d_{x_{n}}F(r)$ with $x_{n}\in
O_{2}$. The limit point of this sequence is $q_{0}$, and the limit vector is
$d_{x}F(r)$ for some $x\in O_{2}$, but the incoming vector is in $-G$, which
contradicts the balanced condition by (6.5).
If $q_{0}\in A_{2}\setminus A_{1}$, then approaching $q_{0}$ with points
$q<q_{n}=F(x_{n})<q_{0}$, we get a new contradiction with the balanced
property. The only possibility is $\mathcal{R}\subset A_{1}\setminus A_{2}$.
As the vector $Z$ is arbitrary, we have indeed $(q+C^{+})\cap U\subset
A_{1}\setminus A_{2}$.
Fix coordinates in $U$, and let $\varepsilon=\frac{1}{2}dist(p,\partial U)$.
Let $B_{\varepsilon}$ be the ball of radius $\varepsilon$ centered at $p$. By
the above, the hypothesis of lemma 3.7 are satisfied, for $A=A_{1}\setminus
A_{2}$, the cone $C^{+}$, the number $\varepsilon$, and $V=B_{\varepsilon}$.
Thus, we learn from lemma 3.7 that $A_{1}\cap A_{2}\cap B_{\varepsilon}$ is
the graph of a Lipschitz function along the direction $Z_{0}$ from any
hyperplane transversal to $Z_{0}$.
∎
The following three lemmas contain more detailed information about the
structure of a balanced split locus near a crossing point. The following is
stated for the case $n>2$, but it holds too if $n=2$, though then $L$ reduces
to a single point $\\{a\\}$.
###### Definition 6.4.
The _normal_ to a subset $X\subset T_{p}^{\ast}\Omega$ is the set of vectors
$Z$ in $T_{p}\Omega$ such that $\omega(Z)$ is the same number for all
$\omega\in X$.
###### Lemma 6.5.
Let $p\in S$ be a crossing point. Let $B\subset T_{p}^{\ast}\Omega$ be the
affine plane spanned by $R_{p}^{\ast}$. Let $L$ be the normal to $B$, which by
hypothesis is a linear space of dimension $n-2$, and let $C$ be a (double-
sided) cone of small amplitude around $L$.
There are disjoint univocal open sets $O_{1},\dots,O_{N}\subset V$ and an open
neighborhood $U$ of $p$ such that $Q_{q}\subset\cup_{i}O_{i}$ for all $q$ in
$U$.
Furthermore, define sets $A_{i}$ as in lemma 6.3, and call
$\mathcal{S}=\cup_{i,j}A_{i}\cap A_{j}$ the _essential part_ of $S$. Define
$\Sigma=\cup_{i,j,k}A_{i}\cap A_{j}\cap A_{k}$ and let
$\mathcal{C}=\mathcal{S}\setminus\Sigma$.
1. (1)
At every $q\in\Sigma$, there is $\varepsilon>0$ such that
$\Sigma\cap(q+B_{\varepsilon})\subset q+C$.
2. (2)
$\Sigma$ itself is contained in $p+C$.
The next lemma describes the intersection of $\mathcal{S}$ with $2$-planes
transversal to $L$.
###### Lemma 6.6.
Let $p\in S$ be a crossing point as above. Let $P\subset T_{p}\Omega$ be a
$2$-plane intersecting $C$ only at the origin, and let $P_{a}=P+a$ be a
$2$-plane parallel to $P$ for $a\in L$.
1. 1.
If $|a|<\varepsilon_{1}$, the intersection of $\mathcal{S}$, the plane
$P_{a}$, and $U$ is a connected Lipschitz tree.
2. 2.
The intersection of $\mathcal{S}$, the plane $P_{a}$, and the annulus of inner
radius $c\cdot|a|$ and outer radius $\varepsilon_{2}$:
$A(c\>|a|,\varepsilon_{2})=\\{q\in U:c\>|a|<|q|<\varepsilon_{2}\\}$
is the union of $N$ Lipschitz arcs separating the sets $A_{i}$.
Figure 2: Two possible intersections of a plane $P_{a}$ with $\mathcal{S}$
#### Remark.
We cannot say much about what happens inside $P_{a}\cap B(P,c\>|a|)$. The
segments in $P_{a}\cap A(c\>|a|,\varepsilon_{2})$ must meet together, yielding
a connected tree, but this can happen in several different ways (see figure
2).
Finally, we can describe the connected components of
$\mathcal{C}=\mathcal{S}\setminus\Sigma$ within $U$:
###### Lemma 6.7.
Under the same hypothesis, for every $i=1,\dots,N$ there is a coordinate
system in $U$ such that:
* •
The set $\partial A_{i}$ is the graph of a Lipschitz function $h_{i}$, its
domain delimited by two Lipschitz functions $f_{l}$ and $f_{r}$, for
$L^{\ast}\subset L$:
$\partial A_{i}=\left\\{(a,t,h_{i}(t)),a\in
L^{\ast},f_{l}(a)<t<f_{r}(a)\right\\}$
* •
A connected component $\mathcal{C}_{0}$ of $\mathcal{C}$ contained in
$\partial A_{i}$ admits the following expression, for Lipschitz functions
$f_{1}$ and $f_{2}$, for $L_{0}\subset L$:
$\mathcal{C}_{0}=\left\\{(a,t,h_{i}(t)),a\in
L_{0},f_{1}(a)<t<f_{2}(a)\right\\}$
###### Corollary 6.8.
$\mathcal{H}^{n-2}(\Sigma)<\infty$.
###### Proof of corollary.
We apply the _general area-coarea formula_ (see [F, 3.2.22]), with $W=\Sigma$,
$Z=L$, and $f$ the projection from $U$ onto $L$ parallel to $P$, and
$m=\mu=\nu=n-2$, to learn:
$\int_{\Sigma}ap\>Jfd\mathcal{H}^{n-2}=\int_{L}\mathcal{H}^{0}(f^{-1}(\\{z\\}))d\mathcal{H}^{n-2}(z)=\int_{L}\mathcal{H}^{0}(\Sigma\cap
P_{a})d\mathcal{H}^{n-2}(a)$
$ap\>Jf|_{\Sigma}$ is bounded from below, so if we can bound
$\mathcal{H}^{0}(\Sigma\cap P_{a})$ uniformly, we get a bound for
$\mathcal{H}^{n-2}(\Sigma)$.
The set $\mathcal{C}\cap P_{a}\cap U$ is a simplicial complex of dimension
$1$, and a standard result in homology theory states that the number of edges
minus the number of vertices is the same as the difference between the
homology numbers of the complex: $h^{1}-h^{0}$. The graph is connected and
simply connected, so this last number is $-1$. The vertices of
$\mathcal{C}\cap P_{a}\cap U$ consist of $N$ vertices of degree $1$ lying at
$\partial U$ and the interior vertices having degree at least $3$. The
_handshaking lemma_ states that the sum of the degrees of the vertices of a
graph is twice the number of edges, so we get the inequality $2e\geq
N+3\bar{v}$ for the number $e$ of edges and the number $\bar{v}$ of interior
vertices. Adding this to the previous equality $e-(N+\bar{v})=-1$, we get
$\bar{v}\leq N-2$. We have thus bounded $\bar{v}=\mathcal{H}^{0}(\Sigma\cap
P_{a})$ with a bound valid for all $a$.
∎
###### Proof of 6.5.
This lemma can be proven in a way similar to 6.3, but we will take some extra
steps to help us with the proof of the other lemmas.
First, recall the map $\Delta$ defined in (3.4). Each point $x$ in
$\Delta^{-1}(R_{p}^{\ast})$ has a univocal neighborhood $\mathcal{O}_{x}$.
Recall $R_{p}^{\ast}$ consists only of covectors of norm $1$. Let $\gamma$ be
the curve obtained as intersection of $B$ and the covectors of norm $1$.
Instead of taking the neighborhoods $\mathcal{O}_{x}$ right away, which would
be sufficient for this lemma, we cover $R_{p}^{\ast}$ with open sets of the
form $\Delta(\mathcal{O}_{x})\cap\gamma$.
By standard results in topology, we can extract a finite refinement of the
covering of $R_{p}^{\ast}\subset\gamma$ by the sets
$\Delta(\mathcal{O}_{x})\cap\gamma$ consisting of disjoint non-empty intervals
$I_{1},\dots,I_{N}$. Let $\tilde{I}_{i}$ be the set of points $tx$ for
$t\in(1-\varepsilon_{1},1+\varepsilon_{1})$ and $x\in I_{i}$, and choose a
linear space $M_{0}$ of dimension $n-2$ transversal to $B$. Define the sets of
our covering:
$O_{i}=\Delta^{-1}(\tilde{I}_{i}+B(M_{0},\varepsilon_{2}))$
for the ball of radius $\varepsilon_{2}$ in $M_{0}$ ($\varepsilon_{1}$ and
$\varepsilon_{2}$ are arbitrary, and small).
We can assume that $Q_{q}\subset\cup_{i}O_{i}$ for all $q$ in $U$ by reducing
$U$ and the $O_{i}$ further if necessary, hence we only need to prove the two
extra properties to conclude the theorem.
The approximate tangent to $\Sigma$ at a point $q\in\Sigma\cap U$ is contained
in the normal to $R_{q}^{\ast}$ (see the definition of approximate tangent in
[AG] and use the proof of proposition 7.3 there, or merely use the balanced
property). If $R_{q}^{\ast}$ is contained in a sufficiently small neighborhood
of $\gamma$ and contains points from at least three different $I_{i}$, its
normal must be close to $L$. Thus if we chose $\varepsilon_{1}$ and
$\varepsilon_{2}$ small enough, the approximate tangent to $\Sigma\cap U$ at a
point $q\in\Sigma$ is contained in $C$. If property (1) did not hold for any
$\varepsilon$ at a point $q$, we could find a sequence of points converging to
$q$ whose directions from $q$ would remain outside $C$, violating the above
property.
Finally, the second property holds if we replace $U$ by $U\cap
B_{\varepsilon}$, for the number $\varepsilon$ that appears when we apply
property (1) to $p$. ∎
Figure 3: $\mathcal{S}$ near a crossing point
###### Proof of 6.6.
Just like in 6.3, we can assume that each set $\\{\widehat{d_{x}F(r)}\text{
for }x\in O_{i}\\}$ can be separated from the others by a hyperplane (e.g., a
direction $Z_{i}$), such that:
$\widehat{d_{x}F(r)}(Z)<\widehat{d_{x^{\prime}}F(r)}(Z)+\delta\qquad\forall\,x\in
O_{i},\,x^{\prime}\in O_{j},\,i\neq j$ (6.6)
for some $\delta>0$ and any unit vector $Z$ in a neighborhood $G_{i}$ of
$Z_{i}$. Thanks to the care we took in the proof of the previous lemma, we can
assume all $Z_{i}$ belong to the plane $P$ in the statement of this lemma:
indeed the intervals $I_{i}$ can be separated by vectors in any plane
transversal to $L$, and the sets $\Delta(O_{i})$ are contained in
neighborhoods of the $I_{i}$.
Define the one-sided cones $C^{+}_{i}=\\{tZ:t>0,Z\in G_{i}\\}$. The above
implies that the intersection of each $C_{i}^{+}$ with $P$ is a nontrivial
cone in $P$ that consists of rays from the vertex.
By the same arguments in 6.3, we can be sure that whenever $q\in A_{i}$, then
$(q+C^{+}_{i})\cap U\subset A_{i}$. This implies that $\partial A_{i}$ is the
graph of a Lipschitz function along the direction $Z_{i}$ from any hyperplane
transversal to $Z_{i}$. We notice $\partial A_{i}$ is (Lipschitz) transversal
to $P$, so for any $a\in L$, $\partial A_{i}\cap P$ is a Lipschitz curve. As
the cone $C$ is transversal to $P$, and the tangent to $\Sigma$ is contained
in $C$, we see $\Sigma\cap P_{a}$ consists of isolated points.
Thus $\mathcal{S}\cap P_{a}$ is a Lipschitz graph and $\Sigma\cap P_{a}$ is
the set of its vertices. If it were not a tree, there would be a bounded open
subset of $P_{a}\cap U\setminus\mathcal{S}$ with boundary contained in
$\mathcal{S}$. An interior point $q$ belongs to some $A_{k}$. Then the cone
$q+C^{+}_{k}$ is contained in $A_{k}$, but on the other hand its intersection
with $P_{a}$ contains a ray that must necessarily intersect $\mathcal{S}$,
which is a contradiction.
We notice $P_{a}\cap(p+C^{+}_{i})\subset A_{i}$. This set is a cone in $P_{a}$
(e.g. a circular sector) with vertex at most a distance $c_{1}|a|$ from $p+a$,
where $c_{1}>0$ depends on the amplitude of the different $C_{i}$.
If $a=0$, the $N$ segments departing from $p$ with speeds $Z_{i}$ belong to
each $A_{i}$ respectively. Let us assume that the intervals $I_{i}$ appearing
in the last proof are met in the usual order $I_{1},I_{2}\dots,I_{N}$ when we
run along $\gamma$ following a particular orientation, and call $P^{i}$ the
region delimited by the rays from $p$ with speeds $Z_{i}$ and $Z_{i+1}$ (read
$Z_{1}$ instead of $Z_{N+1}$).If there is a point $q\in P^{i}\cap A_{k}\cap
B(\varepsilon_{2})$ for sufficiently small $\varepsilon_{2}$, then
$(q+C^{+}_{k})\cap U$ would intersect either $p+C^{+}_{i}$ or $p+C^{+}_{i+1}$,
and yield a contradiction if $k$ is not $i$ or $i+1$. Thus $P^{i}\subset
A_{i}\cup A_{i+1}$. Clearly there must be some point $q$ in $P_{i}\cap
A_{i}\cap A_{i+1}$, to which we can apply lemma 6.3. $A_{i}\cap A_{i+1}$ is a
Lipschitz curve near $q$ transversal to $Z_{i}$ (and to $Z_{i+1}$), and it
cannot turn back. The curve does not meet $\Sigma$, and it cannot intersect
the rays from $p$ with speeds $Z_{i}$ and $Z_{i+1}$, so it must continue up to
$p$ itself. For any $q\in A_{i}\cap A_{i+1}$, the cone $q+C^{+}_{i}$ is
contained in $A_{i}$, and the cone $q+C^{+}_{i+1}$ is contained in $A_{i+1}$.
This implies there cannot be any other branch of $A_{i}\cap A_{i+1}$ inside
$P_{i}$.
This is all we need to describe $\mathcal{S}\cap P\cap B(\varepsilon_{2})$: it
consists of $N$ Lipschitz segments starting at $p$ and finishing in
$P\cap\partial B(\varepsilon_{2})$. The only multiple point is $p$.
For small positive $|a|$, we know by condition (2) of the previous lemma that
$P_{a}\cap\Sigma\subset C\cap P_{a}=B(c_{2}|a|)\cap P_{a}$ for some $c_{2}>0$.
Similarly as above, define regions $P_{a}^{i}\subset P_{a}\cap
A(c|a|,\varepsilon_{2})$ delimited by the rays from $a$ with directions
$Z_{i}$ and $Z_{i+1}$, and the boundary of the ring $A(c|a|,\varepsilon_{2})$,
for a constant $c>\max(c_{1},c_{2})$. Take $c$ big enough so that for any
$q\in P_{a}^{i}$ and any $k\neq i,i+1$ , $q+C^{+}_{k}\cap U\cap P_{a}$
intersects either $p+C^{+}_{i}$ or $p+C^{+}_{i+1}$. The same argument as above
shows that $A_{i}\cap A_{i+1}\subset P_{a}^{i}\subset A_{i}\cup A_{i+1}$. We
conclude there must be a Lipschitz curve of points of $A_{i}\cap A_{i+1}$,
which starts in the inner boundary of $A(c|a|,\varepsilon_{2})$, and ends up
in the outer boundary.
∎
Figure 4: A neighborhood of a crossing point (this view is rotated with
respect to figure 3)
###### Proof of 6.7.
First we assume $U$ has a product form $U=L^{\ast}\times P^{\ast}$ for open
discs $L^{\ast}\subset L$ and $P^{\ast}=B(P,\varepsilon_{2})\subset P$.
Recall $\partial A_{i}$ is the graph of a Lipschitz function along the
direction $Z_{i}$ from any hyperplane transversal to $Z_{i}$. Let $H_{i}=L+W$
be one such hyperplane that contains the subspace $L$ and the vector line
$W\subset P$, and construct coordinates $L\times W\times<Z_{i}>$. It follows
from the previous lemma that $\partial A_{i}\cap P^{\ast}_{a}$ is a connected
Lipschitz curve. In these coordinates $\partial A_{i}$ is the graph of a
Lipschitz function $h_{i}$. Its domain, for fixed $a$, is a connected
interval, delimited by two functions $f_{l}:L^{\ast}\rightarrow W$ and
$f_{r}:L^{\ast}\rightarrow W$. Condition (1) of lemma 6.5 assures they are
Lipschitz.
A connected component $\mathcal{C}_{0}$ of $\mathcal{C}$ is contained in only
one $A_{i}\cap A_{j}$. We can express it in the coordinates defined above for
$\partial A_{i}$. The intersection of $\mathcal{C}_{0}$ with each plane
$P_{a}$ is either empty or a connected Lipschitz curve. The second part
follows as before. ∎
### 6.3 Conclusion
Using lemma 6.2, we show without much effort that $\partial T$ vanishes near
edge points. Using the structure results from the previous section, we show
also that it vanishes at cleave points (including degenerate ones) and
crossing points.
###### Proposition 6.9.
Let $p\in S$ be an _edge_ point. Then the boundary of $T$ vanishes near $p$.
###### Proof.
Let $p$ be an edge point with $Q_{p}=\\{x\\}$. Let $O$ be a univocal
neighborhood of $x$. It follows by a contradiction argument that there is an
open neighborhood $U$ of $p$ such that $Q_{q}\subset O$ for all $q\in U$.
Recall the definition of $T$:
$T(\phi)=\sum_{j}\int_{\mathcal{C}_{j,1}}(h_{1}-h_{2})\phi$
For any cleave point $q\in U$ with $Q_{q}=\\{x_{1},x_{2}\\}$,
$h_{i}(q)=\tilde{u}(x_{i})$. By the above, both $x_{1}$ and $x_{2}$ are in
$O$. As $O$ is univocal, we see $h_{1}=h_{2}$ at $q$. The integrand of $T$
vanishes near $p$, and thus $\partial T=0$. ∎
###### Proposition 6.10.
Let $p\in S$ be a (possibly degenerate) _cleave_ point. Then $\partial T$
vanishes near $p$.
###### Proof.
Use the sets $U$, $A_{1}$ and $A_{2}$ of lemma 6.3.
Whenever $\phi$ is a $n-1$ differential form with support contained in $U$, we
can compute:
$T(\phi)=\int_{A_{1}\cap A_{2}}(h_{1}-h_{2})\phi$
The components of cleave points inside either $A_{1}$ or $A_{2}$ do not
contribute to the integral, for the same reasons as in the previous lemma.
Recall the definition of $\partial T$, for a differential $n-2$ form $\sigma$:
$\partial T(\sigma)=T(d\sigma)=\int_{A_{1}\cap A_{2}}(h_{1}-h_{2})d\sigma$
We can apply a version of Stokes theorem that allows for Lipschitz functions.
We will provide references for this later:
$T(d\sigma)=\int_{A_{1}\cap A_{2}}d(h_{1}-h_{2})\sigma$
The balanced condition imposes that for any vector $v$ tangent to $A_{1}\cap
A_{2}$ at a non-degenerate cleave point $q$ with $Q_{q}=\\{x_{1},x_{2}\\}$.
$\hat{X}^{1}(v)=\hat{X}^{2}(v)$
for the incoming vectors $X^{i}=d_{x_{i}}F(r)$. Recall that
$\mathcal{H}^{n-1}$-almost all points are cleave, and $dh_{i}$ is dual to the
incoming vector $X^{i}$, so $T(d\sigma)=0$. ∎
###### Proposition 6.11.
Let $p\in S$ be a _crossing_ point. Then the boundary of the current $T$
(defined in 5.2) vanishes near $p$.
###### Proof.
We use lemma 6.7 to describe the structure of connected components of
$\mathcal{C}$ near $p$. Let $\Sigma_{T}$, the set of _higher order points_ ,
be the set of those points such that $R_{q}^{\ast}$ spans an affine subspace
of $T^{\ast}_{q}\Omega$ of dimension greater than $2$.
Take any connected component $\mathcal{C}_{0}$ of $\mathcal{C}$ contained in
$\partial A_{i}$. $\partial\mathcal{C}_{0}$ decomposes into several parts:
* •
The regular boundary, consisting of two parts $D_{1}$ and $D_{2}$:
$D_{1}=\\{(a_{1},\dots,a_{n-2},f_{1}(a),h_{i}(f_{1}(a))),\forall a\in
L^{\ast}\text{ such that }f_{l}(a)<f_{1}(a)<f_{2}(a)\\}$
$D_{2}=\\{(a_{1},\dots,a_{n-2},f_{2}(a),h_{i}(f_{2}(a))),\forall a\in
L^{\ast}\text{ such that }f_{1}(a)<f_{2}(a)<f_{r}(a)\\}$
* •
The points of higher order, or $\partial\mathcal{C}_{0}\cap\Sigma_{T}$.
* •
The singular boundary, or those points
$q=(a_{1},\dots,a_{n-2},f_{1}(a),h_{i}(f_{1}(a)))$ where $f_{1}(a)=f_{2}(a)$
and $R_{q}$ is contained in an affine plane.
* •
A subset of $\partial U$.
Using a version of Stokes theorem that allows for Lipschitz functions, we see
that
$\int_{\mathcal{C}_{0}}vd\sigma=\int_{\mathcal{C}_{0}}d(v\sigma)-\int_{\mathcal{C}_{0}}(dv)\sigma=\int_{D_{1}}v\sigma-\int_{D_{2}}v\sigma-\int_{\mathcal{C}_{0}}(dv)\sigma$
for any function $v$ and $n-2$ form $\sigma$ with compact support inside $U$.
Indeed, the last coordinate of the parametrization of $\mathcal{C}_{0}$ is
given by a Lipschitz function, so we can rewrite the integral as one over a
subset of $L\times W$, and only Gauss-Green theorem is needed. We can apply
the version in [F, 4.5.5], whose only hypothesis is that the current
$\mathcal{H}^{n-1}\lfloor\partial\mathcal{C}_{0}$ must be representable by
integration. Using [F, 4.5.15] we find that it is indeed, because its support
is contained in a rectifiable set Here we are assuming that $D_{1}$ is
oriented as the boundary of $\mathcal{C}_{0}$, while $D_{2}$ is oriented in
the opposite way, to match the orientation of $D_{1}$.
Notice we have discarded several parts of $\partial\mathcal{C}_{0}$:
* •
A subset of $\partial\mathcal{C}_{0}$ inside $\partial U$ does not contribute
to the integral because $supp(\sigma)\subset\subset U$.
* •
$\partial\mathcal{C}_{0}\cap\Sigma_{T}$ does no contribute because it has
Hausdorff dimension at most $n-3$.
* •
The singular boundary does not contribute either, because the normal to
$\widetilde{\mathcal{C}_{0}}$ at a point of the singular boundary does not
exist (see [F, 4.5.5]).
We now prove that $\partial T=0$.
For a form $\sigma$ of dimension $n-2$ and compact support inside $U$:
$T(d\sigma)=\sum_{i}\int_{\mathcal{C}_{i}}(h_{l}-h_{r})d\sigma=\sum_{i}\int_{\mathcal{C}_{i}}d(h_{l}-h_{r})\sigma+\sum_{i}\left(\int_{D_{i,1}}(h_{l}-h_{r})\sigma-\int_{D_{i,2}}(h_{l}-h_{r})\sigma\right)$
where $D_{i,1}$ and $D_{i,2}$ are the two parts of the regular boundary of
$\mathcal{C}_{i}$.
The first summand is zero and the remaining terms can be reordered (the sum is
absolutely convergent because $h$ is bounded and $\mathcal{H}^{n-2}(\Sigma)$
is finite):
$\sum_{i}\left(\int_{D_{i,1}}(h_{l}-h_{r})\sigma-\int_{D_{i,2}}(h_{l}-h_{r})\sigma\right)=\int_{\Sigma\setminus\Sigma_{T}}\sum_{(i,j)\in
I(q)}(h_{i,j,l}-h_{i,j,r})\sigma dq$
where every point $q\in\Sigma\setminus\Sigma_{T}$ has a set $I(q)$ consisting
of those $i$ and $j=1,2$ such that $q$ is in the boundary part $D_{j}$ of the
component $\mathcal{C}_{i}$. The integrand at point $q$ is then:
$\sigma\sum_{(i,j)\in I(q)}(h_{i,j,l}-h_{i,j,r})$
where $h_{i,j,l}$ is the value of $\tilde{u}(x)$ coming from the side $l$ of
component $\mathcal{C}_{i}$ and boundary part $D_{j}$.
By the structure lemma 3.5, we can restrict the integral to crossing points.
Let $O_{1},\dots,O_{N}$ be the disjoint univocal sets that appear when we
apply 6.5 to $p$. For a crossing point $q$, $I(q)$ is in correspondence with
the set of indices $k$ such that $O_{k}\cap Q_{p}\neq\emptyset$. Indeed, the
intersection of $\mathcal{S}$ with the plane $P_{a}$ containing $q$ is a
Lipschitz tree, and $q$ is a vertex, and belongs to the regular boundary of
the components that intersect $P_{a}$ in an edge. The $h_{i,j,l}$ in the sum
appear in pairs: one is the value from the left coming from one component
$\mathcal{C}_{i}$ and the value from the right of another component
$\mathcal{C}_{i^{\prime}}$. Each one comes from a different side, so they
carry opposite signs, and they cancel. The integrand at $q$ vanishes
altogether, so $\partial T=0$.
∎
## References
* [AG] P. Angulo, L. Guijarro, Cut and singular loci up to codimension 3, http://arxiv.org/abs/0806.2229
* [BCS] D. Bao, S. S. Chern, Z. Shen, An introduction to Riemann-Finsler geometry, Graduate Texts in Mathematics, 200, Springer.
* [CS] P. Cannarsa, C. Sinestrari, Semiconcave Functions, Hamilton-Jacobi Equations, and Optimal Control, Progress in Nonlinear Differential Equations and Their Applications, 58, Birkhäuser
* [CR] M. Castelpietra, L. Rifford, Regularity properties of the distance functions to conjugate and cut loci for viscosity solutions of Hamilton-Jacobi equations and applications in Riemannian geometry arXiv:0812.4107v1 (2008)
* [D] J. J. Duistermaat, Fourier Integral Operators, Birkhäuser, Boston, Progress in Mathematics, 130 (1996)
* [F] H. Federer, Geometric measure theory, Springer-Verlag New York Inc., New York, 153 (1969).
* [GG] M. Golubitsky, V. Guillemin Stable Mappings and their Singularities Graduate Texts in Mathematics, 14 Springer-Verlag
* [Ha] A. Hatcher. Algebraic topology. Cambridge University Press. http://www.math.cornell.edu/$\sim$hatcher
* [H] J. Hebda, Parallel translation of curvature along geodesics, Trans. Amer. Math. Soc., Vol. 299, No 2 (Feb., 1987), 559-572.
* [IT] J. Itoh, M. Tanaka. The Lipschitz continuity of the distance function to the cut locus. Transactions of the A.M.S. 353 (2000), no. 1, 21-40
* [LN] YY.Li, L. Nirenberg, The distance function to the boundary, Finsler geometry, and the singular set of viscosity solutions of some Hamilton-Jacobi equations. Comm. Pure Appl. Math. 58 (2005), no. 1, 85-146.
* [L] P. L. Lions, Generalized solutions of Hamilton-Jacobi equations Research Notes in Mathematics 69, Pitman, London (1982).
* [MM] C. Mantegazza, A. C. Mennucci, Hamilton-Jacobi Equations and Distance Functions on Riemannian Manifolds Appl. Math. Optim. 47 (2003), pp.1-25
* [Me] A. C. Mennucci, Regularity And Variationality Of Solutions To Hamilton-Jacobi Equations. Part I: Regularity (2nd Edition), ESAIM Control Optim. Calc. Var. 13 (2007), no. 2, 413–417
* [W] F.W.Warner, The conjugate locus of a Riemannian manifold, Amer.J. of Math. 87 (1965) 575-604. MR 34:8344
| arxiv-papers | 2008-07-13T18:04:30 | 2024-09-04T02:48:56.747944 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Pablo Angulo Ardoy, Luis Guijarro",
"submitter": "Pablo Angulo Ardoy",
"url": "https://arxiv.org/abs/0807.2046"
} |
0807.2130 | ## 0.1 Direct observation of the effect of isotope-induced-disorder on
exciton binding energy in LiHxD1-x mixed crystals.
V.G. Plekhanov
Computer Science College, Erika 7a, Tallinn, 10416, ESTONIA
Abstract. The results of the quantitative study of the renormalization of
binding energy the Wannier - Mott exciton by the isotope effect are present
for the first time. For this purpose accurate measurements of the intrinsic
luminescence and mirror reflection spectra of LiHxD1-x mixed crystals with a
clean surface in the temperature range 2 - 100 K were carried. Nonlinear
dependence of exciton binding energy Eb on the isotope mass E${}_{b}\sim$ f(x)
is caused by the bands fluctuation broadening which is connected with the
isotope-induced-disorder. Temperature dependence of exciton binding energy is
briefly discussed. The extrapolation of the asymptotic linear behavior of the
exciton maximum energy to T = 0 K enables to estimate the zero - point
renormalization of the exciton binding energy.
PACS: 32.10B; 71.20P; 78.20; 78.40F.
A wide variety of novel isotope effects have been discovered on last four
decades [1-6] owing to the availability of high-quality bulk semiconductor and
insulator crystals with controlled isotopic composition (see, also, reviews
[7-10]). Recent high resolution spectroscopic studies of excitonic and
impurity transitions in high-quality samples of isotopically enriched Si have
discovered the broadening bound excitons emission (absorption) lines connected
with isotope-induced-disorder as well as the dependence of their binding
energy on the isotope mass [11-13]. The last effect was early observed on the
bound excitons in diamond [14-15] and more earlier on the free excitons in
LiHxD1-x mixed crystals [16].
As is well-known (see, for example [16, 17, 4]) the band gap energy Eg in the
T$\longrightarrow$0 limit has a dependence on the average isotopic mass
$\overline{M\text{ }}$ due to two effects: a) the renormalization of Eg by the
electron-phonon interaction coupled with the dependence of the zero-point
amplitudes on $\overline{M}$ (see, also [18]) and b) the dependence of the
lattice constant on $\overline{M}$, leading to a change in Eg through the
hydrostatic deformation potential. The electron-phonon term is dominant [19,
20] and in the case of semiconductor crystals (C; Ge; Si) with a weak isotope
scattering potential is varied approximately as $\overline{M}^{-1/2}$. The
value of the T =0 electron-phonon renormalization energy contribution to Eg
can be independly determined from an extrapolation of the high temperature
linear dependence of E${}_{g}\sim$ f(T) to T=0 but again it should be
remembered that E${}_{g}\sim$ f(T) also has a small contribution from change
in volume with temperature. Therefore we should distinguish between effects of
the average atomic mass (which imply the virtual crystal approximation - VCA)
and effects of the mass fluctuations (randomness) superimposed onto the
virtual crystal.
In this communication we report the first results of the quantitative study
the dependence of free exciton’s binding energy on isotope mass as well as on
the temperature in LiHxD1-x mixed crystals. We’ll show that the exciton
binding energy increase by 10$\pm$1 meV from 7LiH to 7LiD. Moreover, the
dependence of Eb $\sim$ f(x) is nonlinear. The last effect, as will be shown
below, is caused by the isotope-induced-disorder of LiHxD1-x mixed crystals.
Briefly part of these results have been published in [21].
Specimens of LiH, LiD LiHxD1-x (0$\leq$ x $\leq$ 1) as well as LiHxF1-x were
grown from melt using the modified Bridgman - Stockbarger method (see, also
[22]). This technique is described many times early (see, e.g. [18, 2]). To
improve the stoichiometric composition with respect to hydrogen (deuterium),
the crystals grown were additionally annealed in an atmosphere of hydrogen or
deuterium at a gas pressure of 3 to 5 atm. and a temperature 500 to 5500C
(melting point is 961 and 964 K for LiH and LiD, respectively). For some
crystals the time of heat treatment was as long as 20 days (for more details
see [2,7]).
Given the high reactivity of freshly cleaved LiH crystals in the atmosphere,
we had to develop a procedure of cleaving which would not only ensure an
initially clean surface, but also allow us to keep it fresh for a few hours
(the time taken to complete an experiment). These requirements are satisfied
by the well-known method of cleaving directly in the helium chamber of the
optical cryostat under liquid or superfluid helium, first tried in Ref. [23].
We did not notice any changes in the spectra of reflection or luminescence
while working for 10 t0 16 hours with surface prepared in this way. The device
for cleaving the crystals had three degrees of freedom and rotated through
900, which greatly helped in carrying out the experiments. As a rule,
specimens for experiments were cleaved off bulk high - quality crystals. The
experimental setup for measuring the low - temperature reflection and
luminescence spectra has been described more than once (see, e.g. [7,18]) and
consists of a double grating or prism monochromators, an immersion helium
cryostat, and a photovoltaic detector (in photon counting mode). The results
presented in this paper were obtained from a clean crystal surface cleaved, as
described above, directly in the bath of helium cryostat. The spectra of
mirror reflection were measured using an the angle of incidence of 450. For
our studies we selected specimens which exhibited low dependence of exciton
spectra of reflection and luminescence on surface features.
As demonstrated earlier (see, e.g. [7]) most low-energy electron excitations
in LiH (LiD) crystals are large-radius excitons. The spectrum of exciton
photoluminescence of LiH crystals cleaved in liquid helium consists of a
narrow (in the best crystals, its half-width is the $\Delta$E $\leqslant$
10meV) phononless emission line and its broader phonon repetitions, which
arise due to radiative annihilation of excitons with the production of one to
five longitudinal (LO) phonons (see, Fig. 1). The phononless emission line
coincides in an almost resonant way with the reflection line (see, insert Fig.
1) of the exciton ground state which is another indication of direct electron
transition (X1 \- X4 [2]). The lines of phonon replicas form an equidistant
series biased toward lower energies from the resonance emission line of
excitons. The energy difference between these lines is about 140 meV, which is
close to the calculated energy of the LO phonon in the middle of the Brillouin
zone [21] and measured in [19]. The isotopic shift of the zero-phonon emission
line of LiH crystals equals that in reflection spectra, 103 meV (see, also,
insert in Fig.1). As the deuterium concentrations increases, the width of the
long-wave maximum in reflection (and the width of the phononless line in
luminescence) spectra broadens and maximum shifts towards the shorter
wavelengths. As can clearly see in Fig. 1, all spectra exhibit a similar long-
wave structure. This circumstance allows us to attribute this structure to the
excitation of the ground (1s) and the first excited (2s) exciton states [2].
Three effects are distinctly shown in the reflection spectra by an increase in
deuterium concentration:
1\. The shortwavelength shift of the reflection (as luminescence) spectrum as
a whole;
2\. The different velocity shift of the exciton maximum of the ground and the
first excited states on the temperature (see, Fig. 2);
3\. Broadening of the long-wavelength maximum due to excitation of the ground
exciton state.
By the way we should note that the first excited exciton state is very clearly
seen in the luminescence spectra too (see Fig. 3). Although two distinct
contribution to $\Delta$Eg (and respectively Eb) are present that due to the
effect of the vibration on the lattice constant and the direct effect of the
electron-phonon interaction [19, 20] for the present task of the dependence of
the exciton binding energy on the isotope mass it will be considered only the
mechanism of exciton-phonon interaction.
In accordance with the second effect in reflection spectra (see above point
2), as is to be expected, there is nonlinear dependence of the exciton binding
energy (E${}_{b\text{ }}$ = 4/3$\Delta_{12}$) on the isotope concentration.
Displayed curve in Fig. 4 is the Wannier - Mott exciton binding energy value
which are determined by the hydrogenlike expression Eb =
$\frac{e^{4}\mu}{\text{2}\hslash^{2}\varepsilon^{2}\text{n}^{2}}$ using for
this E1 and E2 values from the reflection and luminescence spectra (see Figs.
1 and 3). The nonlinear nature of this dependence is similar to the
theoretical results [25] , where the influence of the chemical disorder of the
crystal lattice on the Wannier - Mott exciton binding energy was seen and
obtained a qualitative (not quantitative!) agreement with experimental results
on the GaAsxP1-x [26] mixed crystals.
Before the comparison of our experimental results with the theory developed by
Elliott and Kanehisa [25], it would be prudent to briefly review main
properties of their theoretical model. One of the principal result of paper
[25] is the nonlinear dependence of exciton binding energy Eb on the
concentration. As a consequence, the binding energy at half-and-half
concentrations is less than the value derived from the crystal virtual model.
According to Ref. [25] this model considers an exciton with a direct gap of a
semiconductor alloy. Such a system consists of an electron (particle 1) in the
conduction band (c) with mass m${}_{\text{c}}$ and a hole (particle 2) in the
valence band (v) with mass m${}_{\text{v}}$. The problem of the exciton in
disordered systems is to solve the Hamiltonian
H = $\overrightarrow{\text{p}}^{\text{2}}$/2m${}_{\text{c}}$
+$\overrightarrow{\text{p}}^{\text{2}}$/2m${}_{\text{v}}$\+
u($\overrightarrow{\text{r}}_{\text{1}}$ \-
$\overrightarrow{\text{r}}_{\text{2}}$) +
V${}_{\text{c}}$($\overrightarrow{\text{r}}_{\text{1}}$) +
Vv($\overrightarrow{\text{r}}_{\text{2}}$), (1)
with both the Coulomb interaction u and the potential Vv due to disorder
($\nu$ = c,v). Reference [25] neglected disorder - induced interband mixing.
As it is well known, in place of the electron-hole coordinates,
($\overrightarrow{\text{r}}_{\text{1}}$,
$\overrightarrow{\text{p}}_{\text{1}}$) and
($\overrightarrow{\text{r}}_{\text{2}}$,
$\overrightarrow{\text{p}}_{\text{2}}$), one may introduce the center-of-mass
and relative coordinates, ($\overrightarrow{\text{R}}$,
$\overrightarrow{\text{P}}$) and ($\overrightarrow{\text{r}}$,
$\overrightarrow{\text{p}})$ to rewrite (1) as
H = $\overrightarrow{\text{p}}^{\text{2}}$/2$\mu_{\text{r}}$ \+
u($\overrightarrow{\text{r}}$) + $\overrightarrow{\text{P}}^{\text{2}}$/2M +
V${}_{\text{c}}$($\overrightarrow{\text{R}}$ \+
m${}_{\text{v}}\overrightarrow{\text{r}}$/M) +
V${}_{\text{v}}$($\overrightarrow{\text{R}}$ \-
m${}_{\text{c}}\overrightarrow{\text{r}}$/M), (2)
where $\mu_{r}$ and M are the reduced and total masses of excitons,
respectively. Because of the random potential, the translational and relative
degrees of freedom cannot be decoupled. This is essentially difficult when
considering the two-body problem in a disordered system (see [25] and
references therein). However, when the exciton state in question is well
separated from other states so that the energy spacing is much larger than the
translational width and disorder, one can forget about the relative motion (Hr
= $\frac{\overrightarrow{p}^{2}}{\text{2m}_{r}}$ \+ u($\overrightarrow{r}$))
and just apply any single-particle alloy theory solely to their translational
motion. For each exciton state the translational part of Hamiltonian in this
case is
Ht = $\overrightarrow{\text{P}}^{\text{2}}$/2M +
V̄${}_{\text{c}}$($\overrightarrow{\text{R}}$) + V̄${}_{\text{v}}$(R̄) (3).
Here V̄${}_{\text{c}}$ and V̄${}_{\text{v}}$ are averages of V${}_{\text{c}}$
and V${}_{\text{v}}$ with respect to the relative state $\phi$, for example:
$\overline{V_{c}}$($\overrightarrow{R}$) =
$\mathop{\displaystyle\int}$d${}^{3}\overrightarrow{r}\left|\phi\left(\overrightarrow{r}\right)\right|$2Vc[$\overrightarrow{R}$
\+ $\frac{\text{m}_{\text{v}}}{\text{M}}\overrightarrow{r}$] (4).
This approach is very similar to the Born-Oppenheimer adiabatic approximation.
Such situations hold in some mixed alkali halide crystals and probably
A${}_{\text{2}}$B${}_{\text{6}}$ crystals. On the contrary, when the exciton
binding energy is comparable to the disorder energy, the adiabatic
approximation breaks down, and it is essential to take into account the effect
of disorder on both the translational and relative motions. This is the case
with the Wannier-Mott exciton in A${}_{\text{3}}$B${}_{\text{5}}$ alloys, for
which the Elliott and Kanehisa model was developed. In this case the solution
task is to start from the independent electron and hole by neglecting u in (2)
and then to take into consideration the Coulomb interaction between the
average electron and average hole. A further simplified approach adopted in
the literature (see, for example [27] and references therein) in solving the
Bethe-Salpeter equation [28] is to suppose a free-electron-like one particle
Green’s function with a built-in width to allow for the random potential due
to disorder. In the cited theoretical model [25], the average (or ”virtual
crystal”) gap is given by
E${}_{\text{g}}^{\text{vc}}$(x) = E${}_{\text{0}}$ \+ ($\delta_{\text{c}}$ \-
$\delta_{\text{v}}$)(x - 1/2), (5)
where E${}_{\text{0}}$ is average gap, $\delta_{\text{c}}$,
$\delta_{\text{v}}$ are the values of the fluctuation broadening of the
conduction and valence bands, respectively. Reference [25] also assumed the
Hubbard density of states for both the conduction and valence bands with width
W${}_{\text{c}}$ and W${}_{\text{v}}$, respectively, as well as similar
dispersion in both bands. With this assumption the exciton binding energy has
been calculated according to the coherent potential approximation CPA model.
As is well-known, the main idea of the coherent potential methods in the
introduction of an auxiliary medium with a regular , i.e., spatially periodic
potential. By definition in the model this potential is also complex. The
formalism of the coherent potential model, convenient for performing
calculation and does not include fitting parameters, because of using of the
density of phonon states from virtual crystal model which used virtual crystal
approximation (VCA). It should be added here the key feature of the model
developed in Ref. 25 is the short-range nature of the Coulomb potential (for
details, see e.g. [7, 29]).
The data from Fig. 1 and other published sources [2, 7, 21] were used for
plotting the energy E${}_{\text{b}}$ as a function of isotopic concentration x
in Fig. 4. The binding energy (defined as the band edge minus the exciton
energy) is given by [25]:
E${}_{b}^{crys}$ = U${}_{\text{0}}+\frac{\text{W}}{\text{2U}_{\text{0}}}$ \+ W
(6).
In the last relation U0 is the coupling constant at the total exciton momentum
$\overrightarrow{q}$ = 0.
Theoretical description of the binding energy of Wannier- Mott excitons as a
function of concentration x was based on the polynomial derived by Elliott and
cowokers [25]:
E${}_{\text{b}}$ = E${}_{\text{b}}^{\text{crys}}$\-
E${}_{\text{bow}}\left[\frac{\text{1-W}}{\text{2U}_{\text{0}}}\right]$ \-
E${}_{\text{eff}}$, (7)
E${}_{\text{eff}}$ = x $\left(\text{1-x}\right)$
$\frac{\delta_{\text{c}}\delta_{\text{v}}}{\text{W}}$, (8)
where W = W${}_{\text{c}}$ \+ W${}_{\text{v}}$, and W${}_{\text{c}}$ and
W${}_{\text{v}}$ are the widths of the conduction band and the valence band
which are equal to 21 eV [30] and 6 eV [31, 32] respectively. Here
E${}_{\text{bow}}$ is the curvature parameter found from the function
E${}_{\text{g}}$ $\propto$ f(x) (Ebow = 0.046 eV [7]); $\delta_{\text{c}}$ and
$\delta_{\text{v}}$ are the magnitudes of the fluctuation smearing of the
valence band and the conduction band edges , $\delta_{\text{c}}$ = 0.103 eV
and $\delta_{\text{v}}$ = - 0.331 eV. As follows from Fig. 4, these values of
the parameters give a good enough description of the nonlinear dependence of
the binding energy of Wannier-Mott exciton in disordered medium isotope -
mixed crystals LiHxD1-x. This agreement between theory and experiment once
again proves the inherent consistency of the model proposed by Kanehisa and
Elliott, since the isotopic substitution affects the short-range part of the
interaction potential.
In this way, the nonlinear dependence of the binding energy of Wannier-Mott
exciton is caused by isotopic disordering of the crystal lattice. As is seen
from Fig. 4 the exciton binding energy decreasing (relative linear law (VCA)
\- see dashed line in Fig. 4) in the vicinity of the middle meaning
concentration really calls out the fluctuated broadening of the edge of the
conduction and valence bands. In accordance with the theoretical model the
last reason gives rise to the reduced E${}_{\text{g}}$ and there by the
shallowing of the exciton levels and, respectively, the reduction of
E${}_{\text{b}}$.
As follows from Fig. 1, the addition of deuterium leads not only to the short-
wave shift of the entire exciton structure (with different rates for 1s and 2s
states), but also to a significant broadening of the long-wave exciton
reflection line. This line is broadened 1.5 - 3-fold upon transition from pure
LiH to pure LiD. The measure of broadening was the halfwidth of the line
measured in the standard way (see e.g. [33]) as the distance between the
maximum and the minimum in the dispersion gap of the reflection spectrum,
taken at half-height. The concentration dependence of the halfwidth
($\Delta$E${}^{\text{R}}$) of the long-wave band in the exciton reflection
spectrum at 2 K is shown in Fig. 5. Despite the large spread and the very
limited number of concentrations used, one immediately recognizes the
nonlinear growth of $\Delta$E${}^{\text{R}}$ with decreasing x. A similar
concentration dependence of $\Delta E^{\text{R}}$ in the low-temperature
reflection spectra of solid solutions of semiconductor compounds
A${}_{\text{2}}$B${}_{\text{6}}$ and A${}_{\text{3}}$B${}_{\text{5}}$ has been
reported more than once (see e.g. the review of Elliott and Ipatova [34] and
references therein). The observed broadening of exciton lines is caused by the
interaction of excitons with the potential of large-scale fluctuations in the
composition of the solid solution. Efros and colleagues (see e.g. [35]) used
the Lifshitz method of optimal fluctuation [36] to express the formula for the
concentration dependence of the broadening of exciton reflection lines:
$\Delta$E${}^{\text{R}}$ = 0.5$\alpha$
$\left[\frac{\text{x}\left(\text{1-x}\right)}{\text{Nr}_{\text{ex}}}\right]^{\text{1/2}}$.
(9)
where N - the concentration of sublattices nodes where the isotope substitutes
are placed, $\alpha$ = dE${}_{\text{g}}$/ dx; r${}_{\text{ex}}$ is the exciton
radius which varies from 47 Å to 42 Å upon transition from LiH to LiD [2]. The
results of calculation according to Eq. (9) are shown in Fig. 5 by a full
curve.
The experimental results lie much closer to this curve than to the straight
line plotted from the virtual crystal model. At the same time it is clear that
there is only qualitative agreement between theory and experiment at x $>$
0.5. Nevertheless, even this qualitative analysis clearly points to the
nonlinear dependence of broadening on the concentration of isotopes, and hence
to the isotopic disordering. Since isotopic substitution only affects the
energy of optical phonon for the first time, and, as a consequence, the
constant of exciton-phonon interaction (in the first place, the Fröhlich
interaction g${}_{\text{F}}^{\text{2}}$), the nonlinearity of functions
$\Delta$E${}_{\text{b}}$ $\propto$ f(x), $\Delta$E${}^{\text{R}}$ $\propto$
f(x) is mainly related to the nonlinear behavior of
g${}_{\text{F}}^{\text{2}}\propto$ f(x) . In this way, the experimental study
of the concentration dependence of the exciton-phonon interaction constant may
throw light on the nature and mechanism of the large-scale fluctuations of
electron potential in isotopically disordered crystals.
Returning to the results of Fig. 2, let us add that the different temperature
dependence of exciton peaks of n = 1s and 2s exciton states leads to the
temperature dependence of the binding energies of Wannier - Mott excitons
Eb $\sim$ f (T) (10).
This problem has not received any adequate treatment. More specifically, the
energy of the exciton binding Eb in LiH crystals (as well as in mixed crystals
LiHxF1-x (LiDxF${}_{\text{1-x}}$ )) decreases with increasing temperature,
whereas Eb increases for excitons of the green and yellow series in Cu2O
crystals [34]. A linear approximation of the exciton binding energy in
LiD${}_{\text{0.995}}$F${}_{\text{0.005}}$ (see curve 3, Fig. 2) representing
Eb at T=0 K gives Eb(0) $\cong$ 55 meV. From this value we can see that
renormalization of the binding energy by the zero-point vibrations equals
approximately $\simeq$ 10 % from this value, that is, on the other hand,
composes only half of renormalized exciton binding energy by isotope effect
($\simeq$ 10 meV, see above). It is not excluded that the other part of
renormalized exciton binding energy is caused by exciton-polar phonon
interaction.
In conclusion, the exciton luminescence and reflection spectra are used in a
quantitative study of the isotopic and temperature effects in LiHxD1-x mixed
crystals with a clean surface. It was shown that the short-range character of
the potential of a disordered crystal lattice with isotope substitution is
responsible for the broadening of the valence and conduction bands. Nonlinear
dependence of the exciton binding energy on the isotope mass Eb $\sim$ f(x) is
due the isotope-induced-disorder of LiHxD1-x mixed crystals. Temperature
dependence of exciton binding energy is briefly discussed. The extrapolation
of the asymptotic linear behavior of the exciton maximum energy to T = 0 K
enables to estimate the zero - point renormalization of the exciton binding
energy.
Acknowledgements. I would like to express my deep thanks to Prof. K. Wadler
for improving my English.
References.
1\. F.I. Kreingold, K.F. Lider, L.E. Solov’ev, Sov. Phys. JETP Let., 23 (1976)
624.
2\. V.G. Plekhanov, T.A. Betenekova, V.A. Pustovarov, Sov. Phys. Solid State
18 (1976) 1422;
V.G. Plekhanov, Physics - Uspekhi 40 (1997) 553.
3\. V.F. Agekyan, V.M. Asnin, A.M. Kryukov, Sov. Phys. Solid State 31 (1989)
2082.
4\. A.T. Collins, S.C. Lawson, G. Davies and H. Kanba, Phys. Rev. Lett. 65
(1990) 891.
5\. T.R. Anthony, W.F. Banholzer, J.F. Fleisher, L.-H Wei, P.K. Kuo, Phys.
Rev. B42 (1990) 1104.
6\. H.D. Fuchs, C.H. Grein, C. Thomsen, M. Cardona, Phys. Rev. B43 (1991)
4835.
7\. V.G. Plekhanov, Isotope Effects in Solid State Physics, Academic Press,
New York, 2001.
8\. M. Cardona, Phys. Stat. Solidi (b) 220 (2000) 5; Phys. Stat. Solidi (a)
188 (2001) 1209.
9\. E.E. Haller, J. Appl. Phys. 77 (1995) 2857; Solid State Commun. 133 (2005)
693.
10\. M.L.W. Thewalt, Solid State Commun. 133 (2005) 715.
11\. D. Karaiskaj,M.L.W. Thewalt, T. Ruf, M. Cardona, M. Konuma, Solid State
Commun. 123 (2002) 87; M. Cardona and M.L.W. Thewalt, Rev. Mod. Phys. 77,
(2005) 1173.
12\. D. Karaiskaj, T.A. Meyer, M.L.W. Thewalt and M. Cardona, Phys. Rev. B68
(2003) 121201 (R).
13\. S. Tsoi, H. Alawadhi, X. Lu, J.W. Ager III, C.Y. Liao, H. Rieman, Phys.
Rev. B70 (2004) 193201.
14\. H. Kim, S. Rodriguez, T.R. Anthony, Solid State Commun. 102 (1997) 861.
15\. M. Cardona, Solid State Commun. 121 (2002) 7.
16\. A.A. Klochikhin and V.G. Plekhanov, Sov. Phys. Solid State 22 (1980) 342;
V.G. Plekhanov, Fiz. Tverd. Tela 38 (1996) 1159 (in Russian).
17\. F.I. Kreingold, Fiz. Tverd. Tela 20 (1978) 3138 (in Russian).
18\. V.G. Plekhanov, Progress in Materials Science 51 (2006) 287.
19\. V.G. Plekhanov, Phys. Solid State (St-Petersburg) 35 (1993) 1493; J.
Nuclear Science and Technology 43 (2006) 375.
20\. S. Zollner, M. Cardona and S. Gopalan, Phys. Rev. B45 (1992) 3376.
21\. V.G. Plekhanov, Progr. Solid State Chem. 29 (2001) 77.
22\. O.I. Tytyunnik, V.I. Tyutyunnik, B.V. Shulgin, F.F. Gavrilov, and G.I.
Pilipenko, J. Crystal Growth, 68 (1984) 741.
23\. V.G. Plekhanov, A.V. Emelyanenko, A.U. Grinfelds, Phys. Lett. A101 (1984)
291.
24\. J.L. Verble, J.L. Warren, J.L. Yarnell, Phys. Rev. 168 (1968) 980.
25\. M.A. Kanehisa and R.J. Elliott, Phys. Rev. B35 (1987) 2228.
26\. R.J. Nelson, in Excitons, ed. by E.I. Rashba and M.D. Sturge (North-
Holland, Amsterdam, 1982), 319.
27\. A.A. Klochikhin, Sov. Phys. Solid State 22 (1980) 1690.
28\. H.A. Bethe, E. Salpiter, Quantum theory of one and two electron atoms,
AQcademic Press, New York, 1957.
29\. R.J. Elliott, J.A. Krumhansl, P.L. Leath, Rev. Mod. Phys. 46 (1974) 465.
30\. J. Kama and N. Kawakami, Phys. Lett. A126 (1988) 348.
31\. T.A. Betenekova, I.M. Shabanova, F.F. Gavrilov, Sov. Phys. Solid State 20
(1978) 820.
32\. K. Ichikawa, N. Susuki, K. Tsutsumi, J. Phys. Soc. Japan 50 (1981) 3650.
33\. N.N. Ablyazov, A.G. Areshkin, V.G. Melekhin, L.G. Suslina and D.L.
Fedorov, Phys. Stat. Solidi (b) 135 (1986) 217.
34\. R.J. Elliott and I.P. Ipatova (Eds), Optical Properties of Mixed Crystals
(North-Holland, Amsterdam, 1988).
35\. A.L. Efros and M.E. Raikh, in [34] Chapter 5.
36\. I.M. Lifshitz, Selected works, Science, Moscow, 1987 (in Russian).
37\. T. Itoh, S.J. Narita, J. Phys. Soc. Japan 39 (1975) 132.
Figure captions.
1\. Fig. 1. Luminescence spectra of free excitons at 2 K in LiH and LiD
crystals cleaved in liquid helium. In insert: mirror reflection spectra of
crystals. Curve 1: LiH; curve 2: LiHxD1-x and curve 3: LiD. Curve 4 is the
light source without crystals.
2\. Fig. 2. Temperature dependence of the distance between the long-wavelength
peaks ($\Delta_{12}$) in specular reflection spectra of pure and mixed
crystals: 1 - LiH; 2 - LiD; 3 - LiD0.995F0.005.
3\. Fig. 3. The reflection (1) and luminescence (2) spectra of LiD crystal at
2 K.
4\. Fig. 4. Concentration dependence of the binding energy of a Wannier - Mott
exciton at 2 K in LiHxD1-x mixed crystals: 1 - VCA approximation model; 2 -
calculation according to equation (8); experimental points indicated by
triangles.
5\. Fig. 5. Concentration dependence of the half-width of the ground state
line of the exciton in the mirror reflection spectrum at 2 K. 1 - VCA
approximation model; 2 - calculation according to equation (9); experimental
points are indicated by crosses.
| arxiv-papers | 2008-07-14T11:14:14 | 2024-09-04T02:48:56.757372 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "V.G. Plekhanov",
"submitter": "Vladimir Plekhanov",
"url": "https://arxiv.org/abs/0807.2130"
} |
0807.2152 | ## 1 Isotope dependence of band-gap energy.
### 1.0.1 V.G. Plekhanov∙ and N.V. Plekhanov
Computer Science College, Erika Street 7a,Tallinn 10416, ESTONIA
Abstract. The results of the quantitative investigations of the
renormalization of the absorption edge of different compounds by the isotope
effect are described. The obtained dependence of the band gap energy versus
isotope mass is in reasonable agreement with estimated value obtained from the
temperature dependence of Eg in LiH crystals. For the first time it was shown
that the dependence Eg = f(M) for the different materials has a parabolic
character ln($\partial$E${}_{g}/\partial M$) = 6.105(lnEg)2 \- 7.870(lnEg) +
0.565. Nonlinear character of this dependence may be indicated on the
nonlinear dependence of the potential scattering on the isotope mass.
PACS: 32.10.B + 71.20.P + 78.20 + 78.40.F
Isotope substitution is a well-defined and easily controllable method to
investigate intrinsic renormalization mechanisms of elementary excitations of
solids. First of all, phonon frequencies are directly affected by changes of
the average mass of the whole crystal or its sublattices (in the case of
virtual crystal approximation) even when we look upon them as noninteracting
particles, i.e. as harmonic oscillators. The renormalization of the
fundamental electronic gap by electron-phonon interaction also depends on the
isotope mass. Measuring the energy gaps in samples with different isotopic
composition then yields the difference in the changes of the valence- and
conduction-band renormalization.
In this communication we report the first results of the quantitative study
the dependence of the band-gap energy on the isotope effect for different
compounds. As was mentioned above isotopic substitution only affects the
wavefunctions of phonons; therefore, the energy values of electron levels in
the Schrodinger equation ought to have remained the same. This, however, is
not so, since isotopic substitution modifies not only the phonon spectrum, but
also the constant of electron-phonon interaction. It is for this reason the
energy values of purely electron transitions in molecules of hydride and
deuteride are found to be different (see, e.g. [1]). This effect is even more
prominent when we are dealing with a solid [2]. This conclusion was confirmed
on a qualitative level as early as the 1930s in Ref. [3]. Intercomparison of
absorption spectra for thin films of LiH and LiD at room temperature revealed
[3] that the longwave maximum (as we now know to be the exciton peak (see,
e.g. [2])) moves 64.5 meV towards the shorter wavelengths when H is replaced
by D. As will be shown below, this effect becomes even more pronounced at low
temperatures.
∙Correspondence to V.G. Plekhanov; $<$e-mail$>$ vgplekhanov@iati.ee
The mirror reflection spectra of mixed and pure LiD crystals cleaved in liquid
helium are presented in figure 1. For comparison, on the same diagram we have
also plotted the reflection spectra of LiH crystals with clean surfaces. All
spectra have been measured with the same apparatus under the same conditions.
As the deuterium concentration increases, the long-wave maximum broadens and
shifts towards shorter wavelengths. As can clearly be seen in figure 1, all
spectra exhibit a similar long-wave structure. This allows us to attribute
this structure to the excitation of the ground (1s) and the first excited (2s)
exciton states. The energy values of exciton maxima for pure and mixed
crystals at 2K are presented in Table I. The binding energies of excitons Eb,
calculated by the hydrogen-like formula, and the energies of interband
transitions Eg (details see [4]) are also given in Table I.
The ionization energy, found from the temperature quenching of the peak of
reflection spectrum [4] of the 2s state in LiD is 12 meV. This value agrees
fairly well with the value of $\Delta$E2s calculated by the hydrogen-like
formula. Moreover, Eb = 52 meV for LiD agrees well with the energy of
activation for thermal quenching of free-exciton luminescence in these
crystals [2].
At the weak scattering potential the mean-square vibrational amplitude
$<$u2$>$ of an atom dependes on the phonon frequencies and the eigenvectors,
the atomic masses (at low temperatures), as well as the temperature (at high
temperature). Isotope substitution results in a slightly different vibrational
amplitudes (epecially at low temperature) and phonon frequencies and in first
approximation equivalent to changing the temperature (see also [5]). The mass
dependence of $<$u2$>$ becomes vanishingly small at temperature on the order
or higher than the Debye temperature. Changes in either the isotope masses or
temperature thus lead to changes in the band gap (see above Fig. 1) via the
electron-phonon interaction, even at zero temperature in the case of mass
changes. In general, the renormalization of the band gap [4] and its
temperature dependence result from a complicated interplay of 1) first- and
second-order electron-phonon interactions that contribute to the energy of
conduction and valence bands; 2) changes due to thermal or isotopic lattice
expansion, and 3) changes in the phonon occupation numbers (see [5] and
references therein).
The dependence of the band gap energy on isotopic composition has already been
observed for insulators and lowest (indirect-direct) gap of different
semiconductors (see also [6]). It has been shown to result primarily from the
effect of the average isotopic mass on the electron-phonon interaction, with a
smaller contribution from the change in lattice constant. This simplest
approximation, in which crystals of mixed isotopic composition are treated as
crystals of identical atoms having the average isotopic mass is reffered to as
virtual crystal approximation (VCA) [7]. Going beyond the VCA, in isotopically
mixed crystals one would also expect local fluctuations in the band gap energy
from statistical fluctuations in local isotopic composition within some
effective volume, such as that of an exciton. As follows from Fig. 1 excitons
in LiHxD1-x crystals display a unimodal character, which facilitates the
interpretation of their concentration dependence. Fig. 2 shows the
concentration dependence of the energy of interband transition Eg. As can be
seen from Fig. 2 VCA method cannot describe observed experimental results. As
will shown below this deviation from linear low (VCA approximation) is
connected with isotope-induced-disorder in isotope mixed crystals
LiHxD${}_{1-x}.$
The temperature and isotopic mas dependence of a given energy gap Eg(T,Mi) can
be described by average Bose-Einsteinm statistical factor nB corresponding to
an average phonon frequency $\theta_{i}$ as (see also [5; 8]
Eg(T,Mi) = Ebar \- ar($\frac{M_{nat}}{M_{i}}$)${}^{1/2}\left[\text{1 +
2n}_{\text{B}}\right]$, (1)
where nB = 1/[exp$\left(\frac{\theta_{i}}{\text{T}}\right)$ -1] and Ebar and
ar the unrenormalized (bare) gap and the renormalization parameter,
respectively. In the low-temperature limit, T$<$$<$$\theta_{i}$, equation (1)
reduces
Eg(T,Mi) = Ebar \- ar($\frac{M_{nat}}{M_{i}}$)1/2 (2)
Here Eg(T,Mi) is independent of temperature and proportional to (1/Mi)1/2,
whereas ar $is$ the energy difference between the unrenormalized gap (Mi
$\rightarrow\infty$) and the renormalized value [5].
In the high-temperature limit, T$>$$>$$\theta_{i}$ and Eq. (1) can be written
as
Eg(T,Mi) = Ebar \- 2T$\frac{a_{r}}{\theta}$, (3)
and Eg(T,Mi) is independent of Mi [5]. The extrapolation of Eq. (3) to T = 0K
can be used to determine the unrenormalized gap energy Ebar , i.e., the value
that corresponds to atoms in fixed lattice position without vibrations (frozen
lattice [4]), from the measured temperature dependence of Eg(T) in the high-
temperature (i.e. linear in T) region.
Using Eq. (2) it can be written the difference in energy $\Delta$Eg between a
given energy gap in isotopically pure material (LiH) and its isotope analogue
(LiD)
$\Delta$Eg = Eg(Mi) - Eg(Mnat) = ar $\left[\text{1 -
}\left(\frac{M_{nat}}{M_{i}}\right)^{1/2}\right]$, (4)
As can be seen from Table 1 and results of [5] $\Delta$Eg at 2K equals
$\Delta$Eg = 0.103 eV and Eg(LiH, T = 0K) =5.004 eV (linear approximation and
Eg(LiH, T = 300K) = 4.905 eV then using Eq. (3) we obtain ar = 0.196 eV. This
magnitude is very close (approximately 84%) to the value of 0.235 eV of zero
vibration renormalization of the energy band gap in LiH crystals (details see
[4;5]). Using Eq. (4) we obtain $\Delta$Eg(theor) = 0.134 eV that is very
close, on the other hand, to observed experimental value equals 0.103 eV. The
discrepancy between these values may be caused by the negligible contribution
of the isotopic lattice expansion to the band gap renormalization.
The single-mode nature of the exciton reflection spectra of mixed crystals
LiHxD1-x agrees qualitatively with the results obtained with the virtual
crystal model (see, e.g. [7; 9]), being at the same time its extreme
realization, since the difference between ionization potentials ($\zeta$) for
this compounds is zero. According to the virtual crystal model, $\zeta$ = 0
implies that $\Delta$Eg = 0, which in contradiction with the experimental
results for LiHxD1-x crystals (details see, also [6]). In the light of
obtained dependence Eg = f(T, Mi) for LiHxD1-x it is very interested to study
such dependence for different materials. Although the details of the
renormalization process of Eg are not known, observation of the isotope
renormalization of Eg for many materials force to pay a great attention on
this effect.
By now the change in Eg caused by isotopic substitution has been observed for
many broad-gap and narrow-gap compounds. Below we briefly discuss the
variation of the electronic gap (Eg) of different crystals with its isotopic
composition. In the last decade the whole row of different semiconducting
crystals was grown. These crystals are diamond [10], copper halides [11; 12],
germanium [13;14], Si [8] and GaAs [15]. All enumerated crystals show the
dependence of the electronic gap on the isotope masses. It should be noted
that the indicated effect (the variation of Eg and Eb, see Table I and II)
have maximum values in LiH crystals, although this effect in other crystals
with isotopic composition are currently being reliably measured and
investigated well. All numerated crystals show the dependence of the
electronic gap on the isotope masses.
Before we complete the analysis of these results we should mention that before
these investigations, studies were carried out on the isotopic effect on
exciton states for a whole range of crystals by Kreingol’$d$ and coworkers
[16-21]. First of all we should name the classic crystals Cu2O [17-19] with
the substitution O16 $\rightarrow$ O18and Cu63 $\rightarrow$ Cu65. Moreover,
there have been some detailed investigations of the isotopic effect on ZnO
crystals [6] where Eg was seen to increase by 55 cm-1 (O${}^{16}\rightarrow$
O18) and 12 cm-1 ( at Zn64 $\rightarrow$ Zn68) [16;20]. In the paper of
Kreingol’d and coworkers [21] it was shown that the substitution of a heavy
S34 isotope for a light S32 isotope in CdS crystals resulted in a decrease in
the exciton Rydberg constant (Eb ).
More detailed investigations of the exciton reflectance spectrum in CdS
crystals were done by Zhang et al [22]. Zhang et al to directly determine the
binding energy (Eb) and the corresponding band gaps ($E_{g}$) from a
hydrogenlike model.
In this manner it was obtain binding energies of A excitons of 26.4$\pm$0.02
meV and 26.8$\pm$0.02 meV in 112CdS and natCdS, respectively. The
corresponding band gaps Eg($\Gamma_{7c}$ \- $\Gamma_{9v}$) are 2.5806(2) eV
and 2.5809(2) eV, respectively. In the case of B excitons, these values are Eb
= 27.1$\pm$0.2 meV (27.1$\pm$0.2 meV) and Eg ($\Gamma_{7c}$ \- $\Gamma_{7v}$)
= 2.5964(2) $\left[2.5963(2)\text{eV}\right]$ for the 112 CdS
$\left[{}^{nat}\text{CdS}\right]$ sample. Unfortuntealy, the n =2 excited
states of the A and B excitons could not be observed in other isotopic CdS
samples. Better samples are required for such measurements.
For GaAs or ZnSe, isotope substituents of either type should lead to shifts of
the E0 gap which have been calculated to be 430 (420) and 310 (300)
$\mu$eV/amu for cation (anion) mass replacement, respectively [15;23]. These
values are in reasonable agreement with data measured for GaAs $\left[\partial
E_{0}\text{/}\partial\text{M}_{Ga}\text{ = 390 (60) }\mu eV/amu\right]$ and
preliminary results for isotopic ZnSe obtained by Zhang et al [22] based on
photoluminescence measurements of the bound exciton (neutral acceptor I1)
$\left[\partial\text{E/}\partial\text{M}_{Se}\text{ = 140}\pm\text{ 40
}\mu\text{eV/amu and }\partial\text{E/}\partial M_{Zn}\text{ = 240}\pm\text{
40 }\mu eV\text{/amu}\right].$
Such behavior, however, is not found in wurtzite CdS. A previous reflectivity
and photoluminescence study of natCd32S and natCd34S shows [21] that for anion
isotope substitution the ground state (n = 1) energies of both A and B
excitons have a positive energy shift with the large rate of
$\partial$E/$\partial M_{S}$ = 740$\pm$100 $\mu$eV/amu. This value is more
than one order of magnitude larger than $\partial$E/$\partial$MCd obtained by
Zhang et al [22].
Several groups have conducted low-temperature studies of the direct and
indirect band gaps of natural and isotopically controlled Ge single crystals.
For the first time Agekyan et al [13] used photoluminescence, infrared
absorption, and Raman spectroscopy with a Ge crystals of natural composition
and a crystals with 85 % 76 Ge and 15% 74Ge. They found an indirect band-gap
change $\Delta$Eg = 0.9 meV and a direct band-gap change $\Delta$Eg = 1.25 meV
with an error of $\pm$0.05 meV. Etchegoin et al [24] and Davies et al [25]
reported photoluminescence studies of natural and several highly enriched ,
high quality single crystals of Ge. Measurement of the energies of impurity-
bound excitons permits by Davies et al the direct determination of band-gap
shifts with the crystal isotope mass because the radiative recombination does
not require phonon participation. As may be expected from the very large Bohr
orbit of the excitons (see Davies et al [25] and reference therein), their
binding energy only depends on the average isotope mass and not on the
isotopic disorder (see, however [4]). The rate of band-gap energy change with
isotope mass as determined by Davies et al is dEIG/dA = dENP/dA =
0.35$\pm$0.02 meV/amu. Etchegoin et al [24] obtained a very similar value.
Zollner et al.[25] have performed a numerical calculation of the electronic
bands using an empirical pseudopotential method including the necessary
lattice dynamics. They found for Ge (dEIG/dA)e-p = 0.41 meV. The total
calculated shift of the indirect band-gap energy with isotope mass adds up to
(dEIG/dA)total = 0.48 meV. This result compares favorably with the
experimental values stated above by Davies et al and by Etchegoin et al who
reported (dEIG/dA)total = 0.37$\pm$0.01 meV/amu.
Measurements of the direct band gap at the $\Gamma$ point ($\vec{k}$ = 0) in
the Brillouin zone have also been performed. Though the direct band gap is
technologically less important than the minimum indirect band gap, determing
the dependence of this gap on isotope mass is of the same fundamental
significance as the indirect band-gap studies. Davies et al [27] used low-
temperature optical-absorption measurements of very thin samples of Ge single
crystals with natural composition and three different, highly enriched
isotopes. They found dE/dA = 0.49 $\pm$ 0.03 meV/amu for the temperature
extrapolated to zero. Parks et al [28] have used piezo- and photomodulated
reflectivity spectra of four monoisotopic and one natural Ge crystals. These
techniques do not require the extreme sample thinning which is necessary for
optical-absorption measurements and the derivative nature of the spectra
emphasizes the small changes. The excellent signal-to- noise ratio and the
superb spectral resolution allowed a very accurate determination of the
dependence of EDG on isotopic mass. At very low temperatures an inverse
square-root dependence accurately describes band-gap dependence EDG =
E${}_{DG}^{\infty}$ \+ $\frac{C}{\sqrt{M}}$ .
A fit through five data points yields E${}_{DG}^{\infty}$ = 959 meV and C =
-606 meV/amu1/2. Written as a linear dependence for the small range of
isotopic masses, Parks et al find dEDG/dA = 0.49 meV/amu, in perfect agreement
with the results of Davies et al [27]. Parks et al also determined the isotope
mass dependence of the sum of the direct gap and the split-off valence band
$\left(\Delta_{0}\right)$ and find d(EDG \+ $\Delta_{0}$)/dA = 0.74 meV/amu.
The experimental results can be compared to the Zollner et al [26]
calculations which are found to be of the correct order of magnitude. The
theoretical estimates for the contributions of the linear isotope shifts of
the minimum, indirect gaps which are caused by electron-phonon interaction,
are too large by a factor of $\sim 1.7$ and for the smallest direct gap they
are too large by a factor $\sim$ 3.2.
Substitution of Ga70 on the Ga76 increases the band gap in GaAs [23] on 10.5
cm-1. The interested results were communicated in papers of Cardona and
coworkers [15; 12], where it was studied the dependence of Eg on the isotope
effect in CuCl crystals. When the Cu64 on the Cu65 is substituted the value of
Eg in CuCl crystals decreased on 1.24 cm-1, e.g. the isotope effect on the
electronic excitation has an opposite sign.
Considering the series of Ge, GaAs, ZnSe, CuBr, for example, the 3d states of
the first constituent play an increasing role in determining the band
structure. In Ge these states can be considered as localized core states
(atomic energy level $\approx$ -30 eV). Already, however, in GaAs they have
moved up in energy by 10 eV, and their hybridization with the top of the
valence band affects the gap . Proceeding further in the series, this effect
becomes more important, and in CuBr and Cu 3d states even overlap in energy
with halogen p-states, with which they strongly hybridize. Therefore, we
cannot exclude that the main reason for the opposite sign of the isotopic
effect in these compounds may be connected to the different character of the
d-electron-phonon interactions in these semiconductors [6].
Analogous investigations of the direct absorption edge of Si (Eg = 3.562 eV)
have been performed in paper [8] made the value of the coefficient
$\partial$Eg/$\partial$M = 2.0 meV/amu. The measurements of
$\partial$Eg/$\partial$M for isotope effect in CsH (CsH$\rightarrow$CsD) gave
the value $\partial$Eg/$\partial$M = 60 meV/amu [29]. The change of the
indirect gap of diamond between pure C12 and C13has been determined by Collins
et al [10], using for this purpose the cathodoluminescence spectra of diamond.
From the results of Collins et al it was concluded that the dominant
contribution arises from electron-phonon coupling, and that there is a smaller
contribution due to a change in volume of the unit cell produced by changing
the isotope. These two terms were calculated as 13.5 $\pm$ 2.0 and 3.0 $\pm$
1.3 meV respectively (details see, also [30]).
All of these results are documented in Table II, where the variation of Eg and
$\partial$Eg/$\partial$M are shown at the isotope effect. We should
highlighted here that the most prominent isotope effect is observed in LiH
crystals, where the dependence of Eb = f (CH) is also observed and
investigated (see, also [2]). Using the least-squares method it was found the
empirical dependence of ln$\partial$Eg/$\partial$M $\sim$ f(lnEg), which is
depicted on Fig. 3. As can be seen the indicated dependence has a parabolic
character:
ln($\partial$E${}_{g}/\partial M$) = 6.105(lnEg)2 \- 7.870(lnEg) + 0.565. (5)
From this figure it can be concluded also that the small variation of the
nuclear mass causes the small changes in Eg also. When the nuclear mass
increases it causes the large changes in Eg (C; LiH; CsH) [30]. Moreover as
can be seen from Fig. 3 in last case the empirical dependence
ln$\partial$Eg/$\partial$M $\sim$ f(lnEg) is very close to the linear one and
in ordinary coordinate system it has a next expression: $\sqrt{\partial
E_{g}/\partial M\text{ }}$ = E${}_{g}^{\sqrt{6.105}}$ ($\partial
E_{g}/\partial M$ = E${}_{g}^{3.0525}$). By the way it should be noted that at
the large changes Eg result in the changes of the force constants at the
isotope effect from the large variation of nuclear mass. Observable in Fig. 3
rather large scattering data in the rate of change Eg on the isotope mass in
the first step cause the different degree of the isotope-induced-disorder. The
last effect, as is well-known (see e.g.[4]), due the different magnitude of
the scattering potential at the isotope substitution.
We conclude the discussion of the band-gap shifts with isotope mass by
observing that the effects are exceedingly small in the most cases (see Table
2) and most likely will not have any consequences any semiconductors
technological applications. However, this exercise is an excellent
demonstration of the advanced state of our quantative theoretical
understanding of the subtle effects of temperature, pressure and isotope mass
on the electronic band structure.
Thanks are due to Dr. M. Segall for a critical reading of the manuscript.
### 1.0.2 References
1\. G. Herzberg, Molecular Spectra and Molecular Structure. I. Diatomic
Molecules (Van Nos Reinhold, NY) 1939.
2\. V.G. Plekhanov, T.A. Betenekova, V.A. Pustovarov, Sov. Phys. Solid State
18, 1422 (1976); V.G. Plekhanov, Physics-Uspekhi (Moscow) 40, 553 (1997).
3\. A.F. Kapustinsky, L.M. Shamovsky and K.S. Baushkina, Physicochim. (USSR)
7, 799 (1937).
4\. V.G. Plekhanov, Phys. Rev. B54, 3869 (1996).
5\. V.G. Plekhanov, Phys. Solid State (St-Petersburg) 35, 1493 (1993).
6\. V.G. Plekhanov, Isotope Effects in Solid State Physics, Academic Press, NY
- London 2001.
7\. Y. Onodera and Y. Toyozawa, J. Phys. Soc. Japan 24, 341 (1968).
8\. L.F. Lastras-Martinez, T. Ruf, M. Konuma, M. Cardona, Phys. Rev. B61,
12946 (2000).
9\. R.J. Elliott, J.A. Krumhansl and P.L. Leath, Rev. Mod. Phys. 46, 465
(1974).
10\. A.T. Collins, S.C. Lawson, G. Davies and H. Kanda, Phys. Rev. Lett. 65,
891 (1990).
11\. N. Garro, A. Cantarrero, M. Cardona, T. Ruf, Phys. Rev. B54, 4732 (1996).
12\. A. Gobel, T. Ruf, Ch. Lin, M. Cardona, Phys. Rev. B56, 210 (1997).
13\. V.F. Agekyan, V.M. Asnin, A.M. Kryukov, Fiz. Tverd. Tela 31, 101 (1989)
(in Russian).
14\. E.E. Haller, J. Appl. Phys. 77, 2857 (1995).
15\. N. Garro, A. Cantarrero, M. Cardona, T. Ruf, Solid State Commun. 98, 27
(1996).
16\. F.I. Kreingol’d, Fiz. Tverd. Tela 20, 3138 (1978) (in Russian).
17\. F.I. Kreingol’d, ibid, 27, 2839 (1985) (in Russian).
18\. F.I. Kreingol’d, K.F. Lider, and L.E. Solov’ev, JETP Lett. 23, 679 (1976)
(in Russian).
19\. F.I. Kreingol’d, K.F. Lider, and V.F. Sapega, Fiz. Tverd. Tela 19, 3158
(1977) (in Russian).
20\. F. I. Kreingol’d and B.S. Kulinkin, ibid 28, 3164 (1986) (in Russian).
21\. F.I. Kreingol’d, K.F. Lider, and M.B. Shabaeva, ibid 26, 3940 (1984) (in
Russian).
22\. J.M. Zhang, T. Ruf, R. Lauck and M. Cardona, Phys. Rev. B57, 9716 (1998).
23\. N. Garro, A. Cantarero, M. Cardona, Phys. Rev. B54, 4732 (1996).
24\. P. Etchegoin, J. Weber, M. Cardona, Solid State Commun. 83, 843 (1992).
25\. G. Davies, E.C. Lightowlerst, K. Itoh, E.E. Haller, Semicond. Sci.
Technol. 7, 1271 (1992).
26\. S. Zollner, M. Cardona and S. Gopalan, Phys. Rev. B45, 3376 (1992).
27\. G. Davies, E.C. Lightowlerst, E. E. Haller, Semicond. Sci. Technol. 8,
2201 (1993).
28\. C. Parks, A.K. Ramdas, S. Rodriguez, E.E. Haller, Phys. Rev. B49, 14244
(1994).
29\. K. Ghanehari, H. Luo, A.L. Ruoff, Solid State, Commun. 95, 385 (1995);
100, 777 (1996).
30\. V.G. Plekhanov, Rep. Prog. Phys. 61, 1045 (1998).
### 1.0.3 Figure Captions.
Fig. 1. Mirror reflection spectra of crystals: LiH, curve 1; LiHxD1-x, curve
2; and LiD, curve 3 at 4.2 K. Light source without crystal, curve 4. Spectral
resolution of instrument indicated in the diagram. All curves have the same
energy scale.
Fig. 2. Dependence of the interband transition energy Eg in mixed LiHxD1-x
crystals on the hydrogen concentration x. The stright dashed line is the
linear dependence Eg = f(x) in the virtual crystal model. The solid line
corresponds to calculations using the polynom of second degree [2] Points
derived from reflection spectra indicated by crosses, and those from
luminescence spectra by triangles.
Fig. 3. The dependence of ln($\partial$Eg/$\partial$M) $\sim$f(lnEg); points
are experimental data from Table 2 and continuous line - calculation on the
fotmulae (5).
Table 1 .Values of the energy of maxima in exciton reflection spectra of pure
and mixed crystals at 2K, and energies of exciton binding Eb band-to-band
transitions Eg .
Energy, meV | LiH | LiH0.82D0.18 | LiH0.40D0.60 | LiD | Li6H (78K)
---|---|---|---|---|---
E1s | 4950 | 4967 | 5003 | 5043 | 4939
E2s | 4982 | 5001 | 5039 | 5082 | 4970
Eb | 42 | 45 | 48 | 52 | 41
Eg | 4992 | 5012 | 5051 | 5095 | 4980
Table 2. Values of the coefficients $\partial$E/$\partial$M (meV) and energies
0f the band-to-band transitions Eg(eV) according to indicated references.
Substance | $\partial$E${}_{g}/\partial$M (meV) | Eg (eV)
---|---|---
13C$\rightarrow^{12}$C | 14.6 [10] | 5.4125 [10]
7LiH$\rightarrow^{7}$LiD | 103 [2,4] | 4.992$\rightarrow$5.095 [2,4]
7LiH$\rightarrow^{6}$LIH | 12 [5] | 4.980 [2,4]
CsD$\rightarrow$CsH | 60 [29] | 4.440 [29]
30Si$\rightarrow^{28}$Si | 2 [8] | 3.652 [8]
68ZnO$\rightarrow^{64}$ZnO | 0.372 [16,20] | 3.400 [16,20]
Zn18O$\rightarrow$Zn16O | 3.533 [16,20] | 3.400 [16,20]
65CuCl$\rightarrow^{63}$CuCl | -0.076 [11,12] | 3.220 [11,12]
Cu37Cl$\rightarrow$Cu35Cl | 0.364 [11,12] | 3.220 [11,12]
Cd34S$\rightarrow$Cd32S | 0.370 [21] | 2.580 [21]
110CdS$\rightarrow^{116}$CdS | 0.040$\div$0.068 [22] | 2.580 [22]
Cu${}_{2}^{18}$O$\rightarrow$Cu${}_{2}^{16}$O | 1.116 [17-19] | 2.151 [17-19]
71GaAs$\rightarrow^{69}$GaAs | 0.39 [15] | 1.53 [15]
76Ge$\rightarrow^{72}$Ge | 0.225 [13] | 1.53 [13]
${}^{76}\rightarrow^{73}\rightarrow^{70}$Ge | 0.37 [24-28] | 0.74 [24-28]
| arxiv-papers | 2008-07-14T12:52:49 | 2024-09-04T02:48:56.761971 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "V.G. Plekhanov, N.V. Plekhanov",
"submitter": "Vladimir Plekhanov",
"url": "https://arxiv.org/abs/0807.2152"
} |
0807.2158 | # Universally-composable privacy amplification from causality constraints
Lluís Masanes ICFO-Institut de Ciencies Fotoniques, Mediterranean Technology
Parck, 08860 Castelldefels (Barcelona), Spain
###### Abstract
We consider schemes for secret key distribution which use as a resource
correlations that violate Bell inequalities. We provide the first security
proof for such schemes, according to the strongest notion of security, the so
called universally-composable security. Our security proof does not rely on
the validity of quantum mechanics, it solely relies on the impossibility of
arbitrarily-fast signaling between separate physical systems. This allows for
secret communication in situations where the participants distrust their
quantum devices.
In an experimental set-up where a Bell inequality Bell is violated, one has
the certainty that the outcomes of some local measurements are not determined
beforehand. This limits the amount of correlations between such outcomes and
other systems not involved in the experiment. It also limits the knowledge
about these outcomes that a distant party can have. This fundamental piece of
our understanding of physical reality can be exploited for implementing
information-theoretic tasks. For instance, in this letter we show that a
secret key generated from the outcomes of Bell-violating measurements is
secure. This reasoning is independent of quantum mechanics, the only key
assumption is the impossibility of arbitrarily-fast signaling between separate
systems.
The first scheme for generating secret key from Bell-violating correlations
was presented in BHK , and was followed by others AGM ; SGBMPA ; AMP . All
these schemes where presented with partial security proofs. The results
presented in this letter, complemented with the ones in us , provide a general
security proof without assumptions (apart from no signaling) for all these
schemes. We use the strongest security criterion, the so-called universally-
composable security C , which warrants that key distribution is secure in any
context. Our methods are very general, and can be adapted to other Bell
inequality-based key-distribution schemes.
No signaling Consider two parties, Alice and Bob, each having a physical
system which can be measured with different observables. Let $a(b)$ be the
outcome when Alice(Bob)’s system is measured with one of the observables
parametrized by $x(y)$, with joint conditional probability distribution
denoted by $P_{a,b|x,y}$. We say that $P_{a,b|x,y}$ is a nonsignaling
distribution if the marginals depend only on their corresponding observables,
that is $P_{a|x,y}=P_{a|x}$ and $P_{b|x,y}=P_{b|y}$ for all $a,b,x,y$ MAG . It
is clear that if one of these conditions is not satisfied, then arbitrarily-
fast signaling is possible.
Nonlocality The distributions that can be written as
$P_{a,b|x,y}=\mbox{$\sum_{\lambda}$}P_{\lambda}\,P_{a|x,\lambda}\,P_{b|y,\lambda}$
(1)
are called local, and satisfy all Bell inequalities MAG . In the binary case
($a,b,x,y\in\\{0,1\\}$) all Bell inequalities are equivalent to the CHSH
inequality CHSH . For what follows, it is convenient to write the CHSH
inequality as $\mbox{$\langle\mbox{\small
CHSH}|$}P_{a,b|x,y}\rangle\geq\sqrt{2}$, where the vector
$\mbox{$|\mbox{\small CHSH}\rangle$}=\frac{1}{4\sqrt{2}}\
\begin{array}[]{|cc|cc|}\hline\cr 1&5&1&5\\\ 5&1&5&1\\\ \hline\cr 1&5&5&1\\\
5&1&1&5\\\ \hline\cr\end{array}$ (2)
contains the coefficients of the inequality, and the vector
$\mbox{$|P_{a,b|x,y}\rangle$}=\ \begin{array}[]{|cc|cc|}\hline\cr
P_{0,0|0,0}&P_{0,1|0,0}&P_{0,0|0,1}&P_{0,1|0,1}\\\
P_{1,0|0,0}&P_{1,1|0,0}&P_{1,0|0,1}&P_{1,1|0,1}\\\ \hline\cr
P_{0,0|1,0}&P_{0,1|1,0}&P_{0,0|1,1}&P_{0,1|1,1}\\\
P_{1,0|1,0}&P_{1,1|1,0}&P_{1,0|1,1}&P_{1,1|1,1}\\\ \hline\cr\end{array}$ (3)
contains the probabilities for all experimental settings. [We arrange the
components of these vectors in a table for the sake of clarity.] Notice that
in this form, the lower the quantity $\mbox{$\langle\mbox{\small
CHSH}|$}P_{a,b|x,y}\rangle$ the larger the violation. The distribution
attaining maximal violation ($\mbox{$\langle\mbox{\small
CHSH}|$}P_{a,b|x,y}\rangle=1/\sqrt{2}$) is the so-called PR-box PR , which can
be considered the maximally nonlocal (nonsignaling) distribution. The
correlations generated by measuring quantum systems are constrained by
Cirel’son’s bound $\mbox{$\langle\mbox{\small CHSH}|$}P_{a,b|x,y}\rangle\geq
2^{-1/2}3-1\approx 1.121$ TB .
Privacy amplification (PA) is the procedure by which a partially secret
$N_{\mathrm{r}}$-bit string $\mathbf{a}$ (the raw key) is transformed into a
highly-secret $N_{\mathrm{s}}$-bit string $\mathbf{k}$ (the secret key) gpa .
Usually, the secret key is shorter than the raw key
($N_{\mathrm{s}}<N_{\mathrm{r}}$), which is the price for the gain in privacy.
The function implementing this transformation $h(\mathbf{a})=\mathbf{k}$ is
called hash function. It is usually the case that the hash function has to be
generated randomly after the raw key $\mathbf{a}$ has been obtained, but in
our scheme, $h$ is fixed from the beginning and known to everybody, including
the eavesdropper (Eve). An ideal secret key is a uniformly-distributed random
variable $\mathbf{k}$ which is uncorrelated with the rest of the universe
(Eve). The information held by Eve is encoded in the state of a physical
system, which can be measured with one of many different observables,
parametrized by $z$. If $P_{e|z}$ is the distribution for the outcomes when
this system is measured with the observable $z$, then the distribution of an
ideal secret key is
$P^{\mathrm{ideal}}_{\mathbf{k},e|z}=2^{-N_{\mathrm{s}}}P_{e|z}$. Usually, the
real secret key generated by PA is not guaranteed to be an ideal secret key,
$P_{\mathbf{k},e|z}\neq 2^{-N_{\mathrm{s}}}P_{e|z}$.
In general, PA constitutes a sub-routine within cryptographic protocols, which
use secret key as an ingredient (an example being the encryption of messages).
It is desirable that the result obtained when any of these protocols is fed
with the real secret key, is the same as if fed with an ideal secret key, with
arbitrarily high probability. If this is the case, then we say that PA is
universally composable, because it is secure in any context. Clearly, this
happens if the real and ideal secret keys are indistinguishable.
The most general strategy for distinguishing the bipartite states
$P_{\mathbf{k},e|z}$ (the real key) and $2^{-N_{\mathrm{s}}}P_{e|z}$ (the
ideal key) consists of performing joint measurements on the key and Eve’s
system. The no signaling formalism alone does not say anything about joint
measurements. However, the key is a classical system which can be observed
without disturbing the global state. Therefore, the most general strategy is
to read $\mathbf{k}$ and chose an observable $z$ depending on its value. It is
well known that the probability of guessing correctly with the optimal
strategy is
$p_{\mathrm{correct}}=\frac{1}{2}+\frac{1}{4}\sum_{\mathbf{k}}\max_{z}\sum_{e}\left|P_{\mathbf{k},e|z}-2^{-N_{\mathrm{s}}}P_{e|z}\right|\
.$ (4)
Notice that the maximization on $z$ depends on $\mathbf{k}$. When (4) is close
to $1/2$, the optimal strategy for distinguishing the real from the ideal key
is as good as a random guess—this is the security condition that we consider.
In key distribution from Bell-violating correlations, Alice has $N$ systems,
Bob has $N$ systems and, without loss of generality, Eve has one “big” system,
jointly distributed according to an arbitrary (unknown)
$P_{\mathbf{a},\mathbf{b},e|\mathbf{x},\mathbf{y},z}$. [Bold symbols
correspond to bit-string variables.] It is usually assumed that this is a
$(2N+1)$-partite nonsignaling distribution us (i.e. the marginals only depend
on their corresponding observables), however, we are able to proceed with a
weaker assumption. If the secret key is a function of Alice’s string
$\mathbf{k}=h(\mathbf{a})$, then Bob’s $N$ systems can be considered as a
single “big” system, that is, no-signaling within Bob’s systems is not
required in our proof. We refer to this assumption as “$(N+2)$-partite no
signaling”. According to WR , the even weaker assumption of 3-partite no
signaling (where Alice’s $N$ systems are also considered as single one) is
insufficient to warrant security. Of these $N$ pairs of systems,
$N_{\mathrm{r}}\ (N_{\mathrm{r}}<N)$ are used for generating the raw key, and
the rest are used to estimate how much nonlocality is shared by Alice and Bob
us . In the large-$N$ limit, $N_{\mathrm{r}}$ is equal to $N$ up to terms
sublinear in $N$—this is denoted by $N_{\mathrm{r}}\approx N$.
The following result establishes the security of Alice’s key
$\mathbf{k}=h(\mathbf{a})$ when $\mathbf{a}$ is generated by measuring
$N_{\mathrm{r}}$ of Alice’s systems with the observable $x=0$. Of course, it
is necessary that the correlations shared by Alice and Bob
$P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}$ have a sufficiently small
value of $\mbox{$\langle\mbox{\small CHSH}|$}^{\otimes
N_{\mathrm{r}}}\mbox{$|P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}\rangle$}$,
or in other words, are sufficiently nonlocal. However, the goal of key
distribution is that both, Alice and Bob, hold the secret key $\mathbf{k}$.
Later we address this problem.
Main result For almost all functions
$h:\\{0,1\\}^{N_{\mathrm{r}}}\rightarrow\\{0,1\\}^{N_{\mathrm{s}}}$ and any
$(N_{\mathrm{r}}+2)$-partite nonsignaling distribution
$P_{\mathbf{a},\mathbf{b},e|\mathbf{x},\mathbf{y},z}$, the random variable
$\mathbf{k}=h(\mathbf{a})$ satisfies
$\displaystyle\sum_{\mathbf{k}}\max_{z}\sum_{e}\left|P_{\mathbf{k},e|\mathbf{x}=\mathbf{0},z}-2^{-N_{\mathrm{s}}}P_{e|z}\right|$
(5) $\displaystyle\leq$
$\displaystyle\sqrt{2}^{N_{\mathrm{s}}+\sqrt{N_{\mathrm{r}}}}\
\mbox{$\langle\mbox{\small CHSH}|$}^{\otimes
N_{\mathrm{r}}}\mbox{$|P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}\rangle$}\
,$
where $\mathbf{0}$ is the zero vector.
Here and in the rest of the letter we say that “almost all functions have a
particular property” if when randomly picking a function $h$ with uniform
distribution over all functions
$h:\\{0,1\\}^{N_{\mathrm{r}}}\rightarrow\\{0,1\\}^{N_{\mathrm{s}}}$ then the
probability that $h$ does not have that particular property is lower than
$2\exp{\\!(5N_{\mathrm{r}}-2^{\sqrt{N_{\mathrm{r}}}}/4)}$. The above result is
also true for any $\mathbf{x}\neq\mathbf{0}$, but for simplicity we only
consider the case $\mathbf{x}=\mathbf{0}$, which is sufficient for key
distribution.
When the given correlations $P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}$
are generated by measuring quantum systems Cirel’son’s bound implies
$\mbox{$\langle\mbox{\small CHSH}|$}^{\otimes
N_{\mathrm{r}}}\mbox{$|P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}\rangle$}>1$,
which prevents the right-hand side of (5) to be small. Hence, this simple
scheme does not work with quantum correlations. This problem is solved by the
BHK protocol, which yields large secure secret keys. The BHK protocol is
analyzed below. Now, we proceed to prove the main result, and start by stating
two lemmas which are proven in the Appendix.
Lemma 1 For any $(N_{\mathrm{r}}+1)$-partite nonsignaling distribution
$P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}$ we have
$P_{\mathbf{a}|\mathbf{x}=\mathbf{0}}=\langle\Gamma_{\mathbf{a}}\mbox{$|P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}\rangle$}$,
where
$\mbox{$|\Gamma_{\mathbf{a}}\rangle$}=\mbox{$|\gamma_{a_{1}}\rangle$}\otimes\cdots\otimes\mbox{$|\gamma_{a_{N_{\mathrm{r}}}}\rangle$}$
and
$\mbox{$|\gamma_{0}\rangle$}=\frac{1}{8}\ \begin{array}[]{|rr|rr|}\hline\cr
1&-3&1&5\\\ 5&1&-3&1\\\ \hline\cr 1&-3&5&1\\\ 5&1&1&-3\\\
\hline\cr\end{array}\ \ ,\ \ \mbox{$|\gamma_{1}\rangle$}=\frac{1}{8}\
\begin{array}[]{|rr|rr|}\hline\cr 1&5&1&-3\\\ -3&1&5&1\\\ \hline\cr
1&5&-3&1\\\ -3&1&1&5\\\ \hline\cr\end{array}\ .$
Lemma 2 For any given function
$h:\\{0,1\\}^{N_{\mathrm{r}}}\rightarrow\\{0,1\\}^{N_{\mathrm{s}}}$ and any
$\mathbf{k}\in\\{0,1\\}^{N_{\mathrm{s}}}$, define ${\cal
A}_{\mathbf{k}}=h^{-1}(\mathbf{k})$ and
$\mbox{$|\Gamma_{\mathcal{A}_{\mathbf{k}}}\rangle$}=\sum_{\mathbf{a}\in\mathcal{A}_{\mathbf{k}}}\mbox{$|\Gamma_{\mathbf{a}}\rangle$}$.
Almost all functions $h$ satisfy
$\big{|}2^{N_{\mathrm{s}}}\,\mbox{$|\Gamma_{\mathcal{A}_{\mathbf{k}}}\rangle$}-4^{-N_{\mathrm{r}}}\mbox{$|1\mathrm{s}\rangle$}\big{|}\preceq\sqrt{2}^{N_{\mathrm{s}}+\sqrt{N_{\mathrm{r}}}}\
\mbox{$|\mbox{\small CHSH}\rangle$}^{\otimes N_{\mathrm{r}}}\ ,$ (6)
for all $\mathbf{k}$, where the symbol $|\cdot|$ denotes entry-wise absolute
value, the symbol $\preceq$ denotes entry-wise less or equal than, and
$\mbox{$|1\mathrm{s}\rangle$}\in\mathbb{R}^{16^{N_{\mathrm{r}}}}$ has all
entries equal to one.
Proof of the main result Let $h$ be any of the functions which satisfies (6),
and for each $\mathbf{k}$, let $|\Gamma_{{\cal A}_{\mathbf{k}}}\rangle$ be the
vector defined in Lemma 2. Using
$P_{\mathbf{k}|\mathbf{x}=\mathbf{0}}=\langle\Gamma_{\mathcal{A}_{\mathbf{k}}}|P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}\rangle$,
the convexity of the absolute-value function, the inequality (6), and the fact
that the marginal for $\mathbf{a},\mathbf{b}$ cannot depend on $z$, we have
$\displaystyle\sum_{\mathbf{k}}\max_{z}\sum_{e}P_{e|z}\Big{|}P_{\mathbf{k}|\mathbf{x}=\mathbf{0},e,z}-2^{-N_{\mathrm{s}}}\Big{|}$
(7) $\displaystyle\leq$
$\displaystyle\sum_{\mathbf{k}}\max_{z}\sum_{e}P_{e|z}\,\Big{|}\mbox{$\langle\Gamma_{\mathcal{A}_{\mathbf{k}}}|$}-2^{-N_{\mathrm{s}}-2N_{\mathrm{r}}}\mbox{$\langle
1\mathrm{s}|$}\Big{|}\mbox{$|P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y},e,z}\rangle$}$
$\displaystyle=$
$\displaystyle\sqrt{2}^{N_{\mathrm{s}}+\sqrt{N_{\mathrm{r}}}}\mbox{$\langle\mbox{\small
CHSH}|$}^{\otimes
N_{\mathrm{r}}}\mbox{$|P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}\rangle$}\
,$
which is precisely (5). $\Box$
Error correction and public communication It is usually the case that the
given distribution $P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}$ does not
provide perfect correlations between $\mathbf{a}$ and $\mathbf{b}$. Hence, if
$\mathbf{a}$ is the raw key, Bob has to correct the errors in $\mathbf{b}$
before applying the hash function $h$. This can be done by Alice publishing
some information about $\mathbf{a}$, and Bob using it for correcting his
errors. This is a standard procedure in quantum key distribution, which is
detailed in us or Renner . Other procedures within the key distribution
protocol may also require public communication. Let the $N_{\mathrm{c}}$-bit
string ${\bf c}$ be all the information about $\mathbf{a}$ that Alice has
published during the protocol. Because ${\bf c}$ is a function of
$\mathbf{a}$, we can still use the main result (5) in this new setting if we
let both, $\mathbf{k}$ and ${\bf c}$, to be the outcomes of the function
$h:\\{0,1\\}^{N_{\mathrm{r}}}\rightarrow\\{0,1\\}^{N_{\mathrm{c}}}\times\\{0,1\\}^{N_{\mathrm{s}}}$.
However, $\mathbf{k}$ and ${\bf c}$ play different roles: $\mathbf{k}$ is the
secret key and ${\bf c}$ is part of the information owned by Eve. Hence, the
extension of the security condition (5) to the present setting is
$\displaystyle\sum_{\mathbf{k},{\bf
c}}\max_{z}\sum_{e}\big{|}P_{\mathbf{k},{\bf
c},e|z}-2^{-N_{\mathrm{s}}}P_{{\bf c},e|z}\big{|}$ (8) $\displaystyle\leq$
$\displaystyle
2\,\sqrt{2}^{N_{\mathrm{c}}+N_{\mathrm{s}}+\sqrt{N_{\mathrm{r}}}}\
\mbox{$\langle\mbox{\small CHSH}|$}^{\otimes
N_{\mathrm{r}}}\mbox{$|P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}\rangle$}\
,\quad$
where here and in the rest, the conditioning on $\mathbf{x}=\mathbf{0}$ is
implicit. This inequality is obtained by taking (5) and using the triangular
inequality with the third distribution
$2^{-N_{\mathrm{c}}-N_{\mathrm{s}}}P_{e|z}$. The secret key is secure if the
right-hand side of (8) can be made arbitrarily small (as $N_{\mathrm{r}}$
grows). This happens when the length of the final key is
$N_{\mathrm{s}}\approx\log_{2}\\!\Big{[}\mbox{$\langle\mbox{\small
CHSH}|$}^{\otimes
N_{\mathrm{r}}}\mbox{$|P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}\rangle$}^{-2}\Big{]}-N_{\mathrm{c}}\
,$ (9)
up to sub-linear terms.
Parameter estimation In the unconditional-security scenario, the honest
parties are given $N$ pairs of systems in a completely unknown global
distribution. To perform a key distribution protocol, and in particular to set
the numbers $N_{\mathrm{s}}$ and $N_{\mathrm{c}}$, they need to bound some
quantities, like for instance $\mbox{$\langle\mbox{\small CHSH}|$}^{\otimes
N}\mbox{$|P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}\rangle$}$. In order
to do so, they invest some of the given pairs to obtain information about the
distribution $P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}$ of the
$N_{\mathrm{r}}$ remaining pairs. More precisely, they compute the bounds for
$N_{\mathrm{s}},N_{\mathrm{c}}$ for another distribution
$P_{\mathbf{a},\mathbf{b},e|\mathbf{x},\mathbf{y},z}^{\prime}$, which is
warranted to be close to the real (unknown) one
($\sum_{\mathbf{a},\mathbf{b},e}|P_{\mathbf{a},\mathbf{b},e|\mathbf{x},\mathbf{y},z}^{\prime}-P_{\mathbf{a},\mathbf{b},e|\mathbf{x},\mathbf{y},z}|\leq\epsilon$
for all $\mathbf{x},\mathbf{y}$). This is explained with full detail in us .
It is shown in arxiv that
$\displaystyle\sum_{\mathbf{k},{\bf
c}}\max_{z}\sum_{e}\big{|}P_{\mathbf{k},{\bf
c},e|z}-2^{-N_{\mathrm{s}}}P_{e,{\bf c}|z}\big{|}$ $\displaystyle\leq$
$\displaystyle
2\,\sqrt{2}^{N_{\mathrm{s}}+N_{\mathrm{c}}+\sqrt{N_{\mathrm{r}}}}\
\mbox{$\langle\mbox{\small CHSH}|$}^{\otimes
N_{\mathrm{r}}}\mbox{$|P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}^{\prime}\rangle$}+2\epsilon\
,$
which provides the security bound for the real (unknown) distribution in terms
of properties of any $\epsilon$-close primed distribution.
The BHK protocol introduced in BHK and analyzed in AMP ; us gives a rate of
one secret bit per singlet ($\mbox{$|00\rangle$}+\mbox{$|11\rangle$}$). It is
remarkable that this protocol, where the adversary is only constrained by no
signaling, gives the same rate as if the adversary is constrained by no
signaling plus quantum mechanics. The essential novelty of the BHK protocol is
to measure each system with $m\geq 2$ observables, $x\in\\{1,\ldots m\\}$. In
this case, instead of the CHSH, we use the Braunstein-Caves Bell inequality BC
, which can be expressed as $\langle\mbox{\small
BC}\mbox{$|P_{a,b|x,y}\rangle$}\geq\sqrt{2}$, with
$\mbox{$|\mbox{\small
BC}\rangle$}=\frac{1}{2\sqrt{2}\,m}\,\begin{array}[]{|cc|cc|cc|cc|cc|}\hline\cr
1&\alpha&1&\alpha&&&&&&\\\ \alpha&1&\alpha&1&&&&&&\\\
\hline\cr&&1&\alpha&1&\alpha&&&&\\\ &&\alpha&1&\alpha&1&&&&\\\
\hline\cr&&&&\ddots&&\ddots&&&\\\ &&&&&&&&&\\\
\hline\cr&&&&&&\ddots&&1&\alpha\\\ &&&&&&&&\alpha&1\\\
\hline\cr\alpha&1&&&&&&&1&\alpha\\\ 1&\alpha&&&&&&&\alpha&1\\\
\hline\cr\end{array}\ \ ,$ (11)
where $\alpha=2m+1$, and the empty entries represent zeroes. Notice that for
$m=2$ this is equivalent to the CHSH inequality (2). Following the same
methods as above, one can prove inequalities analogous to (5), (8),
(Universally-composable privacy amplification from causality constraints), and
obtain a key rate as in (9) but with the Braunstein-Caves Bell inequality
$N_{\mathrm{s}}\approx\log_{2}\\!\Big{[}\mbox{$\langle\mbox{\small
BC}|$}^{\otimes
N_{\mathrm{r}}}\mbox{$|P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}\rangle$}^{-2}\Big{]}-N_{\mathrm{c}}\
.$ (12)
This rate formula can be improved by modifying $|\mbox{\small BC}\rangle$ in
the following way: take the expression (11) and substitute $\alpha$ by
$\sqrt{1+4m^{2}}$. The security of this rate will be proven somewhere else.
If Alice and Bob share singlets or something close to it, in the estimation
process they measure them with all the observables corresponding to points in
the equator of the block sphere (see BHK ; AMP ; us for details), the
generated correlations have $\mbox{$\langle\mbox{\small BC}|$}^{\otimes
N_{\mathrm{r}}}\mbox{$|P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}\rangle$}\approx
1/\sqrt{2}$, for large $m$. The raw keys $\mathbf{a},\mathbf{b}$ are generated
by measuring all systems with the same observable $x=0$, then
$\mathbf{a}=\mathbf{b}$ and $N_{\mathrm{c}}\approx 0$. Formula (12) tells that
the secret key rate is one secret bit per singlet: $N_{\mathrm{s}}\approx
N_{\mathrm{r}}$. This rate cannot be improved because it is also the optimal
rate achievable against a much weaker (quantum) adversary.
Conclusions We show, for the first time, that key distribution from Bell-
violating correlations is secure according to the strongest notion of
security, the so called universally-composable security. This provides the
possibility of implementing secure cryptographic protocols with untrusted
quantum devices AGM ; no trust . In this model, Alice and Bob have to trust
some of their apparatuses (classical computers and the random number
generator), but can distrust the devices for preparing and measuring the
quantum systems sent through the channel. The efficiency rate is slightly
lower than the one obtained in standard quantum key distribution, where
trusting the quantum devices is necessary.
Interestingly, in our scheme, Bell-inequality violation plays the same role as
the min entropy Renner does in standard quantum key distribution.
Specifically, equations (5) and (9) have a quantum counterpart, obtained with
the exchange
$\log_{2}\\!\Big{[}\mbox{$\langle\mbox{\small CHSH}|$}^{\otimes
N}\mbox{$|P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}\rangle$}^{-2}\Big{]}\
\ \leftrightarrow\ \ H_{\mathrm{min}}(\mathbf{a}|e)\ .$ (13)
A novelty of our scheme is that randomness extraction, or equivalently PA, can
be performed with a constant hash function. This contrasts with previous
methods for extracting randomness (two-universal hashing gpa , extractors,
etc.), which need random functions. However, we still lack an explicit
construction for one of such hash functions.
Acknowledgements The author is grateful to Renato Renner and Andreas Winter
for valuable comments. This work is supported by Caixa Manresa, the spanish
MEC (FIS2005-04627, FIS2007-60182, Consolider QOIT), and the EU-IP programme
SCALA.
## References
* (1) J. S. Bell; Physics 1(3), 195 (1964).
* (2) J. Barrett, L. Hardy, A. Kent; Phys. Rev. Lett. 95, 010503 (2005).
* (3) A. Acin, N. Gisin, Ll. Masanes; Phys. Rev. Lett. 97, 120405 (2006).
* (4) V. Scarani, N. Gisin, N. Brunner, Ll. Masanes, S. Pino, A. Acin; Phys. Rev. A 74, 042339 (2006).
* (5) A. Acin, S. Massar, S. Pironio; New J. Phys. 8, 126 (2006).
* (6) Ll. Masanes, R. Renner, A. Winter, J. Barrett, M. Christandl; arXiv:quant-ph/0606049
* (7) R. Canetti; Proc. 42nd IEEE Symposium on Foundations of Computer Science (FOCS), 136 (2001).
* (8) Ll. Masanes, A. Acín, N. Gisin; Phys. Rev. A 73, 012112 (2006).
* (9) J. F. Clauser, M. A. Horne, A. Shimony, R. A. Holt; Phys. Rev. Lett. 23, 880 (1969).
* (10) S. Popescu, D. Rohrlich; Found. Phys. 24, 379 (1994).
* (11) B. Cirel’son; Lett. Math. Phys. 4, 93 (1980).
* (12) C. H. Bennett, G. Brassard, C. Crépeau, U. M. Maurer; IEEE Trans. Inf. Theory, vol. 41, no. 6 (1995).
* (13) T. M. Cover, J. A. Thomas; Elements on Information Thoery, John Wiley & Sons, Inc. (2006).
* (14) E. Hänggi, R. Renner, S. Wolf; in preparation.
* (15) Ll. Masanes; arXiv:0807.2158.
* (16) R. Renner; Security of Quantum Key Distribution, PhD thesis, quant-ph/0512258.
* (17) S. Braunstein, C. Caves; Ann. Phys. 202, p. 22 (1990).
* (18) A Acín, N. Brunner, N. Gisin, S. Massar, S. Pironio, V. Scarani; Phys. Rev. Lett. 98, 230501 (2007).
## I Appendix
Here we show the two lemmas stated above.
Proof of Lemma 1. Here we use the same tools as in the proof of Lemma 16 from
us . By definition we can write
$P_{\mathbf{a}|\mathbf{x}=\mathbf{0}}=\langle\Gamma^{\prime}_{\mathbf{a}}\mbox{$|P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}\rangle$}$,
where $\mbox{$|\Gamma^{\prime}_{\mathbf{a}}\rangle$}=$
$\mbox{$|\gamma^{\prime}_{a_{1}}\rangle$}\otimes\cdots\otimes\mbox{$|\gamma^{\prime}_{a_{N_{\mathrm{r}}}}\rangle$}$
and
$\mbox{$|\gamma^{\prime}_{0}\rangle$}=\left[\begin{array}[]{rr|rr}1&1&0&0\\\
0&0&0&0\\\ \hline\cr 0&0&0&0\\\ 0&0&0&0\end{array}\right]\
,\quad\mbox{$|\gamma^{\prime}_{1}\rangle$}=\left[\begin{array}[]{rr|rr}0&0&0&0\\\
1&1&0&0\\\ \hline\cr 0&0&0&0\\\ 0&0&0&0\end{array}\right]\ .$
The fact that Bob (when considered as a single system) cannot signal to Alice
can be expressed as
$P_{\mathbf{a}|\mathbf{x},\mathbf{y}}=P_{\mathbf{a}|\mathbf{x},\mathbf{y}^{\prime}}$
for any $\mathbf{y},\mathbf{y}^{\prime}$. This implies that
$P_{\mathbf{a}|\mathbf{x}=\mathbf{0}}=\langle\Gamma^{\prime\prime}_{\mathbf{a}}\mbox{$|P_{\mathbf{a},\mathbf{b}|\mathbf{x},\mathbf{y}}\rangle$}$
where
$\mbox{$|\Gamma^{\prime\prime}_{\mathbf{a}}\rangle$}=\mbox{$|\gamma^{\prime\prime}_{a_{1}}\rangle$}\otimes\cdots\otimes\mbox{$|\gamma^{\prime\prime}_{a_{N_{\mathrm{r}}}}\rangle$}$
and
$\displaystyle\mbox{$|\gamma^{\prime\prime}_{0}\rangle$}=\frac{1}{4}\left[\begin{array}[]{rr|rr}2&2&2&2\\\
0&0&0&0\\\ \hline\cr-1&-1&1&1\\\ 1&1&-1&-1\end{array}\right]\ ,$
$\displaystyle\mbox{$|\gamma^{\prime\prime}_{1}\rangle$}=\frac{1}{4}\left[\begin{array}[]{rr|rr}0&0&0&0\\\
2&2&2&2\\\ \hline\cr 1&1&-1&-1\\\ -1&-1&1&1\end{array}\right]\ .$
The fact that each of the $N_{\mathrm{r}}$ Alice’s systems cannot signal to
the rest, together with Bob’s systems, implies the statement of the lemma.
$\Box$
Proof of Lemma 2 Within this proof, the entries of any vector
$\mbox{$|\Phi\rangle$}\in\mathbb{R}^{16^{N_{\mathrm{r}}}}$ are labeled as
$\Phi(\mathbf{a},\mathbf{b},\mathbf{x},\mathbf{y})$. Also, for any pair of
bit-strings $\mathbf{x},\mathbf{y}$: (i) the string
$\mathbf{x}\cdot\mathbf{y}$ is the bit-wise product, (ii) the string
$\mathbf{x}\oplus\mathbf{y}$ is the bit-wise xor, and (iii) the integer
$\|\mathbf{x}\|$ is the number of ones in $\mathbf{x}$. Using this notation we
can write the entries of the vector $\mbox{$|\mbox{\small
CHSH}\rangle$}^{\otimes N_{\mathrm{r}}}$ as $\mbox{\small CHSH}^{\otimes
N_{\mathrm{r}}}(\mathbf{a},\mathbf{b},\mathbf{x},\mathbf{y})=2^{-5N_{\mathrm{r}}/2}5^{\|\mathbf{a}\oplus\mathbf{b}\oplus\mathbf{x}\cdot\mathbf{y}\|}$.
Next we prove inequality (6) for a given $\mathbf{k}$ and a given entry
$(\mathbf{a}_{0},\mathbf{b}_{0},\mathbf{x}_{0},\mathbf{y}_{0})$. Let
$V_{\mathbf{a}}=1$ if the string $\mathbf{a}$ belongs to ${\cal
A}_{\mathbf{k}}$, and $V_{\mathbf{a}}=0$ otherwise. If we pick a random
function $h$ with uniform distribution over the set of all functions, then the
random variables $V_{\mathbf{a}}$ are independent and distributed according to
$\mbox{Prob}\\{V_{\mathbf{a}}=1\\}=2^{-N_{\mathrm{s}}}$, for all $\mathbf{a}$.
Let
$\mu_{\mathbf{a}}=\Gamma_{\mathbf{a}}(\mathbf{a}_{0},\mathbf{b}_{0},\mathbf{x}_{0},\mathbf{y}_{0})$,
$M=\|\mathbf{a}_{0}\oplus\mathbf{b}_{0}\oplus\mathbf{x}_{0}\cdot\mathbf{y}_{0}\|$,
and note that $|\mu_{\mathbf{a}}|\leq 5^{M}8^{-N_{\mathrm{r}}}$ for all
$\mathbf{a}$. Following Bernstein’s contruction, for any $J$ and $\beta\geq 0$
we have
$\displaystyle\mathrm{Prob}\left\\{\sum_{\mathbf{a}}\mu_{\mathbf{a}}V_{\mathbf{a}}\geq
J\right\\}$ $\displaystyle\leq$ $\displaystyle\mathrm{Prob}\left\\{e^{-\beta
J+\beta\sum_{\mathbf{a}}\mu_{\mathbf{a}}V_{\mathbf{a}}}\geq 1\right\\}$
$\displaystyle\leq$ $\displaystyle e^{-\beta J}\left\langle
e^{\beta\sum_{\mathbf{a}}\mu_{\mathbf{a}}V_{\mathbf{a}}}\right\rangle$
$\displaystyle\leq$ $\displaystyle e^{-\beta
J}\prod_{\mathbf{a}}\left[2^{-N_{\mathrm{s}}}e^{\beta\mu_{\mathbf{a}}}+\left(1-2^{-N_{\mathrm{s}}}\right)\right]$
$\displaystyle\leq$ $\displaystyle e^{-\beta
J}\prod_{\mathbf{a}}\left[1+2^{-N_{\mathrm{s}}}\left(\beta\mu_{\mathbf{a}}+\beta^{2}\mu_{\mathbf{a}}^{2}\right)\right]$
$\displaystyle\leq$ $\displaystyle\exp{\\!\big{[}-\beta
J2^{-N_{\mathrm{s}}}\mbox{$\sum_{\mathbf{a}}$}(\beta\mu_{\mathbf{a}}+\beta^{2}\mu_{\mathbf{a}}^{2})\big{]}}$
where in (I) we need $|\beta\,5^{M}8^{-N_{\mathrm{r}}}|\leq 1$. In this step
we have used the expansion $e^{x}\leq 1+x+x^{2}$, which holds if $x\leq 1$.
With a little work one obtains
$\sum_{\mathbf{a}}\mu_{\mathbf{a}}=4^{-N_{\mathrm{r}}}$ and
$\sum_{\mathbf{a}}\mu_{\mathbf{a}}^{2}\leq 2^{-5N_{\mathrm{r}}}5^{2M}$.
Substituting this two expressions,
$J=2^{-N_{\mathrm{s}}-2N_{\mathrm{r}}}+2^{(\sqrt{N_{\mathrm{r}}}-N_{\mathrm{r}}-N_{\mathrm{s}})/2}\,4^{-N_{\mathrm{r}}}\,5^{M}$
and
$\beta=2^{(\sqrt{N_{\mathrm{r}}}+N_{\mathrm{r}}+N_{\mathrm{s}})/2}\,4^{N_{\mathrm{r}}}\,5^{-M}$
we get
$\displaystyle\mathrm{Prob}\big{\\{}\mbox{$\sum_{\mathbf{a}}$}\mu_{\mathbf{a}}V_{\mathbf{a}}\geq
2^{-N_{\mathrm{s}}-2N_{\mathrm{r}}}+2^{(\sqrt{N_{\mathrm{r}}}-N_{\mathrm{r}}-N_{\mathrm{s}})/2}\,4^{-N_{\mathrm{r}}}\,5^{M}\big{\\}}$
$\displaystyle\leq\ e^{-2^{\sqrt{N_{\mathrm{r}}}}/4}\ .\ \ $
Note that the chosen value for $\beta$ satisfies the required constraint. The
expression obtained when replacing “$\geq$” with “$\leq$” above, can be
derived in a similar way. Then
$\displaystyle\big{|}2^{N_{\mathrm{s}}}\,\Gamma_{\mathcal{A}_{\mathbf{k}}}(\mathbf{a}_{0},\mathbf{b}_{0},\mathbf{x}_{0},\mathbf{y}_{0})-4^{-N_{\mathrm{r}}}\big{|}$
(17) $\displaystyle>$
$\displaystyle\sqrt{2}^{N_{\mathrm{s}}+\sqrt{N_{\mathrm{r}}}}\ \mbox{\small
CHSH}^{\otimes
N_{\mathrm{r}}}(\mathbf{a}_{0},\mathbf{b}_{0},\mathbf{x}_{0},\mathbf{y}_{0})$
holds with probability $2\,e^{-2^{\sqrt{N_{\mathrm{r}}}}/4}$. However, we want
this to not hold for all $\mathbf{k}$ and all entries
$(\mathbf{a},\mathbf{b},\mathbf{x},\mathbf{y})$. The number of different
values of $\mathbf{k}$ is $2^{N_{\mathrm{s}}}$, and the number of different
entries is $16^{N_{\mathrm{r}}}$, then the probability for (6) being not true
is upper-bounded by $2\exp{\\!(5N_{\mathrm{r}}-2^{\sqrt{N_{\mathrm{r}}}}/4)}$.
$\Box$
| arxiv-papers | 2008-07-14T13:39:16 | 2024-09-04T02:48:56.767230 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Lluis Masanes",
"submitter": "Lluis Masanes",
"url": "https://arxiv.org/abs/0807.2158"
} |
0807.2200 | # Differential forms on locally convex spaces and the Stokes formula
Evelina Shamarova
###### Abstract
We prove a version of the Stokes formula for differential forms on locally
convex spaces announced in [10]. The main tool used for proving this formula
is the surface layer theorem proved in the paper [6] by the author. Moreover,
for differential forms of a Sobolev-type class relative to a differentiable
measure [1], we compute the operator adjoint to the exterior differential in
terms of standard operations of calculus of differential forms and the
logarithmic derivative. Previously, this connection was established under
essentially stronger assumptions on the space [8], the measure [7], or
smoothness of differential forms [5]. See also [4].
Grupo de Física Matemática, Universidade de Lisboa.
---
E-mail: evelina@cii.fc.ul.pt
## 1\. Calculus on a Sobolev-type class of differential forms on a locally
convex space
Let $H$ be a Hilbert space with the scalar product $(\,\cdot\,,\,\cdot\,)$,
$\\{e_{n}\\}_{n=1}^{\infty}$be an orthonormal basis of $H$. Let $\Gamma(n)$
denote the set of increasing sequences of natural numbers of length
$n\in{\mathbb{N}}$, $\Gamma(0)=\\{0\\}$. If $\gamma_{1}\in\Gamma(n)$,
$\gamma_{2}\in\Gamma(m)$, $n\geqslant m$, then we consider the sequences
$\gamma_{1}\cup\gamma_{2}$ and $\gamma_{1}\diagdown\gamma_{2}$ as elements of
$\Gamma(m+n)$ and $\Gamma(n-m)$ respectively, putting them in the increasing
order, if necessarily. For every $n\in{\mathbb{N}}$ and
$\gamma=(i_{1},\dots,i_{n})\in\Gamma(n)$, the symbol $e_{\gamma}$ denotes
$e_{i_{1}}\wedge\dots\wedge e_{i_{n}}$, $e_{0}=1$, where the vectors
$e_{i_{1}},\dots,e_{i_{n}}$ are considered as linear continuous functionals on
$H$ (the operation $\wedge$ is defined, for example, in [2]). By $L_{n}(H)$ we
denote the space of antisymmetric $n$-linear Hilbert-Schmidt functionals. Note
that $L_{n}(H)$ is a Hilbert space with the scalar product
$(g_{1},g_{2})_{n}=\sum\limits_{i_{1}<\dots<i_{n}}g_{1}(e_{i_{1}},\dots,e_{i_{n}})\,g_{2}(e_{i_{1}},\dots,e_{i_{n}})$,
and $\\{e_{\gamma}\\}_{\gamma\in\Gamma(n)}$ is the orthonormal basis in
$L_{n}(H)$. Let $\|\,\cdot\,\|_{n}$ denote the norm which corresponds to
$(\,\cdot\,,\,\cdot\,)_{n}$. Let us show that if $f\in L_{n}(H)$, $g\in
L_{m}(H)$, then $f\wedge g\in L_{m+n}(H)$. Indeed, let
$e_{\gamma}=e_{i_{1}}\wedge\dots\wedge e_{i_{n}}$, and let $g(e_{\gamma})$
denote $(g,e_{\gamma})_{n}=g(e_{i_{1}},\dots,e_{i_{n}})$. We obtain:
$\sum\limits_{\gamma\in\Gamma(m+n)}(f\wedge
g)^{2}(e_{\gamma})=\sum\limits_{\gamma\in\Gamma(m+n)}\Bigl{(}\sum\limits_{\gamma_{1}\in\Gamma(n)}\varepsilon(\sigma)\,f(e_{\gamma_{1}})\,g(e_{\gamma\diagdown\gamma_{1}})\Bigr{)}^{2}\\\
\leqslant
C_{m+n}^{n}\sum\limits_{\gamma\in\Gamma(m+n)}\sum\limits_{\gamma_{1}\in\Gamma(n)}f^{2}(e_{\gamma_{1}})\,g^{2}(e_{\gamma\diagdown\gamma_{1}})=C_{m+n}^{n}\sum\limits_{\begin{subarray}{c}\gamma_{1}\in\Gamma(n),\\\
\gamma_{2}\in\Gamma(m):\\\
\gamma_{1}\cup\gamma_{2}=\gamma\end{subarray}}f^{2}(e_{\gamma_{1}})\,g^{2}(e_{\gamma_{2}})\\\
\leqslant
C_{m+n}^{n}\sum\limits_{\gamma_{1}\in\Gamma(n)}f^{2}(e_{\gamma_{1}})\sum\limits_{\gamma_{2}\in\Gamma(m)}g^{2}(e_{\gamma_{2}})<\infty.$
Analogously to the finite dimensional case [11], for elements $f\in L_{m}(H)$
and $g\in L_{n}(H)$, $m>n$, one can define the element
$g\,\text{\LARGE$\lrcorner$}\,f\in L_{m-n}(H)$ by the formula
$(g\,\text{\LARGE$\lrcorner$}\,f,h)_{m-n}=(f,g\wedge h)_{n}$, which holds for
all $h\in L_{m-n}(H)$. Let us show that the operation $\lrcorner$ is well
defined for $f\in L_{m}(H)$ and $g\in L_{n}(H)$. Specifically, we have to show
that
$\sum\limits_{\gamma\in\Gamma(m-n)}(g\,\text{\LARGE$\lrcorner$}\,f,e_{\gamma})^{2}_{m-n}=\sum\limits_{\gamma\in\Gamma(m-n)}(f,g\wedge
e_{\gamma})_{m}^{2}<\infty.$
Note that
$(g\wedge
e_{\gamma})(e_{\gamma^{\prime}})=\sum_{\begin{subarray}{c}\gamma_{1}\in\Gamma(n),\\\
\gamma_{2}\in\Gamma(m-n):\\\
\gamma_{1}\cup\gamma_{2}=\gamma\end{subarray}}\varepsilon(\sigma)\,g(e_{\gamma_{1}})\,e_{\gamma}(e_{\gamma_{2}})=\begin{cases}0,&\gamma\not\subset\gamma^{\prime}\\\
\varepsilon(\sigma)g(e_{\gamma^{\prime}\diagdown\gamma}),&\gamma\subset\gamma^{\prime}\end{cases}$
where $\varepsilon(\sigma)$ is the permutation parity of $\sigma$. Also, we
used here the definition of the exterior multiplication [2]. Taking into
account the latter relation, we obtain:
$\sum\limits_{\gamma\in\Gamma(m-n)}(f,g\wedge
e_{\gamma})_{m}^{2}=\sum\limits_{\gamma\in\Gamma(m-n)}\Bigl{(}\sum\limits_{\gamma^{\prime}\in\Gamma(m)}f(e_{\gamma^{\prime}})\,(g\wedge
e_{\gamma})(e_{\gamma^{\prime}})\Bigr{)}^{2}=$
$=\sum\limits_{\gamma\in\Gamma(m-n)}\Bigl{(}\sum_{\begin{subarray}{c}\gamma^{\prime}\in\Gamma(m):\\\
\gamma\subset\gamma^{\prime}\end{subarray}}\varepsilon(\sigma)\,f(e_{\gamma^{\prime}})\,g(e_{\gamma^{\prime}\diagdown\gamma})\Bigr{)}^{2}\\\
\leqslant\sum\limits_{\gamma\in\Gamma(m-n)}\Bigl{(}\sum_{\begin{subarray}{c}\scriptstyle\gamma_{1}\in\Gamma(n):\\\
\scriptstyle\gamma\cap\gamma_{1}=\varnothing\end{subarray}}f^{2}(e_{\gamma\cup\gamma_{1}})\sum_{\begin{subarray}{c}\gamma_{1}\in\Gamma(n):\\\
\gamma\cap\gamma_{1}=\varnothing\end{subarray}}g^{2}(e_{\gamma_{1}})\Bigr{)}\\\
\leqslant\sum\limits_{\gamma_{1}\in\Gamma(n)}g^{2}(e_{\gamma_{1}})\sum_{\begin{subarray}{c}\gamma\in\Gamma(m-n),\\\
\gamma_{1}\in\Gamma(n):\\\
\gamma\cap\gamma_{1}=\varnothing\end{subarray}}f^{2}(e_{\gamma\cup\gamma_{1}})\leqslant
C_{m}^{n}\sum\limits_{\gamma_{1}\in\Gamma(n)}g^{2}(e_{\gamma_{1}})\sum\limits_{\gamma_{2}\in\Gamma(m)}f^{2}(e_{\gamma_{2}})<\infty.$
Now let $X$ be a locally convex space, and the Hilbert space $H$ be a vector
subspace of $X$.
###### Definition 1.
A mapping $f~{}:\,X\to L_{n}(H)$ is called a differential form of degree $n$
(or differential $n$-form) on $X$.
Note that every differential form $f$ can be presented as:
$f\\!=\\!\sum\limits_{\gamma\in\Gamma(n)}f_{\gamma}\,e_{\gamma}$, where
$f_{\gamma}$ are real-valued functions. The operations of exterior and
interior multiplications are defined for differential forms pointwise.
Let $f$ be a differential form of degree $n$.
###### Definition 2.
We say that $f$ possesses a differential if its coefficients $f_{\gamma}$ are
differentiable in each direction $e_{p}$ for $p\notin\gamma$, and for all
$x\in X$,
$\sum\limits_{\gamma\in\Gamma(n),\,p\notin\gamma}d_{e_{p}}f_{\gamma}(x)\,e_{p}\wedge
e_{\gamma}\in L_{n+1}(H)$. The differential $(n+1)$-form
$df=\sum\limits_{\gamma\in\Gamma(n),\,p\notin\gamma}d_{e_{p}}f_{\gamma}\,e_{p}\wedge
e_{\gamma}$
is called the differential of $f$.
###### Definition 3.
We say that $f$ possesses a codifferential if its coefficients $f_{\gamma}$
are differentiable in each direction $e_{p}$ for $p\in\gamma$, and for all
$x\in X$,
$\sum\limits_{p\in\gamma\in\Gamma(n)}d_{e_{p}}f_{\gamma}(x)\,e_{p}\,\text{\LARGE$\lrcorner$}\,e_{\gamma}\in
L_{n-1}(H)$. The differential $(n-1)$-form
$\delta
f=\sum\limits_{p\in\gamma\in\Gamma(n)}d_{e_{p}}f_{\gamma}\,e_{p}\,\text{\LARGE$\lrcorner$}\,e_{\gamma}$
is called the codifferential of $f$.
Let $\mathfrak{B}_{\scriptscriptstyle X}$ be the $\sigma$-algebra of Borel
subsets of the space $X$. A measure on $X$ means a $\sigma$-additive Hilbert
space valued function on $\mathfrak{B}_{\scriptscriptstyle X}$.
###### Definition 4.
A $\sigma$-additive $L_{n}(H)$-valued measure on $X$ is called a differential
form of codegree $n$.
Every differential form $\omega$ of codegree $n$ can be decomposed as:
$\omega=\sum\limits_{\gamma\in\Gamma(n)}\omega_{\gamma}e_{\gamma}$ where
$\omega_{\gamma}$ are real-valued $\sigma$-additive measures.
###### Definition 5.
Let $g$ be a bounded differential form of degree $m$, $\omega$ be a
differential form of codegree $n\geqslant m$ which is a measure of bounded
variation. The differential form $g\wedge\omega$ of codegree $n-m$ defined as
$(g\wedge\omega)(A)=\int_{A}g(x)\\!\,\text{\LARGE$\lrcorner$}\,\omega(dx)$
is called the exterior product of $g$ and $\omega$.
The differential form $g\wedge\omega$ is well defined. Indeed, the
differential form $\omega$ can be presented in the form
$\omega=f\cdot|\omega|$ (see [3]), where $|\omega|$ denotes the variation of
$\omega$, and $f$ is a differential form of degree $n$ such that
$\|f(x)\|_{n}=1$ for $|\omega|$-almost all $x$. We have
$(g\wedge\omega)(A)=\int_{A}\bigl{(}g(x)\\!\,\text{\LARGE$\lrcorner$}\,f(x)\bigr{)}|\omega|(dx).$
Further,
$\displaystyle\sum\limits_{\gamma\in\Gamma(n-m)}((g\wedge\omega)(A),e_{\gamma})_{n-m}^{2}=\sum\limits_{\gamma\in\Gamma(n-m)}\left(\int_{A}\bigl{(}g(x)\\!\,\text{\LARGE$\lrcorner$}\,f(x),e_{\gamma}\bigr{)}_{n-m}|\omega|(dx)\right)^{2}$
$\displaystyle\leqslant|\omega|(A)\sum\limits_{\gamma\in\Gamma(n-m)}\int_{A}\bigl{(}g(x)\\!\,\text{\LARGE$\lrcorner$}\,f(x),e_{\gamma}\bigr{)}^{2}_{n-m}|\omega|(dx)$
$\displaystyle=|\omega|(A)\int_{A}\|g(x)\\!\,\text{\LARGE$\lrcorner$}\,f(x)\|_{n-m}^{2}|\omega|(dx)\leqslant
C_{n}^{m}\,\bigl{(}|\omega|(A)\bigr{)}^{2}\sup_{x}\|g(x)\|_{m}<\infty.$
###### Definition 6.
We say that the differential form $\omega$ of codegree $n$ possesses a
differential if its coefficients $\omega_{\gamma}$ are differentiable in all
directions $e_{p}$ for $p\in\gamma$, and
$\sum\limits_{p\in\gamma\in\Gamma(n)}d_{e_{p}}\omega_{\gamma}(A)\,e_{p}\,\text{\LARGE$\lrcorner$}\,e_{\gamma}\in
L_{n-1}(H)$ for all $A\in\mathfrak{B}_{\scriptscriptstyle X}$. The
differential form $d\omega$ of codegree $n-1$ defined by
$d\omega=(-1)^{n-1}\sum\limits_{p\in\gamma\in\Gamma(n)}d_{e_{p}}\omega_{\gamma}\,e_{p}\,\text{\LARGE$\lrcorner$}\,e_{\gamma},$
is called the differential of $\omega$.
###### Lemma 1.
Let $g$ and $\omega$ be differential forms of degree $m$ and codegree $n+1>m$
respectively, both possess differentials. Further let $g$ and $\omega$ be such
that $g$ and $dg$ are bounded, $\omega$ and $d\omega$ are of bounded
variation. Then the differential form $g\wedge\omega$ possesses a
differential, and
$d(g\wedge\omega)=g\wedge d\omega+(-1)^{n}dg\wedge\omega.$ (1)
The equality (1) can be easily obtained. Indeed, one should use the
definitions of differentials for $g$ and $\omega$, the definition of the
operation $\wedge$, and compare the coefficients at each $e_{\gamma}$.
## 2\. The operator adjoint to the differential
Now we compute the operator adjoint to the operator $d$ for differential forms
of a Sobolev-type class relative to a real- or complex-valued
$\sigma$-additive measure on $X$. Let $\mu$ be such a measure. We assume that
$\mu$ is differentiable in each direction $e_{p}$ [1], and for all $x$,
$\|\beta^{\mu}(x)\|_{\scriptscriptstyle H}<\infty$, where
$\beta^{\mu}(x)=\sum_{p}\beta_{e_{p}}^{\mu}(x)\,e_{p}$, and
$\beta_{e_{p}}^{\mu}$ is the logarithmic derivative of the measure $\mu$ in
the direction $e_{p}$ [1, 9]. Further let the numbers $p>1$ and $q>1$ be such
that $1/p+1/q=1$. By $\Omega_{p}^{n}$, we denote the vector space of
differential $n$-forms $f$ satisfying the condition
$\int_{X}\|f(x)\|_{n}^{p}\,\mu(dx)<\infty$. Define a norm on $\Omega_{p}^{n}$
by
$\|f\|_{n,p}=\left(\int_{X}\|f(x)\|_{n}^{p}\,\mu(dx)\right)^{1\\!\left.\right/p}.$
For elements
$f=\sum\limits_{\gamma\in\Gamma(n)}f_{\gamma}(x)\,e_{\gamma}\in\Omega_{p}^{n}$
and
$\omega=\sum\limits_{\gamma\in\Gamma(n)}\omega_{\gamma}(x)\,e_{\gamma}\in\Omega_{q}^{n}$,
we define the bilinear operation
$\langle\omega,f\rangle_{n}=\int_{X}\bigl{(}\omega(x),f(x)\bigr{)}_{n}\,\mu(dx).$
(2)
The integral on the right-hand side exists by Hölder’s inequality and by the
definition of $\Omega_{p}^{n}$. By the definition of the scalar product
$(\cdot,\cdot)_{n}$ and by Lebesgue’s theorem, we rewrite (2):
$\displaystyle\langle\omega,f\rangle_{n}=\sum\limits_{\gamma\in\Gamma(n)}\int_{X}f_{\gamma}(x)\,\omega_{\gamma}(x)\,\mu(dx).$
Let $A_{p}^{n}$ be the vector subspace of $\Omega_{p}^{n}$, consisting of
differential forms $f$ possessing the codifferential $\delta
f\in\Omega_{p}^{n-1}$, satisfying the inequality
$\int_{X}\|\beta^{\mu}(x)\|_{\scriptscriptstyle
H}^{p}\,\|f(x)\|_{n}^{p}\,\mu(dx)<\infty,$ (3)
and such that the following condition (i) is fulfilled. Condition (i): for
every $\gamma\in\Gamma(n)$, there exists a $\delta>0$ and non-negative
functions $g_{\gamma}(x)$, $g_{1\gamma}(x)$, and $g_{2\gamma}(x)$, such that
$g_{\gamma}(x)$ is $d_{e_{p}}\mu$-summable for every $p\notin\gamma$,
$g^{2}_{1\gamma(x)}$ and $g^{2}_{2\gamma}(x)$ are $\mu$-summable, and for all
$p\notin\gamma$, for $|t|<\delta$,
$|f_{\gamma}(x+te_{p})|<\min\\{g_{\gamma}(x),g_{1\gamma}(x)\\}$ and
$|d_{e_{p}}f_{\gamma}(x+te_{p})|<g_{2\gamma}(x)$. On $A_{p}^{n}$ we define the
norm
$\|f\|_{A_{p}^{n}}=\|f\|_{n,p}+\|\delta
f\|_{n-1,p}+\left(\int_{X}\|\beta^{\mu}(x)\|^{p}_{\scriptscriptstyle
H}\,\|f(x)\|_{n}^{p}\,\mu(dx)\right)^{1\\!\left.\right/p}.$
Further let $B_{q}^{n}$ be the vector subspace of $\Omega_{q}^{n}$ consisting
of differential forms $\omega$ possessing the differential
$d\omega\in\Omega_{q}^{n+1}$ and satisfying the following condition (ii).
Condition (ii): for every $\gamma\in\Gamma(n)$ there exists a $\delta>0$ and
non-negative functions $g_{\gamma}(x)$, $g_{1\gamma}(x)$, $g_{2\gamma}(x)$,
such that $g_{\gamma}^{2}(x)$ $d_{e_{q}}\mu$-summable for all $q\in\gamma$,
$g_{1\gamma}^{2}(x)$ and $g_{2\gamma}^{2}(x)$ are $\mu$-summable, and for all
$q\in\gamma$, for $|t|<\delta$,
$|\omega_{\gamma}(x+t_{q})|<\min\\{g_{\gamma}(x),g_{1\gamma}(x)\\}$ and
$|d_{e_{q}}\omega_{\gamma}(x+te_{q})|<g_{2\gamma}(x)$. On $B_{q}^{n}$ we
define the norm
$\|\omega\|_{B_{q}^{n}}=\|\omega\|_{n,q}+\|d\omega\|_{n+1,q}.$
It is clear that $d:\,B^{n}_{q}\to\Omega_{q}^{n+1}$ is a linear continuous
operator. Conditions (i) and (ii) are necessary to satisfy the assumptions of
the integration by parts formula proved in [1] which we apply to compute the
operator $d^{*}$.
###### Theorem 1.
For every pair of elements $f\in A_{p}^{n+1}$ and $\omega\in B_{q}^{n}$,
$1/p+1/q=1$, the element $-\beta^{\mu}\\!\,\text{\LARGE$\lrcorner$}\,f-\delta
f$ belongs to $\Omega_{p}^{n}$, and
$\langle
d\omega,f\rangle_{n+1}=\langle\omega,-\beta^{\mu}\\!\,\text{\LARGE$\lrcorner$}\,f-\delta
f\rangle_{n},$
i.e. the adjoint operator $d^{*}:\,A_{p}^{n+1}\to\Omega_{p}^{n}$ is
represented by the formula:
$d^{*}=-(\beta^{\mu}\\!\,\text{\LARGE$\lrcorner$}\,+\delta).$
###### Proof.
Let
$\omega=\\!\\!\sum\limits_{\gamma\in\Gamma(n)}\\!\\!\omega_{\gamma}e_{\gamma}$,
$f=\\!\\!\\!\sum\limits_{\gamma_{1}\in\Gamma(n+1)}\\!\\!\\!f_{\gamma_{1}}e_{\gamma_{1}}$.
We have
$\|\beta^{\mu}(x)\,\text{\LARGE$\lrcorner$}\,f(x)\|_{n}\\\
=\Bigl{\|}\sum\limits_{p\in\gamma_{1}\in\Gamma(n+1)}\beta_{e_{p}}^{\mu}(x)\,f_{\gamma_{1}}(x)\,e_{p}\,\text{\LARGE$\lrcorner$}\,e_{\gamma_{1}}\Bigr{\|}_{n}=\Bigl{\|}\sum\limits_{\gamma\in\Gamma(n)}\Bigl{(}\sum\limits_{p\notin\gamma}\beta_{e_{p}}^{\mu}(x)\,f_{\gamma\cup
p}(x)\Bigr{)}e_{\gamma}\Bigr{\|}_{n}\\\
=\sqrt{\sum\limits_{\gamma\in\Gamma(n)}\Bigl{(}\sum\limits_{p\notin\gamma}\beta_{e_{p}}^{\mu}(x)\,f_{\gamma\cup
p}(x)\Bigr{)}^{2}}\leqslant\sqrt{\sum\limits_{\gamma\in\Gamma(n)}\Bigl{(}\sum\limits_{p\notin\gamma}(\beta_{e_{p}}^{\mu}(x))^{2}\sum\limits_{p\notin\gamma}f^{2}_{\gamma\cup
p}(x)\Bigr{)}}\\\
\leqslant\sqrt{\sum\limits_{p=1}^{\infty}(\beta_{e_{p}}^{\mu}(x))^{2}\sum\limits_{\gamma\in\Gamma(n)}\sum\limits_{p\notin\gamma}f^{2}_{\gamma\cup
p}(x)}\leqslant\sqrt{n+1}\,\|\beta^{\mu}\|_{\scriptscriptstyle
H}\,\sqrt{\sum\limits_{\gamma_{1}\in\Gamma(n+1)}f^{2}_{\gamma_{1}}(x)}\\\
=\sqrt{n+1}\,\|\beta^{\mu}(x)\|_{\scriptscriptstyle H}\,\|f(x)\|_{n+1}.$ (4)
From this and (3) it follows that
$\|\beta^{\mu}\\!\,\text{\LARGE$\lrcorner$}\,f\|_{n,p}<\infty$. By (3),
$\|\delta f\|_{n,p}<\infty$, and hence,
$\|\beta^{\mu}\\!\,\text{\LARGE$\lrcorner$}\,f+\delta f\|_{n,p}<\infty$, i.e.
$-\beta^{\mu}\\!\,\text{\LARGE$\lrcorner$}\,f-\delta f\in\Omega_{p}^{n}$.
Next,
$d\omega=\sum\limits_{\gamma\in\Gamma(n),\,p\in\gamma}d_{e_{p}}\omega_{\gamma}\,e_{p}\wedge
e_{\gamma}=\sum\limits_{\gamma\in\Gamma(n),p\notin\gamma}(-1)^{k_{p}-1}\,d_{e_{p}}\omega_{\gamma}\,e_{\gamma\cup
p}\\\
=\sum_{\gamma_{1}\in\Gamma(n+1)}\Bigl{(}\sum_{\begin{subarray}{c}\gamma\subset\gamma_{1},\\\
p\in\gamma_{1}\diagdown\gamma\end{subarray}}(-1)^{k_{p}-1}\,d_{e_{p}}\omega_{\gamma}\Bigr{)}\,e_{\gamma_{1}},$
where $k_{p}$ is the number of $p$ in the sequence $\gamma_{1}$. Applying the
integration by parts formula [1] (one can easily verify the conditions under
which this formula holds), we obtain:
$\displaystyle\langle d\omega,f\rangle_{n}$
$\displaystyle=\sum\limits_{\gamma_{1}\in\Gamma(n+1)}\int\limits_{X}f_{\gamma_{1}}(x)\Bigl{(}\sum\limits_{\begin{subarray}{c}\gamma\subset\gamma_{1},\\\
p\in\gamma_{1}\diagdown\gamma\end{subarray}}(-1)^{k_{p}-1}\,d_{e_{p}}\omega_{\gamma}(x)\Bigr{)}\,\mu(dx)$
$\displaystyle=\sum_{\begin{subarray}{c}\gamma_{1}\in\Gamma(n+1),\\\
\gamma\subset\gamma_{1},\,p\in\gamma_{1}\diagdown\gamma\end{subarray}}(-1)^{k_{p}-1}\,\int\limits_{X}f_{\gamma_{1}}(x)\,d_{e_{p}}\omega_{\gamma}(x)\,\mu(dx)$
$\displaystyle=\\!-\\!\\!\sum_{\begin{subarray}{c}\gamma_{1}\in\Gamma(n+1),\\\
\gamma\subset\gamma_{1},\,p\in\gamma_{1}\diagdown\gamma\end{subarray}}(-1)^{k_{p}-1}\\!\int\limits_{X}\omega_{\gamma}(x)\,(d_{e_{p}}f_{\gamma_{1}}(x)+f_{\gamma_{1}}(x)\,\beta_{e_{p}}^{\mu}(x))\,\mu(dx)=$
$\displaystyle=-\int\limits_{X}\sum\limits_{\gamma\in\Gamma(n),\,p\notin\gamma}(-1)^{k_{p}-1}\omega_{\gamma}(x)\,(d_{e_{p}}f_{\gamma\cup
p}(x)+f_{\gamma\cup p}(x)\,\beta_{e_{p}}^{\mu}(x))\,\mu(dx).$ (5)
We changed the order of summation and integration when passing to the latter
expression in (5), and applied Lebesgue’s theorem. Clearly, the sequence of
partial sums under the last integral sign in (5) is majorized by an integrable
function. This follows from the definition of the spaces $A^{n+1}_{p}$ and
$B^{n}_{q}$, from Cauchy-Bunyakovsky-Schwarz’s inequality, and from the
inequality
$\sqrt{\sum\limits_{\gamma\in\Gamma(n)}\Bigl{(}\sum\limits_{p\notin\gamma}\beta_{e_{p}}^{\mu}(x)\,f_{\gamma\cup
p}(x)\Bigr{)}^{2}}\leqslant\sqrt{n+1}\,\|\beta^{\mu}(x)\|_{\scriptscriptstyle
H}\,\|f(x)\|_{n+1},$
which was proved, in fact, together with the estimate (4). By the same
argument, all the series in (5) converge absolutely, and hence the order of
summation of these series can be chosen arbitrary. We rewrite the expressions
for $\delta f$ and $\beta^{\mu}\\!\,\text{\LARGE$\lrcorner$}\,f$:
$\displaystyle\delta
f(x)=\sum\limits_{p\in\gamma_{1}\in\Gamma(n+1)}d_{e_{p}}f_{\gamma_{1}}(x)\,e_{p}\,\text{\LARGE$\lrcorner$}\,e_{\gamma_{1}}=\sum\limits_{\gamma\in\Gamma(n)}\Bigl{(}\sum\limits_{p\notin\gamma}(-1)^{k_{p}-1}\,d_{e_{p}}f_{\gamma\cup
p}(x)\Bigr{)}e_{\gamma},$
$\displaystyle\begin{split}\beta^{\mu}(x)\,\text{\LARGE$\lrcorner$}\,f(x)&=\sum\limits_{p\in\gamma_{1}\in\Gamma(n+1)}f_{\gamma}(x)\,\beta_{e_{p}}^{\mu}(x)\,e_{p}\,\text{\LARGE$\lrcorner$}\,e_{\gamma_{1}}\\\
&=\sum\limits_{\gamma\in\Gamma(n),\,p\notin\gamma}(-1)^{k_{p}-1}f_{\gamma\cup
p}(x)\beta_{e_{p}}^{\mu}(x)\,e_{\gamma}.\end{split}$
This and (5) imply the statement of the theorem. ∎
## 3\. The Stokes formula
### 3.1 Assumtions and notation
As before, let $X$ be a locally convex space, $H$ be its vector subspace which
is a Hilbert space relative to the scalar product $(\,\cdot\,,\,\cdot\,)$, and
let $\\{e_{n}\\}_{n=1}^{\infty}$ be an orthonormal basis of $H$. We assume
that $H$ is dense in $X$ and the identical embedding of $H$ into $X$ is
continuous. Let $\Xi_{n}$, $n\in N$, denote the vector space of bounded
differential forms of degree $n$ differentiable along $H$ and possessing
bounded differentials. Let $S_{n}$ denote the space of differential forms of
codegree $n$ which are Radon measures differentiable along $H$. Also, we
assume that the differential forms from $S_{n}$ and their differentials are
measures of bounded variation. Let $\bar{S}_{n}$ and $\bar{\Xi}_{n}$,
$n\in{\mathbb{N}}$, denote pseudo-topological vector spaces of linear
continuous functionals on $\Xi_{n}$ and $S_{n}$, respectively. We assume that
$\bar{S}_{n}$ and $\bar{\Xi}_{n}$ contain $S_{n}$ and $\Xi_{n}$ as dense
subsets.
In addition to that, we assume that the mapping $d:\,\Xi_{0}\to\Xi_{1}$ can be
extended to a continuous mapping $\bar{\Xi}_{0}\to\bar{\Xi}_{1}$, and the
mapping $d:\,S_{1}\to S_{0}$, to a continuous mapping
$\bar{S}_{1}\to\bar{S}_{0}$. Further we assume that for every measure $\nu\in
S_{0}$, the mapping $\Xi_{1}\to S_{1},\;f\mapsto f\nu$ can be extended to a
continuous mapping $\bar{\Xi}_{1}\to\bar{S}_{1}$, and the mapping
$\Xi_{1}\times S_{1}\to S_{0},\,(f,\omega)\mapsto f\wedge\omega$, to a
continuous mapping $\bar{\Xi}_{1}\times\bar{S}_{1}\to\bar{S}_{0}$. We denote
the extended mappings by the same symbols. When the functional
$f\in\bar{S}_{1}$ acts on the element $g\in\Xi_{1}$, we write $\langle
f,g\rangle$. Further let us assume that every sequence of elements from
$\Xi_{n}$, $n=1,2$, converging pointwise to an element from $\bar{\Xi}_{n}$,
converges to this element also with respect to the $\bar{\Xi}_{n}$-topology.
Let $V$ be a domain in $X$ such that its boundary $\partial V$ can be covered
with a finite number of surfaces $\mathcal{U}_{i}$ of codimension $1$.
Everywhere below, a surface of codimension $1$ is the object defined in [6],
p. 552. We assume that the indicator $\operatorname{\mathbb{I}}_{V}$ of the
domain $V$ is an element of the space $\bar{\Xi}_{0}$. Let the sets
$\,\mathcal{U}_{i}$ covering $\partial V$ and their intersections possess the
property ($*$) formulated in [6], p. 559. In what follows, we will use the
notations introduced in [6]. Here we briefly repeat their meaning:
$n^{{}_{\partial V}}:\partial V\to H$ is the normal vector (with respect to
the $H$-topology) to $\partial V$; let $B\subset\partial V$ be a Borel subset,
then $B^{\varepsilon}=\\{y\in X:y=x+tn^{{}_{\partial V}}(x),\,x\in
B,|t|<\varepsilon\\}$ is the $\varepsilon$-layer of $B$; $\varepsilon_{b}$ is
the maximal number for which $\varepsilon_{b}$-layers are well defined;
$\nu^{\varepsilon}$ is a measure on $\partial V$,
$\nu^{\varepsilon}(B)=\frac{\nu(B^{\varepsilon})}{2\varepsilon}$;
$P_{\scriptscriptstyle\partial V}:(\partial V)^{\varepsilon_{b}}\to\partial
V:x+tn^{{}_{\partial V}}(x)\mapsto x$ means the projector of
$\varepsilon$-layers to the surface; $\nu^{{}_{\partial V}}$ is the surface
measure generated by the measure $\nu$ ([12], [6]). Rigorous definitions of
these objects as well as lemmas proving their existence are given in [6]. Note
that by Theorem 2 of [6], $\lim_{\varepsilon\to 0}\nu^{\varepsilon}(\partial
V)=\nu^{{}_{\partial V}}(\partial V)$.
### 3.2 A connection between a measure and the generated surface measure in
terms of differential forms
###### Theorem 2.
Let $\nu\in S_{0}$. Then $d\operatorname{\mathbb{I}}_{V}\cdot\,\nu$ is an
$H$-valued Radon measure on $X$ concentrated on $\partial V$. Moreover,
$n^{{}_{\partial V}}\cdot\nu^{{}_{\partial V}}\in\bar{S}_{1}$, and the
measures $\nu$ and $\nu^{{}_{\partial V}}$ are related through the identity:
$d\operatorname{\mathbb{I}}_{V}\cdot\,\nu=-n^{{}_{\partial
V}}\cdot\nu^{{}_{\partial V}}.$ (6)
Note that by assumption, $d\operatorname{\mathbb{I}}_{V}\in\bar{\Xi}_{1}$ and
$d\operatorname{\mathbb{I}}_{V}\cdot\,\nu\in\bar{S}_{1}$.
###### Proof.
Let $h^{\varepsilon}:(-\varepsilon_{b},\varepsilon_{b})\to[0,1]$,
$\varepsilon<\varepsilon_{b}$,
$h^{\varepsilon}(\tau)=\begin{cases}-\frac{\tau}{2\varepsilon}+\frac{1}{2}~{},&\text{if}\;\tau\in(-(\varepsilon-\varepsilon^{2}),\varepsilon-\varepsilon^{2}),\\\
1,&\text{if}\;\tau\in(-\varepsilon_{b},-\varepsilon),\\\
0,&\text{if}\;\tau\in(\varepsilon,\varepsilon_{b}),\end{cases}$
be ${\rm C}^{\infty}$-smooth functions such that on the intervals
$(-\varepsilon,-(\varepsilon-\varepsilon^{2}))$ and
$(\varepsilon-\varepsilon^{2},\varepsilon)$, the absolute values of their
derivatives change monotonically from $0$ to $\frac{1}{2\varepsilon}$, and
from $\frac{1}{2\varepsilon}$ to $0$, respectively. For
$\varepsilon<\varepsilon_{b}$, we define the functions
$f^{\varepsilon}:X\to{\mathbb{R}}$,
$\displaystyle
f^{\varepsilon}(x)=\begin{cases}h^{\varepsilon}(\tau),&\text{if}\;x=y+\tau
n^{{}_{\partial V}}(y),\,y\in\partial
V,\,\tau\in(-\varepsilon_{b},\varepsilon_{b}),\\\ 1,&\text{if}\;x\in
V\diagdown(\partial V)^{\varepsilon_{b}},\\\ 0,&\text{if}\;x\notin
V\cup(\partial V)^{\varepsilon_{b}}.\end{cases}$ (7)
Let us calculate
$d_{e_{p}}f^{\varepsilon}(x+\tau n^{{}_{\partial
V}}(x))=\left.\frac{d}{dt}f^{\varepsilon}(x+\tau n^{{}_{\partial
V}}(x)+te_{p})\right|_{t=0}$
for $\tau\in(-(\varepsilon-\varepsilon^{2}),\varepsilon-\varepsilon^{2})$ and
$x\in\partial V$. If $t$ is sufficiently small, then there exist $x_{t}$ and
$\tau_{t}$, such that
$x+\tau n^{{}_{\partial V}}(x)+te_{p}=x_{t}+\tau_{t}n^{{}_{\partial
V}}(x_{t}).$ (8)
Let $n_{p}^{{}_{\partial V}}$ be coordinates of the vector $n^{{}_{\partial
V}}$ in the basis $\\{e_{p}\\}_{p=1}^{\infty}$. Subtracting $x_{t}$ from the
both sides of (8), and multiplying by $n^{{}_{\partial V}}(x)$, we obtain:
$(x-x_{t},n^{{}_{\partial V}}(x))+\tau+tn_{p}^{{}_{\partial
V}}(x)=\tau_{t}(n^{{}_{\partial V}}(x),n^{{}_{\partial V}}(x_{t})).$
Hence,
$\tau_{t}=\frac{tn^{{}_{\partial V}}_{p}(x)}{(n^{{}_{\partial
V}}(x),n^{{}_{\partial V}}(x_{t}))}+\frac{\tau+(x-x_{t},n^{{}_{\partial
V}}(x))}{(n^{{}_{\partial V}}(x),n^{{}_{\partial V}}(x_{t}))}.$ (9)
Note that $x_{t}=P_{\scriptscriptstyle\partial V}(x+\tau n^{{}_{\partial
V}}(x)+te_{p})$. We can prove that the derivative
$\left.\frac{d}{dt}x_{t}\right|_{t=0}$ exists with respect to the $H$-topology
similarly to how it was done in the proof of Lemma 6 of [6]. From the results
of [6] (Lemmas 1 and 2), it follows that the derivative
$\left.\frac{d}{dt}n^{{}_{\partial V}}(x_{t})\right|_{t=0}$ exists with
respect to the $H$-topology as well. Taking into account this, we show that
$\left.\frac{d}{dt}(n^{{}_{\partial V}}(x),n^{{}_{\partial
V}}(x_{t}))\right|_{t=0}=0$ and $\left.\frac{d}{dt}(x-x_{t},n^{{}_{\partial
V}}(x))\right|_{t=0}=0$. The latter identity is obvious since
$\left.\frac{d}{dt}x_{t}\right|_{t=0}\in H_{x}$, and $n^{{}_{\partial V}}(x)$
is orthogonal to $H_{x}$, where $H_{x}$ is the intersection of the tangent
space at $x\in\partial V$ with $H$ (see [6]). Further we have:
$0=\left.\frac{d}{dt}\|n^{{}_{\partial
V}}(x_{t})\|^{2}\right|_{t=0}=2\Bigl{(}n^{{}_{\partial
V}}(x),\left.\frac{d}{dt}n^{{}_{\partial V}}(x_{t})\right|_{t=0}\Bigr{)}\\\
=2\left.\frac{d}{dt}(n^{{}_{\partial V}}(x),n^{{}_{\partial
V}}(x_{t}))\right|_{t=0}.$
From this and from (9), it follows that
$\left.\frac{d}{dt}\tau_{t}\right|_{t=0}=n_{p}^{{}_{\partial V}}(x)$. Taking
into account that $f^{\varepsilon}(x+\tau n^{{}_{\partial
V}}(x)+te_{p})=-\frac{\tau_{t}}{2\varepsilon}+\frac{1}{2}$, we obtain that
$d_{e_{p}}f^{\varepsilon}(x+\tau n^{{}_{\partial
V}}(x))=-\frac{n_{p}^{{}_{\partial V}}(x)}{2\varepsilon}$, and hence,
$df^{\varepsilon}(x+\tau n^{{}_{\partial V}}(x))=-\frac{n^{{}_{\partial
V}}(x)}{2\varepsilon}.$ (10)
For $\tau$ which belongs to one of the intervals
$(-\varepsilon,-(\varepsilon-\varepsilon^{2}))$ or
$(\varepsilon-\varepsilon^{2},\varepsilon)$, the differential
$df^{\varepsilon}(x+\tau n^{{}_{\partial V}}(x))$ can be calculated in the
same way. Indeed, $f^{\varepsilon}(x+\tau n^{{}_{\partial
V}}(x)+te_{p})=h^{\varepsilon}(\tau_{t})$, and
$df^{\varepsilon}(x+\tau n^{{}_{\partial
V}}(x))=(h^{\varepsilon})^{\prime}(\tau)n^{{}_{\partial V}}(x).$ (11)
This implies that
$\displaystyle\|df^{\varepsilon}(x)\|<\frac{1}{2\varepsilon}\quad\text{for
all}\;x\in X.$
Note that as $\varepsilon\to 0$,
$f_{\varepsilon}\to\operatorname{\mathbb{I}}_{V}$ pointwise, and hence with
respect to the $\bar{\Xi}_{0}$-topology. By assumption, $df_{\varepsilon}\to
d\operatorname{\mathbb{I}}_{V}$ in the $\bar{\Xi}_{1}$-topology. Let
$g\in\Xi_{1}$. We have:
$\langle d\operatorname{\mathbb{I}}_{V}\cdot\nu,g\rangle=\lim_{\varepsilon\to
0}\langle df_{\varepsilon}\cdot\,\nu,g\rangle=\lim_{\varepsilon\to
0}\int_{X}(df_{\varepsilon}(x),g(x))\nu(dx)\\\ =\lim_{\varepsilon\to
0}\int_{(\partial V)^{\varepsilon}}(df_{\varepsilon}(x),g(x))\nu(dx).$
Let $x\in\partial V$. The function $[0,\varepsilon_{b})\to{\mathbb{R}}$,
$t\mapsto(n^{{}_{\partial V}}(x),g(x+tn^{{}_{\partial V}}(x)))$ is
differentiable. By assumption, $g$ has a bounded derivative, say by a constant
$M$, along $H$. For all $x\in\partial V$, $t\in(-\varepsilon,\varepsilon)$, we
obtain:
$|(n^{{}_{\partial V}}(x),g(x+tn^{{}_{\partial V}}(x)))-(n^{{}_{\partial
V}}(x),g(x))|\\\ \leqslant\Bigl{|}\Bigl{(}n^{{}_{\partial
V}}(x),\frac{d}{dt}g(x+tn^{{}_{\partial
V}}(x))|_{t=t_{0}}\Bigr{)}\Bigr{|}\cdot
t\leqslant\|g^{\prime}(x+t_{0}n^{{}_{\partial V}}(x))n^{{}_{\partial
V}}(x)\|\cdot t\\\ \leqslant\|g^{\prime}(x+t_{0}n^{{}_{\partial
V}}(x))\|_{2}\cdot t<M\varepsilon,$
where $t_{0}<t$. Define a function $\tilde{g}:\,(\partial
V)^{\varepsilon_{b}}\to H$ in the following way: for $x\in\partial V$,
$t\in(-\varepsilon_{b},\varepsilon_{b})$, we set $\tilde{g}(x+tn^{{}_{\partial
V}}(x))=g(x)$. Then, taking into account the above sequence of inequalities,
formulas (10), (11), and the definition of $h^{\varepsilon}$, for all
$x\in(\partial V)^{\varepsilon}$ we obtain that
$|(df^{\varepsilon}(x),g(x))-(df^{\varepsilon}(x),\tilde{g}(x))|<\frac{M}{2}.$
This implies:
$\displaystyle\begin{split}\langle(d\operatorname{\mathbb{I}}_{V})\cdot\nu,g\rangle&=\lim_{\varepsilon\to
0}\int_{(\partial
V)^{\varepsilon}}(df_{\varepsilon}(x),\tilde{g}(x))\nu(dx)\\\
&=-\lim_{\varepsilon\to 0}\int_{(\partial
V)^{\varepsilon}}\frac{1}{2\varepsilon}\,(n^{{}_{\partial
V}}(P_{\scriptscriptstyle\partial V}x),g(P_{\scriptscriptstyle\partial
V}x))\nu(dx)\\\ &\quad+\lim_{\varepsilon\to 0}\int_{(\partial
V)^{\varepsilon}\diagdown(\partial
V)^{\varepsilon-\varepsilon^{2}}}\bigl{(}df_{\varepsilon}(x)-\frac{1}{2\varepsilon}\,n^{{}_{\partial
V}}(P_{\scriptscriptstyle\partial V}x),\tilde{g}(x)\bigr{)}\nu(dx)\\\
&=-\lim_{\varepsilon\to 0}\int_{\partial V}(n^{{}_{\partial
V}}(x),g(x))\nu^{\varepsilon}(dx).\end{split}$ (12)
Indeed,
$\displaystyle\int_{(\partial
V)^{\varepsilon}}\frac{1}{2\varepsilon}\,(n^{{}_{\partial
V}}(P_{\scriptscriptstyle\partial V}x),g(P_{\scriptscriptstyle\partial
V}x))\nu(dx)=\int_{\partial V}(n^{{}_{\partial
V}}(x),g(x))\nu^{\varepsilon}(dx),$
and as $\varepsilon\to 0$,
$\Bigl{|}\int_{(\partial V)^{\varepsilon}\diagdown(\partial
V)^{\varepsilon-\varepsilon^{2}}}(df_{\varepsilon}(x)-\frac{1}{2\varepsilon}n^{{}_{\partial
V}}(P_{\scriptscriptstyle\partial V}x),\tilde{g}(x))\nu(dx)\Bigr{|}\\\
\leqslant M\,\frac{\nu((\partial V)^{\varepsilon}\diagdown(\partial
V)^{\varepsilon-\varepsilon^{2}})}{\varepsilon}=2\,\Bigl{(}\frac{\nu((\partial
V)^{\varepsilon})}{2\varepsilon}-\frac{\nu((\partial
V)^{\varepsilon-\varepsilon^{2}})}{2(\varepsilon-\varepsilon^{2})}\frac{\varepsilon-\varepsilon^{2}}{\varepsilon}\Bigr{)}\to
0.$
Further, we fix an arbitrary $\sigma>0$, and let
$\sigma^{\prime}=\frac{\sigma}{2(M+\nu^{\scriptscriptstyle\partial V}(\partial
V))}$. Since $\nu^{{}_{\partial V}}$ is a Radon measure (see [12], [14]), then
there exists a compact $K_{\sigma}\subset\partial V$, such that
$\nu^{{}_{\partial V}}(\partial V\diagdown K_{\sigma})<\sigma^{\prime}$. For
each point $x_{0}\in K_{\sigma}$ we fix a neighborhood $U_{x_{0}}$, which is
contained in one of $\mathcal{U}_{i}$, possesses the property ($*$) formulated
in [6], and such that for the function $\varphi(x)=(n^{{}_{\partial
V}}(x),g(x))$, the inequality $|\varphi(x)-\varphi(x_{0})|<\sigma^{\prime}$
holds for all $x\in U_{x_{0}}$. We choose a finite number of neighborhoods
$U_{x}$, $x\in K_{\sigma}$, covering $K_{\sigma}$ (let them be neighborhoods
$U_{i}$ of points $x_{i}$), and denote their union by $O_{\sigma}$. It is
clear that $\nu^{{}_{\partial V}}(\partial V\diagdown
O_{\sigma})<\sigma^{\prime}$, and by the construction of $O_{\sigma}$, there
exists the limit $\lim_{\varepsilon\to
0}\nu^{\varepsilon}(O_{\sigma})=\nu^{{}_{\partial V}}(O_{\sigma})$. Hence the
limit $\lim_{\varepsilon\to 0}\nu^{\varepsilon}(\partial V\diagdown
O_{\sigma})=\nu^{{}_{\partial V}}(\partial V\diagdown O_{\sigma})$ exists too.
Further let $B_{i}=U_{i}\diagdown\textstyle\bigcup\limits_{j=1}^{i-1}U_{j}$,
and $\varphi_{\sigma}:\,O_{\sigma}\to{\mathbb{R}}$ be such that
$\varphi_{\sigma}=\sum\limits_{i}\varphi(x_{i})\operatorname{\mathbb{I}}_{B_{i}}$.
It is clear that on $O_{\sigma}$,
$|\varphi(x)-\varphi_{\sigma}(x)|<\sigma^{\prime}$. We have:
$\lim_{\varepsilon\to 0}\int_{\partial
V}\varphi(x)\nu^{\varepsilon}(dx)=\lim_{\varepsilon\to
0}\int_{O_{\sigma}}\varphi_{\sigma}(x)\nu^{\varepsilon}(dx)\\\
+\lim_{\varepsilon\to
0}\Bigl{(}\int_{O_{\sigma}}(\varphi(x)-\varphi_{\sigma}(x))\nu^{\varepsilon}(dx)+\int_{\partial
V\diagdown O_{\sigma}}\varphi(x)\nu^{\varepsilon}(dx)\Bigr{)}.$ (13)
By the definition of $\varphi_{\sigma}$,
$\int_{O_{\sigma}}\varphi_{\sigma}(x)\nu^{\varepsilon}(dx)=\sum\limits_{i}\varphi(x_{i})\nu^{\varepsilon}(B_{i})$,
where the sum contains a finite number of terms. We observe that for every set
$B_{i}$, $\lim_{\varepsilon\to
0}\nu^{\varepsilon}(B_{i})=\nu^{\scriptscriptstyle\partial V}(B_{i})$ by the
construction of $B_{i}$ and by Theorem 2 of [6]. Hence $\lim_{\varepsilon\to
0}\int_{O_{\sigma}}\varphi_{\sigma}(x)\nu^{\varepsilon}(dx)=\int_{O_{\sigma}}\varphi_{\sigma}(x)\nu^{\scriptscriptstyle\partial
V}(dx)$. The limit of the second term in (13) exists by the existence of the
two other limits. We continue (13):
$\lim_{\varepsilon\to 0}\int_{\partial
V}\varphi(x)\nu^{\varepsilon}(dx)=\int_{\partial
V}\varphi(x)\nu^{\scriptscriptstyle\partial V}(dx)\\\ -\int_{\partial
V\diagdown O_{\sigma}}\varphi(x)\nu^{\scriptscriptstyle\partial
V}(dx)-\int_{O_{\sigma}}(\varphi(x)-\varphi_{\sigma}(x))\nu^{\scriptscriptstyle\partial
V}(dx)\\\ +\lim_{\varepsilon\to
0}\Bigl{(}\int_{O_{\sigma}}(\varphi(x)-\varphi_{\sigma}(x))\nu^{\varepsilon}(dx)+\int_{\partial
V\diagdown O_{\sigma}}\varphi(x)\nu^{\varepsilon}(dx)\Bigr{)}.$
Let us estimate the last three terms. We have:
$\Bigl{|}\int_{O_{\sigma}}(\varphi(x)-\varphi_{\sigma}(x))\nu^{\varepsilon}(dx)+\int_{\partial
V\diagdown
O_{\sigma}}\varphi(x)\nu^{\varepsilon}(dx)\Bigr{|}\leqslant\sigma^{\prime}\nu^{\varepsilon}(O_{\sigma})+M\nu^{\varepsilon}(\partial
V\diagdown O_{\sigma}).$
Passing to the limit as $\varepsilon\to 0$ in the both sides of this
inequality, and taking into account that the limit on the left-hand side
exists, we obtain:
$\Bigl{|}\lim_{\varepsilon\to
0}\Bigl{(}\int_{O_{\sigma}}(\varphi(x)-\varphi_{\sigma}(x))\nu^{\varepsilon}(dx)+\int_{\partial
V\diagdown
O_{\sigma}}\varphi(x)\nu^{\varepsilon}(dx)\Bigr{)}\Bigr{|}\leqslant\sigma^{\prime}\nu^{\scriptscriptstyle\partial
V}(\partial V)+M\sigma^{\prime}.$
Analogously, we prove the two other estimates:
$\Bigl{|}\int_{\partial V\diagdown
O_{\sigma}}\varphi(x)\nu^{\scriptscriptstyle\partial
V}(dx)\Bigr{|}<M\sigma^{\prime},\qquad\Bigl{|}\int_{O_{\sigma}}(\varphi(x)-\varphi_{\sigma}(x))\nu^{\scriptscriptstyle\partial
V}(dx)\Bigr{|}<\sigma^{\prime}\nu^{\scriptscriptstyle\partial V}(\partial V).$
From this it follows that
$\Bigl{|}\lim_{\varepsilon\to 0}\int_{\partial
V}\varphi(x)\nu^{\varepsilon}(dx)-\int_{\partial
V}\varphi(x)\nu^{\scriptscriptstyle\partial V}(dx)\Bigr{|}<\sigma.$
Since $\sigma>0$ was chosen arbitrary, we conclude that
$\lim_{\varepsilon\to 0}\int_{\partial V}(n^{{}_{\partial
V}}(x),g(x))\nu^{\varepsilon}(dx)=\int_{\partial V}(n^{{}_{\partial
V}}(x),g(x))\nu^{\scriptscriptstyle\partial V}(dx).$
Together with (12) this implies that
$\langle(d\operatorname{\mathbb{I}}_{V})\cdot\nu,g\rangle=-\int_{\partial
V}(n^{{}_{\partial V}}(x),g(x))\nu^{\scriptscriptstyle\partial V}(dx)$
which is equivalent to (6). The theorem is proved. ∎
###### Corollary 1.
Let $\omega=\sum_{p=1}^{\infty}\omega_{p}e_{p}\in S_{1}$, and
$\omega^{{}_{\partial V}}=\sum_{p=1}^{\infty}\omega_{p}^{{}_{\partial
V}}e_{p}\in\bar{S}_{1}$, where $\omega_{p}^{{}_{\partial V}}$ are the surface
measures generated by the measures $\omega_{p}$. Then $n^{{}_{\partial
V}}\in\bar{\Xi}_{1}$, and the measures $\omega$ and $\omega^{{}_{\partial V}}$
are related through the identity:
$\displaystyle d\operatorname{\mathbb{I}}_{V}\wedge\,\omega=-\,n^{{}_{\partial
V}}\wedge\,\omega^{{}_{\partial V}}.$
###### Proof.
Let us prove that $n^{{}_{\partial V}}\in\bar{\Xi}_{1}$. Indeed,
$n^{{}_{\partial V}}$ originally defined on $\partial V$ can be extended to
$X$ by setting $n^{{}_{\partial V}}(x)=0$ for $x\notin\partial V$. Let us
consider the functions $f_{\varepsilon}$ defined by (7). By (10),
$-2\varepsilon\,df_{\varepsilon}$ converges to $n^{{}_{\partial V}}$
pointwise, and hence with respect to the $\bar{\Xi}_{1}$-topology by
assumption. Note that by assumption, the operation $\wedge$ can be extended
from $\Xi_{1}\times S_{1}$ to $\bar{\Xi}_{1}\times\bar{S}_{1}$ so that
$(f,\omega)\mapsto f\wedge\omega$ is a continuous mapping
$\bar{\Xi}_{1}\times\bar{S}_{1}\to\bar{S}_{0}$. Applying Theorem 2 to each
pair of real-valued measures $\omega_{p}$ and $\omega_{p}^{{}_{\partial V}}$
we obtain:
$\displaystyle
d\operatorname{\mathbb{I}}_{V}\wedge\,\omega=\sum_{p=1}^{\infty}d\operatorname{\mathbb{I}}_{V}\cdot\,\omega_{p}\wedge\,e_{p}=-\sum_{p=1}^{\infty}(n^{{}_{\partial
V}}\cdot\,\omega^{{}_{\partial V}}_{p})\wedge\,e_{p}=-\,n^{{}_{\partial
V}}\wedge\,\omega^{{}_{\partial V}}.$
∎
### 3.3 Derivation of the Stokes formula
###### Definition 7.
Let $\omega\in S_{1}$ and $\omega^{{}_{\partial V}}\in\bar{S}_{1}$. We define
the integral of $\omega$ over the surface $\partial V$ by the identity:
$\int_{\partial V}\omega=\int_{\partial V}(n^{{}_{\partial
V}}(x),\omega^{{}_{\partial V}}(dx)).$
###### Theorem 3 (The Stokes formula).
Let $\omega\in S_{1}$ and $\omega^{{}_{\partial V}}\in\bar{S}_{1}$, then
$\int_{\partial V}\omega=\int_{V}d\omega.$
###### Proof.
Corollary 1 and Definition 7 imply:
$\displaystyle\int_{\partial V}\omega=\int_{\partial V}(n^{{}_{\partial
V}}(x),\omega^{{}_{\partial V}}(dx))=-\langle
d\operatorname{\mathbb{I}}_{V}\wedge\,\omega,1\rangle.$
Let us consider again the functions $f_{\varepsilon}$ defined by (7). We
proved that $f_{\varepsilon}\to\operatorname{\mathbb{I}}_{V}$ pointwise and in
the $\bar{\Xi}_{0}$-topology, and that $df_{\varepsilon}\to
d\operatorname{\mathbb{I}}_{V}$ in the $\bar{\Xi}_{1}$-topology. By Lemma 1,
$0=d(f_{\varepsilon}\wedge\,\omega)(X)=(df_{\varepsilon}\wedge\,\omega)(X)+(f_{\varepsilon}\wedge\,d\omega)(X).$
Hence,
$\int_{V}d\omega=\lim_{\varepsilon\to
0}(f_{\varepsilon}\wedge\,d\omega)(X)=-\lim_{\varepsilon\to
0}(df_{\varepsilon}\wedge\,\omega)(X)=-\langle
d\operatorname{\mathbb{I}}_{V}\wedge\,\omega,1\rangle=\int_{\partial
V}\omega.$
The theorem is proved. ∎
## References
* [1] Averbukh, V. I.; Smolyanov, O. G.; Fomin, S. V., Generalized functions and differential equations in linear spaces. I: Differentiable measures. (Russian) Tr. Mosk. Mat. Obshch. 24, 133-174 (1971).
* [2] Cartan, H., Differential forms, (English) Paris: Hermann, Publishers in Arts and Science; London: Kershaw Publishing Company, Ltd., 166 (1970).
* [3] Diestel, J., Uhl, Jr., J. J., Vector measures. Mathematical Surveys. No. 15. Providence, R.I.: American Mathematical Society. XIII, 322 p. (1977).
* [4] Leandre, R.; Smolyanov, O. G.; Weizsäcker, H. v., Algebraic properties of infinite-dimensional differential forms of finite codegree, (Russian, English) Dokl. Math. 60, No.3, 412-415 (1999); translation from Dokl. Akad. Nauk, Ross. Akad. Nauk 369, No.6, 727-731 (1999).
* [5] Shamarov, N. N., Some formulas for the calculus of differential forms of finite codegree on a locally convex space. (Russian) Vestnik Moskov. Univ. Ser. I Mat. Mekh. 1996, , no. 2, 26–33, 103–104; translation in Moscow Univ. Math. Bull. 51, no. 2, 23–28 (1996).
* [6] Shamarova, E. Yu., Approximation of surface measures in a locally convex space, (Russian, English), Math. Notes 72, No. 4, 551-568 (2002); translation from Mat. Zametki 72, No. 4, 597-616 (2002).
* [7] Shigekawa, I., De Rham–Hodge–Kodaira’s decomposition on an abstract Wiener space, J. Math. Kyoto Univ. ,vol. 26, 191–202 (1986).
* [8] Smolyanov, O. G., De Rham currents and Stokes’ formula in a Hilbert space, (Russian, English) Sov. Math., Dokl. 33, 140-144 (1986); translation from Dokl. Akad. Nauk SSSR 286, 554-558 (1986).
* [9] Smolyanov, O. G., Weizsäcker, H. v., Differentiable families of measures, Journal of Functional Analysis, vol.118, no. 2, p. 455-476 (1993).
* [10] Smolyanov, O. G., Weizsäcker, H. v., Differential forms on infinite-dimensional spaces and an axiomatic approach to the Stokes formula, (Russian, English) Dokl. Math. 60, No.1, 22-25 (1999); translation from Dol. Akad. Nauk, Ross. Akad. Nauk 367, No.2, 151-154 (1999).
* [11] Sternberg, S., Lectures on differential geometry. Prentice-Hall, Englewood Cliffs, N. J. (1964).
* [12] Uglanov, A. V., Surface integrals in linear topological spaces. (Russian, English) Dokl. Math. 52, No.2, 227-230 (1995); translation from Dokl. Akad. Nauk, Ross. Akad. Nauk 344, No.4, 450-453 (1995).
* [13] Uglanov, A. V., Surface integrals in a Banach space, (English) Math. USSR, Sb. 38, 175-199 (1981).
* [14] Uglanov, A. V., Integration on infinite-dimensional surfaces and its applications, (English) Mathematics and its Applications (Dordrecht). 496. Dordrecht: Kluwer Academic Publishers, 262 p., (2000).
* [15] Yakhlakov, V. Yu., Surface measures on surfaces of finite codimension in a Banach space, (Russian, English) Math. Notes 47, No.4, 414-421 (1990); translation from Mat. Zametki 47, No.4, 147-156 (1990).
| arxiv-papers | 2008-07-14T16:45:56 | 2024-09-04T02:48:56.775173 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Evelina Shamarova",
"submitter": "Evelina Shamarova",
"url": "https://arxiv.org/abs/0807.2200"
} |
0807.2220 | # Quantum chromodynamics with advanced computing
Andreas S Kronfeld Fermi National Accelerator Laboratory, Batavia, IL 60510,
USA
(for the USQCD Collaboration) ask@fnal.gov
###### Abstract
We survey results in lattice quantum chromodynamics from groups in the USQCD
Collaboration. The main focus is on physics, but many aspects of the
discussion are aimed at an audience of computational physicists.
## 1 Introduction and background
Quantum chromodynamics (QCD) is the modern theory of the strong nuclear force.
The physical degrees of freedom are quarks and gluons. The latter are the
quanta of gauge fields, in some ways analogous to photons. The crucial
difference is that gluons couple directly to each other, whereas photons do
not. A consequence of the self-coupling is that the force between quarks does
not vanish at large distances (as the Coulomb force $e^{2}/(4\pi r^{2})$ does)
but becomes a constant $F_{\rm QCD}\approx
800~{}\textrm{MeV}\,\textrm{fm}^{-1}\approx 15,000~{}\textrm{N}$. It would,
thus, require a vast amount energy to separate quarks out to a macroscopic
distance. Indeed, long before the separation becomes large enough to measure,
the energy stored in the gluon field “sparks” into quark-antiquark pairs. This
phenomenon of QCD explains why freely propagating quarks are not observed in
nature, and it is called confinement.
QCD is a quantum field theory, which means that from the outset one must deal
with mathematical objects that are unfamiliar even to many physicists. The
most widely used theoretical tool for quantum field theories is relativistic
perturbation theory, as developed for quantum electrodynamics (QED) by
Feynman, Schwinger, Tomonaga and others sixty years ago Schwinger:QED .
Perturbation theory provides the key connection between the mathematical
theory with experiments in QED Kinoshita:1996vz and the Glashow-Weinberg-
Salam theory merging QED with the weak nuclear force Alcaraz:2007ri .
A textbook example of perturbative quantum field theory is to calculate how
virtual pairs induce a distance dependence on the coupling in quantum gauge
theories. In QED, one considers the fine structure constant,
$\alpha=\frac{e^{2}}{4\pi\hbar c},$ (1)
where $-e$ is the charge of the electron. In QCD, one has the strong coupling
$\alpha_{s}$, related to the gauge coupling $g$ by analogy with equation (1).
When virtual pairs are taken into account, $\alpha$ and $\alpha_{s}$ are not
constant but depend on distance; they are said to “run.” The running, depicted
in figure 1, is completely different in the two theories.
(a) (b)
Figure 1: Running of (a) the QED and (b) the QCD coupling with distance. For
QCD the red curve shows the perturbative result, and the (higher) blue show
the effect of confining forces.
In QED $\alpha$ becomes constant at distances $r>\hbar/m_{e}c$ (the electron’s
Compton wavelength), but at short distances it grows. In QCD $\alpha_{s}$
grows at long distance; perturbative running no longer makes sense once $r$ is
in the confining regime. At short distances—or, equivalently, high
energies—the QCD coupling is small enough that perturbation theory can again
be used. For example, the cross section for an electron-positron pair to
annihilate and produce a quark-antiquark pair can be reliably computed as a
power series in $\alpha_{s}$. The quark and antiquark each fragment into a jet
of hadrons—mostly pions, but some protons and neutrons, too—and general
properties such as the energy flow and angular distributions of these jets can
be traced back to the quark and antiquark. In this way, the experiments can
measure quark properties, and these are found to agree with theoretical
calculations Dixon:2007hh .
QCD and the electroweak theory form the foundation of the Standard Model of
elementary particles. They are so well established that many particle
physicists consider the chromodynamic and electroweak gauge symmetries, and
the associated quantum-number assignments of the quarks and leptons, to be
laws of nature. We cannot conceive of a more comprehensive theory that does
not encompass them. But there is more to the Standard Model. We know that
something must spontaneously break the electroweak symmetry and that something
(perhaps the same thing) must generate masses for the quarks and leptons. The
Standard Model contains interactions that _model_ these phenomena. The model
interactions are consistent with observations yet also incomplete. It is hoped
that insights obtained with the Large Hadron Collider (LHC), to commence
operations at CERN later this year, will enable particle physicists to solve
many of the puzzles raised by the Standard Model.
The purpose of this paper is to cover QCD, particularly in the strong-coupling
regime where perturbation theory is insufficient. It is worth bearing in mind,
however, that there is a close connection between many of the results
discussed below and some of the big questions in particle physics. A simple
example is the masses of the quarks. To determine the quark masses (which
because of confinement cannot be measured), we have to calculate the relation
between quark masses and measurable quantities, such as hadron masses. Quark
masses are interesting for many reasons, the most unsettling of which is as
follows.
Quark flavors in the Standard Model:
$\begin{array}[]{ccc}\left(\begin{array}[]{c}u\\\
d\end{array}\right)&\left(\begin{array}[]{c}c\\\
s\end{array}\right)&\left(\begin{array}[]{c}t\\\
b\end{array}\right)\\\\[10.00002pt] m_{u}<m_{d}&m_{c}>m_{s}&m_{t}>m_{b}\\\
!!!&&\end{array}$ What explains the pattern of masses? Figure 2: The six
_flavors_ of quarks come the three _generations_ with a doublet in each
generation. The doublet structure plays a role in the electroweak
interactions. For example, the upper (lower) entries have electric charge
$+2/3$ ($-1/3)$.
The mass of the top quark is larger than that of the bottom quark; see figure
2. Similarly, the mass of the charmed quark is larger than that of the strange
quark. If this simple pattern held with the up and down quarks—the ones that
make up proton and neutrons—then the up quark would be more massive than the
down. But then protons would decay to neutrons, positrons, and neutrinos
($p\to ne^{+}\nu$). The positrons would find electrons and annihilate to
photons.
This universe would consist of neutron stars surrounded by a swarm photons and
neutrinos, and nothing else. Our universe is not at all like this, because
neutrons are more massive than protons. The allowed decay reaction is $n\to
pe^{-}\bar{\nu}$, leading to an abundance of protons and electrons, and making
possible chemistry and biochemistry. With QCD we trace the neutron-proton mass
difference back to the down-up quark masses and then to the deeper origin of
quark masses. The pattern of quark masses is necessarily not simple, and the
search for simple explanation of a messy pattern occupies many particle
theorists. Below we give further examples where QCD is essential for working
out the mysteries of particle physics.
QCD is also the cornerstone of modern nuclear physics. The simplest nucleus is
nothing but the proton, a bound state of two up quarks and one down quark. The
neutron is a bound state of two down quarks and one up quark. The confining
force is very, very strong. By comparison, the force traditionally called the
strong nuclear force is a residue of the fundamental chromodynamic force. This
is similar to van der Waals forces among molecules, which are electromagnetic
forces between neutral objects with structure leading to a distribution of
electric charge. Some basic problems are, therefore, to understand nucleon
structure directly from QCD, to study the excitation spectrum of nucleons (and
their cousins with strangeness and other flavors), and to see how few-nucleon
(or, more generally, few-hadron) systems interact.
Another active area of research in nuclear QCD is the properties of quark
matter at high temperature and density. This regime is important to cosmology,
because an epoch of the early universe consisted of hot quark matter, and to
astrophysics, because, for example, neutron stars are composed of dense quark
matter. In the laboratory, nuclear physicists create hot, dense quark matter
by colliding heavy ions, at present with Brookhaven’s RHIC, soon with (the
heavy-ion mode of) CERN’s LHC, and someday with GSI’s FAIR.
At strong coupling it is necessary to go beyond perturbation theory. In a
relativistic quantum field theory, such as QCD, this is not easy. An important
obstacle is that a quantum field associates one (or perhaps a few) degrees of
freedom with every point in space-time. Hence, the number of degrees of
freedom is uncountably infinite. Furthermore, the fluctuations at short
distances, if added up naively, contribute an ultraviolet divergence to
anything of interest. These fluctuations are treated by a set of ideas known
as _renormalization_ Lepage:1989hf . In short, one must find a tool that
(correctly) handles both strongly coupled fields and the renormalization of
short-distance fluctuations.
This tool is lattice field theory or, for gauge theories like QCD, lattice
gauge theory. The idea, which goes back at least as far as Heisenberg, is to
replace continuous space (or, in practice, space-time) with a discrete grid,
or lattice. Then the number of degrees of freedom is countable. If the lattice
is of finite extent, it occupies a finite physical volume, and, moreover, the
number of degrees of freedom is finite. In 1974, Kenneth Wilson showed how to
maintain gauge symmetry with a discrete lattice Wilson:1974sk . In this and
subsequent work, he and others elaborated how the lattice provides a
mathematically rigorous definition of quantum field theory via functional
integrals Glimm:1987ng .
The functional integral for quantum fields on a finite lattice is a familiar
object, a multi-dimensional integral, that can be evaluated on a computer.
Before downloading some publicly available code usqcd and trying to run it on
a laptop, let us pause to contemplate how many variables of integration are
reasonable (for physics). One wants to have the spacing between sites on the
lattice small enough to resolve hadron structure, and one wants the volume
large enough that the boundary conditions do not modify the structure. If one
supposes that the ratio of these two lengths is 32, then the dimensions of the
integrals of interest are
${\tt gluon~{}d.o.f.}=8\times 4\times 32^{3}\times 128>10^{8},$ (2)
because gluons come in 8 colors with 4 polarization states each. The factor
128 presumes a space-time lattice with a time extent long (in relativistic
units) compared to the spatial extent; this is common practice.
To cope with integrals of such large dimension, the only practical technique
is Monte Carlo integration with importance sampling. The weight guiding the
importance sampling is $e^{-S}$, where $S$ is the classical action for the
random sample of gluon variables. In functional integral formalisms for
quantum mechanics, the expression for the action is the defining equation of
the physical system. The development of algorithms to generate these samples
is a vibrant subject, covered in part by Bálint Joó’s poster here at the
SciDAC conference. This paper will not dwell on algorithms, but it is
important to mention one more complication.
Quarks are fermions and, as such, must satisfy the Pauli exclusion principle.
In the functional integral formalism, this is handled by introducing
anticommuting Grassmann variables for fermions. The integration is no longer
Riemann (or Lebesgue) integration but a formal procedure called Berezin
integration. To make a long story short, in lattice QCD we always can and do
carry out the Berezin integration by hand. The outcome is that the weight for
the importance sampling becomes $\det M\exp(-S_{\rm gluons})$, where $M$ is a
sparse matrix with space-time indices. The matrix $M$ is $N\times N$, where
$N={\tt quark~{}d.o.f.}=n_{f}\times 3\times 2\times 2\times 32^{3}\times
128>10^{8}.$ (3)
The quark field represents $n_{f}$ flavors with 3 colors and 2 spin states of
quark and antiquark for every flavor and color. The matrix $M$ is sparse
because it is a lattice version of a differential operator, the Dirac operator
(generalized to QCD). Incorporating $\det M$ into the weight is
computationally demanding, increasing by two or three orders of magnitude the
amount of floating-point operations needed to generate a new configuration of
gluons.
The lattice is a helpful device mathematically and computationally, but it is
not physical. To obtain a real result of the chromodynamics of continuum
spacetime, one must work out the integrals for a sequence of lattices, and
take the limit $a\to 0$ in a way that respects renormalization. Taking $a$
smaller and smaller increases the computing burden as onerously as
$a^{-(4+z)}$. The exponent $z$ depends on the algorithm and is typically
around 1 or 2. The 4 in the exponent is the key reason why numerical lattice
QCD requires the most advanced computing facilities available. No algorithm
can reduce it, because it is the dimension of space-time.
Most lattice-QCD calculations are carried out in two steps. The first is to
generate an ensemble of gluon fields, with a certain lattice spacing and quark
masses. This requires the computationally super-demanding $\det M$. To
generate an ensemble of useful size, one needs several hundred, perhaps even a
few thousand, samples of the gluon field. Computers of the highest available
capability are needed for this step. In recent years, lattice gauge theorists
have carried out this step with the special-purpose computer QCDOC, large
clusters of PCs, and, more recently, leadership-class machines constructed
under the auspices of the DOE’s Office of Advanced Scientific Computing
Research (ASCR). These ensembles of gluon fields are valuable: within the U.S.
they are generated by and for particle and nuclear physicists and shared under
various agreements. File-sharing is facilitated by formats and software
developed by the International Lattice Data Grid DeTar:2007au .
The second step is to mine these ensembles for physics. A simple example is to
compute the quantum mechanical amplitude for a proton to propagate from one
point to another, and study the behavior of the amplitude as the separation
varies: this is what one does to compute the proton mass, in fact. The key
ingredients are a few rows of the inverse of the matrix $M$, introduced above.
With the same ensemble one can study protons, pions, and more uncommon hadrons
such as the charmed strange pseudoscalar meson (or $D_{s}$ for short).
Different rows of $M^{-1}$ (i.e., different flavors, colors, spins, and space-
time separations) are combined in various ways to obtain the quantum numbers
of the hadrons in the problem at hand. Because each problem is different, and
there are so many of them, we must attack this problem with the highest
capacity computers. The job mix here is heterogeneous, with many users whose
job streams range from many large jobs to extremely many medium-sized jobs.
The challenge is to develop systems that scale well (for the large jobs),
while maintaining the flexibility needed to handle the heterogeneity. The
USQCD Collaboration has designed clusters of PCs that deliver this capacity in
a cost-effective way Holmgren:2004nk .
The remainder of this paper focuses on physics. It is organized according to
four principal themes of lattice gauge theory:
* •
Determination of fundamental parameters of the Standard Model of particle
physics
* •
Study of nucleon structure, the hadron spectrum, and hadron interactions
* •
Simulation of the thermodynamic properties of QCD at nonzero temperature and
density
* •
Lattice gauge theories beyond QCD, such as those appearing in models of
electroweak symmetry breaking
Roughly speaking, the first and last are particle physics, and the second and
third are nuclear physics. But the lines are blurry: for example, some of the
simplest calculations of proton structure (the second item) may play a role
more accurate calculations of $pp$ scattering cross sections at the LHC. The
way that scientific goals drive the computing needs of lattice gauge theory
are discussed further, along the lines of these themes, in public whitepapers
of the USQCD Collaboration usqcd .
## 2 Standard Model particle physics
In quantum electrodynamics, the fundamental parameters are the masses of
electrically charged particles and the fine structure constant $\alpha$.
Similarly, in quantum _chromo_ dynamics, the fundamental parameters are the
masses of colored particles (the quarks) and the strong coupling $\alpha_{s}$.
Just as $\alpha$ can be determined in many ways Kinoshita:1996vz ,
$\alpha_{s}$ can be determined at high energies from jet cross sections with
perturbative QCD, and at low energies from the hadron spectrum with lattice
QCD. Two recent results are (evaluated at distance $r=\hbar/m_{Z^{0}}c$)
$\alpha_{s}=\left\\{\begin{array}[]{llll}0.1172\pm 0.0022&\textrm{perturbative
QCD}&\textrm{jet shapes}&\textrm{Ref.~{}\cite[cite]{\@@bibref{Authors
Phrase1YearPhrase2}{Becher:2008cf}{\@@citephrase{(}}{\@@citephrase{)}}}}\\\
0.1170\pm 0.0012&\textrm{lattice
QCD}&\textrm{hadrons}&\textrm{Ref.~{}\cite[cite]{\@@bibref{Authors
Phrase1YearPhrase2}{Mason:2005zx}{\@@citephrase{(}}{\@@citephrase{)}}}}\end{array}\right.,$
(4)
where both error bars reflect experimental and theoretical uncertainties. The
first result is determined at high energies, around 100 GeV, in the regime
where quarks can be treated as nearly free. The second is determined at low
energies, of a few GeV and lower, where quarks are confined. Both are linked
soundly to the defining equations of QCD. The agreement demonstrates the
richness and broad validity of QCD and should make any enthusiastic scientist
say “Wow!”
A straightforward application of lattice QCD is to work out how hadron masses
depend on quark masses. Quark masses are interesting for several reasons. As
fundamental constants of nature, they are interesting in their own right. As
discussed in the introduction, the pattern of quark masses is puzzling and,
hence, a motivation to search for extensions of the Standard Model that would
contain a simple explanation.
Three flavors of quarks—top, bottom, and charm—have masses large enough to
influence some high-energy scattering and decay processes. Their masses can be
determined with perturbative QCD as well as lattice QCD. (The results agree
Yao:2006px .) The other three quarks—up, down, and strange—have masses so
small that the full bound-state problem must be considered. Lattice QCD,
therefore, provides the best information. Some recent results are the
following Mason:2005bj ; Bernard:2007ps :
$\displaystyle m_{u}=1.9$ $\displaystyle\pm$ $\displaystyle
0.2~{}\textrm{MeV}/c^{2},$ (5) $\displaystyle m_{d}=4.6$ $\displaystyle\pm$
$\displaystyle 0.3~{}\textrm{MeV}/c^{2},$ (6) $\displaystyle m_{s}=88$
$\displaystyle\pm$ $\displaystyle 5~{}\textrm{MeV}/c^{2}.$ (7)
One remarkable aspect of these results is that the strange quark’s mass is
somewhat smaller than had been thought on the basis of less reliable
techniques for attacking strongly-coupled QCD. This point is, perhaps, of
interest mostly to particle physicists, but a further one is of broader
interest. The up and down masses—the constituents of protons and neutrons and,
thus, everyday matter—are very small, within an order of magnitude of the
electron mass ($0.511~{}\textrm{MeV}/c^{2}$). Only about 1–2% of the proton
mass ($938.3~{}\textrm{MeV}/c^{2}$) or neutron mass
($939.6~{}\textrm{MeV}/c^{2}$) is accounted for by the quark masses.
Consequently, most of their masses—and the mass of everyday matter—arises from
confinement, namely, the binding energy of the gluons and the (relativistic)
kinetic energy of bound quarks. This is a stunning insight, to be revisited in
Section 5.
The caption of figure 2 states that the pairings of quarks in each doublet
stem from the electroweak interactions. That is not the full story. There is
no reason for the eigenstates of mass to be the eigenstates of the interaction
term with $W$ gauge bosons. The two are related by a unitary transformation,
$V_{\rm CKM}$, namely,
$\left(\begin{array}[]{c}d\\\ s\\\ b\end{array}\right)_{W}=V_{\rm
CKM}\left(\begin{array}[]{c}d\\\ s\\\ b\end{array}\right)_{\rm mass},\quad
V_{\rm CKM}=\left(\begin{array}[]{ccc}V_{ud}&V_{us}&V_{ub}\\\
V_{cd}&V_{cs}&V_{cb}\\\ V_{td}&V_{ts}&V_{tb}\end{array}\right).$ (8)
Here the initials CKM stand for Cabibbo-Kobayashi-Maskawa; Cabibbo introduced
equations (8) for two generations, and Kobayashi and Maskawa were the first to
study the ramifications for three generations.
The left-hand side of the first equation denotes the electroweak basis. The
labels on the elements of $V_{\rm CKM}$ are the weak isopartner and the mass
eigenstate, because the transition amplitude for $W\to\bar{b}u$, for example,
is proportional to $V_{ub}$. The study of physical processes related to
$V_{\rm CKM}$ is called flavor physics, because the central issues are to
understand why there are several flavors of quarks and what lends them their
separate identities.
As a unitary matrix, $V_{\rm CKM}$ contains complex entries. Some of the
phases can be absorbed into unobservable phases of the quark wave functions.
For only two generations, one physical parameter remains, and it is real. For
three generations, as in nature, three physical parameters are real and one
phase remains. The complex couplings imparted by the phase lead to physical
processes that proceed at different rates for particles and the corresponding
antiparticles. Such reactions are needed to explain the abundance of matter
and the dearth of antimatter in the universe. Among particle physicists, this
phenomenon is called $CP$ violation. It is clearly intriguing to know whether
the $CP$ violation of the CKM matrix is enough to account for the matter-
antimatter asymmetry of the universe. According to present measurements and
theoretical understanding, it is insufficient, so one would like to know what
other interactions violate $CP$.
In the Standard Model, there is a scalar boson called the Higgs boson. Section
5 will explain more about it. For now let us note that both the quark masses
and the CKM matrix have their origin in matrices of couplings between quark
fields and the Higgs field. Therefore, the puzzles of $CP$ violation and the
pattern of the quark masses are tied together by the Standard Model.
To make progress in flavor physics, lattice-QCD calculations are needed to
interpret the experimental measurements. In a schematic form the rate $\Gamma$
is given by
$\Gamma=\left(\begin{array}[]{c}\textrm{known}\\\
\textrm{factor}\end{array}\right)\left(\begin{array}[]{c}\textrm{CKM}\\\
\textrm{factor}\end{array}\right)\left(\begin{array}[]{c}\textrm{QCD}\\\
\textrm{factor}\end{array}\right),$ (9)
where the “known factor” consists of well-measured physical constants and
numerical factors like $4\pi$. The paradigm is to measure as many flavor-
changing processes as possible, and use the Standard-Model formulae to over-
determine the CKM matrix. If all the determinations are consistent, then one
can set limits on non-Standard sources of flavor and $CP$ violation. To do so,
the QCD factor must be computed, and often the only way to do so is with
lattice QCD.
There are dozens of relevant measurements, but because the CKM matrix is
unitary, the elements of the CKM matrix must satisfy constraints. A particular
vivid one is the unitarity triangle that stems from the orthogonality of
columns of a unitary matrix:
$V_{ud}^{*}V_{ub}+V_{cd}^{*}V_{cb}+V_{td}^{*}V_{tb}=0,$ (10)
which traces out a triangle in the complex plane. The triangle and the present
uncertainties from various constraints is shown in figure 3, dividing each
side by $V_{cd}^{*}V_{cb}$.
Figure 3: Constraints on the CKM unitarity triangle from experiments and QCD
calculations, highlighting those where lattice QCD plays a role. The upper
panel shows the current status, and the lower a forecast of how the
uncertainties can be reduced with lattice QCD VandeWater:2007zz . In the lower
panel, one can envision tension between the lime-green hyperbola and the blue
wedge, depending on how the central values evolve and measurements and
lattice-QCD computations improve. The current status of the angles $\alpha$
and $\gamma$ is omitted, because better measurements are needed; they are
expected from LHC$b$.
Except for the wedge labelled $\sin\beta$ the uncertainties can be reduced
with lattice calculations of the QCD factor.
Lattice calculations for a wide range of observables relevant to the CKM
matrix have appeared over the past few years, and further improvements are
under way. These include calculations of properties of strange kaons
Bernard:2007ps ; Aubin:2004fs ; Beane:2006kx ; Gamiz:2006sq ; Antonio:2007pb ;
Boyle:2007qe , charmed $D$ and $D_{s}$ mesons Aubin:2004ej ; Aubin:2005ar ;
Follana:2007uv , and beautiful $B$ and $B_{s}$ mesons Okamoto:2004xg ;
Dalgic:2006dt ; Gray:2005ad ; Bernard:2006zz ; Dalgic:2006gp ; Laiho:2007pn .
For recent reviews, including references to work in Asia, Australia, and
Europe, see DellaMorte:2007ny ; Juttner:2007sn .
The past several months witnessed an unexpected development in flavor physics,
in the field of leptonic decays of pseudoscalar mesons. (Pseudoscalar bosons
have spin 0 and negative parity; the most common examples are the pion and
kaon.) One can consider the measured decay rates to determine a QCD matrix
element called the decay constant, and denoted $f_{\pi}$ (for the pion),
$f_{K}$ (for the kaon), and so on. Table 1 shows the results of some SciDAC-
supported calculations of decay constants and the corresponding measurements.
Table 1: Decay constants $f_{P}$ of pseudoscalar mesons (here $P\in\\{\pi,K,D,D_{s}\\}$). Values in MeV. Experimental determinations assume a suitable element of the CKM matrix element consistent with all flavor data. Meson | MILC Bernard:2007ps | FNAL Aubin:2005ar | HPQCD Follana:2007uv | Experiment | (Ref.) | Deviation
---|---|---|---|---|---|---
$\pi$ | $128\pm 3$ | – | $132\pm 2$ | $130.7\pm 0.4$ | Yao:2006px | $0.4\sigma$
$K$ | $154\pm 3$ | – | $157\pm 2$ | $159.8\pm 1.5$ | Yao:2006px | $1.7\sigma$
$D^{+}$ | – | $201\pm 19$ | $207\pm 4$ | $206\pm 9$ | Eisenstein:2008sq | $0.1\sigma$
$D_{s}$ | – | $249\pm 17$ | $241\pm 3$ | $277\pm 9$ | Dobrescu:2008er | $3.8\sigma$
The calculations agree very well with the measurements, except in the case of
the charmed strange pseudoscalar meson $D_{s}$.
This set of circumstances is odd, because the computation is easiest for the
$D_{s}$. Algorithms for the propagators of light quarks slow down as the quark
mass $m_{q}$ is decreased. As a consequence, properties of hadrons containing
up or down quarks are reached via a (controlled) extrapolation in $m_{q}$.
This introduces a source of uncertainty that is absent for the $D_{s}$, which,
as the lightest bound state of a charmed quark and a strange antiquark, can be
computed by working directly at the physical quark masses.
There is no obvious explanation for this discrepancy within the Standard
Model. Unless there is a real blunder somewhere, the discrepancy points to a
non-Standard particle mediating the decays $D_{s}\to\mu\nu$ and
$D_{s}\to\tau\nu$, thereby changing the interpretation of the “measured”
$f_{D_{s}}$. It turns out to be rather easy to devise models of such
interactions Dobrescu:2008er . Given the agreement of CLEO’s new measurement
of $f_{D^{+}}$ Eisenstein:2008sq , new $W^{\prime}$ bosons and charged Higgs
bosons seem unlikely, making leptoquarks—particles with baryon and lepton
number—are the most plausible agents of decay.
## 3 Nucleon structure and nuclear physics
Let us turn now to nuclear physics, where the importance of lattice QCD is
difficult to overstate. Many nuclear laboratories, such as Jefferson
Laboratory in the U.S., have turned their attention to the structure of the
proton (the simplest nucleus!) Hagler:2007hu , the excitation spectrum of
baryons McNeile:2007fu , and processes that may produce glueballs. On the
theoretical side, the previous decade witnessed dramatic growth in the
application of effective field theories to understand nuclear interactions
Kaplan:2006sv . These techniques connect a wealth of nuclear and hadronic data
to the QCD Lagrangian. Lattice QCD is an irreplaceable tool to interpret the
new and upcoming experiments at JLab and to compute the so-called low-energy
constants of the effective theory Beane:2008dv .
A topic at the interface between nuclear and particle physics concerns the
distributions of partons (a generic term for quarks and gluons) inside the
proton. In high-energy collisions of $p\bar{p}$ (at the Tevatron) or $pp$ (at
the LHC), the transverse motion of partons can be neglected. To compute cross
sections (to produce, say, a Higgs boson), the key property needed is the
distribution of the longitudinal momentum. From a nuclear perspective, this is
also one of the most basic aspects of proton structure. Usually one studies
parton distributions $q$ as a function of the fraction $x$ of longitudinal
momentum carried by a parton, $0<x<1$. Fits to scattering data can determine
$q(x)$ only over a range, $x_{\rm min}<x<x_{\rm max}$, limited by kinematics
and statistics. Lattice QCD can compute moments of $q$: $\langle
x^{n}\rangle_{q}=\int_{0}^{1}dx\,x^{n}q(x)$. Clearly the two kinds of
information are complementary. The experimental range can be extended with _ad
hoc_ functions possessing the right asymptotic forms at the endpoints, and in
this way an Ansatz-guided experimental determination of $\langle
x^{n}\rangle_{q}$ is possible. One finds good agreement with lattice
calculations Hagler:2007xi . To explore nucleon structure further, the next
step is to probe the transverse structure and compute the so-called
generalized parton distributions Hagler:2007xi .
The axial charge of the nucleon is another topic of keen interest, because it
appears in the description of $\beta$ decay. It is also related to the
relative spin polarization of up and down quarks in the proton. Some recent
results with lattice QCD are
$g_{A}=\left\\{\begin{array}[]{rl}1.21\pm
0.08&\textrm{Ref.~{}\cite[cite]{\@@bibref{Authors
Phrase1YearPhrase2}{Edwards:2005ym}{\@@citephrase{(}}{\@@citephrase{)}}}}\\\
1.20\pm 0.07&\textrm{Ref.~{}\cite[cite]{\@@bibref{Authors
Phrase1YearPhrase2}{Yamazaki:2008py}{\@@citephrase{(}}{\@@citephrase{)}}}}\end{array}\right.,$
(11)
showing good agreement with experiment ($g_{A}=1.269\pm 0.003$). Work is under
way to reduce the theoretical uncertainty.
In all fields of physics, spectroscopy is a time-honored approach to learn
about the dynamics of the underlying phenomena of interest. In nuclear and
hadronic physics, therefore, an important area of investigation is the
excitation spectrum of hadrons, particularly baryons. Excited states present
additional challenges for lattice QCD, but the tools needed are available and
have been demonstrated to work Sasaki:2001nf ; Basak:2007kj . This subject
extends to include topics such as the intriguing Roper resonance Mathur:2003zf
and electromagnetic transitions of the form $\gamma N\to N^{*}$, where $N^{*}$
is an excited nucleon Lin:2008qv .
Because the gluons couple to each other, it is conceivable that there are
hadrons consisting essentially of gluons, with no valence quarks. (Virtual
quark-antiquark pairs appear in all hadrons.) The most persuasive evidence
that glueballs do indeed exist come from lattice-QCD calculations. In an
approximation omitting virtual quark pairs, the evidence is clear that stable
gluon-only bound states arise Vaccarino:1999ku ; Morningstar:1999rf . There
are two challenges with glueballs. One, obviously, is to extend the lattice
calculations to full QCD with virtual quark-antiquark pairs. The other is to
identify them unambiguously in the laboratory Sexton:1995kd . Experiments with
$\gamma p$ collisions are expected to produce such states and are planned at
the 12 GeV upgrade of CEBAF, the accelerator at Jefferson Lab. To plan this
program it is essential to know the photocouplings of the mesons (exotic or
not). It is possible to compute these photocouplings in lattice QCD, as has
been demonstrated for charmonium ($\bar{c}c$ mesons) Dudek:2006ej . An
extension of this work to light mesons, including exotics and glueballs, is
under way.
A long-term goal of nuclear physics is to compute many-body interactions from
QCD. In addition to providing a better understanding of the nucleus, these
calculations are relevant to mesonic atoms, namely, those containing a
positively and negatively charged meson, and to strangeness in neutron stars.
Although in some cases lattice QCD results can be compared to measurements,
the more interesting situation is where this cannot be done. The aim here is
to gain confidence in the reliability of these difficult calculations, so that
one can use them (with robust uncertainty estimates) in applications where it
is otherwise impossible to acquire the needed information any other way.
An interesting example is the lattice calculation of the $\pi K$ scattering
lengths, of which there are two, $a_{1/2}$ and $a_{3/2}$, depending on whether
the $\pi K$ system has isospin $I=\frac{1}{2}$ or $\frac{3}{2}$. A recent
lattice-QCD calculation finds Beane:2006gj
$\displaystyle a_{1/2}m_{\pi}c/\hbar$ $\displaystyle=$
$\displaystyle+0.1725\pm 0.0017^{+0.0023}_{-0.0156},$ (12) $\displaystyle
a_{3/2}m_{\pi}c/\hbar$ $\displaystyle=$ $\displaystyle-0.0574\pm
0.0016^{+0.0024}_{-0.0058},$ (13)
in convenient dimensionless units. These are in rough, but not spectacular,
agreement with other determinations using chiral perturbation theory or the
Roy-Steiner equations. For example, an analysis based on the Roy-Steiner
method Buettiker:2003pp finds
$\displaystyle a_{1/2}m_{\pi}c/\hbar$ $\displaystyle=$ $\displaystyle+0.224\pm
0.022,$ (14) $\displaystyle a_{3/2}m_{\pi}c/\hbar$ $\displaystyle=$
$\displaystyle-0.0448\pm 0.0077,$ (15)
employing an extrapolation from high energies down to the threshold (where the
scattering length is defined); see Beane:2008dv for more discussion. A more
direct experimental determination is proposed by the DiRAC Collaboration,
which would form $\pi^{-}K^{+}$ atoms DiRAC . In this sense, Eqs. (12) and
(13) represent something exciting: a theoretical prediction awaiting
definitive experimental confirmation.
Multimeson systems are proving to be a fruitful area of research, with further
work on $\pi\pi$ Beane:2007xs , $KK$ Beane:2007uh , and even states with as
many as twelve pions Detmold:2008fn . Once methods have been vetted in mesonic
systems, the next step is to consider nucleon-nucleon Beane:2006mx and
nucleon-hyperon Beane:2006gf systems. This is more difficult, because the
signal-to-noise ratio in the Monte Carlo estimate of baryonic correlation
functions is worse. Here the methods of effective field theories can be used
to extend feasible lattice-QCD calculations to a wider range of nuclear
phenomena. The hyperon-nucleon interaction is of interest, because there is
speculation that the large Fermi energy may make it energetically favorable
for some nucleons to transmute to hyperons Kaplan:1986yq .
## 4 QCD thermodynamics
The discussion in the introduction of the computational scope of numerical
lattice QCD mentions that, but does not explain why, the spacetime lattice
usually has a time extent much longer than the spatial dimension, for example,
$32^{3}\times 128$. The reason is that the functional integral actually
describes a grand canonical ensemble (in the sense of thermodynamics) at
temperature $T$,
$k_{B}T=\hbar c/N_{t}a,$ (16)
where $N_{t}$ is the number of sites in the time direction, $a$ is the lattice
spacing, and $k_{B}$ is Boltzmann’s constant. In the applications to particle
and nuclear physics discussed in Sections 2 and 3, one wants $T=0$ and, hence,
takes $N_{t}$ large enough for the Boltzmann suppression
$e^{-mc^{2}/k_{B}T}=e^{-N_{t}amc/\hbar}$ to eliminate thermal effects, where
$m$ is a hadron mass.
There are, however, many areas of research where a nonzero temperature and a
nonzero chemical potential arise. For nonzero temperature, one simply makes
$N_{t}$ smaller, typically $4\leq N_{t}\leq 12$ in current work. Such lattices
take somewhat less memory, but not less computing because thermodynamics
obviously requires calculations at several temperatures. It is conceptually
easy to introduce a nonzero chemical potential into lattice gauge theory, but
the conceptually clean approach is computationally beyond present resources.
For small chemical potential, several approaches are available and efficient
Schmidt:2006us .
The physical phenomena of interest are summarized by the phase diagram in
figure 4.
Figure 4: Phase diagram of QCD, showing conjectured phases and phenomena of
physical interest. Accelerator experiments at RHIC and LHC probe small
chemical potential, around the crossover regime. At the planned Facility for
Antiproton and Ion Research (FAIR) experiments can reach larger chemical
potential. Image from the CBM Collaboration cbm .
The key feature is a phase transition from hadronic matter to a substance
called the quark-gluon plasma. The transition is expected to be first order
and end at a critical point at chemical potential $\mu\neq 0$. Between the
critical point and the $\mu=0$ axis the transition is a rapid crossover.
Lattice-QCD calculations, as well as measurements of heavy-ion collisions at
RHIC and LHC, probe the crossover, which is relevant to the cooling of the
early universe. Neutron stars appear at high baryonic density, namely, high
$\mu$. At even higher densities, several phases with color superconductivity
are thought to exist. Apart from heavy-ion collisions (so far at small $\mu$)
much of this phase diagram is unexplored, and lattice gauge theory offers the
only _ab initio_ , quantitative, theoretical tool to see if it really reflects
QCD.
The central concept in QCD thermodynamics is the equation of state (EoS),
namely, the relationship between the pressure $p$ and energy density
$\epsilon$ as the temperature $T$ and chemical potential $\mu$ are varied. The
EoS reveals the relevant degrees of freedom. In QCD, are they hadrons? Or
quarks and gluons? At low temperatures, hadrons are clearly present, and
phenomenological models of hadron gases can describe QCD thermodynamics. At
very high temperatures, perturbation theory in $\sqrt{\alpha_{s}(\hbar
c/k_{B}T)}$ can be used. But near the transition, nonperturbative methods are
needed.
The first step is to find the transition temperature, which has now been done
with 2+1 flavors of realistic sea quarks Bernard:2004je ; Cheng:2006qk . The
different physics below, near, and above the critical temperature $T_{c}$ also
has ramifications for controling systematic errors. Discretization effects
from the lattice, for example, depend on the basic degrees of freedom. (For a
crossover it is, strictly speaking, incorrect to speak of a critical
temperature. But the critical point in the phase diagram is close enough to
the $\mu=0$ axis that thermodynamic variables change rapidly at $T_{c}$.)
The next step is to map out the thermal contribution to the trace of the
energy-momentum tensor, $\tr\Theta=\epsilon-3p$. From the temperature
dependence of $\epsilon-3p$ other thermodynamic variables, such as the entropy
density, can be worked out. This also has now been done with 2+1 flavors of
realistic sea quarks Bernard:2006nj ; Cheng:2007jq , including at small $\mu$
Bernard:2007nm . As the system is heated, even to $2T_{c}$, the behavior of
$\epsilon-3p$ shows that the system does not turn rapidly into weakly
interacting quarks and gluons.
To make contact with heavy-ion collisions, the last step would be to take the
output of lattice QCD as input to hydrodynamic models. With these models one
would then compute experimental observables, such as elliptic flow and
quarkonium suppression. With current computing resources, the EoS has been
pinned down with 20–25% accuracy. To make a big impact on hydrodynamic
modeling, however, it will be necessary to reduce the error to 5%. This is
conceptually straightforward to achieve, but it will be computationally
challenging. Part of the strategy is to reduce discretization effects, which
scale as $a^{2}$ ($a$ is the lattice spacing), while the net cost scales as
$a^{-11}$.
## 5 Non-Standard particle physics
As mentioned above, particle physics awaits a new era when the LHC starts
producing physics results. Commencing later this year, the LHC will collide
two beams of 7 TeV (TeV = teraelectronvolt) protons. (Amusingly, some particle
physicists now call this energy scale the terascale, deliberately alluding to
terascale computing. It would be unkind to tell them that in computing we are
moving on to the petascale.) In particular, the LHC will address the central
problem of particle physics, which is the origin of the $W$ and $Z$ boson
masses. The fact that these masses do not vanish demonstrates that some
mechanism breaks the electroweak gauge symmetry spontaneously. “Spontaneous
symmetry breaking” means simply that, although the equations of motion are
symmetric, the solution chosen is not symmetric. For example, in the Standard
Model a scalar field breaks electroweak symmetry, via a technique invented by
Higgs. Interactions between the Higgs field and the fermions also generate (in
the Standard Model) the quark and charged lepton masses and the CKM matrix.
Therefore, flavor physics will remain interesting in the LHC era, and the
physics program discussed in Section 2 will need to be continued and extended.
In this section, however, I would like to consider applications of lattice
gauge theory to understanding the mechanism of electroweak symmetry breaking.
Since we suspect that the Standard Higgs sector is far from the full story, it
is possible (some would say likely) that strongly coupled fields are involved.
For example, instead of being an elementary field, the Higgs boson (and its
siblings, the longitudinal polarization states of the $W$ and $Z$ bosons)
could be composite. The constituents would then form a whole spectrum of
states, just as quarks bind into many different hadrons. Even if the Higgs is
elementary at the terascale, other agents of electroweak symmetry breaking
could be strongly coupled.
Numerical work on theoretical bounds on the Higgs mass are a long-standing
topic in lattice field theory Dashen:1983ts ; Heller:1993yv . The issue is
that the Higgs mass is proportional to its self-coupling, which runs at short
distances in a way similar to $\alpha$ in QED; see figure 1. The perturbative
running becomes invalid once the coupling is too large, and nonperturbative
methods are required. A subtle relation arises between the Higgs mass and a
scale where new phenomena imply that the Standard Model breaks down. To make a
long story short, one can in this way bound the Higgs mass from above. A
similar analysis applied to the Yukawa coupling between the Higgs field and
the top quark leads to a lower bound on the Higgs mass, as discussed in a
recent comprehensive analysis Fodor:2007fn .
A popular class of non-Standard models is to replace the Higgs sector with a
strongly interacting sector that is, in some respects at least, analogous to
QCD. In fact, absent a stronger source of symmetry breaking, the pions of QCD
would break the electroweak symmetry, leading to $W$ and $Z$ masses around 100
MeV. What one needs, then, is a new quantum number, called technicolor, and a
gauge force to bind the corresponding particles, called techniquarks,
together. These dynamics are posited to take place at terascale energies and,
by analogy with QCD, generate vector boson masses around 100 GeV, as observed.
The earliest models of technicolor followed the example of QCD too closely and
have been ruled out by experimental measurements. One way around the
experimental constraints is to assume that the coupling evolves more slowly
than in QCD. This is called “walking” (i.e., slower than running) technicolor.
For a wide range of scales below the confining scale, the coupling is too
strong for perturbation theory to be relied on, so nonperturbative work with
lattice gauge theory is called for. In the past year, the walking hypothesis
has indeed been tested, by examining the scaling behavior of the technipion
mass, the vector meson mass, and the technipion decay constant
Catterall:2007yx . Even more recently, the evolution of $\alpha_{s}$ in (a
TeV-scale version of) QCD with several flavors has been computed
Appelquist:2007hu , to see how many flavors are needed to having a walking
coupling, and how a walking coupling changes the physics.
The most popular way to relieve the theoretical problems of the Standard Model
is to introduce supersymmetry, a symmetry that transforms fermions into
bosons, and vice versa. In these models, every known particle (quarks,
leptons, gauge bosons) comes with another state known as a superpartner. In
the minimal supersymmetric model, the Higgs sector contains three neutral
bosons, a charged pair, and their superpartners. No superpartner has been
observed, so at the terascale and below supersymmetry must be broken somehow.
Usually supersymmetric models are conceived of as weakly coupled, but, since
they are gauge theories, strong coupling is not to be dismissed out of hand.
It is therefore of great interest to formulate and simulate supersymmetric
gauge theories on a spacetime lattice. This has been an extremely fruitful
theoretical field during the past several years Kaplan:2003uh ; Giedt:2007hz ,
and now first numerical simulations are being carried out Catterall:2007fp ;
Catterall:2008yz . This rapid development yields hope that nonperturbative
studies of supersymmetric models will be possible, including studies of
supersymmetry breaking.
## 6 Perspective
Every now and then one hears the question, “How much computing would it really
take to solve QCD on a computer?” The questioner is (usually) well meaning,
but most experts would agree that the question is not well conceived. The
calculations for some of the simplest hadronic quantities are finally becoming
mature, with analyses that forthrightly address all sources of uncertainty.
This class of observable, sometimes called “gold-plated” Davies:2003ik ,
consists of masses of particles that are stable (with respect to the strong
interactions), not too close to thresholds, and of hadronic transition matrix
elements with one or zero of these stable hadrons in the final and initial
states. Many, but not all, aspects flavor physics (Section 2), and some
aspects of hadronic structure (Section 3), lie in this class, and the first
solid round of serious computation is nearing its end.
But this is the end of the beginning. In these nearly mature fields, it turns
out that precision is important, at least for the foreseeable future. For
example, to gain persuasive evidence of new phenomena in flavor physics, 5–25%
accuracy is not enough. A better target (given the stakes and the quality of
the corresponding experiments) is 1–3%. If the history of kaon physics is a
guide, it will eventually be important to go further. Happily, the need for
and clear path to precision has persuaded the particle-physics community that
the resources needed for lattice flavor physics are worth the cost.
Even a quick glance at Section 3 reveals that the large majority of
interesting problems in nuclear physics require much more computing resources
than gold-plated observables. By definition the excited hadrons are not stable
under the strong interaction. They will be harder to pin down, and it is
possible that we will not know how hard until the work currently under way
reaches some of its milestones. The calculations of scattering lengths, while
impressive, must be repeated with several values of the physical volume to
ensure that the mathematical formalism used to extract them works as it
should. Given the importance of nuclear lattice QCD in guiding nuclear
experimentation, one would hope for full support from experimenters for the
needed computing resources.
The “how much?” question should be rephrased. It is really a class of
questions, “How much computing would it take to compute my favorite
observable?” Of course, the answer depends on your observable and its wider
context. It should be clear, however, that physics with either precision and
complexity leads to questions whose answers will need one or two orders of
magnitude more computing than what is available now.
Another reason the “how much?” question is ill-conceived is that it often
overlooks the heterogeneous job mix needed to attack many problems in lattice
QCD. It is not enough to generate huge ensembles of lattice gauge fields, with
ever smaller lattice spacing and ever smaller quark masses, on the world’s
highest-capability computer. This step is necessary, creating a gold mine of
information. It is not enough, however, merely to open a mine; the gold must
be extracted and, here, we need a large capacity of computing that is flexible
in every way (while having significant capability of its own).
We do not know how broad the future of lattice gauge theory will be. If the
LHC discovers a strongly interacting sector beyond the Standard Model, it is
likely that chiral fermions play a role. This is simply because, at the
terascale, we already know that left-handed fermions and right-handed fermions
are different fields. But chiral fermions possess some conceptual and
computational challenges not found in QCD. A strongly coupled terascale will
bring many more particle theorists to lattice gauge theory, and this larger
community will clamor for computing that is up to the task.
This work has been supported in part by the United States National Science
Foundation and the Office of Science of the U.S. Department of Energy (DOE).
Software development and hardware prototyping within the USQCD Collaboration
are supported by SciDAC. Fermilab is operated by Fermi Research Alliance, LLC,
under Contract DE-AC02-07CH11359 with the US DOE.
## References
* (1) Schwinger J (ed) 1958 Selected Papers in Quantum Electrodynamics (New York: Dover) ISBN 0-486-60444-6
* (2) Kinoshita T 1996 Rept. Prog. Phys. 59 1459–1492
* (3) Alcaraz J et al. (LEP) 2007 Precision electroweak measurements and constraints on the Standard Model (Preprint arXiv:0712.0929 [hep-ex])
* (4) Dixon L J 2007 Hard QCD processes at colliders, in 23rd International Symposium on Lepton-Photon Interactions at High Energy (Preprint arXiv:0712.3064 [hep-ph])
* (5) Lepage G P 1989 What is renormalization? in From Actions to Answers ed DeGrand T and Toussaint D (Singapore: World Scientific) (Preprint arXiv:hep-ph/0506330)
* (6) Wilson K G 1974 Phys. Rev. D 10 2445–2459
* (7) Glimm J and Jaffe A M 1987 Quantum Physics: A Functional Integral Point of View (New York: Springer) ISBN 0-387-96476-2
* (8) USQCD Collaboration URL http://www.usqcd.org/
* (9) Detar C E (ILDG) 2006 PoS LATTICE2007 009 (Preprint arXiv:0710.1660 [hep-lat])
* (10) Holmgren D J 2005 Nucl. Phys. B Proc. Suppl. 140 183–189 (Preprint arXiv:hep-lat/0410049)
* (11) Becher T and Schwartz M D 2008 A precise determination of $\alpha_{s}$ from LEP thrust data using effective field theory (Preprint arXiv:0803.0342 [hep-ph])
* (12) Mason Q et al. (HPQCD) 2005 Phys. Rev. Lett. 95 052002 (Preprint arXiv:hep-lat/0503005)
* (13) Yao W M et al. (Particle Data Group) 2006 J. Phys. G 33 1–1232
* (14) Mason Q, Trottier H D, Horgan R, Davies C T H and Lepage G P (HPQCD) 2006 Phys. Rev. D 73 114501 (Preprint arXiv:hep-ph/0511160)
* (15) Bernard C et al. (MILC) 2006 PoS LATTICE2007 090 (Preprint arXiv:0710.1118 [hep-lat])
* (16) Van de Water R S (USQCD) 2007 J. Phys. Conf. Ser. 78 012079
* (17) Aubin C et al. (MILC) 2004 Phys. Rev. D 70 114501 (Preprint arXiv:hep-lat/0407028)
* (18) Beane S R, Bedaque P F, Orginos K and Savage M J (NPLQCD) 2007 Phys. Rev. D 75 094501 (Preprint arXiv:hep-lat/0606023)
* (19) Gamiz E et al. (HPQCD) 2006 Phys. Rev. D 73 114502 (Preprint arXiv:hep-lat/0603023)
* (20) Antonio D J et al. (RBC and UKQCD) 2008 Phys. Rev. Lett. 100 032001 (Preprint arXiv:hep-ph/0702042)
* (21) Boyle P A et al. (RBC and UKQCD) 2008 Phys. Rev. Lett. 100 141601 (Preprint arXiv:0710.5136 [hep-lat])
* (22) Aubin C et al. (Fermilab Lattice and MILC) 2005 Phys. Rev. Lett. 94 011601 (Preprint arXiv:hep-ph/0408306)
* (23) Aubin C et al. (Fermilab Lattice and MILC) 2005 Phys. Rev. Lett. 95 122002 (Preprint arXiv:hep-lat/0506030)
* (24) Follana E, Davies C T H, Lepage G P and Shigemitsu J (HPQCD) 2008 Phys. Rev. Lett. 100 062002 (Preprint arXiv:0706.1726 [hep-lat])
* (25) Okamoto M et al. (Fermilab Lattice and MILC) 2005 Nucl. Phys. B Proc. Suppl. 140 461–463 (Preprint arXiv:hep-lat/0409116)
* (26) Dalgic E et al. (HPQCD) 2006 Phys. Rev. D 73 074502 (Preprint arXiv:hep-lat/0601021)
* (27) Gray A et al. (HPQCD) 2005 Phys. Rev. Lett. 95 212001 (Preprint arXiv:hep-lat/0507015)
* (28) Bernard C et al. (Fermilab Lattice and MILC) 2006 PoS LAT2006 094
* (29) Dalgic E et al. (HPQCD) 2007 Phys. Rev. D 76 011501 (Preprint arXiv:hep-lat/0610104)
* (30) Laiho J (Fermilab Lattice and MILC) 2006 PoS LATTICE2007 358 (Preprint arXiv:0710.1111 [hep-lat])
* (31) Della Morte M 2007 PoS LATTICE2007 008 (Preprint arXiv:0711.3160 [hep-lat])
* (32) Jüttner A 2007 PoS LATTICE2007 014 (Preprint arXiv:0711.1239 [hep-lat])
* (33) Eisenstein B I et al. (CLEO) 2008 Precision measurement of $B(D^{+}\to\mu^{+}\nu)$ and the pseudoscalar decay constant $f_{D^{+}}$ (Preprint arXiv:0806.2112 [hep-ex])
* (34) Dobrescu B A and Kronfeld A S 2008 Phys. Rev. Lett. 100 241802 (Preprint arXiv:0803.0512 [hep-ph])
* (35) Hägler P 2007 PoS LATTICE2007 013 (Preprint arXiv:0711.0819 [hep-lat])
* (36) McNeile C 2007 PoS LATTICE2007 019 (Preprint arXiv:0710.0985 [hep-lat])
* (37) Kaplan D B 2007 Effective field theory as the bridge between lattice QCD and nuclear physics, in Quark Confinement And The Hadron Spectrum 7 ed Ribeiro J et al. (New York: AIP) (Preprint arXiv:nucl-th/0611025)
* (38) Beane S R, Orginos K and Savage M J (NPLQCD) 2008 Hadronic interactions from lattice QCD (Preprint arXiv:0805.4629 [hep-lat])
* (39) Hägler P et al. (LHPC) 2008 Phys. Rev. D 77 094502 (Preprint arXiv:0705.4295 [hep-lat])
* (40) Edwards R G et al. (LHPC) 2006 Phys. Rev. Lett. 96 052001 (Preprint arXiv:hep-lat/0510062)
* (41) Yamazaki T et al. (RBC and UKQCD) 2008 Phys. Rev. Lett. 100 171602 (Preprint arXiv:0801.4016 [hep-lat])
* (42) Sasaki S, Blum T and Ohta S 2002 Phys. Rev. D 65 074503 (Preprint arXiv:hep-lat/0102010)
* (43) Basak S et al. (LHPC) 2007 Phys. Rev. D 76 074504 (Preprint arXiv:0709.0008 [hep-lat])
* (44) Mathur N et al. 2005 Phys. Lett. B 605 137–143 (Preprint arXiv:hep-ph/0306199)
* (45) Lin H W, Cohen S D, Edwards R G and Richards D G 2008 First lattice study of the $N$-$P_{11}(1440)$ transition form factors (Preprint arXiv:0803.3020 [hep-lat])
* (46) Vaccarino A and Weingarten D 1999 Phys. Rev. D 60 114501 (Preprint arXiv:hep-lat/9910007)
* (47) Morningstar C J and Peardon M J 1999 Phys. Rev. D 60 034509 (Preprint arXiv:hep-lat/9901004)
* (48) Sexton J, Vaccarino A and Weingarten D 1995 Phys. Rev. Lett. 75 4563–4566 (Preprint arXiv:hep-lat/9510022)
* (49) Dudek J J, Edwards R G and Richards D G 2006 Phys. Rev. D 73 074507 (Preprint arXiv:hep-ph/0601137)
* (50) Beane S R et al. (NPLQCD) 2006 Phys. Rev. D 74 114503 (Preprint arXiv:hep-lat/0607036)
* (51) Buettiker P, Descotes-Genon S and Moussallam B 2004 Eur. Phys. J. C 33 409–432 (Preprint arXiv:hep-ph/0310283)
* (52) Dimeson Relativistic Atom Complex Collaboration (DiRAC) URL http://dirac.web.cern.ch/DIRAC/
* (53) Beane S R et al. (NPLQCD) 2008 Phys. Rev. D 77 014505 (Preprint arXiv:0706.3026 [hep-lat])
* (54) Beane S R et al. (NPLQCD) 2008 Phys. Rev. D 77 094507 (Preprint arXiv:0709.1169 [hep-lat])
* (55) Detmold W et al. (NPLQCD) 2008 Multi-Pion States in Lattice QCD and the Charged-Pion Condensate (Preprint arXiv:0803.2728 [hep-lat])
* (56) Beane S R, Bedaque P F, Orginos K and Savage M J (NPLQCD) 2006 Phys. Rev. Lett. 97 012001 (Preprint arXiv:hep-lat/0602010)
* (57) Beane S R et al. (NPLQCD) 2007 Nucl. Phys. A 794 62–72 (Preprint arXiv:hep-lat/0612026)
* (58) Kaplan D B and Nelson A E 1986 Phys. Lett. B 175 57–63
* (59) Schmidt C 2006 PoS LAT2006 021 (Preprint arXiv:hep-lat/0610116)
* (60) Compressed Baryonic Matter Collaboration (CBM) URL http://www.gsi.de/fair/experiments/CBM/
* (61) Bernard C et al. (MILC) 2005 Phys. Rev. D 71 034504 (Preprint arXiv:hep-lat/0405029)
* (62) Cheng M et al. (RBC and Bielefeld) 2006 Phys. Rev. D 74 054507 (Preprint arXiv:hep-lat/0608013)
* (63) Bernard C et al. (MILC) 2007 Phys. Rev. D 75 094505 (Preprint arXiv:hep-lat/0611031)
* (64) Cheng M et al. (RBC and Bielefeld) 2008 Phys. Rev. D 77 014511 (Preprint arXiv:0710.0354 [hep-lat])
* (65) Bernard C et al. (MILC) 2008 Phys. Rev. D 77 014503 (Preprint arXiv:0710.1330 [hep-lat])
* (66) Dashen R F and Neuberger H 1983 Phys. Rev. Lett. 50 1897
* (67) Heller U M, Klomfass M, Neuberger H and Vranas P M 1993 Nucl. Phys. B 405 555–573 (Preprint arXiv:hep-ph/9303215)
* (68) Fodor Z, Holland K, Kuti J, Nogradi D and Schroeder C 2007 PoS LATTICE2007 056 (Preprint arXiv:0710.3151 [hep-lat])
* (69) Catterall S and Sannino F 2007 Phys. Rev. D 76 034504 (Preprint arXiv:0705.1664 [hep-lat])
* (70) Appelquist T, Fleming G T and Neil E T 2008 Phys. Rev. Lett. 100 171607 (Preprint arXiv:0712.0609 [hep-ph])
* (71) Kaplan D B 2004 Nucl. Phys. B Proc. Suppl. 129 109–120 (Preprint arXiv:hep-lat/0309099)
* (72) Giedt J 2006 PoS LAT2006 008 (Preprint arXiv:hep-lat/0701006)
* (73) Catterall S and Wiseman T 2007 JHEP 12 104 (Preprint arXiv:0706.3518 [hep-lat])
* (74) Catterall S and Wiseman T 2008 Black hole thermodynamics from simulations of lattice Yang-Mills theory (Preprint arXiv:0803.4273 [hep-th])
* (75) Davies C T H et al. (HPQCD, MILC, and Fermilab Lattice) 2004 Phys. Rev. Lett. 92 022001 (Preprint arXiv:hep-lat/0304004)
| arxiv-papers | 2008-07-14T17:58:00 | 2024-09-04T02:48:56.780652 | {
"license": "Public Domain",
"authors": "Andreas S. Kronfeld",
"submitter": "Andreas S. Kronfeld",
"url": "https://arxiv.org/abs/0807.2220"
} |
0807.2285 | # Gluino zero-modes for calorons at finite temperature
Margarita García Pérez margarita.garcia@uam.es Antonio González-Arroyo
antonio.gonzalez-arroyo@uam.es Alfonso Sastre alfonso.sastre@uam.es Instituto
de Física Teórica UAM/CSIC Departamento de Física Teórica
Universidad Autónoma de Madrid, E-28049–Madrid, Spain
###### Abstract
We study the solutions of the Dirac equation in the adjoint
representation(gluinos) in the background field of SU(2) unit charge calorons.
Our solutions are forced to be antiperiodic in thermal time and would occur
naturally in a semiclassical approach to ${\cal N}=1$ Super-symmetric Yang-
Mills theory at finite temperature.
###### keywords:
Caloron , gluino zero-modes , finite temperature , super-symmetric Yang-Mills
theory
IFT-UAM/CSIC-08-30
FTUAM-08-09
## 1 Introduction
In this paper we derive analytic expressions for the finite temperature gluino
zero modes of the Dirac operator in the background field of the $Q=1$ SU(2)
calorons. These are self-dual configuration in $R^{3}\times S_{1}$ including
the well known Harrington-Shepard (HS) solution [1] as well as the non-trivial
holonomy calorons [2]-[4]. The periodicity in one direction, to be referred as
thermal-time, occurs naturally in a path-integral approach to finite
temperature Yang-Mills theory and, with the inclusion of spinor fields in the
adjoint representation (gluinos), in its minimal supersymmetric extension.
Calorons are thus the natural objects to be considered in a semiclassical
approach to these theories at finite temperature. They smoothly interpolate
between instantons and BPS monopoles at zero and high temperature respectively
[2]-[5], providing a very interesting link between them. One of the required
ingredients for such semiclassical analysis is the knowledge of fermionic zero
modes in the background of the caloron field. Although those in the
fundamental representation of the gauge group have been known for quite some
time [6]-[10], this is not the case for the gluino zero modes. They have been
derived only recently by two of the present authors [11], and just for the
case of periodic boundary conditions in $S_{1}$. These are the relevant modes
for supersymmetric compactifications but not what is needed when studying
${\cal N}=1$ SUSY Yang-Mills fields at finite temperature. Antiperiodicity in
thermal-time has to be required in that case. The goal of this letter is to
obtain analytic expressions for the antiperiodic solutions, derived here for
the first time even for the trivial holonomy, Harrington-Shepard, case. This
requires a different approach than the one employed in [11] which was based on
the relation between zero modes and self-dual deformations of the gauge field,
providing only periodic solutions.
The paper is organized as follows. In section 2 we will describe the strategy
followed to obtain the antiperiodic zero modes and present the analytic
expressions for the solutions. In section 3 we analyze their properties in
several relevant limits, paying particular attention to the one in which the
caloron dissociates into a pair of static BPS constituent monopoles. The
trivial-holonomy HS zero mode solution and the equal mass constituent monopole
cases are also discussed in some detail. Conclusions and a brief summary of
results are presented in section 4.
## 2 Formalism
As mentioned previously, our goal is that of solving the massless covariant
Dirac equation in the adjoint representation of the group
$\not\\!\\!D\Psi=0$ (1)
in the background field of a Q=1 caloron [2]-[4]. This problem has been
partially addressed in Ref. [11]. The approach that was followed in that paper
was based on the well-known relation between self-dual deformations of the
gauge field and the zero-modes of the Dirac operator in the adjoint
representation. However, the solutions obtained in this way are periodic in
thermal-time with the same period $\beta$ (to be taken equal to 1 in what
follows) as the gauge field itself. Thus, a different strategy has to be set
up to derive the antiperiodic modes relevant for finite temperature. In what
follows we will present the basic idea behind our procedure and the results
obtained with it. In all technical aspects we will rely strongly in the
notation and derivations done in Ref. [11].
The observation that leads to our solution is the fact that antiperiodic
solutions turn out to be periodic in the double period. Thus, the method of
attack developed in Ref. [11] for periodic zero-modes can be carried over if
the whole problem is seen as living in this duplicated space-time. This
replica trick has been used by some of the authors in other works [12, 13] and
is an important source of information when dealing with periodic gauge fields.
In our case, the problem becomes that of finding self-dual deformations of the
$Q=2$ caloron obtained by the replica procedure. Notice that the topological
charge is 2 in this case, so we expect 4 (CP-pairs) of self-dual deformations.
Since the gauge field is periodic in the original period, they can be split
into those which are periodic and those that are antiperiodic in the original
period. The former were studied in our previous paper and correspond to the
ordinary deformations of the $Q=1$ caloron. Since there are 2 pairs of those,
which are periodic in the small torus, we expect to find two pairs of
antiperiodic zero-modes. Unfortunately, although some particular solutions are
known [14], there is no analytic general expression for the $Q=2$ caloron
which would reduce the study of deformations to the differentiation of the
general solution with respect to the parameters of the moduli space. In this
paper we will thus follow an alternative strategy. Incidentally our results
could well prove useful in achieving the goal of obtaining the most general
$Q=2$ caloron solution.
The general formula relating deformations to zero-modes in the adjoint
representation is:
$\Psi=\frac{1}{2}\delta A_{\mu}\gamma_{\mu}(\mathbf{I}\pm\gamma_{5})V\,,$ (2)
with the $+$ or $-$ sign depending on whether the solution is self-dual or
antiself-dual. $V$ is an arbitrary constant spinor and hence, the zero-modes
that we are looking for can be arranged into two-dimensional complex vector
spaces. These spaces are generated by any solution $\Psi$ and its euclidean CP
transform
$\Psi\longrightarrow\Psi^{c}\equiv\gamma_{5}C\Psi^{*}\,.$ (3)
Our formula can easily be shown to satisfy the Dirac equation provided the
deformation satisfies the background Lorentz gauge condition.
$D_{\mu}\delta A_{\mu}=0\,.$ (4)
Using the general ADHM construction one can obtain formulas for the self-dual
deformations in terms of those for the Nahm-ADHM data. The ADHM construction
for SU(2) tells us that a self-dual gauge field can be constructed as [15]
$A_{\mu}(x)=\frac{i}{F}\,(u^{\dagger}\partial_{\mu}u)^{\prime}\,,$ (5)
where $u$ is a vector in quaternions, $F=1+u^{\dagger}u$, and the prime
denotes the traceless part. The vector $u$ is obtained as the solution of the
following equation
$(\widetilde{A}^{\dagger}-x_{\mu}\overline{\sigma}_{\mu})u=q\,,$ (6)
where the quaternionic matrix $\widetilde{A}^{\dagger}$ and the vector $q$ are
x-independent. We have introduced the Weyl matrix
$\overline{\sigma}_{\mu}=(\mathbf{I},i\vec{\tau})$ whose adjoints are
$\sigma_{\mu}$ ($\vec{\tau}$ are the Pauli matrices). In the proof of self-
duality one must demand that the following matrix
$R\equiv(\widetilde{A}^{\dagger}-x_{\mu}\overline{\sigma}_{\mu})(\widetilde{A}-x_{\mu}\sigma_{\mu})+q\otimes
q^{\dagger}\,,$ (7)
is real and invertible. For our purpose it is interesting to write down the
expression of the adjoint zero-modes in terms of the deformations of the ADHM
data
$\delta A_{\mu}=\frac{-i}{2}(\delta
q^{\dagger}-u^{\dagger}\delta\widetilde{A})\overline{\sigma}^{\mu}\sigma^{\nu}\partial_{\nu}\omega+{\rm
h.c.}\,,$ (8)
where $\omega=R^{-1}q$. To guarantee that the deformations $\delta q$ and
$\delta\widetilde{A}$ provide a self-dual deformation $\delta A_{\mu}$
satisfying the background field gauge condition, one must impose certain
conditions. These are best expressed in terms of the matrix with quaternionic
entries ${\cal F}\equiv M^{\dagger}\delta M\equiv{\cal F}_{\mu}\sigma_{\mu}$,
where $M^{\dagger}=(q,\widetilde{A}^{\dagger}-\bar{\sigma}_{\mu}x_{\mu})$. The
condition then reduces to the hermiticity of ${\cal F}_{\mu}$ (${\cal
F}_{\mu}={\cal F}_{\mu}^{\dagger}$).
The previous formulas apply for $Q=1$ calorons by extending the vector $q$ to
become a delta-like functional over the periodic functions in one-variable
$z$, while $2\pi i\widetilde{A}$ is a covariant Weyl operator with respect to
a 1-dimensional abelian gauge field $\hat{A}_{\mu}(z)$, the Nahm-dual gauge
field. After suitable rotations, translations and gauge transformations the
caloron Nahm data can be taken to be [2, 3]
$q^{(0)}(z)=\rho(P_{+}\,\delta(z-\delta_{1})+P_{-}\,\delta(z+\delta_{1}))\,,$
(9)
where $P_{\pm}=(1\pm\tau_{3})/2$. The parameter $\delta_{1}$ parametrizes the
holonomy, becoming trivial for $0$ and $\frac{1}{2}$. Without loss of
generality we will assume in what follows that
$\delta_{1}\leq\delta_{2}\equiv\frac{1}{2}-\delta_{1}$. In the previous
formula the delta functions have to be taken as periodic functions in $z$ with
unit period. The Nahm-dual gauge field of the caloron is given by
$\hat{A}^{(0)}_{\mu}(z)=-2\pi\delta_{\mu
3}(X_{3}^{1}\chi_{1}(z)+X_{3}^{2}\chi_{2}(z))\,,$ (10)
where $X^{a}_{3}$ is the position of the ath constituent monopole on the
z-axis. They can be obtained from the relations
$m_{1}X_{3}^{1}+m_{2}X_{3}^{2}=0$, and $X_{3}^{2}-X_{3}^{1}=\pi\rho^{2}$,
where $m_{a}=4\pi\delta_{a}$ are proportional to the constituent monopole
masses. The function $\chi_{1}$ is the characteristic function of the interval
$[-\delta_{1},\delta_{1}]$ and $\chi_{2}$ that of its complementary.
Eq. (7) implies that the Nahm-dual gauge field is self-dual at all but a
finite number of points. Eq. (6) is then the solution of the Weyl equation
except at those isolated points. As we will see later the conditions on the
deformations $\delta\widetilde{A}$ that enter Eq. (8), are precisely
equivalent to requiring that $\delta\hat{A}_{\mu}$ is again a self-dual
deformation satisfying the background gauge condition. Thus, they can be
obtained as the solution of the adjoint Weyl equation of the Nahm-dual field,
up to delta functions.
Now we should apply this scheme to the replicated caloron taken as a self-dual
solution in the double torus with period $2\beta$ (remember $\beta$ is fixed
to 1). Since this caloron now has charge $Q=2$ its corresponding Nahm-dual
gauge field is now a matrix. Using the general construction of Nahm-dual
replicas given in [12] we obtain
$\hat{A}^{R}_{\mu}(z)=\pmatrix{\hat{A}^{(0)}_{\mu}(z)&0\cr
0&\hat{A}^{(0)}_{\mu}(z+\frac{1}{2})}\,,$ (11)
where $\hat{A}^{(0)}_{\mu}(z)$ is the Nahm data of the ordinary caloron, and
$\hat{A}^{R}_{\mu}(z)$ is the Nahm data of the replicated caloron.
One may now wonder which is the corresponding $q$ for such a replica solution.
We will argue that the solution is actually given by
$q^{R}(z)=\pmatrix{q^{(0)}(z)\cr q^{(0)}(z+\frac{1}{2})}\,,$ (12)
Notice that each of the components of $q$ and $\hat{A}$ are periodic with unit
period, but the whole set is periodic with period $1/2$ with a twist matrix
given by $\tau_{1}$:
$\hat{A}^{R}_{\mu}(z+\frac{1}{2})=\tau_{1}\hat{A}^{R}_{\mu}(z)\tau_{1}\,.$
(13)
The quantity $q$ transforms by periodicity as follows:
$q^{R}(z+1/2)=\tau_{1}q^{R}(z)\,.$ (14)
From here it is possible to use the general formulas of the ADHM construction
to verify that indeed we obtain a replicated solution. In particular we have
that $u^{R}(z)$ is given by:
$u^{R}(z)=\pmatrix{u^{(0)}(z)\cr u^{(0)}(z+\frac{1}{2})}\,.$ (15)
Now
$F^{R}-1=\int_{0}^{\frac{1}{2}}dzu^{R\dagger}(z)u^{R}(z)=\int_{0}^{1}dz\,u^{(0)\dagger}(z)u^{(0)}(z)\,,$
(16)
which coincides with $F-1$ for the caloron. The replicated gauge potential
follows from
$A_{\mu}^{R}(x)=\frac{i}{F^{R}}\int_{0}^{\frac{1}{2}}dz(u^{R\dagger}(z)\partial_{\mu}u^{R}(z))^{\prime}=A_{\mu}^{(0)}(x)\,.$
(17)
One might wonder whether the choice of $q$ and $\hat{A}^{R}_{\mu}$ are
consistent with the condition of self-duality in Nahm-dual space, namely that
$R=M^{\dagger}M$ commutes with the quaternions. To verify that this is so, one
must realize that the condition of self-duality should hold only in the domain
of these operators. These are two-component vectors $\psi(z)$ satisfying
$\psi(z+\frac{1}{2})=\tau_{1}\psi(z)\,.$ (18)
Thus, they should be of the form
$\psi(z)=\pmatrix{\phi(z)\cr\phi(z+\frac{1}{2})}\,.$ (19)
Thus, $qq^{\dagger}$ acting on this vector yields:
$\displaystyle 2\rho\pmatrix{q^{(0)}(z)\cr
q^{(0)}(z+\frac{1}{2})}(P_{+}\phi(\delta_{1})+P_{-}\phi(-\delta_{1}))=$
$\displaystyle
2\rho^{2}\pmatrix{P_{+}\phi(\delta_{1})\delta(z-\delta_{1})+P_{-}\phi(-\delta_{1})\delta(z+\delta_{1})\cr
P_{+}\phi(\delta_{1})\delta(z+\delta_{2})+P_{-}\phi(-\delta_{1})\delta(z-\delta_{2})}\,.$
(20)
The imaginary part of the upper component coincides with
$\rho^{2}\tau_{3}\,(\delta(z-\delta_{1})-\delta(z+\delta_{1}))\,\phi(z)\,,$
(21)
which is what is needed to cancel the self-duality violation.
Now we proceed to study the self-dual deformations of this replicated caloron
satisfying the background field condition. We will make use of our general
formula Eq. (8). The conditions following that equation when translated to our
case become
$\displaystyle\hat{\bar{D}}\psi^{R}\equiv\frac{d\psi^{R}}{dz}-i\bar{\sigma}_{\mu}[\hat{A}^{R}_{\mu},\psi^{R}]=4\pi^{2}i\left(q_{\mu}^{R}\delta
q_{\nu}^{\dagger R}-\delta q_{\nu}^{R}q_{\mu}^{\dagger
R}\right)\bar{\sigma}_{\mu}\sigma_{\nu}\,,$ (22)
where
$\psi^{R}=\delta\hat{A}^{R}_{\mu}\sigma_{\mu}=-\delta\widetilde{A}^{R}/(2\pi)$.
The quantities $\delta q^{R}_{\nu}$ are two-component column vector whose
elements are linear combinations of delta functions with complex coefficients.
The holonomy fixes that the argument of the delta functions must be
$z\pm\delta_{1}$ and $z\pm\delta_{1}+\frac{1}{2}$. Notice that, as anticipated
previously, up to the delta functions in the right-hand side, the equation
adopts the form of the Weyl equation for adjoint zero-modes in Nahm dual
space.
Our next step will then be that of finding the solution of Eq. (22). Notice
that both $\psi^{R}$ and $\delta q^{R}\equiv\delta
q^{R}_{\nu}\bar{\sigma}_{\nu}$ are the unknowns. Without much effort one can
demonstrate that given a solution one can obtain other solutions by the
operation $\psi^{R}\rightarrow\psi^{R}Q$, $\delta q^{R}\rightarrow
Q^{\dagger}\delta q^{R}$, with $Q$ an arbitrary constant quaternion. This
transformation is associated to the double degeneracy of adjoint zero-modes.
We must also point out certain subtleties necessary to understand Eq. (22) and
their solutions. The main idea is that the equation must be understood as one
relating two operators acting on the space two-component functions of the form
Eq. (19). The right-hand side of Eq. (22) acts by multiplication. Thus, the
left-hand side must be equivalent, when acting over our space of functions, to
the multiplication by a linear combination of delta functions. This imposes
non-trivial conditions on the form of $\delta q^{R}$. In what follows we will
give the possible values for $\delta q^{R}$ that follow from the previous
analysis, as well as the resulting form for the equation for $\psi^{R}$,
skipping all the details of the derivation.
Before showing the equations, we recall that $\psi^{R}$ is a $2\times 2$
matrix in (Nahm-dual) colour space
$\psi^{R}(z)=\pmatrix{\psi_{11}(z)&\psi_{12}(z)\cr\psi_{21}(z)&\psi_{22}(z)}\,.$
(23)
The boundary conditions specify that it is enough to know the form of
$\psi_{11}$ and $\psi_{12}$ (the other components can be obtained by
translating in $z$ by $1/2$). The equations for $\psi_{11}$ coincide with
those for the $Q=1$ caloron, and therefore can be associated with deformations
that are periodic in time. Thus, our sought time-antiperiodic zero-modes
should follow from the equation
$\displaystyle\partial_{z}\psi_{12}+\tau_{3}(\Delta\hat{A})\psi_{12}=4\pi^{2}\rho\Big{\\{}\,P_{+}$
$\displaystyle\Big{(}$
$\displaystyle\delta(z-\delta_{1})-\delta(z-\delta_{2})\Big{)}$
$\displaystyle+P_{-}$ $\displaystyle\Big{(}$
$\displaystyle\delta(z+\delta_{2})-\delta(z+\delta_{1})\,\Big{)}\Big{\\}}\,Q\,,$
(24)
where the function $\Delta\hat{A}$ is given by
$\displaystyle\Delta\hat{A}\equiv\hat{A}^{(0)}_{3}(z)-\hat{A}^{(0)}_{3}(z+1/2)=2\pi^{2}\rho^{2}(\chi(-\delta_{1},\delta_{1})-\chi(\delta_{2},1-\delta_{2}))\,.$
(25)
The arbitrary quaternion $Q$ reflects the degeneracy of solutions mentioned
earlier. Keeping that in mind one only needs to solve the equation for $Q=0$
and $Q=1$. A particular solution is all that is needed, since the general
solution can be obtained by linear combinations of these ones with
quaternionic coefficients. The counting matches the predictions of the index
theorem. As for the periodic case there are essentially two CP-pairs of zero-
modes.
After these considerations we proceed to show the two particular solutions
that we will need. The first one corresponds to the inhomogeneous equation
($Q=1$) and is given by
$\psi_{12}=4\pi^{2}\rho\left(P_{+}\chi(\delta_{1},\delta_{2})+P_{-}\chi(1-\delta_{2},1-\delta_{1})\right)\,.$
(26)
The value of $\delta q^{R}$ associated to it is
$\delta
q^{R}=iP_{+}\pmatrix{\delta(z+\delta_{2})\cr\delta(z-\delta_{1})}-iP_{-}\pmatrix{\delta(z-\delta_{2})\cr\delta(z+\delta_{1})}$
(27)
These expressions can now be introduced into the general formula Eq. (8) to
obtain the first solution
$\displaystyle\delta
A_{\mu}^{(1)}=-\frac{1}{2}\Big{(}P_{+}\bar{\sigma}_{\mu}\hat{\partial}\omega(-\delta_{2})-P_{-}\bar{\sigma}_{\mu}\hat{\partial}\omega(\delta_{2})\Big{)}$
$\displaystyle-i\pi\rho\Big{(}\int_{\delta_{1}}^{\delta_{2}}u^{\dagger}(z+\frac{1}{2})P_{-}\bar{\sigma}_{\mu}\hat{\partial}\omega(z)+\int_{1-\delta_{2}}^{1-\delta_{1}}u^{\dagger}(z+\frac{1}{2})P_{+}\bar{\sigma}_{\mu}\hat{\partial}\omega(z)\Big{)}+{\rm
h.c.}\,.$ (28)
The quantities $u$ and $\omega$ are the ones associated to the $Q=1$ caloron.
The analytic expressions needed to do the calculation were explicitly given in
our previous paper [11].
Now we investigate the other solution, associated to $\delta q^{R}=0$. One has
to solve the homogeneous equation (2) for vanishing right hand side. A
particular solution is given by
$\displaystyle\psi_{12}(z)=\exp\\{-\tau_{3}\int_{0}^{z}dz^{\prime}\,\Delta\hat{A}(z^{\prime})\\}\equiv\phi_{s}(z)-\tau_{3}\phi_{a}(z)\,.$
(29)
Since $\Delta\hat{A}(z^{\prime})$ is constant at intervals, the integral in
the exponent is trivial to perform. We leave the explicit form of
$\phi_{s}(z)$ and $\phi_{a}(z)$ to the reader. It is interesting to point out
nonetheless, that $\phi_{s}(z)$ is periodic in $z$ with period $\frac{1}{2}$
and $\phi_{a}(z)$ antiperiodic.
From the previous expression we can compute the corresponding self-dual
deformation using Eq. (8). The result is given by
$\displaystyle\delta A^{(2)}_{\mu}=\frac{-i}{4\pi}\int_{0}^{1}dz\
(u^{\dagger}(z+\frac{1}{2})(\phi_{s}(z)+\tau_{3}\phi_{a}(z))\bar{\sigma}_{\mu}\hat{\partial}\omega(z))+{\rm
h.c.}\,.$ (30)
Again, the integration over $z$ can be performed analytically using the
formulas of our previous paper [11].
We have arrived to the general solution our problem. The adjoint zero-modes of
the (self-dual) caloron which are antiperiodic in time are
$\Psi=\frac{1}{2}\delta
A^{(1)}_{\mu}\gamma_{\mu}(\mathbf{I}+\gamma_{5})V_{1}+\frac{1}{2}\delta
A^{(2)}_{\mu}\gamma_{\mu}(\mathbf{I}+\gamma_{5})V_{2}\,,$ (31)
where $V_{a}$ are arbitrary constant spinors and $\delta A^{(a)}_{\mu}$ are
given in Eqs.(28)-(30).
It is interesting to mention that the general investigation of the possible
values of $\delta q^{R}$ has led to another solution having a fairly simple
form. The expression of the left-handed Weyl spinor,
$\Psi^{(3)}\equiv\Psi_{a}^{(3)}\tau_{a}$, is:
$\Psi_{a}^{(3)}=\sigma_{\mu}\partial_{\mu}T^{a}\sigma_{a}V\,,$ (32)
where $a$ labels a colour component, $\sigma_{\alpha}$ acts on the spin
indices and $V$ denotes an arbitrary constant 2-spinor. The functions $T_{a}$
depend on the colour index as: $T^{1}=T^{2}=-1/F$ and
$T^{3}=P_{+}\chi+P_{-}\bar{\chi}$. The function $\chi$ is essentially the
function with the same name given in Ref. [2, 3]. Curiously this solution
interpolates between the non-supersymmetric periodic adjoint zero-mode for
$m_{1}=0$ ($\delta_{1}=0$) and one of our antiperiodic solutions (Eq. (28))
for $m_{1}=m_{2}$ ($\delta_{1}=1/4$). Using the formulas of the next section
it can be proven that the solution is neither periodic non antiperiodic for
other values of the mass $m_{1}$.
## 3 Properties of the solutions
In this section we will investigate the general properties of the solutions
found in the previous section.
### 3.1 Periodicity in time
Here we will explicitly verify the required antiperiodicity in time of our
general solution. In our gauge the caloron vector potential satisfies
$A_{\mu}^{(0)}(x_{0}+1)=e^{i{m_{1}\tau_{3}\over
2}}A_{\mu}^{(0)}(x_{0})e^{-i{m_{1}\tau_{3}\over 2}}$ (33)
Thus, the required antiperiodicity of the adjoint zero-modes amounts to:
$\Psi(x_{0}+1)=-e^{i{m_{1}\tau_{3}\over 2}}\Psi(x_{0})e^{-i{m_{1}\tau_{3}\over
2}}$ (34)
This property follows easily from the form of our solutions and the
periodicity behaviour of $u$:
$u(z,x_{0}+1)=e^{i2\pi z}u(z,x_{0})e^{-i{m_{1}\tau_{3}\over 2}}$ (35)
and an identical relation for $\omega$ and $q$.
### 3.2 Far-field limit and Normalization
The reader might question whether our general solution Eq. (31) is
normalizable. One can investigate the behaviour at points whose distance to
the location of the constituent monopoles ($r_{1}$ and $r_{2}$) is much larger
that $\beta$ and that $\pi\rho^{2}$. For the unequal mass case the zero-mode
density goes to zero exponentially as $e^{-(m_{2}-m_{1})r_{2}}$. The equal
mass case ($m_{1}=m_{2}=\pi$) is more subtle since both solutions decay in
power-like fashion. The non-homogeneous solution Eq. (28) coincides with the
additional solution Eq. (32) in this case. In the limit under consideration
$\chi$ goes to zero exponentially and
$F=(r_{1}+r_{2}+\pi\rho^{2})/(r_{1}+r_{2}-\pi\rho^{2})$. Thus the density
behaves as $1/r^{4}$.
An alternative approach to normalizability of the solutions is to compute the
norm of the solutions. In fact there exist a general formula [2, 19] which
allows one to compute the norm and the scalar products of the solutions in
terms of Nahm-data directly. This is also useful in checking if the real
dimensionality of the space of solutions is 8 (4 complex dimensions, 2
quaternionic dimensions), as indicated by the index theorem. Using this
formula we obtain
$\displaystyle\left|\delta
A_{\mu}^{(1)}\right|^{2}=4\pi^{2}+8\pi^{4}\rho^{2}\,(\delta_{2}-\delta_{1})\,,$
(36) $\displaystyle\left|\delta
A_{\mu}^{(2)}\right|^{2}=\frac{\sinh(4\pi^{2}\rho^{2}\delta_{1})}{2\pi^{2}\rho^{2}}+(\delta_{2}-\delta_{1})\,\cosh(4\pi^{2}\rho^{2}\delta_{1})\,,$
(37) $\displaystyle\left<\delta A_{\mu}^{(1)},\delta
A_{\mu}^{(2)}\right>=2\pi^{2}\rho\,\e^{-2\pi^{2}\rho^{2}\delta_{1}}\,(\delta_{2}-\delta_{1})\,.$
(38)
### 3.3 Profile of the zero-mode density
In this subsection we will describe the qualitative properties of the zero-
mode densities. For that purpose we developed two independent programs to draw
these profiles. Both programs give matching results. In Fig. 1 we give the
contour plot in a z-y plane of the solution $\delta A^{(2)}_{\mu}$ (top) and
an orthogonal CP-pair (bottom) for $\rho=1$ (giving an intermediate size
caloron separation) and two representative values of the masses. The $z$ axis
is the line joining the constituent monopoles and is represented horizontally.
The vertical axis denotes the $y$ axis (the density is axially symmetric).
For the equal mass case ($m_{1}=m_{2}=\pi$) the mode following from Eq. (30)
has an approximately constant higher density along the line joining both
calorons (top left). This can be interpreted as a string. In contrast, the
other solution associated to Eq. (28) has a region of small density located
along the line joining the two monopoles (bottom left). As the masses become
unequal, the most massive monopole dominates the densities. The right contour
plots show the situation for $\delta_{1}=0.23$.
Figure 1: Contour plots of the density of the two antiperiodic zero modes in
the $y-z$ plane. Constituent monopoles are localized at $y=0$ and separated
along the z axis which is drawn horizontally. Left: For $m_{1}=m_{2}=\pi$ and
$\rho=1$. Right: For $\delta_{1}=0.23$ and $\rho=1$.
### 3.4 Limiting cases
The caloron is an interesting solution which interpolates between the gauge
potential of an instanton and that of a BPS monopole. It is interesting then
to see how our antiperiodic zero-modes behave in these extreme cases. We will
first concentrate in the situation corresponding to the trivial holonomy,
Harrington-Shepard, caloron: $\delta_{1}=0$. In that case one of the
constituent monopoles is massless and pushed to infinity. The $\rho$ parameter
of the solution does no longer control the separation between the monopoles
but is still a free parameter. For small $\rho$ the HS caloron approaches an
ordinary, zero temperature, instanton. From our general formulas, it is easy
to check that in that limit and close to the center of the caloron the time
periodicity becomes irrelevant and the two zero mode CP-pairs approach the
periodic zero modes of the instanton. In the opposite,
$\rho\rightarrow\infty$, limit the HS caloron becomes a BPS monopole with time
independent action density. Despite the time independence of the background
there are still 4 non-trivial antiperiodic zero-modes. They can be easily
derived from Eqs. (28) and (30) by taking the appropriate $\delta_{1}=0$ and
$\rho\rightarrow\infty$ limits. Up to a gauge transformation we obtain:
$\displaystyle\delta A^{(1)^{\prime}}_{\mu}(x)$ $\displaystyle=$
$\displaystyle\eta_{3\mu}^{\alpha}\,\pi\rho\,\Big{(}\bar{e}_{1}^{2}(x)\,E_{\alpha}^{\rm
bps}(r)-\bar{e}_{2}^{2}(x)\,\tilde{E}_{\alpha}(r)\Big{)}+{\rm h.c.}\,,$ (39)
$\displaystyle\delta A^{(2)}_{\mu}(x)$ $\displaystyle=$ $\displaystyle{1\over
4\pi}\,\Big{(}\bar{e}_{2}^{2}(x)\,E_{\mu}^{\rm
bps}(r)+\bar{e}_{1}^{2}(x)\,\tilde{E}_{\mu}(r)\Big{)}+{\rm h.c.}\,,$ (40)
where $\delta A^{(2)}_{\mu}$ is directly derived from Eq. (30) and $\delta
A^{(1)^{\prime}}_{\mu}$ is the combination of Eqs. (28) and (30) orthogonal to
$\delta A^{(2)}_{\mu}$. In the expression above, $E_{\alpha}^{\rm bps}$ is the
electric field of the BPS monopole:
$E_{\alpha}^{\rm bps}(x)=-i\,{g^{2}(2\pi r)-1\over
2r^{2}}\,P_{\alpha}^{+}-i\,{\pi g^{\prime}(2\pi r)\over r}\,P_{\alpha}^{-}\,,$
(41)
and we have introduced the time independent quantity:
$\tilde{E}_{\alpha}(x)={\tanh(\pi r)\over 2\cosh(\pi r)}\,\Big{(}i\,{g(\pi
r)-\cosh(\pi r)\over r^{2}}\,P_{\alpha}^{+}-i\,{\pi g^{\prime}(\pi r)\over
r}\,P_{\alpha}^{-}\Big{)}\,,$ (42)
with $g(u)=u/\sinh(u)$, $g^{\prime}(u)$ its derivative with respect to $u$,
and $P_{\mu}^{\pm}=(\bar{\sigma}_{\mu}\pm\hat{n}\bar{\sigma}_{\mu}\hat{n})/2$,
$\hat{n}=x_{i}\tau_{i}/r$. The antiperiodicity of the solution is encoded in
the time dependent quaternions $\bar{e}_{1}^{2}$ and $\bar{e}_{2}^{2}$ defined
through:
$e^{-i\pi x_{\mu}\bar{\sigma}_{\mu}}=i(e_{1}^{2}(x)+ie_{2}^{2}(x))\,.$ (43)
For non-trivial holonomy there is also an interesting limit in which the
caloron solution tends to the BPS monopole. It corresponds to making the
separation of the constituent monopoles tend to infinity
($\rho\rightarrow\infty$). Our adjoint zero-modes lead to those of the BPS
monopole if the appropriate limit is taken ($r_{1}<<\pi\rho^{2}$, $\rho>>1$).
For example, for the equal mass case ($m_{1}=m_{2}=\pi$) the first solution
Eq. (28) follows quite simply by applying the appropriate limit to Eq. (32).
Computing the density we obtain:
$2(h^{\prime 2}(\pi r)+1-2h^{\prime}(\pi r)\cos\theta)+g^{2}(\pi r)+g^{\prime
2}(\pi r)+2g(\pi r)g^{\prime}(\pi r)\cos\theta$ (44)
where $h^{\prime}(u)$ is the derivative of $h(u)\equiv u\coth(u)$. This
profile has axial symmetry depending explicitly on the azimuthal angle
$\theta$. Notice also that the solution is non-normalizable.
### 3.5 Comparison with numerical results
We have crosschecked our results with a direct evaluation of adjoint zero-
modes on the torus obtained by lattice methods using Neuberger’s overlap
operator [17] in the adjoint representation. One expects that the spatial
profile of the torus solutions approaches our analytical formulas as the box
size becomes much larger than all scales of the problem ( $\beta$ and
$\pi\rho^{2}$). To make a quantitative comparison we computed the zero-mode
density integrated in time along the line $x=y=0$ joining both constituent
monopoles. It is not possible a priori to construct numerical zero-modes with
a prescribed value of $\rho$ and $\delta_{1}$, although some tuning is
possible [18]. For the numerical comparison displayed in Fig. 2 we slightly
tuned by hand these parameters to improve the agreement ($\rho=0.79$,
$\delta_{1}=0.172$). A technical point which one has to address is how to
guarantee that the same linear combinations are selected for the numerical and
analytical data. We chose to define the two linearly independent modes by
imposing that at the center of mass ($x=y=z=0$) one has maximal and the other
minimal density (integrated over time).
Figure 2: Comparison between numerical (circles) and analytic (lines) zero
modes for $\rho=0.79$ and $\delta_{1}=0.172$. We display the density of the
zero modes, integrated in time, along the line joining the two monopoles.
## 4 Conclusions
In this paper we have obtained analytic formulas for the zero-modes of the
Dirac equation for gluinos in the background field of Q=1 SU(2) calorons with
antiperiodic boundary conditions in thermal-time. Our formulas are valid for
non-trivial holonomy as well as for the Harrington-Shepard caloron and include
as a limiting case those of BPS monopoles. The solutions have finite norm and
decay exponentially with distance if the masses of the constituent monopoles
differ. Their density profile contrasts with the case of periodic zero-modes.
For example, as the monopoles are pulled apart the density does not decouple
into independent lumps centered at the monopoles, but rather describes a
string joining the monopoles. Nonetheless, the number of normalizable zero-
modes matches in both cases.
Our work has methodological interest since our approach is applicable to other
cases including the extension to SU(N), and might be instrumental in finding
formulas for calorons of higher charge. From a physical viewpoint our work
provides a first step towards a semiclassical study of N=1 SUSY Yang-Mills at
finite temperature. There are interesting issues at stake such as that of
supersymmetry breaking at finite temperature, which has been a subject of
debate since early times [20]-[26]. It is our intention to address these
questions in future work.
## Acknowledgments
We would like to thank Falk Bruckmann for discussions at the initial stages of
this work. We acknowledge financial support from Comunidad Autónoma de Madrid
under the program HEPHACOS P-ESP-00346. A.S. is supported by an FPU fellowship
of Spanish Research Ministry (MEC). M.G.P. and A.G-A acknowledge financial
support from grants from CICYT FPA2006-05807, FPA2006-05485 and FPA2006-05423
. The authors participate in the Consolider-Ingenio 2010 CPAN (CSD2007-00042).
We acknowledge the use of the IFT cluster for part of our numerical results.
## References
* [1] B. J. Harrington and H. K. Shepard, Phys. Rev. D 17, 2122 (1978). Phys. Rev. D 18 (1978) 2990.
* [2] T. C. Kraan and P. van Baal, Nucl. Phys. B 533 (1998) 627 [arXiv:hep-th/9805168].
* [3] T. C. Kraan and P. van Baal, Phys. Lett. B 428 (1998) 268 [arXiv:hep-th/9802049]. T. C. Kraan and P. van Baal, Phys. Lett. B 435 (1998) 389 [arXiv:hep-th/9806034].
* [4] K. M. Lee, Phys. Lett. B 426 (1998) 323 [arXiv:hep-th/9802012]. K. M. Lee and C. h. Lu, Phys. Rev. D 58, 025011 (1998) [arXiv:hep-th/9802108].
* [5] P. Rossi, Phys. Rept. 86 (1982) 317.
* [6] B. Grossman, Phys. Lett. A 61 (1977) 86
* [7] N. Bilic, Phys. Lett. B 97 (1980) 107.
* [8] A. Gonzalez-Arroyo and Yu. A. Simonov, Nucl. Phys. B 460 (1996) 429 [arXiv:hep-th/9506032].
* [9] M. García Pérez, A. González-Arroyo, C. Pena and P. van Baal, Phys. Rev. D 60 (1999) 031901 [arXiv:hep-th/9905016].
* [10] M. N. Chernodub, T. C. Kraan and P. van Baal, Nucl. Phys. Proc. Suppl. 83 (2000) 556 [arXiv:hep-lat/9907001]. F. Bruckmann, D. Nogradi and P. van Baal, Nucl. Phys. B 666 (2003) 197 [arXiv:hep-th/0305063].
* [11] M. García Pérez and A. González-Arroyo, JHEP 0611 (2006) 091 [arXiv:hep-th/0609058].
* [12] A. González-Arroyo, Nucl. Phys. B 548 (1999) 626 [arXiv:hep-th/9811041].
* [13] A. González-Arroyo and C. Pena, JHEP 9809 (1998) 013 [arXiv:hep-th/9807172]. M. García Pérez, A. González-Arroyo, C. Pena and P. van Baal, Nucl. Phys. B 564 (2000) 159 [arXiv:hep-th/9905138].
* [14] F. Bruckmann and P. van Baal, Nucl. Phys. B 645, 105 (2002) [arXiv:hep-th/0209010]. F. Bruckmann, D. Nogradi and P. van Baal, Nucl. Phys. B 698 (2004) 233 [arXiv:hep-th/0404210].
* [15] M. F. Atiyah, N. J. Hitchin, V. G. Drinfeld and Y. I. Manin, Phys. Lett. A 65 (1978) 185.
* [16] W. Nahm, Phys. Lett. B 90 (1980) 413; “All Selfdual Multi - Monopoles For Arbitrary Gauge Groups,” CERN-TH-3172 Presented at Int. Summer Inst. on Theoretical Physics, Freiburg, West Germany, Aug 31 - Sep 11, 1981
* [17] H. Neuberger, Phys. Rev. Lett. 81 (1998) 4060 [arXiv:hep-lat/9806025]; Phys. Lett. B 427 (1998) 353 [arXiv:hep-lat/9801031]; Phys. Lett. B 417 (1998) 141 [arXiv:hep-lat/9707022].
* [18] M. García Pérez, A. González-Arroyo, A. Montero and P. van Baal, JHEP 9906 (1999) 001 [arXiv:hep-lat/9903022].
* [19] T. C. Kraan, Commun. Math. Phys. 212 (2000) 503 [arXiv:hep-th/9811179].
* [20] A. K. Das and M. Kaku, Phys. Rev. D 18 (1978) 4540.
* [21] L. Girardello, M. T. Grisaru and P. Salomonson, Nucl. Phys. B 178 (1981) 331.
* [22] L. Van Hove, Nucl. Phys. B 207 (1982) 15.
* [23] H. Aoyama and D. Boyanovsky, Phys. Rev. D 30 (1984) 1356.
* [24] D. Boyanovsky, Phys. Rev. D 29 (1984) 743.
* [25] H. Matsumoto, M. Nakahara, Y. Nakano and H. Umezawa, Phys. Lett. B 140 (1984) 53; Phys. Rev. D 29 (1984) 2838.
* [26] D. Buchholz and I. Ojima, Nucl. Phys. B 498 (1997) 228 [arXiv:hep-th/9701005].
| arxiv-papers | 2008-07-15T00:08:23 | 2024-09-04T02:48:56.788295 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Margarita Garcia Perez, Antonio Gonzalez-Arroyo and Alfonso Sastre",
"submitter": "Antonio Gonzalez-Arroyo",
"url": "https://arxiv.org/abs/0807.2285"
} |
0807.2322 |
On phase transition signal in inelastic collision
J.Manjavidze
JINR, Dubna, Russia111 On leave in absence: Andronikashvili Institute of
Physics, Tbilisi, Georgia
###### Abstract
The paper is devoted to retrieval of the first order phase transition signal
in the inelastic collisions. The primary intent is to show that the
experimentally observable signal exist iff the multiplicity is sufficiently
large. We discuss corresponding phenomenology from the point of view of
experiment.
I
The question of possibility to observe first order phase transition in the
hadron and heavy ion collisions is discussed widely at present time [1]. The
aim of paper is discuss the phenomenology of that problem.
The first order phase transition in statistics appears in the result of
creation of local critical fluctuations (e.g. bubbles of vapor if the
(liquid$\to$gas) transition is considered), in contrast to the second order
phase transition where the whole system undergo the transition. The dimension
of such fluctuations increase, if they are under-critical, and the whole
system in result undergo the transition.
It is hard to imagine that exactly such picture appears in the hadron or ion
inelastic collisions. The point is that in the event-by-event measurement the
dimension of bubbles, if they exist, may be both smaller or larger critical
one. Therefore, in the best case, we observe the mixture of two-phase medium
and the question how one may increase on experiment the weight of events with
under-critical bubbles is the first problem. It is evident that the density,
usually used in statistics, can not be introduced as the ”order parameter”
since only the result of particles $production$ process is observed.
In the paper [2] the ”chemical potential”, $\mu(n,s)$, was offered as the
”order parameter”. It is the work which is necessary for creation of one
particle and it was shown that if the role of under-critical babbles dominate
then $\mu(n,s)$ must decrease with number of produced particles $n$. In
another words, it was offered to define the boiling, i.e. creation of under-
critical bubbles, through $intensity$ of the process of particles production
(evaporation).
This is the main idea why the very high multiplicity (VHM) processes were
considered. It is not important what additional criterium is used searching
first order phase transition. In any case we must consider the VHM domain to
have intensive production of particles. In addition, the kinetic degrees of
freedom must be suppressed in the VHM region.
Therefore, the special attention will be given to the VHM processes. Then
exist approximation [2]:
$\mu(n,s)\simeq-\frac{T(n,s)}{n}\ln\sigma_{n}(s)$ (1)
in this multiplicity region. Here $T$ is the mean energy, including mass, of
produced particles, i.e. $T$ is associated with temperature, and $\sigma_{n}$
is the normalized to unite multiple production cross section which can be
considered in the VHM region as the ”partition function” of the $equilibrium$
system. The equilibrium condition will be defined later, see inequality (17),
definition (18) and [3] where the detailed explanation was given. See also the
footnote 5.
Continuing the analogy with thermodynamics one can say that
$(-T\ln\sigma_{n})/n$ is the Gibbs free energy per one particle. Then $\mu$
can be interpreted as the ”chemical potential” measured with help of
$observed$ free particles222Notice that one may consider $n$ as the
multiplicity in the experimentally observable range of phase space..
The definition (1) is quiet general. It can be used both for hadron-hadron and
ion-ion collisions, both for low and high energies, both for ”boiling” media
of coloured partons and colorless hadrons. Definition (1) is model free and
operates only with ”external” directly measurable parameters. The single
indispensable condition: we work in the VHM region.
We will discuss in the paper the chance of experimental measurement of
$\mu(n,s)$ defined by (1), what kind uncertainties hides it from
$experimental$ point of view noting the the cross sections in VHM domain are
small. The correction to (1) are not essential in the VHM region but
nevertheless the field-theoretical definition of chemical potential in using
Wigner functions formalism [4] will be published, see also [3].
II.
It is necessary to remind main steps toward (1) to understand hidden
phenomenological uncertainties. The starting point [2] was the generating
function
$\rho(z,s)=\sum_{n=0}^{\infty}z^{n}\sigma_{n}(s),~{}\rho(1,s)=1,~{}\sigma_{n}=0~{}{\rm
at}~{}n>n_{max}=\sqrt{s}/m,$ (2)
where $m$ is the hadron mass. One may use inverse Mellin transformation:
$\sigma_{n}(s)=\frac{1}{2\pi i}\oint\frac{dz}{z^{n+1}}\rho(z,s)$ (3)
to find $\sigma_{n}$ if $\rho(z,s)$ is known. One may calculate integral (3)
by saddle point method. The equation (of state):
$n=z\frac{\partial}{\partial z}\ln\rho(z,s)$ (4)
defines mostly essential value $z=z(n,s)$. Therefore, only
$z<z_{max}=z(n_{max},s)$ (5)
have the physical meaning.
One may write $\rho(z,s)$ in the form:
$\rho(z,s)=\exp\left\\{\sum_{l=0}^{\infty}z^{l}b_{l}(s)\right\\},$ (6)
where the ”Mayer group coefficient” $b_{l}$ can be expressed through
correlators $c_{k}(s)$:
$b_{l}(s)=\sum_{k=l}^{\infty}\frac{(-1)^{(k-l)}}{l!(k-l)!}c_{k}(s).$
Let us assume now that we have Poisson distribution, i.e. if in the sum:
$\ln\rho_{(}z,s)=\sum_{k}\frac{(z-1)^{k}}{k!}c_{k}(s)$ (7)
one may leave first term, then it is easy to see that
$z(n,s)=n/c_{1}(s),~{}c_{1}(s)\equiv\bar{n}(s),$ (8)
are essential and in the VHM region:
$\ln\sigma_{n}(s)=-n\ln\frac{n}{c_{1}(s)}(1+O(1/\ln n))=-n\ln z(n,s)(1+O(1/\ln
n)).$ (9)
Therefore, in considered case with $c_{k}=0,~{}k>1,$ exist following
asymptotic estimation for $n>>1$:
$\ln\sigma_{n}\simeq-n\ln z(n,s),$ (10)
i.e. $\sigma_{n}$ is defined in VHM region mainly by the solution of Eq.(4)
and the correction can not change this conclusion. It will be shown that that
kind estimation is hold for arbitrary asymptotics of $\sigma_{n}$. The
definition (1) based on this observation.
If we understand $\sigma_{n}$ as the ”partition function” in the VHM region
then $z$ is the $activity$ usually introduced in statistical physics.
Correspondingly the chemical potential $\mu$ is defined trough $z$:
$\mu=T\ln z.$ (11)
Combining this definition with estimation (10) we define $\sigma_{n}$ through
$\mu$. But, $if$ this estimation does not depend from asymptotics of
$\sigma_{n}$ over $n$, i.e. if it has general meaning, then it can be used for
definition of $\mu(n,s)$ through $\sigma_{n}(s)$ and $T(n,s)$ at $n>>1$. Just
this formal idea is realized in (1): it can be shown in Sec.III that (1) is
correct at the asymptotical value of $n$.
III.
Now we will make the important step. To put in a good order our intuition it
is useful to consider $\rho(z,s)$ as the $nontrivial$ function of $z$. In
statistical physics the thermodynamical limit is considered for this purpose.
In our case the finiteness of energy $\sqrt{s}$ and of the hadron mass $m$ put
obstacles on this way since the system of produced particles necessarily
belongs to the energy-momentum surface333It must be noted that the canonical
thermodynamic system belongs to the energy-momentum shell because of the
energy exchange, i.e. interaction, with thermostat. The width of the shell is
defined by the temperature. But in particle physics there is no thermostat and
the physical system completely belongs to the energy momentum surface.. But we
can continue $theoretically$ $\sigma_{n}$ to the range $n>n_{max}$ and
consider $\rho(z,s)$ as the nontrivial function of $z$. This step hides the
assumption that nothing new appear at $n>n_{max}$, i.e. the VHM interval
$\bar{n}<<n<n_{max}$ is sufficiently wide to represent main physical
processes.
Let us consider the analog generating function which has the first $n<n_{max}$
coefficient of expansion over $z$ equal to $\sigma_{n}$ and higher
coefficients for $n\geq n_{max}$ are deduced from continuation of theoretical
value of $\sigma_{n}$ to $n\geq n_{max}$. Then the inverse Mellin
transformation (3) gives a good estimation of $\sigma_{n}$ through this
generating function if the fluctuations near $z(n,s)$ are Gaussian or, it is
the same, if
$\left.\frac{|2n-z^{3}\partial^{3}\ln\rho(z,s)/\partial
z^{3}|}{|n+z^{2}\partial^{2}\ln\rho(z,s)/\partial
z^{2}|^{3/2}}\right|_{z=z(n,s)}<<1.$ (12)
Notice that if the estimation (10) is generally rightful then one can easily
find that l.h.s. of (12) is $\sim 1/n^{1/2}$. Therefore, one may consider
$\rho(z,s)$ as the nontrivial function of $z$ considering $z(n,s)<z_{max}$ if
$\bar{n}<<n<n_{max}$.
Then it is easily deduce that the asymptotics of $\sigma_{n}(s)$ is defined by
the leftmost singularity, $z_{c}$, of function $\rho(z,s)$ since, as it
follows from Eq.(4), the singularity ”attracts” the solution $z(n,s)$ in the
VHM region. In result we may classify asymptotics of $\sigma_{n}$ in the VHM
region if (12) is hold.
Our problem is reduced to the definition of possible location of leftmost
singularity of $\rho(z,s)$ over $z>0$444The singularities in complex $z$ plane
will not be considered since they lead only to oscillations in multiplicity
distribution.. It must be stressed that the character of singularity is not
important for definition of $\mu(n,s)$ in the VHM region at least with
$O(1/\ln n)$ accuracy. One may consider only three possibility at
$n\to\infty$: (I) $z(n,s)\to z_{c}=1$; (II) $z(n,s)\to
z_{c},~{}~{}1<z_{c}<\infty$; (III) $z(n,s)\to z_{c}=\infty$. The structure of
complex $z$ plane is much more complicate but for our purpose the above
described picture is sufficient.
Correspondingly one may consider only three type of asymptotics in the VHM
region: (I) $\sigma_{n}>O(e^{-n})$; (II) $\sigma_{n}=O(e^{-n})$. Such
asymptotics is typical for hard processes with large transverse momenta, like
for jets [3]; (III) $\sigma_{n}<O(e^{-n})$. That asymptotic behavior is
typical for multiperipheral-like kinematics, where the longitudinal momenta of
produced particles are noticeably higher than the transverse ones [3].
Therefore the case (I) is the best candidate for phase transition since in
this case the cross sections are comparatively large in the VHM region, i.e.
particles ”intensively” produced in that case. Notice that if (I) is not
realized in nature then the (II) kind processes would dominate in the VHM
region.
Let us consider now the estimation (1). It follows from (3) that, up to the
preexponential factor,
$\ln\sigma(n,s)\approx-n\ln z(n,s)+\ln\rho(z(n,s),s).$ (13)
We want to show that, in a vide range of $n$ from VHM region,
$n\ln z(n,s)\sim\ln\rho(z(n,s),s).$ (14)
Let as consider now the mostly characteristic examples.
(I) Singularity at $z=1$. The physical meaning of singularity at $z=1$ may be
illustrated by the droplet model [6]. The Mayer’s group coefficient, see (6),
for cluster from $l$ particle is $b_{l}(\beta)\sim\exp\\{-\beta\tau
l^{(d-1)/d}\\},$ where $\tau l^{(d-1)/d}$, $l>>1$, is the surface tension
energy, $d$ is the dimension. Therefore, if $d>1$ the series over $l$ in (6)
diverges at $z=1$.
This case was considered in [2] in details. In the used lattice gas
approximation $\ln{z}(n)\sim n^{-5}$ and $\ln\sigma_{n}\approx-n^{-4}=-n\ln
z(n)(1+O(1/n)).$ Notice that the simplest droplet model predicts unphysical
asymptotics: $\sigma_{n}\to const$ in the VHM region.
(II) Singularity at $1<z_{c}<\infty$. Let us consider one jet contribution:
$\ln\rho(z,s)=-\gamma\ln(1-\bar{n}_{j}(s)(z-1))$. In this case
$z(n,s)=z_{c}(1-\gamma/n),~{}n>>\gamma,$ and $\ln\sigma_{n}=-n\ln
z(n,s)(1+O(\ln n/n)).$
(III) Singularity at $z=\infty$. For $k$ Pomeron exchange:
$\ln\rho(z,s)=c_{k}(s)(z-1)^{k}$. In this case $z(n)=(n/kc_{k})^{1/k}>>1$ and
$\ln\sigma_{n}\approx-n\ln z(n)(1+O(1/\ln n)).$
One can conclude:
(i) The definition (1) in the VHM region is rightful since the correction
falls down with $n$. On this stage we can give only the qualitative estimation
of corrections. Nevertheless (1) gives the correct $n$ dependence in the VHM
region.
(ii) Activity $z(n,s)$ tends to $z_{c}$ from the right in the case (I) and
from the left if we have the case (II) or (III).
(iii) The accuracy of estimation of the chemical potential (1) increase from
(III) to (I).
IV.
The temperature $T$ is the next problem. The temperature is introduced usually
using Kubo-Martin-Schwinger (KMS) periodic boundary conditions. But this way
assumes from the very beginning that the system (a) is equilibrium and (b) is
surrounded by thermostat through which the temperature is determined. The
first condition (a) we take as the simplification which gives the equilibrium
state.
The second one (b) is the problem since there is no thermostat in particle
physics. For this reason we introduce the temperature as the Lagrange
multiplier $\beta=1/T$ of energy conservation law [3]. In such approach the
condition that the system is in equilibrium with thermostat replaced by the
condition that the fluctuations in vicinity of $\beta$ are Gaussian.
The interesting for us $\rho(z,{s})$ we define through inverse Laplace
transform of $\rho(z,\beta)$:
$\rho(z,s)=\int\frac{d\beta}{2\pi i\sqrt{s}}e^{\beta\sqrt{s}}\rho(z,\beta).$
(15)
It is known that if the interaction radii is finite, i.e. the hadron mass is
finite, then the equation (of state):
$\sqrt{s}=-\frac{\partial}{\partial\beta}\rho(z,\beta)$ (16)
have real positive solution $\beta(n,s)$ at $z=z(n,s)$. We will assume that
the fluctuations near $\beta(n,{s})$ are Gaussian. This means that the
inequality [3]:
$\left.\frac{|\partial^{3}\ln\rho(z,\beta)/\partial\beta^{3}|}{|\partial^{2}\ln\rho(z,\beta)/\partial\beta^{2}|^{3/2}}\right|_{z=z(n,s),\beta=\beta(n,s)}<<1$
(17)
is satisfied. Therefore, we prepare the formalism to find ”thermodynamic”
description of the processes of particle production assuming that this
$S$-matrix condition of equilibrium (17) is hold555Introduction of
$\beta(n,s)$ allows to describe the system of large number of degrees of
freedom in terms of single parameter $\beta(n,s)$, i.e. it is nothing but the
useful trick. It is no way for this reason to identify entirely $1/\beta(n,s)$
with thermodynamic temperature where it has the self-contained physical sense.
Nevertheless path-integral representation of $\rho(\beta,z)$ defined from
$S$-matrix coincides with Feynman-Kac representation of grand partition
function [3] if (17) is hold. It must be noted also that the energy spectrum
of produced particle in this case have Boltzmann form, $e^{-\beta\eta}$..
I want to underline that our thermal equilibrium condition (17) have absolute
meaning: if it is not satisfied then $\beta(n,s)$ loses every sense since the
expansion in vicinity of $\beta(n,s)$ leads to the asymptotic series. In this
case only the dynamical description of $S$-matrix can be used.
It is not hard to see [3] that
$\frac{\partial^{l}}{\partial\beta^{l}}\ln
R(z,\beta)|_{z=z(n,s),\beta=\beta(n,s)}=<\prod_{i=1}^{l}(\eta_{i}-<\eta>)>_{n,s}$
(18)
is the $l$-point energy correlator, where $<...>_{n,s}$ means averaging over
all events with given multiplicity and energy. Therefore (17) means
”relaxation of $l$-point correlations”, $l>2$, measured in units of the
dispersion of energy fluctuations, $l=2$. One can note here the difference of
our definition of thermal equilibrium from thermodynamical one [5].
V.
We may conclude that:
(i) The definition of chemical potential (1) was discussed. This important
observable can be measured on the experiment directly. Chemical potential,
$\mu(n,s)$, must decrease in the VHM region if the first order phase
transition occur, case (I), and it rise in opposite case, see (II) and (III),
see Sec.III.
(ii) We are forced to assume that the energy and the multiplicity are
sufficiently large, i.e. the experimental value $z^{exp}(n,s)$ is sufficiently
close to $z_{c}=1$. In opposite case the leading leftmost singularity over $z$
would not be ”seen” on experiment and the production processes constitutes
from the complicated mixture of subprocesses.
(iii) The cross section $\sigma_{n}$ falls down rapidly with $n$ and for this
reason the VHM events are hardly observable. One may avoid this problem
considering the finite energy heavy ion collisions as the most candidates of
processes described by methods of thermodynamics and $z_{c}$ is easier
”reachable” in this case.
(iv) One can define $z(n,s)$ also directly from Eq.(4):
$n=z\frac{\partial}{\partial z}\ln\sum_{n}z^{n}\sigma_{n}^{exp}(s),$ (19)
using experimental values $\sigma_{n}^{exp}(s)$. But comparing (19) with
definition (10),
$n\ln z\simeq-\ln\sigma_{n}^{exp}(s),$ (20)
it seems that last one gives more definite value of $z^{exp}(n,s)$ than the
”integral” equation (19) especially since the statistical errors are large in
the VHM region and the theoretical correction to Eq.(20) are small, $\sim
1/n$.
Summarizing the results we conclude: if the energy is sufficiently large, i.e.
if $z_{max}$ is sufficiently close to $z_{c}=1$, if the multiplicity is
sufficiently large, so that (17) is satisfied and $z(n,s)$ can be sufficiently
close to $z_{c}$, then one may have confident answer on the question:
observable or not the first order phase transition in hadron/ion collisions.
The heavy ion collisions are favorable to observe the phase transition.
Acknowledgements.
I would like to thank participants of 7-th International Workshop on the ”Very
High Multiplicity Physics” (JINR, Dubna) for stimulating discussions. I am
grateful to V.Priezzhev, A.Sissakian, A.Sorin and V.Kekelidze for valuable
attention.
## References
* [1] BNL Report, Hunting the Quark Gluon Plasma, BNL-73847-2005; C.Alt et al., The NA-49 Collaboration, nucl-ex/0710.0118; M.Creutz, Phys. Rev., D15 (1977) 1128; M.Gazdzicki and M.I.Gorenstein, Acta Physica Polonica, B 30, 2705, (1999)
* [2] J.Manjavidze and A.Sissakian, Proc. VHM Physics Workshops, (World Scient., 2008)
* [3] J.Manjavidze and A.Sissakian, Phys. Rep., 346 (2001) 1, hep-ph/0105245
* [4] J.Manjavidze, Phys.Part.Nucl. 30 (1999) 49, hep-ph/9802318
* [5] N.N.Bogolyubov, Studies in Statistical Mechanics, (North-Holland Publ. Co., Amsterdam, 1962)
* [6] T.D.Lee and C.N.Yang, Phys.Rev., 87 (1952) 404, 410; J.S.Langer, Ann.Phys., 41 (1967) 108
| arxiv-papers | 2008-07-15T08:56:34 | 2024-09-04T02:48:56.793305 | {
"license": "Public Domain",
"authors": "J. Manjavidze",
"submitter": "Joseph Manjavidze",
"url": "https://arxiv.org/abs/0807.2322"
} |
0807.2407 |
# First Result of Net-Charge Jet-Correlations from STAR
Quan Wang (for the STAR Collaboration) wang187@purdue.edu Department of
Physics, Purdue University, West Lafayette, IN 47907, USA
###### Abstract
We presented results on azimuthal correlation of net-charge with high $p_{T}$
trigger particles. It is found that the net-charge correlation shape is
similar to that of total-charge. On the near-side, the net-charge and total-
charge $p_{T}$ spectra have similar shape and both are harder than the
inclusives. On the away-side, the correlated spectra are not much harder than
the inclusives, and the net-charge/total-charge ratio increases with $p_{T}$
and is similar to the inclusive ratio.
## 1 Introduction
Two novel phenomena have been observed at intermediate transverse momentum
($2<p_{T}<5$GeV/$c$) in heavy-ion collisions at RHIC. First, the azimuthal di-
hadron correlation with a high $p_{T}$ ($p_{T}>3$ GeV/$c$) trigger particle is
significantly broadened on the away-side (double-peak structure) in central
Au+Au collisions relative to p+p collisions bib:broaden . Second, the bayon to
meson ratio is significantly enhanced bib:star bib:ratio . In order to
understand the physics mechanism(s) for these phenomena, we study two-particle
azimuthal correlation of net-charge and compare to that of total-charge. Since
the net-charge is dominated by net-protons, and the total-charge is the sum of
mesons and bayons, the comparison between net-charge and total-charge
azimuthal correlations may help further our understanding of baryon-meson
effects at RHIC.
## 2 Analysis
We obtain the raw azimuthal correlations from positively ($h^{+}$) and
negatively ($h^{-}$) charged hadrons with a charged high $p_{T}$ trigger
particle ($h$). The trigger and associated particle $p_{T}$ ranges are
$3<p_{T}<4$ GeV/$c$ and $1<p_{T}<3$ GeV/$c$, respectively. Both the trigger
and associate particles are restricted to $|\eta|<1$. Tracking efficiency is
corrected for associated particles. The two-particle acceptance effect is
corrected for by the event-mixing technique. The correlations are normalized
per trigger particle. By taking the difference and the sum, we obtain the net-
charge ($h\mbox{-}\Delta Q$) and total-charge ($h\mbox{-}Q$) correlations:
$h\mbox{-}\Delta Q=h\mbox{-}h^{+}-h\mbox{-}h^{-},$ (1)
$h\mbox{-}Q=h\mbox{-}h^{+}+h\mbox{-}h^{-}.$ (2)
We obtain the corresponding backgrounds by using the same technique from
mixed-events and adding in $v_{2}$ modulation, where $v_{2}$ is presently
taken from the charged hadron measurements bib:rpv2 bib:4pv2 . The final
correlation signal is given by
$signal=raw-a*background$ (3)
where $a$ is the normalization factor from Zero Yield At 1 radian (ZYA1)
method. The systematic uncertainty estimate is done according to Ref.
bib:broaden .
Figure 1: Background subtracted azimuthal correlations of net-charge (data
points) and total-charge (solid histogram) in 12% central Au+Au collisions at
200 GeV. The trigger and associated $p_{T}$ ranges are 3-4 GeV/$c$ and 1-3
GeV/$c$, respectively. Errors shown for the data points are statistical only.
The total-charge result is scaled by 1/15 to compare with the net-charge
result. The dashed histograms show the systematic errors on the total-charge
result, obtained from background subtraction using the modified reaction-plane
$v_{2}$ bib:rpv2 and the 4-particle cumulant $v_{2}$ bib:4pv2 .
## 3 Results and Discussion
Figure 1 shows the background subtracted results. We find the correlation
shapes are similar between total-charge and net-charge. The net-charge
correlation shows a significant away-side broadening (double-peak structure),
similar to previously observed in total-charge bib:broaden . We study the
$p_{T}$ dependence of the correlated yields by dividing the data into four
associated $p_{T}$ bins. In different $p_{T}$ bins, the correlation shapes are
also found to be similar between total-charge and net-charge.
Figure 2: $p_{T}$ spectra of net-charge (triangles) and total-charge
(squares) correlated with a high $p_{T}$ trigger particle ($3<p_{T}<4$
GeV/$c$) on both near-side (filled symbols) and away-side (open symbols) in
AuAu central collisions at 200 GeV. Errors shown are systematic errors. The
inclusive (round points) $p_{T}$ spectrum is also shown for comparison.
We obtain associated particle $p_{T}$ spectra by integrating over the near-
side ($|\Delta\phi|<1$) and the away-side ($|\Delta\phi|>1$). The obtained
spectra are shown in Fig. 2. For comparison, we also show the inclusive
charged hadron $p_{T}$ spectrum (i.e. without requirement of a high $p_{T}$
trigger particle). We fit the spectra by an exponential function:
$e^{-p_{T}/T}$, where $T$ is the inverse slope parameter. The fit results are
listed in Table 1.
Table 1: Inverse slope parameter $T$ in GeV/$c$ from exponential fit to the $p_{T}$ spectra. Errors are statistical only. | Near-side | Away-side
---|---|---
net-charge | $0.491\pm 0.069$ | $0.378\pm 0.032$
total-charge | $0.459\pm 0.003$ | $0.336\pm 0.002$
inclusive | $0.312\pm 0.001$
As seen from the extracted values of the inverse slope parameter, the shapes
of net-charge and total-charge correlated spectra on near-side are similar,
and they are both harder than the inclusive spectrum. The away-side correlated
net-charge spectrum seems to be harder than the total-charge, and neither is
much harder than the inclusive one.
Figure 3: Ratios of correlated yields of net-charge over total-charge on
away-side (squares) and near-side (triangles) in 12% central Au+Au collisions
at 200 GeV. The trigger $p_{T}$ range is from 3-4 GeV/$c$. Vertical bars are
statistical errors. Horizontal bars are the $p_{T}$ bin size. The curve shows
the inclusive ratio of
$\frac{p-\overline{p}}{\pi^{+}+\pi^{-}+p+\overline{p}}$, where the shaded area
is statitical uncertainty (the data at $p_{T}<$2.5 GeV/$c$ are from TOF
measurement, and the $p_{T}>$2.5 GeV/$c$ data are from TPC measurement)
bib:star .
The inclusive bayon to meson ratio is strongly enhanced at intermediate
$p_{T}$ as mentioned earlier. We want to study the bayon/meson ratio of the
correlated hadrons with high $p_{T}$ trigger particles. We take the ratio of
the $p_{T}$ spectra on near- and away-side separately. The obtained results
are shown in Fig. 3, and compared with the inclusive ratio. The ratio of the
correlated net-charge to total-charge yields on away-side is larger than that
on the near-side. The away-side ratio is similar to the inclusive ratio, with
an average difference of $0.34\sigma$. The near-side ratio seems smaller than
the inclusive ratio with an average difference of $0.80\sigma$.
Figure 3 shows that the near-side ratio is independent of $p_{T}$. A fit of
the near-side ratio to a constant yields $0.037\pm 0.005(stat.)$, with a
$\chi^{2}/ndf=1.7/3$. This is to be compared to the inclusive ratio of
$0.050\pm 0.001$. On the other hand, the away-side ratio appears to increase
with $p_{T}$, similar to the inclusives. A constant fit yields $0.055\pm
0.004(stat.)$, with a significantly larger $\chi^{2}/ndf=5.4/3$.
## 4 Summary
We have presented results on azimuthal correlation of net-charge with high
$p_{T}$ trigger particles. The net-charge correlation shape is found to be
similar to that of total-charge. We studied the associate $p_{T}$ spectra of
correlated net-charge, and the net-charge to total-charge ratio. On the near-
side, net-charge and total-charge spectra have similar shape; the net-
charge/total-charge ratio is constant over $p_{T}$. However, they are both
harder than the inclusives, suggesting jet nature for the correlated
particles. On the away-side, the correlated spectra are not much harder than
the inclusives, suggesting partial equilibration of the correlated hadrons
with the medium. The net-charge/total-charge ratio increases with $p_{T}$ and
is similar to inclusives.
## References
* (1) J. Adams _et al._ , STAR Collaboration, Phys. Rev. Lett. 95, 152301 (2005).
* (2) B.I. Abelev _et al._ , STAR collaboration, Phys. Rev. Lett. 97, 152301 (2006).
* (3) B.I. Abelev _et al._ , STAR collaboration, Phys. Lett. B 655, 104 (2007).
* (4) J. Adams et al. STAR Collaboration, Phys. Rev. C 72, 014904 (2005).
* (5) C. Adler et al. STAR Collaboration, Phys. Rev. C 66, 034904 (2002).
| arxiv-papers | 2008-07-15T16:04:54 | 2024-09-04T02:48:56.798410 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Quan Wang",
"submitter": "Quan Wang",
"url": "https://arxiv.org/abs/0807.2407"
} |
0807.2422 | # Frustrated total internal reflection and the illusion of superluminal
propagation
Vera L. Brudny vera@df.uba.ar Departamento de Física, Facultad de Ciencias
Exactas y Naturales, Universidad de Buenos Aires, Cd. Universitaria, Pabellón
1, C1428 EHA Buenos Aires, Argentina W. Luis Mochán mochan@fis.unam.mx
Instituto de Ciencias Físicas, Universidad Nacional Autónoma de México,
Apartado Postal 48-3, 62251 Cuernavaca, Morelos, México
###### Abstract
We analyze the propagation of a pulse across a vacuum gap separating opposite
flat parallel faces of two transparent dielectrics by means of an explicitly
causal and retarded propagator constructed directly from the free-space wave
equation. Nevertheless, our approach yields apparently superluminal
propagation for the case of frustrated total internal reflection (FTIR), that
is, a transmitted wave packet appears on the far side of the gap at the same
time that the corresponding incident packet crosses the front one. Thus, in
this example superluminality is just an illusion, being consistent with both
casuality and classical electrodynamics. We study the origin of the apparent
superluminality in this case, which is inherent to light pulse propagation in
free space and does not depend on the particulars of light-matter interaction,
and find that it is due to propagation from the lateral wings of the incident
pulse to the central part of the transmitted pulse. Thus, notwithstanding
their similarities, FTIR is not equivalent to 1D tunneling. We propose
experiments to test our explanation of superluminality using opaque screens to
block part of the wavefront, although we demonstrate that the propagation of
smooth finite pulses constrained to be made up completely of evanescent
Fourier components is indistinguishable from truly superluminal propagation,
i.e., it may be completely accounted for using an explicitely superluminal and
acausal propagator as well as the causal subluminal one.
###### pacs:
42.25.Bs, 41.20.Jb, 03.65.Xp, 42.25.Gy
## I Introduction
The physics of light propagation is a topic of active research, due to its
relevance to several technological applications and basic research. Recent
research on new materials has shown that it is possible to exercise an
extraordinary control on the propagation of light pulses, which in turn has
generated new interest in both practical and fundamental questions on light
propagation. Moreover, despite the fact that Maxwell completed the formulation
of the classical theory of electromagnetism in 1864 and Einstein’s special
theory of relativity was presented in 1905, some hot controversies continue to
arise on the subject of superluminal propagation Wang et al. (2007); Winful
(2007); Pereyra et al. (2007); Ranfagni et al. (2007); Winful (2005, 2003);
Buttiker et al. (2003); Windul (2003-1); Buttiker et al. (2003); Winful
(2003-2) and its possible implications for both classical and quantum
information theory Peres et al. (2004); De Angelis et al (2007). Recent
reviews on the subject can be found in Refs. Winful (2006); Boyd et al. (2002)
One consequence of the special theory of relativity is that no signal can
cause an effect outside the light cone of its source. Violation of this
principle of relativistic causality leads to paradoxes such as that of an
effect preceding its cause Azbel’ (1994); Garrison et al (1998). When dealing
with light propagation in a material characterized by a given dispersion
relation $\omega(\vec{k})$ between the frequency $\omega$ and the wave vector
$\vec{k}$, several velocities may be defined, such as the phase velocity
$v_{\phi}=\omega/k$ and the group velocity $v_{g}=\nabla_{\vec{k}}\omega$. It
is recognized that under certain conditions Jackson (Jackson) both velocities
can exceed the speed of light in vacuum $c$. This does not contradict the
postulates of the special theory of relativity, for it has been recognized
since the works of Sommerfeld Sommerfeld (1914) and Brillouin Brillouin1960
(Brillouin) that in order not to violate the principle of causality it is the
information velocity that must not exceed $c$. The question that arises then
is what is an appropriate and operative definition of information velocity.
There have been several discussions and proposals on this subject, but the
question is not yet settled Diener (1996, 1997); Kuzmich et al. (2000);
Stenner et al. (2003, 2005); Ranfagni et al. (2006).
The vast majority of the published work concerning superluminal pulse
propagation deals with light propagation in material media, and the usual
analysis attempts to explain how the interaction of electromagnetic radiation
with the medium affects the propagation of light pulses in such a way that
they appear to travel superluminally. In a pioneering paper Icsevgi et al.
(1969), Icsevgi and Lamb performed a theoretical investigation of the
propagation of intense laser pulses through a laser amplifier. Apparent
superluminal light propagation has been reported in gain-assisted systems wang
et al. (2000); Janowicz et al. (2006); Huang et al. (2008) as well as in
birefringent crystals solli et al. (2003); Brunner et al. (2004); halvorsen et
al. (2008), composite media and photonic crystals Kulkarni et al. (2004);
safian et al. (2006) and dispersive media Bigelow et al. (2006); Talukder et
al. (2005).
There have also been claims of evidence of superluminal propagation in free
space Mugnai et al. (2000, 2005) and during optical tunneling in frustrated
total internal reflection (FTIR) configurations Carey et al. (2000); Mochan
and Brudny (2001); Carey et al. (2001); Shaarawi et al. (2002). Optical
tunneling has been studied by several authors, both theoretically and
experimentally Brudny and Mochan (2001); Barbero et al. (2000); Reiten et al.
(2001); Balcou et al. (1997); Resch et al. (2001). Some of the observed
results are still subject of debatable interpretation Winful (2006) and do not
close the subject of whether there are possibilities for superluminal
transmission of information in such systems. We have therefore chosen to
address this subject in way that leads to straightforward interpretation of
the results while resorting only to classical electromagnetic theory. We claim
that although some results may appear to indicate superluminal propagation,
there is no real superluminal transfer of information.
In this paper we analyze mathematically the propagation of a pulse across a
vacuum gap separating opposite flat parallel faces of two transparent
dielectrics by means of an explicitly causal and retarded propagator
constructed directly from the free-space wave equation. Our results yield
indeed an apparent superluminal propagation corresponding to the conditions of
FTIR, but they show explicitly that it is consistent with both casuality and
with classical electrodynamics. Our example shows superluminality effects
inherent to light pulse propagation in free space which therefore does not
depend on the particulars of light-matter interaction. The illusion of
superluminality consists of transmitted pulses arriving to the far side of the
gap in synchrony with the crossing of the front surface by the incident pulse.
We explain this illusion of superluminal behavior in terms of a causal,
subluminal propagation, taking into account the spatial extent of the incident
pulse along its transverse as well as its longitudinal directions and we
propose experiments to demonstrate the retarded nature of propagation in FTIR.
Nevertheless, we find that for constrained pulses fully made up of evanescent
Fourier components, subluminal and superluminal propagation in FTIR
experiments are indistinguishable.
This paper is organized as follows. We first introduce a propagator that will
allow the analyzis of electromagnetic pulse propagation across a vacuum gap
(Section II). This propagator is both casual and retarded and complies with
the classical electromagnetic theory. We then study the propagation across the
gap of pulses that arrive as plane waves with well defined angles of
incidence. In Section III we study the case of subcritical angles, yielding
non-evanescent transmitted waves. In Section IV we study the case of
hypercritical angles, yielding evanescent waves. We conclude that the
propagation of a light pulse in a FTIR configuration may appear superluminal
and acausal but that it is actually subluminal and that propagation has to
account necessarily for the lateral wings of the incident pulse. In Section V
we suggest experiments that might demonstrate the actual causal and subluminal
nature of the apparent superluminal behavior by using sharp opaque screens
that block parts of the incident wave so that its extent becomes finite along
both its propagation and its transverse directions. Nevertheless, as the
borders of these screens produce propagating diffracted waves, in Section VI
we eliminate them and we study incident pulses that are finite along several
spatial directions but that have a smooth profile. We obtain that if they are
comprised of evanescent Fourier components only, they appear to propagate
superluminally through the vacuum gap, even though their behavior is
determined by our causal retarded propagator. In Section VII we construct an
alternative acausal, superluminal propagator, and prove that it is exactly
equivalent to the causal and retarded propagator when applied to fully
evanescent finite pulses. Thus, for such constrained pulses, it is impossible
to distinguish causal subluminal from acausal and superluminal propagation;
the illusion of superluminality appears to be not only a matter of
interpretation of the result of the propagation, but may be also present in
the description of the propagation process itself. We present our conclusions
in Section VIII.
## II Propagator
Consider two transparent dielectrics occupying the regions $z\leq 0$ and
$z\geq d$. In this section we obtain the causal and retarded propagator that
describes the motion of a pulse across a vacuum gap $0<z<d$ spanning from the
interface at $z=0$ to that at $z=d$ (Fig. 1).
Figure 1: Pulse with arbitrary profile incident at an angle
$\theta_{i}<\theta_{c}$ onto the surface $z=0$ of a dielectric with index of
refraction $n=2$. The pulse is transmitted into a vacuum gap and into a second
dielectric at $z=d$. The speed of propagation within each media, $c/n$ and
$c$, are indicated, as well as the speed of propagation of the wavefronts
along the interfaces $v_{\parallel}$.
As we want to consider explicitly the angle of incidence onto the interface
$z=0$, we cannot treat the problem beforehand as if it were 1D. For
simplicity, we will assume full translational symmetry along the $y$
direction, so that our problem becomes 2D. Thus, we start with the scalar wave
equation
$(\nabla^{2}-\frac{1}{c^{2}}\frac{\partial^{2}}{\partial
t^{2}})G_{0}(t,x,z;t^{\prime},x^{\prime},z^{\prime})=\delta(t-t^{\prime})\delta(x-x^{\prime})\delta(z-z^{\prime})$
(1)
with a unit singular point source fired at time $t^{\prime}$ at position
$(x^{\prime},z^{\prime})$, which is solved by the explicitly causal and
retarded free space Green’s function Morse and Feshbach (alguno)
$G_{0}(t,x,z;t^{\prime},x^{\prime},z^{\prime})=\frac{c}{2\pi}\frac{\Theta\left(c(t-t^{\prime})-\sqrt{(x-x^{\prime})^{2}+(z-z^{\prime})^{2}}\right)}{\sqrt{c^{2}(t-t^{\prime})^{2}-(x-x^{\prime})^{2}-(z-z^{\prime})^{2}}},$
(2)
where $\Theta(\ldots)$ is the Heaviside unit step function. Using image theory
we can construct a Green’s function that obeys Dirichlet boundary conditions
at the surface $z=0$,
$G(t,x,z;t^{\prime},x^{\prime},z^{\prime})=G_{0}(t,x,z;t^{\prime},x^{\prime},z^{\prime})-G_{0}(t,x,z;t^{\prime},x^{\prime},-z^{\prime}).$
(3)
In the half-space $z>0$ Green’s theorem yields the solution Morse and Feshbach
(alguno)
$\phi(t,x,z)=\int dx^{\prime}\int
dt^{\prime}\,P(t,x,z;t^{\prime},x^{\prime},0^{+})\phi(t^{\prime},x^{\prime},0^{0})$
(4)
of the homogeneous scalar wave equation that is outgoing as $z\to\infty$ and
is null in the remote past, where $\phi(t^{\prime},x^{\prime},0^{+})$ denotes
its previous values on the boundary $z=0^{+}$, and
$P(t,x,z;t^{\prime},x^{\prime},0^{+})=\frac{\partial}{\partial
z^{\prime}}\left.G(t,x,z;t^{\prime},x^{\prime},z^{\prime})\right|_{z^{\prime}=0^{+}}$
(5)
is the causal retarded propagator of the problem. We may identify the field
$\phi$ with the component $E_{y}$ of the electric field $\vec{E}$ in the case
of a TE or $s$ polarized incoming wave, and with the component $B_{y}$ of the
magnetic field $\vec{B}$ in the case of TM or $p$ polarization. The propagator
(5) does not account for the presence of the two dielectrics bounding the air
gap. Thus, the field (4) has no information about the multiple reflections at
the boundary of the gap. In principle, these can be incorporated by reflecting
the field at the interfaces $z=0,d$ using the appropriate Fresnel coefficient
and propagating it back and forth across the gap with the propagator (5) for
the $z=0$ surface and a similar one for the $z=d$ surface. The total field
would then be the sum of all the multiply reflected fields and would have
information about the electromagnetic properties of the reflecting surfaces.
In this paper we will restrict ourselves to an analysis of the first crossing
of the air gap $0^{+}\to d^{-}$, and thus our results will be unrelated to the
nature of the bounding media.
Substituting Eq. (2) into (3) and (5) we obtain
$P(t,x,z;t^{\prime},x^{\prime},0^{+})=-\frac{c}{\pi}\frac{\partial}{\partial
z}\frac{\Theta\left(c(t-t^{\prime})-\sqrt{(x-x^{\prime})^{2}+z^{2}}\right)}{\sqrt{c^{2}(t-t^{\prime})^{2}-(x-x^{\prime})^{2}-z^{2}}}.$
(6)
The field $\phi$ can then be written in terms of an ancillary function
$\phi(t,x,z)=-\frac{\partial}{\partial z}\psi(t,x,z),$ (7)
where
$\psi(t,x,z)=\frac{c}{\pi}\int dx^{\prime}\int
dt^{\prime}\,\frac{\phi(t^{\prime},x^{\prime},0^{+})}{\sqrt{c^{2}(t-t^{\prime})^{2}-(x-x^{\prime})^{2}-z^{2}}},$
(8)
plays the role of a potential and the integration is performed within the
region $c(t-t^{\prime})>\sqrt{(x-x^{\prime})^{2}+z^{2}}$. Clearly, the
procedure above yields a causal (sub)luminal propagation from the $z=0^{+}$
plane to any point in the $z>0$ vacuum.
## III Non-Evanescent wave transmission
We consider now that an arbitrarily shaped pulse impinges at a well defined
angle $\theta_{i}$ on the inside surface $z=0^{-}$ of a homogeneous non-
dispersive dielectric with index of refraction $n$ (Fig. 1). The incident
pulse is therefore described by an arbitrary function
$\phi_{i}(t,x,z)=f_{i}[t-(n/c)\hat{n}_{i}\cdot\vec{\rho}]$ of a single
variable $t-(n/c)\hat{n}_{i}\cdot\vec{\rho}$, where
$\hat{n}_{i}\equiv(\sin\theta_{i},\cos\theta_{i})$ is a unit vector pointing
along the angle of incidence $\theta_{i}$ and $\vec{\rho}\equiv(x,z)$. Notice
that in this case the incident wavefronts have an infinite extension in the
direction normal to $\hat{n}_{i}$. At $z=0^{+}$ and after being transmitted
into vacuum, the outgoing field can therefore be written as
$\phi(t,x,0^{+})=f_{t}(t-x/v_{\parallel})$ (9)
where $f_{t}$ is related to the arbitrary function $f_{i}$ and the Fresnel
amplitude for transmission from the dielectric into vacuum. As we are
concerned only with propagation across the vacuum gap, we will take $f_{t}$ as
given and we will disregard its relation with $f_{i}$, which would involve the
dielectric properties of the incident medium.
The intersection of the pulse with the interface $z=0$ is therefore seen to
travel along $x$ with velocity $v_{\parallel}=c/(n\sin\theta_{i})$ (not to be
confused with the parallel component of the incident velocity
$\hat{n}_{i}c/n$). For incidence angles smaller than the critical angle
$\theta<\theta_{c}=\sin^{-1}(1/n)$, $v_{\parallel}>c$ and we have normal
transmission, while for $\theta>\theta_{c}$, $v_{\parallel}<c$ and total
internal reflection ensues.
In order to set up a reference with which to compare the evanescent case, in
this section we employ our propagator to study a non-evanescent plane pulse.
Thus, we consider here the case $\theta<\theta_{c}$ and we substitute Eq. (9)
into Eq. (8) to obtain
$\psi(t,x,z)=\frac{c}{\pi}\int dt^{\prime\prime}\,f_{t}(t^{\prime\prime})\int
dx^{\prime}\,\left(c^{2}(t-t^{\prime\prime}-x^{\prime}/v_{\parallel})^{2}-(x-x^{\prime})^{2}-z^{2}\right)^{-1/2}$
(10)
after changing integration variables from $t^{\prime}$ to
$t^{\prime\prime}\equiv t^{\prime}-x^{\prime}/v_{\parallel}$. The integration
(10) has to be performed over the region $c(t-t^{\prime\prime})-\mu
x^{\prime}>\sqrt{(x-x^{\prime})^{2}+z^{2}}$, where $\mu\equiv
c/v_{\parallel}$. Thus, $t^{\prime\prime}$ has an upper bound
$t_{m}=t-(\mu x-\nu z)/c,$ (11)
where $\nu\equiv\sqrt{1-c^{2}/v_{\parallel}^{2}}$, and for each value of
$t^{\prime\prime}<t_{m}$, $x^{\prime}$ is bounded by the limits
$x^{\prime}_{\pm}=-(1/\nu^{2})\left(x-\mu c(t-t^{\prime\prime})\pm\sqrt{[\mu
x-c(t-t^{\prime\prime})]^{2}-\nu^{2}z^{2}}\right).$ (12)
A simple interpretation of Eqs. (11) and (12) can be obtained with the help of
Fig. 2.
Figure 2: Contributions to the non-evanescent wave observed at an event ${\cal
E}=(t,x,z)$. Some world lines
$(ct^{\prime},x^{\prime}=v_{\parallel}(t^{\prime}-t^{\prime\prime}),0)$ of
source points labeled by fixed values of $t^{\prime\prime}$ and thus moving on
the $z=0$ plane along with the incident pulse are shown. The positions
$x^{\prime}_{\pm}$ denote the intersections of the past light cone of $\cal E$
with one of the world lines plotted. We consider the case $v_{\parallel}>c$.
The coordinate and time axes are indicated.
Consider an event ${\cal E}=(ct,x,z)$ defined by the observation of the field
at a given position $(x,z)$ with $z>0$ and at a given time $t$. Causality
requires that only events within the past light-cone of $\cal E$ are able to
influence it. Notice that a given value of $t^{\prime\prime}$ denotes a point
that moves along $x$ keeping a fixed position with respect to the intersection
of the incident pulse with the $z=0$ interface. In Fig. 2 we show the world
lines $(ct^{\prime},x^{\prime},0)$ of a few such points. Since each of them
moves with speed $v_{\parallel}>c$, most of its world line lies outside the
past light cone of $\cal E$. Only if $t^{\prime\prime}<t_{m}$ can it actually
cross the light cone, entering and leaving at positions $x^{\prime}_{-}$ and
$x^{\prime}_{+}$ respectively.
After the change of variables from $x^{\prime}$ to
$\eta\equiv\frac{x^{\prime}-[x-\mu c(t-t^{\prime\prime})]}{\sqrt{[\mu
x-c(t-t^{\prime\prime})]^{2}-\nu^{2}z^{2}}},$ (13)
Eq. (10) simplifies to
$\psi(t,x,z)=\frac{c}{\pi\nu}\int_{-\infty}^{t_{m}}dt^{\prime\prime}\,f_{t}(t^{\prime\prime})\int_{-1}^{1}\frac{d\eta}{\sqrt{1-\eta^{2}}}.$
(14)
The integration over $\eta$ is immediate, so that substituting Eq. (14 in
Eq.(7) we obtain finally
$\phi(t,x,z)=f_{t}(t-\hat{n}_{t}\cdot\vec{\rho}/c).$ (15)
where $\hat{n}_{t}=(\mu,\nu)$. Thus, as illustrated in Fig. 1, the transmitted
wave is a pulse with the same profile as the incident field and propagating
with speed $c$ at the well defined angle
$\theta_{t}=\sin^{-1}\mu=\cos^{-1}\nu$, in accordance with Snell’s law as
could have been expected.
## IV Evanescent wave transmission
We consider now the case $\theta_{i}>\theta_{c}$, for which $v_{\parallel}<c$
and the transmitted wave becomes evanescent. In this case we substitute Eq.
(9) into Eq. (8) to obtain
$\psi(t,x,z)=\frac{c}{\pi}\int
dt^{\prime\prime}f_{t}(t-x/v_{\parallel}+t^{\prime\prime})\int
dx^{\prime\prime}\left(c^{2}(t^{\prime\prime}+x^{\prime\prime}/v_{\parallel})^{2}-(x^{\prime\prime})^{2}-z^{2}\right)^{-1/2},$
(16)
after introducing the variables $x^{\prime\prime}\equiv x^{\prime}-x$ and
$t^{\prime\prime}=t^{\prime}-t-x^{\prime\prime}/v_{\parallel}$. As shown in
Fig. 3, for any observation event $\cal E$ and any value of
$t^{\prime\prime}$, there is exactly one intersection $x^{\prime\prime}_{-}$
between the past light cone of $\cal E$ and the world line
$(ct^{\prime},x^{\prime\prime}=v_{\parallel}(t^{\prime}-t^{\prime\prime}-t),0)$,
where now $x^{\prime\prime}_{-}=-\gamma\beta[\gamma
ct^{\prime\prime}+\sqrt{z^{2}+(\gamma\beta ct^{\prime\prime})^{2}}]$ and we
introduced the definitions $\beta\equiv v_{\parallel}/c$ and $\gamma\equiv
1/\sqrt{1-v_{\parallel}^{2}/c^{2}}$. We have assumed that $v_{\parallel}>0$.
As the world line leaves the past light cone at $x^{\prime\prime}_{-}$, in Eq.
(16) the integration over $t^{\prime\prime}$ is unconstrained and that over
$x^{\prime\prime}$ extends from $-\infty$ to $x^{\prime\prime}_{-}$.
Figure 3: Contributions to the evanescent wave observed at an event ${\cal
E}=(ct,x,z)$ as in Fig. 2, but for $v_{\parallel}<c$. The position
$x^{\prime\prime}_{-}$ denotes the intersections of the past light cone of
$\cal E$ with one of the world lines
$x^{\prime\prime}=v_{\parallel}(t^{\prime}-t^{\prime\prime}-t)$ corresponding
to $t^{\prime\prime}$.
Another change of variable, from $x^{\prime\prime}$ to
$\eta=\frac{1}{\gamma\beta}\frac{x^{\prime\prime}+\gamma^{2}\beta
ct^{\prime\prime}}{\sqrt{z^{2}+(\gamma\beta ct^{\prime\prime})^{2}}},$ (17)
yields
$\psi(t,x,z)=\frac{\gamma\beta
c}{\pi}\int_{-\infty}^{\infty}dt^{\prime\prime}\,f_{t}(t-x/v_{\parallel}+t^{\prime\prime})\int_{-\infty}^{-1}\frac{d\eta}{\sqrt{\eta^{2}-1}}.$
(18)
The integrals over $x^{\prime\prime}$ and over $\eta$ in Eqs. (16) and (18)
respectively yield an infinite value for $\psi$. This divergence is not unlike
that commonly found for the electromagnetic potentials produced by infinitely
extended sources. For example, when calculating the electric field produced by
a uniformly charged plane one cannot simply obtain the corresponding potential
by integrating the Coulomb kernel over the whole surface. However, in that
case the electric field may be obtained either by deriving the Coulomb kernel
first and integrating afterwards or else, by truncating the integrations at a
finite distance, deriving the resulting potential to obtain the field and
afterwards taking the limit of an infinite surface. Here we follow the later
procedure. Thus, we set a finite lower integration limit
$x^{\prime\prime}_{L}$ in Eq. (16), corresponding to a lowest point $\eta_{L}$
in Eq. (18), and we take the limit $x^{\prime\prime}_{L}\to-\infty$,
$\eta_{L}\to-\infty$ after obtaining the field $\phi$.
As $\psi$ depends on $z$ only through $\eta_{L}$, substituting Eq. (18) in (7)
we obtain
$\phi(t,x,z)=\frac{\gamma\beta
c}{\pi}\int_{-\infty}^{\infty}dt^{\prime\prime}\,f_{t}(t-x/v_{\parallel}+t^{\prime\prime})\frac{1}{\sqrt{\eta_{L}^{2}-1}}\frac{\partial\eta_{L}}{\partial
z}.$ (19)
In the limit $x^{\prime\prime}_{L}\to-\infty$ we evaluate
$\xi_{L}\equiv\frac{1}{\sqrt{\eta_{L}^{2}-1}}\frac{\partial\eta_{L}}{\partial
z}\to\frac{z}{z^{2}+(\gamma\beta ct^{\prime\prime})^{2}},$ (20)
and we obtain finally
$\phi(t,x,z)=\frac{1}{\pi}\int_{-\infty}^{\infty}dt^{\prime\prime}\,f_{t}(t-x/v_{\parallel}+t^{\prime\prime})\frac{\gamma|v_{\parallel}|z}{z^{2}+(\gamma
v_{\parallel}t^{\prime\prime})^{2}}.$ (21)
Taking the absolute value of $v_{\parallel}$ in the numerator of Eq. (21)
allows its use also for $v_{\parallel}<0$.
To grasp the meaning of Eq. (21) we evaluate it for an infinitely sharp pulse
$f_{t}(\tau)\equiv f_{0}\delta(\tau).$ (22)
Substitution into Eq. (21) yields
$\phi(t,x,z)=\frac{f_{0}}{\pi}\frac{\gamma|v_{\parallel}|z}{z^{2}+\gamma^{2}(x-v_{\parallel}t)^{2}}.$
(23)
Surprisingly, at any time $t$ the pulse transmitted at a distance $z$ from the
interface is given by a Lorentzian of width $z/\gamma$ centered in front of
the actual position $x=v_{\parallel}t$ of the incident pulse on the $z=0$
surface (Fig. 4).
Figure 4: Infinitely sharp wavefront (heavy solid line) incident at an angle
$\theta=34.5^{\circ}>\theta_{c}$ upon the surface of a dielectric with index
of refraction $n=2$. The pulse widens and diminishes as it is transmitted
across an air gap and into a second dielectric. The thin lines indicate the
nominal pulse width and the dashed line its center.
Thus, the propagation seems to be instantaneous in the direction normal to the
surface, and actually, part of the pulse seems to travel backwards in time
Carey et al. (2000); Resch et al. (2001), as at a position $(x,z)$ it becomes
appreciable at times $t<x/v_{\parallel}$, that is, before the incoming pulse
reaches the corresponding position $(x,0)$. However, our deduction of Eq. (22)
shows that it is completely consistent with a causal and retarded propagation,
and that the field at a $(x,z)$ at time $t$ is not produced instantaneously by
the incoming field at $(x,0)$, but arises from previously excited positions
$(x^{\prime},0)$ with $x^{\prime}<x+x^{\prime\prime}_{-}$.
It is interesting to note that, according to Eq. (23), the height of the
transmitted pulse is inversely proportional to the distance $z$ from the
surface, instead of decaying exponentially as usually found for evanescent
waves. However, Eq. (23) describes the propagation into vacuum of a single
infinitely sharp incident wavefront. In a wavetrain made up of a succession of
incident pulses, the regions excited by neighboring pulses overlap each other,
as the width of each transmitted pulse increases in proportion to $z$, and
therefore their corresponding fields interfere. This interference is at the
origin of the exponential decay of periodic waves, as can be verified by
choosing
$f_{t}(\tau)=Ae^{-i\omega\tau}$ (24)
and substituting into Eq. (21), which yields
$\phi(t,x,z)=Ae^{i(Qx-\omega
t)}\int_{-\infty}^{\infty}\frac{dt^{\prime\prime}}{\pi}e^{-i\omega
t^{\prime\prime}}\frac{\gamma|v_{\parallel}|z}{z^{2}+(\gamma
v_{\parallel}t^{\prime\prime})^{2}}.$ (25)
A simple contour integration closing the integration path with an infinite
semicircle on the lower half complex $t^{\prime\prime}$ plane yields the
familiar result
$\phi(t,x,z)=Ae^{i(Qx-\omega t)-\kappa z},$ (26)
where $Q=\omega/v_{\parallel}=n\sin\theta_{i}\omega/c$ is the parallel
component of the wave vector and
$\kappa=\omega/(\gamma|v_{\parallel}|)=\sqrt{Q^{2}-\omega^{2}/c^{2}}=1/l$ the
inverse of the decay length $l$. As $\omega$ increases, the distance along $x$
between successive maxima and minima decreases, yielding larger interference
effects and a shorter decay length.
## V Screens
The results of the previous section suggest an experiment that could confirm
that propagation of evanescent waves in the FTIR geometry is not superluminal
nor acausal. The experiment could be performed simply by partially covering
the surface of the first interface with a couple of opaque screens as shown in
Fig. 5.
Figure 5: Pulse as in Fig. 1 but impinging on the surface of the first
dielectric at an angle $\theta_{i}>\theta_{c}$. The surface is covered by
semi-infinite opaque screens $S_{a}$ and $S_{b}$ with edges at $x_{a}$ and
$x_{b}$. The leading wavefront of the incoming pulse is about to reach
$x_{a}$. $\cal F$ denotes the foremost wavefront.
If transmission were indeed superluminal, we would expect a non-null
transmitted field $\phi(t,x_{a},d)$ across the gap in front of the edge
$x_{a}$ of the first screen $S_{a}$ as soon as the leading wavefront $\cal F$
of the incident pulse reaches $x_{a}$. Similarly, we would expect that the
field $\phi(t,x_{b},d)$ would be modified as soon as $\cal F$ reaches the edge
$x_{b}$ of the second screen $S_{b}$.
To calculate the field corresponding to Fig. 5 we go back to Eq. (16). The
screens confine the integration region to the interval
$x_{a}<x^{\prime}=x+x^{\prime\prime}<x_{b}$. This inequality has to be obeyed
together with the previous constriction $x^{\prime}<x+x^{\prime\prime}_{-}$.
These conditions can only be satisfied by those wavefronts which have reached
$x_{a}$ and left behind the first screen before $t-T_{a}$, where
$T_{i}=(1/c)\sqrt{(x-x_{i})^{2}+z^{2}}$, $i=a,b$, is the minimum time required
to reach $(x,z)$ from $(x_{i},0)$ moving at speed $c$. Thus, only those points
on the wavefront labeled by $t^{\prime\prime}>T^{\prime\prime}_{a}$ can
contribute to (16), where
$T^{\prime\prime}_{i}\equiv\frac{x-x_{i}}{v_{\parallel}}-\frac{1}{c}\sqrt{(x_{i}-x^{\prime})^{2}+z^{2}},\quad
i=a,b.$ (27)
For those wavefronts which have left $S_{a}$ by time $T_{a}$ but have not
reached $S_{b}$ at time $T_{b}$, namely, those with
$T^{\prime\prime}_{b}<t^{\prime\prime}<T^{\prime\prime}_{a}$, the integral
over $x^{\prime\prime}$ in Eq. (16) has to be performed from
$x^{\prime\prime}_{a}=x+x_{a}$ up to $x^{\prime\prime}_{-}$. Finally, for
those wavefronts which have already been blocked by $S_{b}$ by time $T_{b}$,
namely, those with $t^{\prime\prime}<T^{\prime\prime}_{b}$, the upper limit of
integration has to be replaced by $x^{\prime\prime}_{b}=x+x_{b}$. Therefore,
$\psi(t,x,z)=\frac{\gamma\beta
c}{\pi}\int_{T^{\prime\prime}_{b}}^{T^{\prime\prime}_{a}}dt^{\prime\prime}\,f_{t}(t-x/v_{\parallel}+t^{\prime\prime})\int_{\eta_{a}}^{-1}\frac{d\eta}{\sqrt{\eta^{2}-1}}+\frac{\gamma\beta
c}{\pi}\int_{-\infty}^{T^{\prime\prime}_{b}}dt^{\prime\prime}\,f_{t}(t-x/v_{\parallel}+t^{\prime\prime})\int_{\eta_{a}}^{\eta_{b}}\frac{d\eta}{\sqrt{\eta^{2}-1}}$
(28)
where we used the change of variables (17) and substituted
$x^{\prime\prime}\to x+x_{i}$ in it to define the limits $\eta_{i}$. Notice
that $\psi$ depends on $z$ only through the integration limits $\eta_{i}$, so
that substituting Eq. (28) in (7) we obtain
$\psi(t,x,z)=\frac{\gamma\beta
c}{\pi}\int_{-\infty}^{\infty}dt^{\prime\prime}\,f_{t}(t-x/v_{\parallel}+t^{\prime\prime})[\xi_{a}\Theta(T^{\prime\prime}_{a}-t^{\prime\prime})-\xi_{b}\Theta(T^{\prime\prime}_{b}-t^{\prime\prime})],$
(29)
where
$\xi_{i}\equiv\frac{1}{\sqrt{\eta_{i}^{2}-1}}\frac{\partial\eta_{i}}{\partial
z}\equiv\xi_{L}\zeta_{i},$ (30)
$\xi_{L}$ is given by Eq. (20)and
$\zeta_{i}\equiv\zeta(x^{\prime\prime}_{i},t^{\prime\prime})$ with
$\zeta(x^{\prime\prime},t^{\prime\prime})=-\,\frac{x^{\prime\prime}+\gamma^{2}\beta
ct^{\prime\prime}}{\sqrt{(x^{\prime\prime})^{2}+2\gamma^{2}\beta
cx^{\prime\prime}t^{\prime\prime}-\gamma^{2}\beta^{2}(z^{2}-c^{2}(t^{\prime\prime})^{2})}}.$
(31)
Substituting Eqs. (17) and (30) in (29) we finally obtain
$\phi(t,x,z)=\frac{1}{\pi}\int_{-\infty}^{\infty}dt^{\prime\prime}\,f_{t}(t-x/v_{\parallel}+t^{\prime\prime})\frac{\gamma
v_{\parallel}z}{z^{2}+(\gamma
v_{\parallel}t^{\prime\prime})^{2}}(1-C(t^{\prime\prime})),$ (32)
where
$C(t^{\prime\prime})=1+[\zeta_{b}\Theta(T^{\prime\prime}_{b}-t^{\prime\prime})-\zeta_{a}\Theta(T^{\prime\prime}_{a}-t^{\prime\prime})].$
(33)
Notice that the field $\phi$ in the presence of screens (Eq. (32)) is given by
an expression similar to that corresponding to the field in the absence of
screens (Eq. (21)) but with a correction term $C$ due to the diffraction by
the screen.
As in the previous section, we consider again the case of a sharp incident
pulse Eq. (22). Substituting in (32) we obtain
$\phi(t,x,z)=\frac{\gamma
v_{\parallel}z}{z^{2}+\gamma^{2}(x-v_{\parallel}t)^{2}}\times\left\\{\begin{array}[]{ll}0&\mbox{if
$ct<x_{a}/\beta+\sqrt{(x-x_{a})^{2}+z^{2}}$},\\\
\zeta(x_{a}-x,x/v_{\parallel}-t)-\zeta(x_{b}-x,x/v_{\parallel}-t)&\mbox{if
$ct>x_{b}/\beta+\sqrt{(x-x_{b})^{2}+z^{2}}$},\\\
\zeta(x_{a}-x,x/v_{\parallel}-t)&\mbox{otherwise}.\end{array}\right.$ (34)
We remark that the field is zero until the time $x_{a}/v_{\parallel}$ when the
incident pulse shows up from behind the screen $S_{a}$, and this information
has had enough time $(1/c)\sqrt{(x_{a}-x)^{2}+z^{2}}$ to propagate from the
screen’s edge $(x_{a},0)$ to the observation point $(x,z)$. Similarly, the
information that the pulse has hidden behind screen $S_{b}$ does not reach the
observation point until the time
$x_{b}/v_{\parallel}+(1/c)\sqrt{(x-x_{b})^{2}+z^{2}}$. The field has
singularities due to the passage of the incident pulse through the screen
edges, that propagate at speed $c$ from the events $(x_{a}/\beta,x_{a},0)$ and
$(x_{b}/\beta,x_{b},0)$. Notice that $\zeta(x_{i}-x,0)\to 1$ as $x\to\infty$.
Thus, if we follow the incident pulse, i.e., we take $x\approx
v_{\parallel}t$, then $\phi\to 0$ asymptotically after the pulse hides behind
$S_{b}$. Furthermore, if the screens are very far apart we recover the field
(23) between the screens.
The features above are illustrated in Fig. 6
Figure 6: Field isolines of $\phi(t,x,d)$ produced by an unit delta function
input pulse that propagates along the $z=0$ surface with speed
$v_{\parallel}=c/\sqrt{2}$ and is blocked at $x<x_{a}=-3d$ and $x>x_{b}=3d$ by
opaque screens (shaded regions). The world line $x=v_{\parallel}t$ of the
incident pulse is indicated by a dashed line. The times $t_{a}$ and $t_{b}$
when the incident pulse cross the edges of each screen are indicated by the
horizontal dot-dashed lines. The singularities of the transmitted field are
indicated by the thick solid hyperbolas.
which show the transmitted field at the plane $z=d$. Notice the delay $d/c$
after the incident pulse crosses $x_{a}$ at $t_{a}=x_{a}/v_{\parallel}$ before
a non-null field first appears across the gap at $(x_{a},d)$, and a similar
delay after the incident pulse crosses $x_{b}$ at $t_{b}=x_{b}/v_{\parallel}$
before the field starts to be extinguished at $(x_{b},d)$. Furthermore, notice
that for some time the field penetrates a small distance $\approx d$ beyond
$x_{b}$ as if there were no screen. The field is singular at the hyperbolas
with vertices at $x=x_{a}$, $t=t_{a}+d/c$ and at $x=x_{b}$, $t=t_{b}+d/c$
given by the intersection of the $(ct,x,d)$ hyperplane and the future light
cone of the events $(ct_{a},x_{a},0)$, $(ct_{b},x_{b},0)$. Thus, we have shown
observable consequences of the fact that evanescent waves in FTIR do not
propagate superluminally nor acausally in the direction normal to the
dielectric-vacuum interfaces, but with retardation and obliquely. A graphical
approach to the results of this section and an animation illustrating them may
be found in Refs. Mochan and Brudny, 2001 and Ref. Brudny and Mochan, 2001
respectively.
## VI A smooth transverse profile
Perfectly opaque screens such as those considered in the previous section
introduce sharp discontinuities in the pulse at $z=0$. The truncated pulse no
longer has a well defined propagation direction $\theta_{i}$ but may still be
represented by a superposition of pulses with varying propagation directions.
A sharp truncation leads to the presence of subcritical incident angles
$\theta_{i}<\theta_{c}$, and therefore to the presence of both, evanescent and
non-evanescent transmitted fields. It has been argued Carey et al. (2001) that
the retardation effects discussed in the previous section may be due only to
the non-evanescent contributions, known to be subluminal. The comparatively
slow subluminal contributions would be unable to affect the arrival of the
superluminal signals if the later were actually present. However, any small
non-evanescent wave would dominate the transmitted signal after a wide enough
gap. Thus, it is interesting to study the propagation of pulses with a finite
transverse extension but with a smooth lateral cutoff and built up completely
from hypercritical $\theta_{i}>\theta_{c}$ evanescent contributions.
To explore the propagation of the smoothly truncated pulses discussed above,
we consider an incoming field given by a Fourier integral
$\phi(t,x,0^{+})=\int\frac{d\omega}{2\pi}\int\frac{dQ}{2\pi}\,f_{\omega
Q}e^{i(Qx-\omega t)},$ (35)
where $f_{\omega Q}$ is the amplitude for each parallel component of the wave
vector $Q$ and frequency $\omega$. We can change integration variable from $Q$
to the parallel velocity $v\equiv\omega/Q$,
$\phi(t,x,0^{+})=\int\frac{d\omega}{2\pi}\int\frac{dv}{2\pi}\,f_{\omega
v}e^{-i\omega(t-x/v)}.$ (36)
where we introduced the velocity dependent amplitude $f_{\omega
v}\equiv(\omega/v^{2})f_{\omega,\omega/v}$. The incident field (36) will give
rise to evanescent waves exclusively as long as all non-null components
$f_{\omega v}$ have $v<c$. At this point we could integrate first Eq. (36)
with respect to $\omega$, obtaining thus a superposition of plane pulses, each
of which may be propagated across the air gap according to Eq. (21).
Alternatively, we may propagate each monochromatic component using Eq. (26)
and afterward perform the integrations in Eq. (36). We follow the later
approach and write
$\phi(t,x,z)=\int\frac{d\omega}{2\pi}\int\frac{dv}{2\pi}\,f_{\omega
v}e^{-\omega[z/(\gamma v)+i(t-x/v)]},$ (37)
where $\gamma=1/\sqrt{1-v^{2}/c^{2}}$ as in Sec. IV. For simplicity we assume
that we may factor $f_{\omega v}=f_{\omega}f_{v}$ into frequency and velocity
dependent amplitudes, $f_{\omega}$ and $f_{v}$ respectively; $f_{\omega}$
controls the time duration of the pulse, or equivalently, its longitudinal
extent, while $f_{v}$ controls its transverse extent. We further assume a
narrow Gaussian velocity distribution of width $\Delta v$ around a nominal
velocity $v_{0}<c$,
$f_{v}=\frac{\sqrt{2\pi}}{\Delta v}e^{-u^{2}/2\Delta v},$ (38)
where $u\equiv v-v_{0}$. We assume $\Delta v$ is small enough that the non-
evanescent contributions to the field may be neglected and the exponent in Eq.
(37) may be linearized in $u$. Thus, the transmitted field becomes
$\phi(t,x,z)\approx\int\frac{d\omega}{2\pi}\,f_{\omega}e^{-\omega[z/(\gamma_{0}v_{0})+i(t-x/v_{0})]}\int\frac{du}{\sqrt{2\pi}\Delta
v}\,e^{-u^{2}/2\Delta v^{2}+\omega u[z/\gamma_{0}-ix]/v_{0}^{2}},$ (39)
where $\gamma_{0}=1/\sqrt{1-v_{0}^{2}/c^{2}}$. The integration over $u$ is
immediate and yields
$\phi(t,x,z)=\int\frac{d\omega}{2\pi}\,f_{\omega}e^{-\omega[z/(\gamma_{0}v_{0})+i(t-x/v_{0})]-[\Delta
v\omega(x+iz/\gamma_{0})]^{2}/2v_{0}^{4}}.$ (40)
In Fig. 7 we illustrate the results of applying Eq. (40) to a pulse with a
Gaussian frequency distribution,
$f_{\omega}=A\frac{\sqrt{2\pi}}{\Delta\omega}e^{-(\omega-\omega_{0})^{2}/2\Delta\omega^{2}},$
(41)
of area $A$ and width $\Delta\omega$ centered at $\omega_{0}$
($\omega_{0}=16c/d$, $\Delta\omega=2c/d$, $v_{0}=0.7c$, $\Delta v=0.15c$).
Figure 7: Intensity of a Gaussian pulse localized in both space and time ($x$
and $t$) incident on the front face of an air gap of width $d$ (left) and
after after crossing it (right). The nominal velocity along $x$ is
$v_{0}=0.7c$ with width $\Delta v=0.15c$. The nominal frequency is
$\omega_{0}=16c/d$ with width $\Delta\omega=2c/d$
The pulse is seen to form on the $z=0$ surface at $x\approx-5d$ and at time
$t\approx 7d/c$, it propagates along the $x\approx v_{0}t$ line for a while,
peaks at $x=0$ at time $t=0$ and disappears at $x\approx 5d$, $t\approx 7d/c$.
Its maximum duration $\tau$ and size $L$ are $c\tau\approx L\approx d$ for a
fixed position and fixed observation time respectively, and it contains
altogether about six nodes.
Surprisingly, after crossing the gap, the pulse looks essentially the same! It
appears at the back $z=d$ face of the air gap at roughly the same time and the
same position as at the front $z=0$ face. It also peaks at the origin at $t=0$
and disappears from the back surface in concordance to the incident pulse on
the front face. Thus, it truly appears to propagate instantaneously. The main
difference between the incident and transmitted pulse is that the intensity of
the later is suppressed by 12 orders of magnitude. Another interesting
difference is that the number of visible nodes in the transmitted pulse has
decreased to about 4. This is a consequence of the fact that in FTIR, the
Fourier components with higher frequencies are more damped than those with
lower frequencies. Finally, a more subtle difference is that the speed of
propagation along the back face is slightly but noticeably larger than that on
the front face. This is due to the fact that plane waves incident at angles
closer to $\theta_{c}$ have a larger penetration length than waves incident at
larger angles. Thus, the angle of propagation of the transmitted pulse is
smaller than that of the incident wave Balcou et al. (1997).
In Sec. IV we have argued that the transmission of evanescent plane pulses
across an air gap under FTIR conditions is fully consistent with a retarded
and causal propagation along oblique directions. We have strengthened our
argument by showing that there is a delay before a perturbation, such as
blocking part of the incident wavefront, can produce an effect on the pulse
transmitted across the gap. Furthermore, by truncating an incident plane pulse
producing an abrupt transverse profile, we showed that the transmitted pulse
is shifted along the surface in the direction of propagation. However, in Sec.
VI we showed through an example that if the pulse has a smooth transverse
profile, such that all its Fourier components are evanescent, it is
transmitted as if it were indeed superluminal. To understand this result, in
Fig. 8 we show schematically a pulse smoothly truncated along its transverse
direction, built from narrow plane components propagating along well defined
directions $\theta>\theta_{c}$.
Figure 8: Evanescent transmission of a smoothly truncated pulse made up
initially of a superposition of narrow plane wavefronts (heavy solid lines)
with a distribution of angles above $\theta_{c}$. As it crosses the gap, each
wavefront widens (shaded bands around dashed lines, as in Fig. 4). The
ellipses represent schematically the contour levels of the pulse and are
centered at the region where the different contributions add coherently in
phase, i.e., the regions where the centers of each component coincide.
Snapshots are taken at three different times. The solid arrows illustrate the
nominal propagation of the peak. The heavy arrow indicates the apparent
instantaneous tunneling across the air gap. The dashed arrows illustrate the
actual (sub)luminal propagation.
The incident field peaks at the regions where the directional components add
in phase, indicated in the figure by elliptical regions around the crossing
point of the different incident wavefronts. As each of the components crosses
the gap, it is widened according to Eq. (23). The peak of the transmitted
pulse appears in the regions of largest overlap between the different widened
transmitted components, i.e., at the regions where their centers coincide. The
peak of the transmitted pulse at the back face of the air gap is seen to
appear at the same time as the peak of the incident pulse reaches the front
surface. Thus, it would seem as if the peak tunneled instantaneously in the
direction normal to the gap. However, in the previous sections we have shown
that each of the transmitted components of the transmitted wave originates
causally from regions in the lateral wings of the incident pulse. Thus, the
peak of the transmitted field does not actually originate from the peak of the
incident field; it is formed by contributions from the lateral wings of the
incident field, which reach the front face of the air gap first. The lateral
wings of each component have enough time to cross the gap traveling at speed
$c$ and combine to form the relatively small transmitted peak right at the
time when the larger incident peak reaches the front surface. Similarly, the
different components of the field produced by the peak of the incident pulse
get out of step as they cross the air gap and, therefore, do not contribute to
the peak of the transmitted pulse, but rather, to its lateral wing. This is
illustrated by the dashed arrows in Fig. 8. Thus, it seems that the physical
propagation (retarded and causal) can not be distinguished from the
nonphysical propagation (superluminal and non-causal) as long as we only
consider smooth incident pulses which contain only propagation directions
above the critical angle Carey et al. (2000, 2001).
## VII Evanescent propagator
In Sec. VI we found that a particular pulse made up of only evanescent
components seemed to propagate instantaneously across the air gap, in contrast
to the abruptly truncated pulses considered in Sec. V for which retardation
effects have observable consequences. This was explained graphically in Fig. 8
for incident pulses built up from narrow directional components, each of which
is widened as it is transmitted across the gap. To show that this behavior is
generic, we start from the Fourier decomposition of an arbitrary field
$\phi(t,x,z)=\int\frac{dQ}{2\pi}\int\frac{d\omega}{2\pi}e^{i(Qx-\omega
t)}\phi_{\omega,Q}(z),$ (42)
where
$\phi_{\omega,Q}(z)=\int dt\int dx\,e^{-i(Qx-\omega t)}\phi(t,x,z).$ (43)
The condition that $\phi(t,x,z)$ is made up exclusively of evanescent waves is
equivalent to stating that the integration region in Eq. (35) is given by
$|\omega|<|Q|c$, i.e., $\phi_{\omega,Q}=0$ if $|\omega|>|Q|c$. Using Eq. (26)
we propagate each Fourier component from $z=0^{+}$ to $z>0$ as
$\phi_{\omega,Q}(z)=e^{-\kappa z}\phi_{\omega,Q}(0^{+}),$ (44)
where $\kappa=\sqrt{Q^{2}-\omega^{2}/c^{2}}$. Thus, we can combine Eqs. (42),
(43) and (44) to obtain
$\phi(t,x,z)=\int dx^{\prime}\int
dt^{\prime}\,P^{\prime}(t,x,z;t^{\prime},x^{\prime},0^{+})\phi(t^{\prime},x^{\prime},0^{+}),$
(45)
where
$P^{\prime}(t,x,z;t^{\prime},x^{\prime},0^{+})\equiv
P^{\prime}(t-t^{\prime},x-x^{\prime},z)=\int\frac{dQ}{2\pi}\int\frac{d\omega}{2\pi}e^{i[Q(x-x^{\prime})-\omega(t-t^{\prime})]-\kappa
z}$ (46)
and the integration region is given by $|\omega|<|Q|c$. Comparing Eq. (45)
with (4) we find that $P^{\prime}$ is a propagator that can be used in the
same way as the propagator $P$ defined in Eq. (6) to find the value of the
field at $z>0$ given its values at $z=0^{+}$, provided the field is built up
of evanescent components only, as in the examples of the two previous
sections. We remark that our original causal, retarded and subluminal
propagator $P$ was able to propagate any arbitrary outgoing field. However, by
adding constrains to the field, we gain freedom in our choice of propagator,
as we can chose arbitrarily its effect on fields that do not obey the
constrain. Thus, if we can find any function $P^{\prime\prime}$ such that
$\int dx^{\prime}\int
dt^{\prime}\,[P^{\prime\prime}(t,x,z;t^{\prime},x^{\prime},0^{+})-P(t,x,z;t^{\prime},x^{\prime},0^{+})]f(t^{\prime}-x^{\prime}/v)=0,$
(47)
for an arbitrary flat pulse $f$ moving along $x$ with any velocity $-c<v<c$,
then we could employ $P^{\prime\prime}$ instead of $P$ to propagate an
arbitrary evanescent pulse. $P^{\prime}$ above is just one of the many
possible choices of a propagator for evanescent pulses.
To proceed, we make a change of variable $\omega\to Qv$ to write Eq. (46) as
$P^{\prime}(\tau,\xi,z)=\int\frac{dQ}{2\pi}\int_{-c}^{c}\frac{dv}{2\pi}\,|Q|e^{iQ(\xi-v\tau)}e^{-Qz/\gamma},$
(48)
where $\gamma=1/\sqrt{1-v^{2}/c^{2}}$, $\xi=x-x^{\prime}$, $\tau=t-t^{\prime}$
and we perform the integration over $Q$,
$P^{\prime}(\tau,\xi,z)=\frac{1}{2\pi^{2}}\int_{-c}^{c}dv\,\frac{(z/\gamma)^{2}(\xi-v\tau)^{2}}{[(z/\gamma)^{2}+(\xi-v\tau)^{2}]^{2}}.$
(49)
Notice that $P^{\prime}(t,x,z;t^{\prime},x^{\prime},0^{+})$ is symmetric under
the interchange $\xi\leftrightarrow-\xi$ and also under the interchange
$\tau\leftrightarrow-\tau$, i.e.,
$P^{\prime}(t,x,z;t^{\prime},x^{\prime})=P^{\prime}(t,x^{\prime},z;t^{\prime},x)=P^{\prime}(t^{\prime},x,z;t,x^{\prime})=P^{\prime}(t^{\prime},x^{\prime},z;t,x)$.
Thus, the evanescent propagator $P^{\prime}$ is superluminal and acausal.
To finish the calculation of $P^{\prime}$ we make another change of
integration variable $v=c\sin\alpha$ to write
$P^{\prime}(\tau,\xi)=\frac{c}{2\pi^{2}}\int_{-\pi/2}^{\pi/2}d\alpha\cos\alpha\frac{(z\cos\alpha)^{2}-(\xi-c\tau\sin\alpha)^{2}}{[(z\cos\alpha)^{2}+(\xi-c\tau\sin\alpha)^{2}]^{2}},$
(50)
and we perform the integration
$P^{\prime}(\tau,\xi)=\frac{c}{4\pi^{2}}\frac{1}{s^{3}}\left[z\log\left(\frac{(c^{2}\tau^{2}-z^{2}-zs)^{2}-c^{2}\tau^{2}\xi^{2}}{(c^{2}\tau^{2}-z^{2}+zs)^{2}-c^{2}\tau^{2}\xi^{2}}\right)-4s\right],$
(51)
where $\rho=\sqrt{\xi^{2}+z^{2}}$ is the spatial distance from the source to
the observation point and $s^{2}=\rho^{2}-c^{2}\tau^{2}$ is the squared space-
time interval. This expression may be simplified to
$P^{\prime}(\tau,\xi)=\frac{c}{2\pi^{2}}\times\left\\{\begin{array}[]{ll}z\log(|z+s|/|z-s|)/s^{3}-2/s^{2}&\mbox{if
$s^{2}>0$},\\\ -2z\arctan(|s|/z)/|s|^{3}-2/|s|^{2}&\mbox{if $s^{2}<0$},\\\
\end{array}\right..$ (52)
The evanescent propagator is displayed in Fig. 9.
Figure 9: Propagator $P^{\prime}(t,x,d;t^{\prime},x^{\prime},0^{+})$
normalized to $c/d^{2}$ as a function of the displacement $x-x^{\prime}$ and
the delay $t-t^{\prime}$. Distance is measured in units of $d$ and time in
units of $d/c$. For aid in visualization, the height of the propagator was
truncated at $P^{\prime}=0.1c/d^{2}$.
The figure shows explicitly the temporal and spatial symmetry, and thus the
superluminality and non-causality of $P^{\prime}$. Notice that $P^{\prime}$
has singularities at the projected light-lines $x-x^{\prime}=\pm
c(t-t^{\prime})$ which converge at the origin $x=x^{\prime}$, $t=t^{\prime}$.
Thus, propagation is largest for instantaneous propagation in the direction
normal to the air gap.
We have found that the propagation of evanescent pulses can be described with
either the exact propagator $P$ of the problem, which is causal, retarded and
subluminal, or with an evanescent propagator $P^{\prime}$ which is
superluminal and non-causal. Both yield exactly the same transmitted pulse
when the incident pulse contains only evanescent components. Thus, it seems to
be impossible to distinguish superluminal from subluminal propagation in
experiments performed with purely evanescent pulses Carey et al. (2001).
An explanation for the curious conclusion found in this section can be
obtained by going back to Eq. (42) which we rewrite as
$\phi(t,x,z)=\int\frac{dQ}{2\pi}\phi_{Q}(t,z).$ (53)
Notice that for each finite wave vector $Q$, the time dependent Fourier
component $\phi_{Q}(t,z)$ has a strictly finite spectrum $-|Q|c<\omega<|Q|c$.
Thus, $\phi_{Q}(t,z)$ is an analytical function of $t$ with no singularities
and can be analytically continued to an arbitrary time $t_{2}$ from its values
in an arbitrarily small neighborhood of any other arbitrary time $t_{1}$.
Therefore, $\phi_{Q}(t,z)$ is perfectly predictable in principle. Using
antropomorphic language, we may say that at time $t_{1}$ the system knows from
the present values of $\phi(t\approx t_{1},x,0^{+})$ what its future values
$\phi(t\approx t_{2}\geq t_{1}+z/c,x,0^{+})$ will be, and thus it can use that
knowledge to build subluminally a transmitted pulse $\phi(t\approx t_{2},x,z)$
that will mimic $\phi(t\approx t_{2},x,0^{+})$, giving the impression that
superluminal transmission has taken place at time $t_{2}$. The validity of the
argument above in the presence of thermal or quantum noise has to be
investigated.
## VIII Conclusions
To study the propagation of a light pulse through a vacuum gap between two
parallel dielectrics in a FTIR configuration we constructed a propagator
derived directly from the wave equation resulting from Maxwell’s equations.
This propagator is retarded and complies with the relativistic causality
principle inherent to classical electromagnetism. Therefore, it can only
account for causal, (sub)luminal propagation of light pulses. However, when
this propagator is used to study the propagation of wave packets through the
gap, we find apparent superluminal behavior, that is, a wave packet might
appear on the far side of the gap at the same time that the incident packet
reaches the front one. Therefore this illusion of superluminality, present
within classical electromagnetic theory even in vacuum, is fully consistent
with relativistic causality. We showed explicitly that propagation in FTIR
actually takes place subluminally between the lateral wings of the incident
pulse and the central peak of the transmitted pulse, and we proposed simple
experiments that could verify this statement. Thus, although FTIR has many
similitudes to 1D tunneling, its correct physical interpretation requires a 2D
or 3D analysis. On the other hand, we constructed an explicitly superluminal
and acausal propagator that yields identical results as the retarded causal
one when applied to smooth pulses made up of evanescent contributions only.
Thus, there is a class of pulses for which superluminal and subluminal
propagation would be indistinguishable.
###### Acknowledgements.
We acknowledge partial support from UBACYT and CONICET (VLB) and from DGAPA-
UNAM under project IN111306 (WLM). VLB is a member of CONICET.
## References
* Wang et al. (2007) Zhi-Yong Wang and Cai-Dong Xiong, Phys. Rev. A 75, 042105 (2007).
* Winful (2007) H. G. Winful, Phys. Rev. A 76, 057803 (2007).
* Pereyra et al. (2007) P. Pereyra and H.P. Simanjuntak, Phys. Rev. E 75, 056604 (2007).
* Ranfagni et al. (2007) A. Ranfagni, G. Viliani, C. Ranfagni, R. Mignani, R. Ruggeri and A.M. Ricci, Phys. Lett. A 370, 370 (2007).
* Winful (2005) H. G. Winful, Phys. Rev. E 72, 046608 (2005).
* Winful (2003) H. G. Winful, Phys. Rev. Lett 90(2), 23901 (2003).
* Buttiker et al. (2003) M. Buttiker and S. Washburn, Nature 422, 271 (2003).
* Windul (2003-1) Herbert G. Winful, Nature 424, 628 (2003).
* Buttiker et al. (2003) M. Buttiker and S. Washburn, Nature 424, 638 (2003).
* Winful (2003-2) H. G. Winful, Phys. Rev. E 68, 016615 (2003).
* Peres et al. (2004) A. Peres and D. Terno, Rev. Mod. Phys. 76, 93 (2004).
* De Angelis et al (2007) T. De Angelis, E. Nagali, F. Sciarrino and F. De Martini, Phys. Rev. Lett. 99, 193601 (2007).
* Winful (2006) H. G. Winful, Physics Reports 436, 1 (2006).
* Boyd et al. (2002) R. W. Boyd and D.J. Gauthier, _Progress in Optics, Vol. 43, E. Wolf (ed.)_ (Elsevier, Amsterdam, 2002).
* Garrison et al (1998) J. C. Garrison, M. W. Mitchell, R. Y. Chiao and E. L. Bolda, Phys. Lett. A 245, 19 (1998).
* Azbel’ (1994) Mark Ya. Azbel’, Solid State Commun. 91(6), 439 (1994).
* Jackson (Jackson) J. D. Jackson, _Classical Electrodynamics_ (Wiley, New York, 1975), 2nd. ed.
* Sommerfeld (1914) A. Sommerfeld, Ann. Physik 44, 177 (1914). English translation available in Chap. II of Brillouin1960 (Brillouin).
* Brillouin1960 (Brillouin) L. Brillouin, _Wave Propagtion and Group Velocity_ (Academic Press, New York, 1960).
* Diener (1996) G. Diener, Phys. Lett. A 223, 327 (1996).
* Diener (1997) G. Diener, Phys. Lett. A 235, 118 (1997).
* Kuzmich et al. (2000) A. Kuzmich, A. Dogariu, L.J. Wang, P.W. Milonni and R.Y. Chiao, Phys. Rev. Lett. 86, 3925 (2001).
* Stenner et al. (2003) M.D. Stenner, D.J. Gauthier and M.A. Neifeld, Nature 425, 695 (2003).
* Stenner et al. (2005) M.D. Stenner, D.J. Gauthier and M.A. Neifeld, Phys. Rev. Lett. 94, 053902 (2005).
* Ranfagni et al. (2006) A. Ranfagni, P. Fabeni, G.P. Pazzi, A.M. Ricci, R. Trinci, R. Mignani, R. Ruggeri and F. Cardone, Phys. Lett. A 352, 473 (2006).
* Icsevgi et al. (1969) A. Icsevgi and W.E. Lamb, Phys. Rev. 185(2), 517 (1969).
* wang et al. (2000) L.J. Wang, A. Kuzmich and A. Dogariu, Nature (London) 406, 277 (2000).
* Janowicz et al. (2006) M. Janowicz and J. Mostowski, Phys. Rev. E 73, 046613 (2006).
* Huang et al. (2008) G. Huang, Ch. Hang and L. Deng, Phys. Rev. A 77, 011803(R) (2008).
* solli et al. (2003) D.R. Solli, C.F. McCormick, C. Ropers, J.J. Morehead, R.Y. Chiao and J.M. Hickmann, Phys. Rev. Lett. 91, 143906 (2003).
* Brunner et al. (2004) N. Brunner, V. Scarani, M. Wegmüller, M. Legré and N. Gisin, Phys. Rev. Lett. 93, 203902 (2004).
* halvorsen et al. (2008) T.G. Halvorsen and J.M. Leinaas, Phys. Rev. A 77, 023808 (2008).
* Kulkarni et al. (2004) M. Kulkarni, N. Seshadri, V.S.C. Manga Rao and S. Dutta Gupta, J. Mod. Opt. 51, 549 (2004).
* safian et al. (2006) R. Safian, C.D. Sarris and M. Mojahedi, Phys. Rev. E 73, 066602 (2006).
* Bigelow et al. (2006) M.S. Bigelow, N.N. Lepeshkin, H. Shin and R.W. Boyd, J. Phys.: Condens. Matter 18, 3117 (2006).
* Talukder et al. (2005) A.I. Talukder, T. Haruta and M. Tomita, Phys. Rev. Lett. 94, 223901 (2005).
* Mugnai et al. (2000) D. Mugnai, A. Ranfagni and R. Ruggeri, Phys. Rev. Lett. 84, 4830 (2000).
* Mugnai et al. (2005) D. Mugnai and I. Mochi, Phys. Rev. E 73, 016606 (2006).
* Carey et al. (2000) J. J. Carey, J. Zawadzka, D. A. Jaroszynski and K. Wynne, Phys. Rev. Lett. 84, 1431 (2000).
* Mochan and Brudny (2001) W.L. Mochán and V.L. Brudny, Phys.Rev.Lett 87, 119101 (2001).
* Carey et al. (2001) J. J. Carey, J. Zawadzka, D. A. Jaroszynski and K. Wynne, Phys. Rev. Lett. 87, 119102 (2001).
* Shaarawi et al. (2002) A. M. Shaarawi, B. H. Tawfik and I. M. Besieris, Phys. Rev. E. 66, 046626 (2002).
* Brudny and Mochan (2001) V.L. Brudny and W.L. Mochán, Optics Express 19(11) 561(2001).
* Barbero et al. (2000) A. P. Barbero H. E. Hernández-Figueroa and E. Recami, Phys. Rev. E 62(6), 8628 (2000).
* Reiten et al. (2001) M. T. Reiten D. Grischkowsky and R. A. Cheville, Phys. Rev. E 64, 036604 (2001).
* Balcou et al. (1997) Ph. Balcou and L. Dutriaux, Phys. Rev. Lett. 78(5), 851 (1997).
* Resch et al. (2001) K.J. Resch, J.S. Lundeen and A.M. Steinberg, IEEE J. Quantum Electron. 37(6), 794 (2001).
* Morse and Feshbach (alguno) Phillip M. Morse and Herman Feshbach, _Methods of Theoretical Physics_ (Mc.Graw-Hill, New York, 1953).
| arxiv-papers | 2008-07-15T18:09:41 | 2024-09-04T02:48:56.802077 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Vera L. Brudny and W. Luis Moch\\'an",
"submitter": "Luis Moch\\'an",
"url": "https://arxiv.org/abs/0807.2422"
} |
0807.2456 | # A Coincidence Problem:
How to Flow from ${\cal N}=2$ SQCD to ${\cal N}=1$ SQCD
Stefano Bolognesi
William I. Fine Theoretical Physics Institute, University of Minnesota, 116
Church St. S.E., Minneapolis, MN 55455, USA
bolognesi@physics.umn.edu
###### Abstract:
We discuss, and propose a solution for, a still unresolved problem regarding
the breaking from ${\cal N}=2$ super-QCD to ${\cal N}=1$ super-QCD. A mass
term $W=\mu\hbox{\rm Tr}\,\Phi^{2}/2$ for the adjoint field, which classically
does the required breaking perfectly, quantum mechanically leads to a relevant
operator that, in the infrared, makes the theory flow away from pure ${\cal
N}=1$ SQCD. To avoid this problem, we first need to extend the theory from
${\rm SU}(n_{c})$ to ${\rm U}(n_{c})$. We then look for the quantum
generalization of the condition $W^{\prime}(m)=0$, that is, the coincidence
between a root of the derivative of the superpotential $W(\phi)$ and the mass
$m$ of the quarks. There are $2n_{c}-n_{f}$ of such points in the moduli
space. We suggest that with an opportune choice of superpotential, that
selects one of these coincidence vacua in the moduli space, it is possible to
flow from ${\cal N}=2$ SQCD to ${\cal N}=1$ SQCD. Various arguments support
this claim. In particular, we shall determine the exact location in the moduli
space of these coincidence vacua and the precise factorization of the SW
curve.
Super-QCD, Extended supersymmetry breaking, Seiberg-Witten solution, Seiberg
duality
††preprint: FTPI-MINN-08/26; UMN-TH-2705/08
## 1 Introduction
In this paper, we want to discuss a problem of supersymmetry breaking, from
${\cal N}=2$ to ${\cal N}=1$. An issue, which is not completely understood in
the literature, is how to flow from ${\cal N}=2$ super-QCD to pure ${\cal
N}=1$ super-QCD [2]. We shall explain the problem, why it is still unresolved,
and propose a solution for it.
${\cal N}=1$ super-QCD contains the gauge vector multiplet $W_{\alpha}$ for
the gauge group ${\rm SU}(n_{c})$, and $n_{f}$ flavors of massless quarks $Q$
in the fundamental representation and ${\widetilde{Q}}$ in the anti-
fundamental representation. ${\cal N}=2$ super-QCD is an extension of the
previous one. We must add a chiral adjoint multiplet $\Phi$ that together with
$W_{\alpha}$ forms a ${\cal N}=2$ gauge supermultiplet. The quarks $Q$ and
${\widetilde{Q}}^{\dagger}$ fit together to form an ${\cal N}=2$ matter
hypermultiplet. An opportune superpotential ${\widetilde{Q}}\Phi Q$ should
also be added.
At the classical level, both theories have a moduli space of vacua. The ${\cal
N}=2$ moduli space is divided into two distinct parts: the Coulomb branch and
the Higgs branch. On the Coulomb branch, $\phi\neq 0$ and $q,\widetilde{q}=0$.
The rank of the gauge group is preserved. On the Higgs branch, some of the
quarks $q$’s and $\widetilde{q}^{\,}$’s develop an expectation value. The
Higgs branches develop out of various singular sub-manifolds of the Coulomb
branch where the quarks become effectively massless. For what we are
interested now, we can just consider the origin of the moduli space, $\phi=0$,
$q,\widetilde{q}=0$, which is the meeting point between the Coulomb branch and
the maximal Higgs branch.
The ${\cal N}=1$ moduli space is different. First of all, there is no Coulomb
branch, since there is no $\phi$. Second, the Higgs branch is bigger than the
one of ${\cal N}=2$. Now only the $D$-term determines the expectation value of
$q$ and $\widetilde{q}$. The $F$-term constraint $\widetilde{q}q=0$ is now
absent.
The simplest way to break ${\cal N}=2$ supersymmetry, while preserving ${\cal
N}=1$, is to add a mass term in the superpotential for the adjoint chiral
field: $\mu\hbox{\rm Tr}\,\Phi^{2}/2$. The effect of this perturbation is to
lift all the Coulomb branch except from the origin with its maximal Higgs
branch attached. Another effect happens as $\mu$ becomes very, very large. The
potential for the meson $\widetilde{q}q$, coming from the $F$-term of ${\cal
N}=2$, becomes more and more shallow and vanishes in the limit $\mu\to\infty$.
In this limit, the full Higgs branch of ${\cal N}=1$ is thus recovered.
Figure 1: Classical supersymmetry breaking from ${\cal N}=2$ to ${\cal N}=1$
super-QCD. The breaking is obtained by a mass parameter $\mu$ for the field
$\Phi$. At $\mu=0$, we have the ${\cal N}=2$ theory that has a Coulomb branch
and a maximal Higgs branch that develops from the origin of the moduli space
$\phi=0$, $q$, $\widetilde{q}=0$. The mass term lifts the Coulomb branch
leaving only the Higgs branch. As $\mu$ goes to infinity, the Higgs branch is
enhanced by other flat directions, $\widetilde{q}q\neq 0$, and we recover the
full moduli space of ${\cal N}=1$. Figure 2: The notion of origin of the
moduli space does not exist anymore for the quantum version of ${\rm
SU}(n_{c})$ super-QCD. There are various points of maximal singularity labeled
by an integer $r$ that runs from $\widetilde{n}_{c}=n_{f}-n_{c}$ to
$[n_{f}/2]$. Higgs branches of increasing size (quaternionic $r(n_{f}-r)$)
develop from these singularities.
Quantum mechanically, it is a completely different story. The metric of the
Coulomb branch of ${\cal N}=2$ is modified, and the origin of the moduli
space, strictly speaking, does not exist anymore. It is instead split in many
different points out of which various Higgs branches develop (Figure 2). These
singularities are labeled by an integer $r$ that runs from
$\widetilde{n}_{c}=n_{f}-n_{c}$ to $[n_{f}/2]$.111By $[n_{f}/2]$, here and in
the rest of the paper, we mean the integer part of $n_{f}/2$. Adding a mass
perturbation $\mu\hbox{\rm Tr}\,\Phi^{2}/2$, we lift all the Coulomb branch
with the exception of these singular points and their Higgs branches. Every
one of these points gives a particular ${\cal N}=1$ theory, but none of them
flows to pure ${\cal N}=1$ super-QCD in the $\mu\to\infty$ limit.
The mechanism to flow from ${\cal N}=2$ super-QCD to ${\cal N}=1$ super-QCD,
which is very simple in classical theory, does not work in quantum theory. In
this paper, we shall present a proposal to make this mechanism work. This will
certainly open new possibilities in the study of the ${\cal N}=1$ dynamics and
its rich dualities.
* * *
The ideas we shall present originate, in part, from some recent developments
in the theory of the Abelian and non-Abelian heterotic vortex-string. It is
good to say a few words about this subject since it could help to understand
better what follows.
${\rm U}(n_{c})$ ${\cal N}=2$ theories with quark hypermultiplets with mass
$m$, can be put in a weak Higgs phase with a suitable Fayet-Iliopoulos (FI)
term for the ${\rm U}(1)$ gauge field. This term does not break ${\cal N}=2$
supersymmetry; it only breaks the ${\rm SU}(2)_{R}$ symmetry. There is a
supersymmetric vacuum where $\phi=m$, so that the quarks are effectively
massless and can develop a condensate. At the Higgs breaking energy scale, the
dynamics are essentially transferred to the $1+1$ action of the non-Abelian
vortex. Being the vortex half-BPS, the $1+1$ effective action has ${\cal
N}=(2,2)$ supersymmetries on its worldsheet. The tension of the vortex is
proportional to the central charge $T=4\pi\xi$, where $\xi$ is the FI term.
A superpotential for the adjoint field $\Phi$ can be added in order to break
supersymmetry down to ${\cal N}=1$, for example, the single trace of an
holomorphic function ${\cal W}=\hbox{\rm Tr}\,W(\Phi)$. The vacuum structure
is not changed, but now the string has tension
$4\pi\sqrt{\xi^{2}+|W^{\prime}(m)|^{2}}$ and loses its BPS properties.
$W^{\prime}(m)$ is the value of the superpotential evaluated at the mass of
the quarks [7].
Something very special happens when a zero of $W^{\prime}$ coincides with the
mass $m$. In this case, the tension is back to the central charge $T=4\pi\xi$,
and the vortex effective theory re-acquires part of the supersymmetries. The
vortex is half-BPS, and the worldsheet theory has a ${\cal N}=(2,0)$
supersymmetries inherited by the bulk ${\cal N}=1$. Additional studies show
that the strong dynamics on the $1+1$ theory dynamically break supersymmetry
[4, 5, 6].
These studies, up to now, leave an unresolved important question: is the
coincidence condition $W^{\prime}(m)=0$ quantum-mechanically meaningful? To
explain the question better, let us call $a$ one of the roots of $W^{\prime}$.
Classically we can choose to fine tune the parameters so that $a$ is exactly
equal to $m$. The non-trivial question is if this fine tuning is stable or not
under quantum corrections. Note that both $a$ and $m$ are in general
renormalized by instanton corrections. The quantum stability of this
coincidence condition becomes a non-trivial question in the non-Abelian
theory.
The heterotic string gives an ambiguous answer to these questions. On one
hand, there is the enhancement of supersymmetry in the particular case $a=m$.
Symmetry enhancement is in general what protects from quantum corrections. On
the other hand, from the analysis of the strong dynamics of the worldsheet
theory, we know that ${\cal N}=(2,0)$ supersymmetry is dynamically broken. But
this, for the present paper, is not our concern.
* * *
Now back to the main subject of the paper, the $4d$ dynamics without the FI
term. The present paper originates from the question we previously asked, but
for the present work, we shall restrict ourselves to the case of zero Fayet-
Iliopoulos term. We thus want to answer these questions: 1) Is the coincidence
condition $W^{\prime}(m)=0$ quantum-mechanically meaningful? 2) And if it is
so, how to express this coincidence condition in the quantum setup?
We shall see that the solution of this coincidence problem will give us the
right vacuum and superpotential to flow from ${\cal N}=2$ super-QCD to ${\cal
N}=1$ super-QCD.
The previous digression about the heterotic vortex-string gives us the first
clue on how to solve the problem. First of all, ${\rm SU}(n_{c})$ is not the
right environment in which to search; we should extend the theory to ${\rm
U}(n_{c})$. The only price we pay is the loss of asymptotic freedom for the
global ${\rm U}(1)$. But this is not a big problem. We can always think that
the theory is embedded in a bigger and finite theory (for example, a bigger
${\rm SU}(N_{c})$ theory) that at a certain energy scale is broken to ${\rm
U}(n_{c})$.
The ${\cal N}=2$ ${\rm U}(n_{c})$ has, with respect to ${\rm SU}(n_{c})$, one
more dimension in the Coulomb branch, namely the coordinate $u_{1}=\hbox{\rm
Tr}\,\phi$. We have already said that in the Coulomb branch of ${\rm
SU}(n_{c})$ there is no remnant of the classical origin of the moduli space.
There are instead various points labeled by an integer $r$ from
$\widetilde{n}_{c}$ to $[n_{f}/2]$. Things are different extending the moduli
space with one more dimension, $\hbox{\rm Tr}\,\phi$. It is in this bigger
environment that we shall find the quantum analogs of the origin of the moduli
space; actually we shall find $2n_{c}-n_{f}$ of them.
Our goal is now to find the quantum generalization of the coincidence points
defined by $W^{\prime}(m)=0$. In other terms, when a root $a$ of $W^{\prime}$
coincides with the mass of the quarks $m$. When $a$ and $m$ are generic, we
have two kinds of vacua. One is a Coulomb vacuum, with $\phi=a$, and the quark
has effective mass $a-m$. The other is a Higgs vacuum where $\phi=m$ (color-
flavor locking), and the quarks condense: $\widetilde{q}q=W^{\prime}(m)$. In
the case of coincidence, we have $\phi=m=a$. The quarks are massless and do
not condense.
Quantum mechanically, the quark condensate, in a generic $r$ vacuum, is given
by222The case $r=\widetilde{n}_{c}$ is an exception to this rule. Here there
is an extra massless particle, and the dual-quark can thus have zero
condensate. [3]
$\widetilde{q}q=r\left.\sqrt{{W^{\prime}(z)}^{2}+f(z)}\right|_{z=m}\ .$ (1)
Where the terms in the square root are given by the factorization of the
Seiberg-Witten curve,
$\displaystyle{y}^{2}$ $\displaystyle=$
$\displaystyle\frac{1}{4}P_{n_{c}}(z)^{2}-\Lambda^{2n_{c}-n_{f}}\prod_{i=1}^{n_{f}}(z-m_{i})$
(2) $\displaystyle=$
$\displaystyle\frac{1}{4{g_{k}}^{2}}\left({W{}^{\,\prime}}(z)^{2}+f(z)\right)H_{n_{c}-k}(z)^{2}\
.$
$H(v)^{2}$ is the polynomial containing the double roots corresponding to
particles that become massless. The rest is contained in
${W^{\prime}(z)}^{2}+f(z)$ where $W$ is the classical superpotential and
$f(z)$ is the quantum modification. The effect of $f(z)$ is to split the
roots, otherwise doubled, of $W^{\prime}(z)^{2}$. For the vacua we are
interested in, the $r$ vacua, the superpotential has degree $k=2$. So quantum
mechanically the roots of $W^{\prime}(z)$ are split in two. To find a
generalization of the coincidence points, we should take one of these zeros
and, moving in the $\hbox{\rm Tr}\,\phi$ direction, make it collide with the
bunch of zeros at $z=m$ corresponding to the quark singularity (Figure 3).
Figure 3: In a generic $r$ vacuum (in the example we have the
$n_{c},n_{f}=6,6$ and $r=3$), all the roots of the Seiberg-Witten curve are
paired, apart from two of them. These two roots correspond to the polynomial
${W{}^{\,\prime}}(z)^{2}+f(z)$ in the curve factorization. To obtain the
coincidence point, we need to move one of these two unpaired roots and make it
coincide with the bunch of roots at the quark mass. We can achieve this
changing the value of the coordinate $\hbox{\rm Tr}\,\phi$ in the moduli
space. In Section 5, we shall find the precise factorization of the curve in
these coincidence vacua.
Doing this, we get something else for free: all the $r$ vacua go to the same
coincidence points. This is one of the strongest indications that tells us
that these $2n_{c}-n_{f}$ points are the quantum generalization of the origin
of the moduli space in classical theory.
Figure 4: Moving away from the $\hbox{\rm Tr}\,\phi=0$ section of the moduli
space, it is possible to make all the $r$ vacua to collide in the coincidence
points. The Higgs branch that emanates from these points is the maximal one,
and all the other Higgs branches are incorporated into it. There are
$2n_{c}-n_{f}$ of these points and are situated at $\hbox{\rm
Tr}\,\phi=(2n_{c}-n_{f})\Lambda e^{i2\pi k/(2n_{c}-n_{f})}$.
Every $r$ vacuum is a point in which the curve has a multiple root $z^{2r}$
corresponding to massless quarks and all the other roots, except two, are
doubled. We then change the $\hbox{\rm Tr}\,\phi$ coordinate, keeping all the
pairing of the roots, and we move until one of the unpaired roots collides
with the bunch of quark zeros $z^{2r}$. We can do the same for all the $r$
vacua. They all intersect in the coincidence points, and the Higgs branches
are all included in the maximal one (Figure 4).
* * *
We organize the paper as follows. In Section 2, we discuss the simple example
of super-QED. Now there is no strong dynamics in the infrared, and everything
works out smoothly. In Section 3, we introduce the non-Abelian ${\rm
U}(n_{c})$ theories. We provide the needed information for the rest of the
paper. In Section 4 we describe the quantum coincidence in the first non-
trivial examples, starting from $2$ colors and $2$ flavors. In Section 5, we
describe the coincidence points in the general case. An explicit formula shall
be found for them and for the superpotential that select them. In Section 6,
we take a look from the MQCD perspective. We conclude in Section 7 with
discussion about the renormalization group (RG) flow and summarize the
arguments in favor of our claim.
## 2 Warm-up with the Abelian Theory
A good place to start, in explaining the notion of coincidence points, is the
${\rm U}(1)$ Abelian theory. We shall now introduce these three theories:
${\cal N}=2$ super-QED, broken ${\cal N}=2$ super-QED by a mass term, and pure
${\cal N}=1$ super-QED. These theories are infrared free, so there is no
strong dynamics that could chance the classical vacuum structure and phases.
The notion of coincidence point and its quantum stability are thus
straightforward.
The action for ${\cal N}=2$ super-QED, expressed in ${\cal N}=1$ superfields
notation, is a sum of the Kahler term plus a superpotential term
$\displaystyle{\cal L}$ $\displaystyle=$ $\displaystyle\int d^{2}\theta
d^{2}\bar{\theta}\,\left(\frac{1}{e^{2}}\Phi^{\dagger}\Phi+Q^{\dagger}_{i}e^{V}Q^{i}+{\widetilde{Q}}^{\dagger
i}e^{-V}{\widetilde{Q}}_{i}\right)+$ (3) $\displaystyle+\int
d^{2}\theta\,\left(\frac{1}{4e^{2}}W^{\alpha}W_{\alpha}+{\cal
W}(\Phi,Q,{\widetilde{Q}})\right)+{\rm h.c.}\ ,$
where
${\cal W}(\Phi,Q,{\widetilde{Q}})=\sqrt{2}\left({\widetilde{Q}}\Phi
Q-m{\widetilde{Q}}_{i}Q^{i}\right)\ .$ (4)
The potential for the scalar fields is the sum of $F$ terms and the $D$ term
$V=2|(\phi-m)q^{i}|^{2}+2|(\phi-m)\widetilde{q}_{i}|^{2}+2e^{2}|\widetilde{q}_{i}q^{i}|^{2}+\frac{e^{2}}{2}(|q^{i}|^{2}-|\widetilde{q}_{i}|^{2})^{2}\
.$ (5)
The index $i$ represents the flavor and runs from $1$ to $n_{f}$. The moduli
space consists of a Coulomb branch and a Higgs branch. The Coulomb branch is
parameterized by the expectation value of $\phi$, and the quark condensate
instead vanishes. When $\phi=m$, the effective quark mass vanishes and a Higgs
branch develops from the Coulomb one. The Higgs branch is a manifold
parameterized by the following algebraic equations
$\displaystyle\widetilde{q}_{i}q^{i}$ $\displaystyle=$ $\displaystyle 0\ ,$
$\displaystyle|q^{i}|^{2}-|\widetilde{q}_{i}|^{2}$ $\displaystyle=$
$\displaystyle 0\ .$ (6)
For $n_{f}=1$ there is no Higgs branch. For greater $n_{f}$, it is a manifold
of complex dimension $2n_{f}-2$. Higgs branches, in ${\cal N}=2$ theories, are
conic hyper-Kahler manifolds (in ${\cal N}=1$ theories are only Kahler).
Now we break ${\cal N}=2$ to ${\cal N}=1$ by means of the superpotential
$W(\Phi)$ that is a holomorphic function of $\Phi$
${\cal W}(\Phi,Q,{\widetilde{Q}})=\sqrt{2}\left({\widetilde{Q}}\Phi
Q-m{\widetilde{Q}}_{i}Q^{i}-W(\Phi)\right)\ .$ (7)
The potential for the scalar fields is now modified into
$V=2|(\phi-m)q^{i}|^{2}+2|(\phi-m)\widetilde{q}_{i}|^{2}+2e^{2}|\widetilde{q}_{i}q^{i}-W{}^{\,\prime}(\phi)|^{2}+\frac{e^{2}}{2}(|q^{i}|^{2}-|\widetilde{q}_{i}|^{2})^{2}\
.$ (8)
There are now two kinds of vacua. One is when the value of $\phi$ is equal to
some zero of the derivative of the superpotential. We denote $a_{j}$ these
roots so that $W^{\prime}(z)=\prod_{j=1}^{k}(z-a_{j})$ where $k+1$ is the
degree of the superpotential. There are $k$ vacua where $\phi=a_{i}$. These
are Coulomb vacua because the expectation values of $q$ and $\widetilde{q}$
are zero. Then there is a Higgs vacuum where $\phi$ is locked to the
hypermultiplet mass $m$. Here $q$ and $\widetilde{q}$ develop an expectation
value
$\widetilde{q}q={W}^{\prime}(m)\ ,\qquad q=\widetilde{q}^{\,*}\ .$ (9)
The number of vacua is thus $k+1$. They all preserve supersymmetry.
Things are different when one root of $W^{\prime}$, say $a_{1}$, is exactly
equal to the hypermultiplet mass. From here on, we shall use the word
coincidence to indicate this particular circumstance. In this coincidence
case, we have only $k$ vacua instead of $k+1$. $k-1$ of them are Coulomb
vacua, and as before, they arise when $\phi=a_{j}$ for $j=2,\dots k$. The last
vacuum is when $\phi=m=a_{1}$. It is a kind of hybrid between the Higgs and
the Coulomb vacua previously described. The phase is Coulomb, because the
expectation value of the quark condensate vanishes. But it is a Coulomb phase
of a different nature from the previous one. In the previous case, the quark
hypermultiplet was massive, with effective mass equal to $a_{1}-m$. So at low
energy we had only a gauge vector multiplet (and being ${\rm U}(1)$ the gauge
coupling is finite and the theory is free). In the coincidence case, the
effective quark mass is $a_{1}-m=0$, so at low energy the ${\rm U}(1)$ vector
multiplet is coupled to a massless quark (and being ${\rm U}(1)$ the gauge
coupling runs to zero in the infrared). Now what about the quantum stability?
In principle, we can always tune the parameters of the bare Lagrangian so that
we have a coincidence between the mass and a root of the superpotential. The
question is if this tuning is stable or not under quantum corrections.
Parameters in the superpotential, like the roots $a_{j}$ and the mass $m$, are
subject to the non-renormalization theorem and thus unchanged by perturbative
quantum correction. They could be modified by the non-perturbative dynamics
(that will be the case in the following part of the paper), but for the
Abelian case we do not have to worry about that. This implies the quantum
stability of this coincidence.
We now consider the quadratic superpotential
$W(\phi)=\mu\left(\frac{\phi^{2}}{2}-m\phi\right)\ ,$ (10)
so that the derivative is the linear polynomial $W^{\prime}(z)=2\mu(z-m)$, and
we choose the root to coincide exactly with the quark mass. Integrating out
the superfield $\Phi$, we get
$\Phi=m+\frac{1}{\mu}{\widetilde{Q}}Q\ ,$ (11)
and the effective superpotential
${\cal W}(Q,{\widetilde{Q}})=\sqrt{2}\left(\frac{1}{2}\mu
m^{2}+\frac{1}{2\mu}{\widetilde{Q}}Q\,{\widetilde{Q}}Q\right)\ .$ (12)
The potential for the scalar fields is
$V=\frac{2}{\mu^{2}}|\widetilde{q}q\,q^{i}|^{2}+\frac{2}{\mu^{2}}|\widetilde{q}q\,\widetilde{q}_{i}|^{2}+\frac{e^{2}}{2}(|q^{i}|^{2}-|\widetilde{q}_{i}|^{2})^{2}\
.$ (13)
With a gauge and flavor transformation, we can always bring a generic solution
into the form
$\displaystyle q$ $\displaystyle=$ $\displaystyle(k,0,0,\dots,0)\ ,$
$\displaystyle\widetilde{q}$ $\displaystyle=$ $\displaystyle(0,k,0,\dots,0)\
,\qquad k\in\mathbb{R}^{+}\ .$ (14)
At this point, the potential for the $\widetilde{q}q$ meson field is
$V(\widetilde{q}q)=\frac{4}{\mu^{2}}(\widetilde{q}q)^{2}k^{2}+\dots\ .$ (15)
In the $\mu\to\infty$, we have an enhancement of the moduli space because the
$F$ term condition $\widetilde{q}q=0$ loses progressively its weight. The
direction $\widetilde{q}q\neq 0$, which is generically lifted by the potential
(13), becomes more and more shallow and finally flat in the $\mu\to\infty$
limit. Note that at the origin of the Higgs branch (where it touches the
Coulomb branch), the meson $\widetilde{q}q$ is always massless, but the
$\widetilde{q}q$ direction is still lifted by the higher order potential
quartic in the meson field.
Pure ${\cal N}=1$ super-QED has the following Lagrangian
$\displaystyle{\cal L}$ $\displaystyle=$ $\displaystyle\int d^{2}\theta
d^{2}\bar{\theta}\,\left(Q_{i}^{\dagger}e^{V}Q^{i}+{\widetilde{Q}}^{\dagger
i}e^{-V}{\widetilde{Q}}_{i}\right)+$ (16) $\displaystyle+\int
d^{2}\theta\,\frac{1}{4e^{2}}W^{\alpha}W_{\alpha}+{\rm h.c.}\ ,$
and the potential for the scalar fields is given just by the $D$ term
$V=\frac{e^{2}}{2}(|q^{i}|^{2}-|\widetilde{q}_{i}|^{2})^{2}\ .$ (17)
With a gauge and flavor transformation, we can bring a solution into the form
$\displaystyle q=(k,0,0,\dots,0)\ ,$
$\displaystyle\widetilde{q}=({\widetilde{k}},\lambda,0,\dots,0)\ ,$
$\displaystyle k,{\widetilde{k}},\lambda\in\mathbb{R}^{+}\ ,\qquad
k^{2}={\widetilde{k}}^{2}+\lambda^{2}\ .$ (18)
So to summarize here what happens to the moduli space for various values of
$\mu$:
* •
The ${\cal N}=2$ theory ($\mu=0$) has a Coulomb branch with complex dimension,
one parameterized by $\phi$. A Higgs branch with complex dimension $2n_{f}-2$
develops from the origin $\phi=0$.
* •
Switching on the mass $\mu$, the Coulomb branch is lifted, and only the origin
$\phi=0$ survives with its Higgs branch attached.
* •
In the $\mu\to\infty$ limit, the $F_{\phi}$ condition $\widetilde{q}q=0$ is no
more effective due to the $1/\mu$ suppression. The Higgs branch is enhanced
and becomes $2n_{f}-1$ dimensional. We thus smoothly recover the moduli space
of ${\cal N}=1$ SQED.
## 3 Non-Abelian Theory
The non-Abelian ${\rm U}(n_{c})$ super QCD has the following action333All the
time we use Tr in the paper, we mean a trace in the color space.
$\displaystyle{\cal L}$ $\displaystyle=$ $\displaystyle\int d^{2}\theta
d^{2}\bar{\theta}\,\frac{2}{e^{2}}\hbox{\rm Tr}\,(\Phi^{\dagger}e^{V}\Phi
e^{-V})+\sum_{i=1}^{n_{f}}\,(Q_{i}^{\dagger}e^{V}Q^{i}+\widetilde{Q}_{i}e^{-V}{\widetilde{Q}}^{\dagger
i})+$ (19) $\displaystyle+\int d^{2}\theta\,\left(\frac{1}{2e^{2}}\hbox{\rm
Tr}\,(W^{\alpha}W_{\alpha})+{\cal W}(\Phi,Q,{\widetilde{Q}})\right)+{\rm
h.c.}\ ,$
where the superpotential is
${\cal
W}(\Phi,Q,{\widetilde{Q}})=\sum_{i=1}^{n_{f}}\,\sqrt{2}(\widetilde{Q}_{i}\Phi
Q^{i}-m_{i}\widetilde{Q}_{i}Q^{i})\ ,$ (20)
and $m^{i}$ are the masses for the flavors with the index $i=1,\dots,n_{f}$.
The charges and symmetries of the fields are given in Table 1.
| ${\rm U}(n_{c})$ | $\times$ | ${\rm SU}(n_{f})$ | $\times$ | ${\rm U}(1)_{R}$ | $\times$ | ${\rm U}(1)_{J}$
---|---|---|---|---|---|---|---
$Q$ | $\bf n_{c}$ | | $\bf n_{f}$ | | $0$ | | $1$
${\widetilde{Q}}$ | $\bf\bar{n}_{c}$ | | $\bf\bar{n}_{f}$ | | $0$ | | $1$
$\Phi$ | $\bf adj$ | | $\bf 1$ | | $2$ | | $0$
$\Lambda_{\mbox{\tiny${\cal N}=2$ }}^{2n_{c}-n_{f}}$ | $\bf 1$ | | $\bf 1$ | | $2(2n_{c}{-}n_{f})$ | | $0$
Table 1: Fields, symmetries, and charges for ${\cal N}=2$ SQCD.
The gauge group is ${\rm U}(n_{c})={\rm SU}(n_{c})\times{\rm
U}(1)/\mathbb{Z}_{n_{c}}$. At a very high energy scale, which we call
$\Lambda_{\mbox{\tiny$\mathrm{cutoff}$}}$, the gauge couplings for the Abelian
${\rm U}(1)$ and for the non-Abelian ${\rm SU}(n_{c})$ are the same.444At this
scale, the theory becomes part of a bigger, asymptotically free, theory. An
example could be an ${\rm SU}(n_{c}+1)$ gauge theory that is broken to ${\rm
U}(n_{c})$ at the scale $\Lambda_{\mbox{\tiny$\mathrm{cutoff}$}}$. Then the
gauge couplings run in opposite directions. The non-Abelian running is given
by
$\frac{1}{e_{\mbox{\tiny${\rm
SU}(n_{c})$}}^{2}}=(2n_{c}-n_{f})\log{\left(\frac{\Lambda_{\mbox{\tiny${\cal
N}=2$ }}}{\mu}\right)}\ ,$ (21)
and becomes strong in the infrared. The ${\rm U}(1)_{R}$ symmetry is anomalous
and broken to $\mathbb{Z}_{2n_{c}-n_{f}}$ by instanton zero modes. This fact,
as usual, can be elegantly incorporated giving a ${\rm U}(1)_{R}$ charge to
the dynamical scale $\Lambda_{\mbox{\tiny${\cal N}=2$ }}$ as in Table 1.
We then break supersymmetry adding a superpotential $W(\Phi)$ for the adjoint
field $\Phi$. The superpotential term becomes
${\cal
W}(\Phi,Q,{\widetilde{Q}})=\sum_{i=1}^{n_{f}}\,\sqrt{2}(\widetilde{Q}_{i}\Phi
Q^{i}-m_{i}\widetilde{Q}_{i}Q^{i})-\sqrt{2}\hbox{\rm Tr}\,W(\Phi)\ ,$ (22)
where
$W(z)=\sum_{j=0}^{k}\,\frac{g_{j}}{j+1}z^{j+1}\ ,\qquad
W^{\prime}(z)=g_{k}\prod_{j=1}^{k}\,(z-a_{j})\ .$ (23)
The adjoint scalar field $\phi$, from the $D$ term, is a $n_{c}\times n_{c}$
complex-hermitian matrix that can thus be diagonalized. The eigenvalues follow
the same rule explained for the Abelian case. In the supersymmetric vacua, the
diagonal elements of the adjoint field must be equal to a flavor mass $m_{i}$
(the color-flavor locking case) or to a root $a_{j}$ of $W^{\prime}$. There
can also be some degeneracy in the eigenvalues. A generic solution is
$\langle\phi\rangle=\left(\begin{array}[]{cccc}m_{i}\mathbf{1}_{r_{i}}&&&\\\
&\ddots&&\\\ &&a_{j}{\mathbf{1}}_{n_{j}}&\\\ \end{array}\right)\
,\qquad\sum_{j=1}^{k}\,n_{j}+\sum_{i=1}^{n_{f}}\,r_{i}=n_{c}\ .$ (24)
The adjoint field breaks the gauge group to $\prod_{i=1}^{n_{f}}{\rm
U}(r_{i})\times\prod_{j=1}^{k}{\rm U}(n_{j})$. The quarks locked to the
adjoint field are effectively massless and condense after the perturbation of
$W(\Phi)$. The quark condensate $W{}^{\,\prime}(m_{i})$ breaks the residual
gauge groups $\prod_{i=1}^{n_{f}}{\rm U}(r_{i})$. In the low-energy, we are
thus left with a $\prod_{j=1}^{k}{\rm U}(n_{j})$ super-Yang-Mills theory
without quarks.
In the case of coincidence, we have to consider also another possibility; the
previous analysis is not exhaustive. We call $h$ the number of coincidences.
So we choose the last $h$ roots of $W^{\prime}$ to be equal to the last $h$
masses
$s_{l}=a_{k-h+l}=m_{n_{f}-h+l}\ ,\qquad l=1,\dots,h\ .$ (25)
The other roots $a_{j}$ for $j=1,\dots,k-h$ and the other masses $m_{i}$ for
$i=1,\dots,n_{f}-h$ are generic (no coincidence). Now the eigenvalues of
$\phi$ can be equal to some $s_{l}$ with $l=1,\dots,h$, or to $a_{j}$ with
$j=1,\dots,k-h$ or to $m_{i}$ with $i=1,\dots,n_{f}-h$. The generic structure
is thus
$\langle\phi\rangle=\left(\begin{array}[]{ccccc}s_{l}\mathbf{1}_{n_{l}}&&&&\\\
&\ddots&&&\\\ &&a_{j}\mathbf{1}_{n_{j}}&&\\\ &&&\ddots&\\\
&&&&m_{i}\mathbf{1}_{n_{i}}\end{array}\right)\ ,$ (26)
and the gauge group is broken as
${\rm U}(n_{c})\rightarrow\prod_{l=1}^{h}{\rm
U}(n_{l})\times\prod_{j=1}^{k-h}{\rm U}(n_{j})\times\prod_{i=1}^{n_{f}}{\rm
U}(n_{i})\ .$ (27)
The various residual gauge groups have different fate according to the three
cases. In the first case, we have at low energy ${\rm U}(n_{l})$ SQCD with a
certain number of _massless_ flavors. The second case leaves ${\rm U}(n_{j})$
pure SYM, which has $n_{j}$ discrete supersymmetric vacua where the gauge
group is in the confinement phase. In the third case, the quarks acquire an
expectation value $\widetilde{Q}Q=W^{\prime}(m_{i})$. The gauge group is thus
in the Higgs phase. The fact that the first case has a completely different
phase structure is an indication that the coincidence we talking about should
be taken quantum mechanically seriously.
We now break to ${\cal N}=1$ supersymmetry by turning on a bare mass $\mu$ for
the adjoint superfield $\Phi$. In the microscopic theory, this corresponds to
an ${\cal N}=1$ theory with a superpotential
${\cal W}=\sqrt{2}\,\left({\widetilde{Q}}_{i}\Phi Q^{i}-{\mu\over 2}\hbox{\rm
Tr}\,\Phi^{2}\right).$ (28)
For $\mu\gg\Lambda_{\mbox{\tiny${\cal N}=2$ }}$ we can integrate $\Phi$ out in
a weak-coupling approximation, obtaining an effective quartic superpotential
${\cal W}^{\prime}=\frac{1}{{\mu\sqrt{2}}}\hbox{\rm
Tr}\,(Q{\widetilde{Q}}Q{\widetilde{Q}})\ .$ (29)
Classically, in the limit $\mu{\rightarrow}\infty$ this superpotential becomes
negligible, and we find ${\cal N}=1$ ${\rm U}(n_{c})$ super-QCD with $n_{f}$
flavors and no superpotential. Quantum mechanically, we cannot say that since
(29) is a relevant operator and in general, as we shall see more in detail,
makes the theory to flow away from the right coincidence point.
Pure ${\cal N}=1$ SQCD has no adjoint field $\Phi$, and the Lagrangian is
$\displaystyle{\cal L}$ $\displaystyle=$ $\displaystyle\int d^{2}\theta
d^{2}\bar{\theta}\,\sum_{i=1}^{n_{f}}\,(Q^{\dagger}_{i}e^{V}Q^{i}+\widetilde{Q}_{i}e^{-V}\widetilde{Q}^{\dagger
i})+$ (30) $\displaystyle+\int d^{2}\theta\,\frac{1}{2e^{2}}\hbox{\rm
Tr}\,(W^{\alpha}W_{\alpha})+{\rm h.c.}\ .$
Fields and charges are given in Table 2.
| ${\rm SU}(n_{c})$ | $\times$ | ${\rm SU}(n_{f})_{\rm L}$ | $\times$ | ${\rm SU}(n_{f})_{\rm R}$ | $\times$ | ${\rm U}(1)_{R}$
---|---|---|---|---|---|---|---
$Q$ | $\bf n_{c}$ | | $\bf n_{f}$ | | $\bf 1$ | | $0$
${\widetilde{Q}}$ | $\bf\bar{n}_{c}$ | | $\bf 1$ | | $\bf\bar{n}_{f}$ | | $0$
$\Lambda_{\mbox{\tiny${\cal N}=1$ }}^{3n_{c}-n_{f}}$ | $\bf 1$ | | $\bf 1$ | | $\bf 1$ | | $(3n_{c}-n_{f})$
Table 2: Fields, symmetries and charges for ${\cal N}=1$ SQCD.
Note that now the non-Abelian flavor symmetry is enhanced to ${\rm
SU}(n_{f})_{\rm L}\times{\rm SU}(n_{f})_{\rm R}$. Left and right quarks can be
rotated independently. The running of the non-Abelian gauge coupling, due to
the absence of the field $\phi$, is now a little bit faster
$\frac{1}{e_{\mbox{\tiny${\rm
SU}(n_{c})$}}^{2}}=(3n_{c}-n_{f})\log{\left(\frac{\Lambda_{\mbox{\tiny${\cal
N}=1$ }}}{\mu}\right)}\ .$ (31)
The ${\rm U}(1)_{R}$ symmetry is anomalous and broken to
$\mathbb{Z}_{3n_{c}-n_{f}}$.
The matching of the dynamical scales between the ${\cal N}=2$ and the ${\cal
N}=1$ theory is given by
$\Lambda_{\mbox{\tiny${\cal N}=1$
}}^{3n_{c}-n_{f}}=\mu^{n_{c}}\Lambda_{\mbox{\tiny${\cal N}=2$
}}^{2n_{c}-n_{f}}\ .$ (32)
This can be inferred by the running of the coupling constants and by the ${\rm
U}(1)_{R}$ anomaly charges.
* * *
In the next subsection we review some basic information that shall be needed
in the rest of the paper. In Subsection 3.1, we introduce the moduli space,
and in particular the Higgs branches for the various theories. In Subsection
3.2, we recall the Seiberg-Witten solution for the dynamic on Coulomb branch.
In Subsection 3.3, we describe the particular case of the $r$ vacua and the
formula for the quark condensate.
### 3.1 Moduli Space
The moduli space of vacua is defined by the $D$-term equations
$\displaystyle[\phi,\phi^{\dagger}]=0\ ,$ $\displaystyle q_{a}^{\,i}\
q_{i}^{\,\dagger b}-\widetilde{q}_{a}^{\,\dagger i}\
\widetilde{q}_{i}^{\,b}=0\ ,$ (33)
and the $F$-term equations
$\displaystyle q_{a}^{\,i}\ \widetilde{q}_{i}^{\,b}$ $\displaystyle=$
$\displaystyle 0\ ,$ $\displaystyle\phi_{a}^{\,b}\ q_{b}^{\,i}$
$\displaystyle=$ $\displaystyle 0\ ,$ $\displaystyle\widetilde{q}_{i}^{\,b}\
\phi_{b}^{\,a}$ $\displaystyle=$ $\displaystyle 0\ .$ (34)
From now on, we shall choose all the masses $m^{i}$ to be equal to zero.
The moduli space can be divided into two different branches. The first is the
Coulomb branch where the quarks condensate vanishes and the gauge group
preserves its rank. From the first of (33), it follows that the adjoint field
$\phi$ can be diagonalized by a gauge transformation. The Coulomb branch has
complex dimension $n_{c}$ and can be parameterized by the gauge invariant
coordinates
$u_{j}=\frac{1}{j}\langle\hbox{\rm Tr}\,\phi^{\,j}\rangle\ ,\qquad
j=1,\dots,n_{c}\ .$ (35)
The Coulomb branch is a Kahler manifold and develops singularities where some
particles, vector multiplets or matter hypermultiplets, become massless. The
metric is modified by quantum correction, and the solution is encoded in the
Seiberg-Witten curve.
Higgs branches are conic hyper-Kahler manifolds that develop from some of the
singularities of the Coulomb branch where two or more flavors of quarks become
massless. Two are the important properties of the Higgs branches: 1) the
metric does not have any dependence on $\phi$; 2) the metric does not receive
quantum corrections. Quantum corrections can only change the point on the
Coulomb branch where the Higgs branch develops and the pattern of
intersections between the various branches; they cannot modify the Higgs
branch itself [2].
There is no baryonic branch since we are working in ${\rm U}(n_{c})$ rather
than ${\rm SU}(n_{c})$. There are non-baryonic branches and they develop where
$\phi=(0,\dots,0,\phi_{r+1},\dots,\phi_{n_{c}})\ .$ (36)
After a gauge and flavor rotation, the quarks can be brought in the form
$\displaystyle
q=\left(\begin{array}[]{cccccccc}k_{1}&&&0\phantom{{}_{r}}&&&0&\\\
&\ddots&&&\ddots&&&\ddots\\\ &&k_{r}&&&0\phantom{{}_{r}}&&\\\
&&&&&&&\end{array}\right)\
,\qquad\widetilde{q}^{\,t}=\left(\begin{array}[]{cccccccc}0\phantom{{}_{r}}&&&k_{1}&&&0&\\\
&\ddots&&&\ddots&&&\ddots\\\ &&0\phantom{{}_{r}}&&&k_{r}&&\\\
&&&&&&&\end{array}\right)\ ,$ (45) $\displaystyle k_{i}\in\mathbb{R}^{+}\ ,$
(46)
where $r\leq[n_{f}/2]$. These non-baryonic $r$ branches are the ones that
develop from the sub-manifold (36) of the Coulomb branch. The $r$ Higgs branch
preserves a ${\rm U}(n_{f}-r)$ subgroup of the flavor symmetry. Part of the
flat directions of the Higgs branch are given by the Goldstone bosons of the
broken global symmetries. The complex dimension of the branch is
$2r(n_{f}-r)$. Note that a Higgs branch can contain, as a sub-manifold, all
the other branches with smaller $r$.
For the moduli space of pure ${\cal N}=1$ super-QCD, there is only the
$D$-term
$q_{a}^{\,i}\ q_{i}^{\,\dagger b}-\widetilde{q}_{a}^{\,\dagger i}\
\widetilde{q}_{i}^{\,b}=0\ .$ (47)
Up to gauge and global transformation, the quarks can be brought in the
form555See [8] for a complete review.
$\displaystyle q=\left(\begin{array}[]{ccccccc}k_{1}&&&&0\phantom{{}_{1}}&&\\\
&k_{2}&&&&\ddots&\\\ &&\ddots&&&&0\phantom{{}_{n_{f}-n_{c}}}\\\
&&&k_{n_{c}}&&&\end{array}\right)\
,\qquad\widetilde{q}^{\,t}=\left(\begin{array}[]{ccccccc}{\widetilde{k}}_{1}&&&&\lambda_{1}&&\\\
&{\widetilde{k}}_{2}&&&&\ddots&\\\ &&\ddots&&&&\lambda_{n_{f}-n_{c}}\\\
&&&{\widetilde{k}}_{n_{c}}&&&\end{array}\right)\ ,$ (56) $\displaystyle
k_{i},{\widetilde{k}}_{i},\lambda_{i}\in\mathbb{R}^{+}\ ,$
$\displaystyle|k_{i}|^{2}=|{\widetilde{k}}_{i}|^{2}+|\lambda_{i}|^{2}\
,\quad\forall i\ .$ (57)
The complex dimension is $2n_{c}n_{f}-n_{c}^{2}$. Note that now there is also
a baryonic branch that was absent in the ${\cal N}=2$ case.
As in the Abelian case, we can break ${\cal N}=2$ with a mass term
$\mu\hbox{\rm Tr}\,\Phi^{2}/2$ and flow from ${\cal N}=2$ to pure ${\cal
N}=1$. So to summarize here what happens, this time only classically, to the
moduli space;
* •
The ${\cal N}=2$ theory has a Coulomb branch with complex dimension $n_{c}$
and an equal number of massless vector multiplets. Non-baryonic Higgs branches
labeled by $r$ develop from singularities of the Coulomb branch. They are
hyper-Kahler manifolds. In particular at the origin $\phi=0$, the maximal
Higgs branch $r=[n_{f}/2]$ develops. It is given by the quark expectation
values $q$ and $\widetilde{q}$, minus the $F_{\phi}$ conditions, divided by
the complexified gauge group $G^{\rm c}$
$\mathbb{C}^{2n_{c}n_{f}}-\\{\widetilde{q}^{\,a}_{i}\ q^{\,i}_{b}=0\\}/G^{\rm
c}\ .$ (58)
The quaternionic dimension of the maximal Higgs branch is
$[n_{f}/2]n_{f}-[n_{f}/2]^{2}$.
* •
The mass term $\mu\hbox{\rm Tr}\,\Phi^{2}/2$ in the superpotential lifts all
the Coulomb branch with the exception of the origin $\Phi=0$ and the maximal
Higgs branch attached to it.
* •
At $\mu\to\infty$ the $F_{\phi}$ conditions loses its effectiveness. The Higgs
branch is thus parameterized by
$\mathbb{C}^{2n_{c}n_{f}}/G^{\rm c}\ .$ (59)
The Higgs branch is enhanced to the $2n_{c}n_{f}-n_{c}^{2}$ complex
dimensional space. We thus smoothly recover the moduli space of pure ${\cal
N}=1$ SQCD.
### 3.2 Quantum Dynamics on the Coulomb Branch
We now recall some basic properties of the quantum solution of the Coulomb
branch given by the Seiberg-Witten curve [9]. First we consider ${\rm
SU}(n_{c})$ with $n_{f}$ flavors. The curve $\Sigma_{\mbox{\tiny${\cal N}=2$
}}$ is the Riemann surface
${y}^{2}=\frac{1}{4}\det(z-\Phi)^{2}-\Lambda^{2n_{c}-n_{f}}z^{n_{f}}\ ,$ (60)
double cover of the $z$ plane with genus $g=n_{c}-1$. At low energy, one has
$n_{c}-1$ ${\rm U}(1)$ gauge multiplets, and we call their scalar components
$a_{i}$, where $i=1,\dots,n_{c}-1$. The moduli space is a $n_{c}-1$
dimensional complex manifold ${\cal M}_{\mbox{\tiny${\rm SU}(n_{c})$ }}$,
parameterized by the gauge invariant coordinates $u_{j}$ given in (35). The SW
solution is expressed as the function of the coordinates $s_{k}$666 The
relationship between $u_{j}$ and $s_{k}$, important in what follows, is
encoded in a single one:
$P_{n_{c}}(z)=z^{n_{c}}\exp{\left(-\sum_{j=1}^{\infty}\,\frac{u_{j}}{z^{j}}\right)_{+}}$,
whereby $(\;)_{+}$ we mean that we discard the negative power expansion. given
by
$P_{n_{c}}(z)=\sum_{k=0}^{n_{c}}\,s_{k}z^{n_{c}-k}\ ,$ (61) $s_{0}=1\ ,\quad
s_{1}=0\ ,\quad
s_{k}=(-)^{k}\sum_{i_{1}<\dots<i_{k}}\,\phi_{i_{1}}\dots\phi_{i_{k}}\ .$ (62)
Since $\Sigma_{\mbox{\tiny${\cal N}=2$ }}$ is a genus $n_{c}-1$ Riemann
surface, we can choose $n_{c}-1$ independents holomorphic differentials:
$\lambda_{j}\propto\frac{z^{n_{c}-j}dz}{y}\ ,\qquad j=2,\dots,n_{c}\ .$ (63)
Each $a_{i}$ corresponds to an $\alpha_{i}$ cycle on $\Sigma_{{\cal N}=2}$,
while its dual $a_{Dj}$ corresponds to a $\beta_{j}$ cycle chosen in such a
way that the intersection is $\langle\alpha_{i},\beta_{j}\rangle=\delta_{ij}$.
The SW solution is given by the period integrals
$\frac{\partial a_{i}}{\partial s_{j}}=\oint_{\alpha_{i}}\lambda_{j}\
,\qquad\frac{\partial a_{Di}}{\partial s_{j}}=\oint_{\beta_{i}}\lambda_{j}\ .$
(64)
Now we study the ${\rm U}(n_{c})$ theory with $n_{f}$ flavors, and we will see
that the solution can be easily incorporated into the previous ones, with a
few modifications. The low-energy theory has one more ${\rm U}(1)$ factor that
comes from the decomposition ${\rm U}(n_{c})={\rm SU}(n_{c})\times{\rm
U}(1)/\mathbb{Z}_{n_{c}}$, and we denote its scalar component with
$a_{n_{c}}$. This factor has no strong dynamics: in the $n_{f}=0$ case, it is
completely free, while in the $n_{f}\neq 0$ case, it is infrared free. The
moduli space ${\cal M}_{\mbox{\tiny${\rm U}(n_{c})$ }}$ has one dimension more
and is parameterized by $u_{1}$ in (35). The Riemann surface is the same given
in (60), but here $\phi$ can have non-zero trace and
$\Sigma_{\mbox{\tiny${\cal N}=2$ }}$ also depends on the modulus $u_{1}$. To
complete our task, we must find the cycle $\alpha_{n_{c}}$ that corresponds to
$a_{n_{c}}$ and the differential $\lambda_{1}$ that corresponds to $s_{1}$.
The cycle $\alpha_{n_{c}}$ is the one that encircles all the cuts in the $z$
plane. Note that this is a trivial cycle, and only a meromorphic differential
can be different from zero when it is integrated around it. The differential
that corresponds to $s_{1}=-u_{1}$ is
$\lambda_{1}\propto\frac{z^{n_{c}-1}dz}{y}\ ,$ (65)
and is meromorphic because it has a pole at $\infty$. With these
modifications, the solution is encoded in (64).
The ${\rm U}(n_{c})$ theory can also be considered as part of a bigger,
asymptotically free gauge theory. For example, we can take a ${\rm
SU}(n_{c}+1)$ ${\cal N}=2$ and break it to ${\rm SU}(n_{c})\times{\rm
U}(1)/\mathbb{Z}_{n_{c}}$ at a certain energy scale. This can be achieved
simply by choosing one diagonal element of $\phi$ very distant from the others
and fix it. We have thus a sub-manifold of the ${\rm SU}(n_{c}+1)$ moduli
space, which is exactly like the one of ${\rm U}(n_{c})$. What we achieve with
this is that now we can display also the magnetic cycle $\beta_{n_{c}}$. This
is important if we want to compute the monodromies around the singularities
and consequently the charges of the massless hypermultiplets on these
singularities. We shall use this trick in the next section.
### 3.3 Classical and Quantum $r$ Vacua
We now review the physics of color-flavor locked vacua or, as we like to call
them, $r$ vacua. The adjoint field $\phi$ has $r_{{\rm cl}}$ diagonal elements
locked to the hypermultiplet mass $m$
$\langle\phi\rangle=\left(\begin{array}[]{cccc}m{\bf 1}_{r_{{\rm cl}}}&&&\\\
&a_{1}{\bf 1}_{n_{1}}&&\\\ &&\ddots&\\\ &&&a_{k}{\bf 1}_{n_{k}}\\\
\end{array}\right)\ ,$ (66)
and the remaining diagonal blocks are divided between the roots of
$W^{\prime}$ and $\sum_{j=1}^{k}\,n_{j}+r_{{\rm cl}}=n_{c}$. We use $r_{{\rm
cl}}$ to denote the classical value and $0\leq r_{{\rm cl}}\leq n_{f}$. The
gauge group is broken by $\langle\phi\rangle$ down to ${\rm
U}(r)\times\prod_{j=1}^{k}{\rm U}(n_{j})$. If $r$ colors and flavors are
locked at the same eigenvalue, then in the low energy we also have a massless
hypermultiplet in the fundamental of U$(r)$, that we denote as $q$. Apart from
these perturbative objects, there are also non-Abelian and Abelian monopoles,
each carrying magnetic charge under two of the unbroken gauge groups.
The $F_{\phi}$ term and the $D$ term together yield the potential for the
quark fields
$V=g^{2}\hbox{\rm
Tr}\,|q\widetilde{q}-W^{\prime}|^{2}+\frac{g^{2}}{4}\hbox{\rm
Tr}\,(qq^{\dagger}-\widetilde{q}^{\dagger}\widetilde{q})^{2}\ .$ (67)
The non-Abelian quarks develop thus a condensate
$\widetilde{q}q=r_{{\rm cl}}W{}^{\,\prime}(m)\ .$ (68)
The non-Abelian group ${\rm U}(r)$ is Higgsed, non-Abelian vortices are
formed, and non-Abelian monopoles confined [18].
Quantum mechanically, is a little bit different. One difference is that in the
strong coupling region, where $m$ is of order of the dynamical scale, the
massless quarks are continuously transformed into magnetic degrees of freedom,
dual-quarks. Another difference is that (68) is modified by quantum
corrections. We are now going to compute these quantum corrections as in [3].
The quantum parameter $r$ is given by
${r}={\rm min}(r_{{\rm cl}},n_{f}-r_{{\rm cl}})\ .$ (69)
| ${\rm SU}(n_{f})$ | ${\rm SU}({r})$ | U$(1)_{0}$ | U$(1)_{1,\dots,n_{c}-{r}}$
---|---|---|---|---
| $\mathbf{n_{f}}$ | $\mathbf{r}$ | $1$ | $0$
Table 3: Low energy in $r$ vacua.
There are various reasons for the appearance of this mirror symmetry and the
quantum parameter $r$. We saw from the classical discussion of the moduli
space that the hyper-Kahler dimension of the $r$ Higgs branch is $r(n_{f}-r)$.
The gauge theory is, classically, ${\cal N}=2$ ${\rm SU}(r)$ with $n_{f}$
quarks. Only for $r\leq[n_{f}/2]$, this is not strong in the infrared, and we
can conclude that it remains as a low-energy effective description. Branches
which classically have $r_{{\rm cl}}$ greater than $[n_{f}/2]$, quantum
mechanically falls into the $r$ vacua classification with $r=n_{f}-r_{{\rm
cl}}$. The factorization of the $\Sigma_{\mbox{\tiny${\cal N}=2$ }}$ curve is
[10, 11]
$\qquad y^{2}=\frac{1}{4{g_{n_{c}-{r}}}^{2}}(W^{\prime 2}+f)(z-m)^{2r}\ .$
(70)
It is easy to see, from the curve (60), that the factor $z^{2r}$ can be pulled
out only if $r\leq[n_{f}/2]$. In the low energy, we have a ${\cal N}=2$ ${\rm
SU}({r})\times{\rm U}(1)\times{\rm U}(1)^{n_{c}-{r}}$ gauge theory with
hypermultiplet $\widetilde{D},D$ with charges given in Table 3. The low-energy
superpotential is
${\cal W}_{\rm
low}=\sqrt{2}\left(\sum_{i=1}^{n_{f}}\,(\widetilde{D}_{i}A_{\mathbf{r}}D^{i}+\widetilde{D}_{i}A_{0}D^{i})-W_{\mathrm{eff}}(A_{\mathbf{r}},A_{0},\dots,A_{n_{c}-r})\right)\
,$ (71)
where the effective superpotential that breaks to ${\cal N}=1$ is
$W_{\mathrm{eff}}=\sum_{j=0}^{n_{c}-r}\,g_{j}u_{j+1}(A_{\mathbf{r}},A_{0},\dots,A_{n_{c}-r})\
.$ (72)
We choose a superpotential with $k=n_{c}-r$, so that all the possible points
in the $r$ sub-manifold can in principle be selected. $A$’s are the chiral
superfields of the ${\cal N}=2$ gauge multiplets, $A_{\mathbf{r}}$ is one of
the ${\rm SU}({r})$ gauge multiplet, and $A_{0}$ is one of the U$(1)$
multiplet coupled to the dual quarks.
To compute the quantum condensate $\widetilde{d}d$, it is more convenient to
split the flavors’ masses a little bit and use the Abelianized theory so
obtained. Then we simply make the limit of coincident masses for the result so
obtained. We have $r$ flavors with masses $m_{j}$ with $j=1,\dots,r$. In the
limit if coincident masses $m_{j}\to m$, $\forall j$ and we recover the $r$
vacuum. The chiral superfields that we previously called $A_{\mathbf{r}}$ and
$A_{0}$, are now described by $r$ Abelian fields that we denote as $A_{(0,j)}$
with $j=1,\dots,r$. Each ${\rm U}(1)_{(0,j)}$ is locked to one flavor
$\widetilde{D}_{j}$, $D_{j}$. The low-energy superpotential is now
${\cal W}_{\rm
low}=\sqrt{2}\left(\sum_{j=1}^{r}\,\widetilde{D}_{j}A_{(0,j)}D^{j}-W_{\rm
eff}(A_{(0,1)},\dots,A_{(0,r)};A_{1},\dots,A_{n_{c}-r})\right)\ .$ (73)
For our computation, we need to consider only the $F_{A}$ terms of the
potential:
$\displaystyle
F_{A_{(0,j)}}=2e_{(0,j)}^{2}\left|\widetilde{d}_{j}d^{j}-\frac{\partial W_{\rm
eff}}{\partial a_{(0,j)}}\right|^{2}\ ,$ $\displaystyle j=1,\dots,r\ ,$
$\displaystyle F_{A_{s}}=2e_{s}^{2}\left|\frac{\partial W_{\rm eff}}{\partial
a_{s}}\right|^{2}\ ,$ $\displaystyle s=1,\dots,n_{c}-r\ .$ (74)
The first ones give the condensates, while the second gives the stationary
condition necessary to compute the position in the moduli space:
$\displaystyle\widetilde{d}_{j}d^{j}=\frac{\partial W_{\rm eff}}{\partial
a_{(0,j)}}\ ,$ $\displaystyle j=1,\dots,r\ ,$ $\displaystyle 0=\frac{W_{\rm
eff}}{\partial a_{s}}\ ,$ $\displaystyle s=1,\dots,n_{c}-r\ .$ (75)
We can write a matrix equation like (75) where the couplings vector is
$\left[{\bf g}\right]=(g_{0},\dots,g_{n_{c}-r},0,\dots,0)$ (76)
and the tension vector is
$\left[{\bf\widetilde{d}d}\right]=(\widetilde{d}_{1}d_{1},\dots,\widetilde{d}_{r}d_{r},0,\dots,0)\
.$ (77)
With these conventions, the equation (75) becomes
$\left[{\bf\widetilde{d}d}\right]=\left[\mathbf{\frac{\partial u}{\partial
a}}\right]\left[{\bf g}\right]\qquad\Longrightarrow\qquad\left[{\bf
g}\right]=\left[\mathbf{\frac{\partial a}{\partial
u}}\right]\left[{\bf\widetilde{d}d}\right]\ .$ (78)
where in the last we have simply multiplied by the inverse matrix. The last
version is the one that we shall find more convenient for computing the
condensates.
Let us begin with the simplest example: the $r=1$ vacuum of the
$(n_{c},n_{f})=(2,2)$ theory. The superpotential is quadratic
$W(z)=g_{0}z+\frac{g_{1}}{2}z^{2}\ ,$ (79)
and the factorization of $\Sigma_{\mbox{\tiny${\cal N}=2$ }}$ gives
$\displaystyle y^{2}$ $\displaystyle=$
$\displaystyle\frac{1}{4}P_{2}(z)^{2}-\Lambda^{2}(z-m)^{2}$ (80)
$\displaystyle=$ $\displaystyle\frac{1}{4g_{1}^{2}}(W^{\prime 2}+f)(z-m)^{2}\
.$
The cycle $\alpha_{0}$ is the one that encircles the double roots $z=m$. The
crucial ingredient for the following computation is the residue around the
cycle $\alpha_{0}$:
$\frac{1}{4\pi i}\oint_{\alpha_{0}}\frac{dz}{y}=\frac{g_{1}}{\sqrt{W^{\prime
2}+f}}\ .$ (81)
In this case, the relationship between $s$ and $u$ coordinates is:
$s_{1}=-u_{1}\ ,\qquad s_{2}=-u_{2}+\frac{{u_{1}}^{2}}{2}\ .$ (82)
For our proof, we will need to calculate only
$\frac{\partial a_{0}}{\partial u_{2}}=\frac{\partial a_{0}}{\partial
s_{1}}\frac{\partial s_{1}}{\partial u_{2}}+\frac{\partial a_{0}}{\partial
s_{2}}\frac{\partial s_{2}}{\partial u_{2}}=-\frac{\partial a_{0}}{\partial
s_{2}}.$ (83)
First, we observe that the solution (64) and the residue (81) give777The
proper normalization of the holomorphic differential has been chosen to
reproduce the correct semiclassical result
$\frac{\partial a_{0}}{\partial s_{2}}=-\frac{g_{1}}{\sqrt{W^{\prime 2}+f}}\
.$ (84)
Then, writing the equations (75) in the matrix form suggested in (78), we get
$\left(\begin{array}[]{c}g_{0}\\\ g_{1}\\\
\end{array}\right)=\left(\begin{array}[]{cc}\,\partial a_{0}/\partial
u_{1}&\,\partial a_{1}/\partial u_{1}\\\ \,\partial a_{0}/\partial
u_{2}&\,\partial a_{1}/\partial u_{2}\\\
\end{array}\right)\left(\begin{array}[]{c}\widetilde{d}d\\\ 0\\\
\end{array}\right)\ .$ (85)
The simple passage of multiplying by the inverse matrix has simplified our
work a lot because now (85) is expressed as a function of $\partial
a_{i}/\partial u_{j}$, known through (64). Furthermore, only $\partial
a_{0}/\partial u_{1,2}$, the ones obtained by an integral around the collided
roots, are important because the others are multiplied by zero. From (85), we
need only the second equation
$g_{1}=\frac{\partial a_{0}}{\partial u_{2}}\widetilde{d}d\ ,$ (86)
that, using (84), gives the condensate
$\widetilde{d}d=\left.\sqrt{W^{\prime 2}+f}\right|_{z=m}\ .$ (87)
We give another example ${\rm U}(3)$ with two flavors of mass $m_{1}$ and two
flavors of mass $m_{2}$ (see Figure 5 for the roots and cycles).
Figure 5: Cycles in ${\rm U}(3)\to{\rm U}(1)_{(0,1)}\times{\rm
U}(1)_{(0,2)}\times{\rm U}(1)_{1}$ theory. Displayed are the electric cycles
$\alpha_{(0,1)}$, $\alpha_{(0,2)}$, and $\alpha_{1}$. Two flavors at the poles
$m_{1}$ and $m_{2}$ are massless and charged with respect to ${\rm
U}(1)_{(0,1)}$ and ${\rm U}(1)_{(0,2)}$. In the limit $m_{1}\to m_{2}\to m$,
we recover the $r=2$ vacuum by breaking ${\rm U}(3)\to{\rm SU}(2)\times{\rm
U}(1)_{0}\times{\rm U}(1)_{1}$.
The superpotential is still (79), while the factorization gives
$\Sigma_{\mbox{\tiny${\cal N}=2$ }}$:
$\displaystyle y^{2}$ $\displaystyle=$
$\displaystyle\frac{1}{4}P_{3}(z)^{2}-\Lambda^{2}(z-m_{1})^{2}(z-m_{2})^{2}$
(88) $\displaystyle=$ $\displaystyle\frac{1}{4g_{1}^{2}}(W^{\prime
2}+f)(z-m_{1})^{2}(z-m_{2})^{2}\ .$
The relationship between $s$ and $u$ in this case is given by (82) plus:
$s_{3}=-u_{3}+u_{1}u_{2}-\frac{{u_{1}}^{3}}{6}\ .$ (89)
By using the trick explained before, we write the last of (78) for this
particular case
$\left(\begin{array}[]{c}g_{0}\\\ g_{1}\\\ 0\\\
\end{array}\right)=\left(\begin{array}[]{ccc}\,\partial a_{(0,1)}/\partial
u_{1}&\,\partial a_{(0,2)}/\partial u_{1}&\,\partial a_{1}/\partial u_{1}\\\
\,\partial a_{(0,1)}/\partial u_{2}&\,\partial a_{(0,2)}/\partial
u_{2}&\,\partial a_{1}/\partial u_{2}\\\ \,\partial a_{(0,1)}/\partial
u_{3}&\,\partial a_{(0,2)}/\partial u_{3}&\,\partial a_{1}/\partial u_{3}\\\
\end{array}\right)\left(\begin{array}[]{c}\widetilde{d}_{1}d^{1}\\\
\widetilde{d}_{2}d^{2}\\\ 0\\\ \end{array}\right)\ .$ (90)
Now, as in the previous case, we need to calculate only the residues around
${\widetilde{m}}_{1}$ and ${\widetilde{m}}_{2}$. The last equation of (90) is
$0=\widetilde{d}_{1}d^{1}\frac{1}{\left.\sqrt{W^{\prime
2}+f}\right|_{m_{1}}(m_{1}-m_{2})}+\widetilde{d}_{2}d^{2}\frac{1}{\left.\sqrt{W^{\prime
2}+f}\right|_{m_{2}}(m_{2}-m_{1})}\ .$ (91)
using also the second equation of (90). Using
$\frac{\partial a}{\partial u_{2}}=-\frac{\partial a}{\partial
s_{2}}+u_{1}\frac{\partial a}{\partial s_{3}}\ ,$ (92)
the second equation of (90) leads to another independent equation
$1=\widetilde{d}_{1}d^{1}\frac{m_{1}}{\left.\sqrt{W^{\prime
2}+f}\right|_{m_{1}}(m_{1}-m_{2})}+\widetilde{d}_{2}d^{2}\frac{{\widetilde{m}}_{2}}{\left.\sqrt{W^{\prime
2}+f}\right|_{m_{2}}(m_{2}-m_{1})}\ .$ (93)
This equation together with (91) are enough to establish the solution
$\widetilde{d}_{1}d^{1}=\left.\sqrt{W^{\prime 2}+f}\right|_{m_{1}}\
,\qquad\widetilde{d}_{2}d^{2}=\left.\sqrt{W^{\prime 2}+f}\right|_{m_{2}}\ .$
(94)
Thus we obtain, in the limit of coincident masses $m_{1}=m_{2}=m$, the little
formula for the dual-quark condensate
$\displaystyle\widetilde{d}d=r\left.\sqrt{W^{\prime}(z)^{2}+f(z)}\right|_{z=m}\
.$ (95)
This is the quantum generalization of (68). The crucial ingredient for the
computation has been the residue around the mass poles. We have for simplicity
restricted ourselves to the specific case of $n=k=n_{c}-r$, but the formula
(95) still holds in the generic case. We refer to [3] for more details about
the derivation. In Section 6, we shall re-derive it a simpler way in the MQCD
setup.
## 4 The First Examples of Quantum Coincidence
We now describe the first example of coincidence vacua. We begin with the
easiest case, $n_{c}=2$. After diagonalization, we have $\phi={\rm
diag}(\phi_{1},\phi_{2})$. So the gauge invariant coordinates of the moduli
space are $u_{1}=\hbox{\rm Tr}\,\phi=\phi_{1}+\phi_{2}$ and
$u_{2}=\frac{1}{2}\hbox{\rm
Tr}\,\phi^{2}=\frac{1}{2}(\phi_{1}^{2}+\phi_{2}^{2})$. The SW curve is:888When
we write $\Lambda$ without any subscript, we always mean
$\Lambda_{\mbox{\tiny${\cal N}=2$ }}$.
$\displaystyle{y}^{2}$ $\displaystyle=$
$\displaystyle\frac{1}{4}(z-\phi_{1})^{2}(z-\phi_{2})^{2}-\Lambda^{4-n_{f}}z^{n_{f}}$
(96) $\displaystyle=$
$\displaystyle\frac{1}{4}\left(z^{2}-u_{1}z+\frac{u_{1}^{2}}{2}-u_{2}\right)^{2}-\Lambda^{4-n_{f}}z^{n_{f}}$
$\displaystyle=$
$\displaystyle\left(\frac{1}{2}(z-\phi_{1})(z-\phi_{2})-\Lambda^{2-n_{f}/2}z^{n_{f}/2}\right)\left(\frac{1}{2}(z-\phi_{1})(z-\phi_{2})+\Lambda^{2-n_{f}/2}z^{n_{f}/2}\right)\
.$
The last passage can be done only for $n_{f}$ even.
We start now with the specific case of $n_{f}=2$. Thanks to the factorization
(96), we can now can explicitly compute the four roots of the polynomial. We
call them $z^{-}_{\,1,2}$ and $z^{+}_{\,1,2}$, respectively, for the left and
right factors. Their value is
$\displaystyle z^{-}_{\,1,2}$ $\displaystyle=$
$\displaystyle\frac{1}{2}\left(u_{1}+2\Lambda\pm\sqrt{(u_{1}+2\Lambda)^{2}-4\left(\frac{u_{1}^{2}}{2}-u_{2}\right)}\right)\
,$ $\displaystyle z^{+}_{\,1,2}$ $\displaystyle=$
$\displaystyle\frac{1}{2}\left(u_{1}-2\Lambda\pm\sqrt{(u_{1}-2\Lambda)^{2}-4\left(\frac{u_{1}^{2}}{2}-u_{2}\right)}\right)\
.$ (97)
It is good to begin to study the problem in different sections of the moduli
space. First, at $\hbox{\rm Tr}\,\phi=0$, where it corresponds to the ${\rm
SU}(2)$ moduli space. The curve is
$\displaystyle{y}^{2}$ $\displaystyle=$
$\displaystyle\frac{1}{4}(z^{2}-u_{2})^{2}-\Lambda^{2}z^{2}$ (98)
$\displaystyle=$ $\displaystyle\left(\frac{1}{2}(z^{2}-u_{2})-\Lambda
z\right)\left(\frac{1}{2}(z^{2}-u_{2})+\Lambda z\right)\ .$
There are two singular points. One is $u_{2}=0$ where
$z^{-}_{\,1,2}=2\Lambda,0$ and $z^{+}_{\,1,2}=0,-2\Lambda$ (Figure 7). This is
the root of the $r=1$ branch. The other singularity is at $u_{2}=-\Lambda^{2}$
and $z^{-}_{\,1,2}=\Lambda,\Lambda$ and $z^{+}_{\,1,2}=-\Lambda,-\Lambda$
(Figure 7). This is the root of the $r=0$ branch (this is what is called “root
of the baryonic branch” in [2]). Note that the singularity is doubled, and
looking at the complete moduli space, it will be clear why.
Figure 6: Roots near the $r=1$ vacuum: $u_{1}=0$, $u_{2}=.16$.
Figure 7: Roots near the $r=0$ vacuum: $u_{1}=0$, $u_{2}=-1.006$.
Now we take a section where $\hbox{\rm Tr}\,\phi=M\gg\Lambda$. One singularity
is at $\phi={\rm diag}(0,M)$, where one quark becomes massless. It corresponds
to $u_{2}=M^{2}$. The curve is
$\displaystyle{y}^{2}$ $\displaystyle=$
$\displaystyle\frac{1}{4}\left(\det\mbox{\small$\left(\begin{array}[]{cc}0&\\\
&M\end{array}\right)$ }\right)^{2}-\Lambda^{2}z^{2}$ (101) $\displaystyle=$
$\displaystyle z^{2}\left(\frac{1}{4}(z-M)^{2}-\Lambda^{2}\right)\ .$ (102)
The other singularities are “near” $\phi={\rm diag}(M/2,M/2)$ and are strong
coupling singularities of a pure ${\rm SU}(2)$ gauge theory. They correspond
to, respectively, to the coincidence of $z^{-}_{1}=z^{-}_{2}$ at
$u_{2}=-\Lambda^{2}/4-M\Lambda/2$, and the coincidence of
$z^{+}_{1}=z^{+}_{2}$ at $u_{2}=-\Lambda^{2}/4+M\Lambda/2$. They are the usual
monopole and dyon singularity of pure ${\rm SU}(2)$.
Figure 8: The $\hbox{\rm Im}\,u_{1}=0$ section of the moduli space. The lines
are the co-dimension two singularities. The four points are the $r=0,1$ vacua
at $u_{1}=0$, and the two coincidence points.
We can now draw a picture with the singularities (Figure 8). The space
$u_{1},u_{2}$ is two complex dimensional. Singularities are one complex
dimensional surfaces, objects of co-dimension one, where two roots of the SW
curve coincide. We draw only three real dimensions, the plane $u_{2}$ and the
real part of $u_{1}$. In this plot, the singularities are lines (real co-
dimension two). The three singular curves are
$\displaystyle u_{2}$ $\displaystyle=$ $\displaystyle-\Lambda^{2}\pm
u_{1}\Lambda+\frac{u_{1}^{2}}{4}\ ,$ (103) $\displaystyle u_{2}$
$\displaystyle=$ $\displaystyle\frac{u_{1}^{2}}{2}\ .$ (104)
There are three special points where the curves intersect, and there are more
coincidences of roots. One has already been discussed and is the root of $r=0$
branch at $u_{1}=0$. The other two special points intersections are at
$u_{1}=\pm 2\Lambda$ and $u_{2}=2\Lambda^{2}$. Where the curve is
$y^{2}=z^{3}(z\pm 4\Lambda)\ .$ (105)
Figure 9: Roots near the first coincidence vacuum: $u_{1}=1.9$, $u_{2}=1.68$.
Figure 10: Roots near the second coincidence vacuum: $u_{1}=-1.9$,
$u_{2}=1.68$.
Clearly something special happens in these two singularities (see Figures 10
and 10 for the roots in these vacua). These are the coincidence points. To
understand what is happening, it is good to start from the $r=1$ vacuum at
$u_{1}=0$ of Figure 7. Here we have two roots in zero, and they correspond to
a massless quark. The other two roots are in $\pm 2\Lambda$. A mass
perturbation $W=\mu\hbox{\rm Tr}\,\Phi^{2}/2$ selects this $r=1$ vacuum, and
the two unpaired roots correspond to the polynomial $W^{\prime}(z)^{2}+f(z)$.
The splitting is due to the quantum effect of $f(z)$. We can now change the
$u_{1}$ coordinate, while remaining on the singularity $u_{2}=u_{1}^{2}/2$,
and so change the position of the unpaired roots. To achieve this, we can use
a generic quadratic superpotential
$W(\Phi)=\mu\hbox{\rm Tr}\,\left(\frac{\Phi^{2}}{2}-\alpha\Phi\right)\ .$
(106)
The parameter $\alpha$ determines the level of $u_{1}$ that this
superpotential selects. It selects all the maximal singularity points at the
level $u_{1}=\alpha$.999Consider the quark singularity $u_{2}=u_{1}^{2}/2$. If
we want to select a generic point on this singularity then $W=\mu(u_{2}-\alpha
u_{1})=\mu(u_{1}^{2}/2-\alpha u_{1})$. The vacuum selected will be thus at
$u_{1}=\alpha$ and $u_{2}=\alpha^{2}/2$. Changing $\alpha$, we can thus move
along $u_{1}$ while staying on the singular lines. In this way, the two
unpaired roots change their position while the two at the quark singularity
remain fixed. When we reach $u_{1}=2\Lambda$, the root $z^{+}_{\,2}$ collides
with the quark roots, and thus we have the coincidence point, exactly as was
anticipated in Figure 3. At $u_{1}=-2\Lambda$ is $z^{-}_{\,1}$ that coincides
with the quark roots, at $u_{1}=+2\Lambda$ is $z^{+}_{\,2}$ that coincides
with the quark roots. So we have that the two superpotentials
$W(\Phi)=\mu\hbox{\rm Tr}\,\left(\frac{\Phi^{2}}{2}\mp 2\Lambda\Phi\right)$
(107)
select respectively the two coincidence vacua.
Now we discuss the low-energy effective action in these various vacua. First
the two $u_{1}=0$ cases. The root of the $r=1$ non-baryonic Higgs branch (see
Table 5) has a flavor doublet $\widetilde{D},D$ (the dual-quark) charged under
only one of the ${\rm U}(1)$’s factors. The superpotential is
${\cal W}_{r=1}=\sqrt{2}\widetilde{D}A_{1}D\ ,$ (108)
The Higgs branch emanating from it is exactly the non-baryonic branch $r=1$.
$\displaystyle\widetilde{d}_{i}d^{i}=0\ ,$
$\displaystyle|\widetilde{d}_{i}|^{2}-|d^{i}|^{2}=0\ .$ (109)
This correctly reproduces the $r=1$ non-baryonic Higgs branch. The low-energy
effective action, after the mass perturbation $W=\mu\hbox{\rm
Tr}\,\Phi^{2}/2$, becomes
${\cal W}_{r=1}^{\prime}=\sqrt{2}\left(\widetilde{D}A_{1}D-\mu
u_{2}(A_{1},A_{2})\right)\ .$ (110)
The dual quark develops a condensate, the ${\rm U}(1)_{1}$ is Higgsed, and the
only thing that remains in the infrared is the free ${\rm U}(1)_{2}$ theory.
| | ${\rm U}(1)_{1}$ | $\times$ | ${\rm U}(1)_{2}$
---|---|---|---|---
${\bf 2}\times D$ | | $1$ | | $0$
Table 4: Low energy in the $r=1$ vacuum.
| | ${\rm U}(1)_{1}$ | $\times$ | ${\rm U}(1)_{2}$
---|---|---|---|---
$E_{1}$ | | $1$ | | $0$
$E_{2}$ | | $0$ | | $1$
Table 5: Low energy in the $r=0$ vacuum.
Then we have the $r=0$ vacuum whose particles and charges are given in Table
5. The low-energy effective action is
${\cal
W}_{r=0}=\sqrt{2}\left(\widetilde{E}_{1}A_{1}E_{1}+\widetilde{E}_{2}A_{2}E_{2}\right)\
.$ (111)
There is no Higgs branch emanating from it. This point is again selected by
$W=\mu\hbox{\rm Tr}\,\Phi^{2}/2$, and the low-energy effective action becomes
${\cal
W}_{r=0}^{\prime}=\sqrt{2}\left(\widetilde{E}_{1}A_{1}E_{1}+\widetilde{E}_{2}A_{2}E_{2}-\mu
u_{2}(A_{0},A_{1})\right)\ .$ (112)
Both hypermultiplets condense, and thus we have a mass gap.
We finally come to the coincidence points. In these vacua, mutual non-local
particles become massless at the same time. We have to determine exactly the
charges of these particles with respect to a common basis (note that the
previous two Tables 5 and 5 were not in the same basis). First of all, we
choose a basis of cycles as in Figure 11. Note that we displayed the roots
$z_{\,1,2}^{-}$, $z^{+}_{\,1,2}$ near to the first coincidence point. There
are other two roots in the figure; we need to embed the theory in ${\rm
SU}(3)\to{\rm SU}(2)\times{\rm U}(1)$ in order to show the magnetic cycle
$\beta_{2}$. Cycle $\alpha_{1}$ winds around the roots $z^{-}_{2}$ and
$z^{+}_{1}$. Its dual $\beta_{1}$ winds around the roots $z^{-}_{2}$ and
$z^{+}_{2}$. Cycle $\alpha_{2}$ winds around all the four roots, and its dual
$\beta_{2}$ winds around $z^{-}_{1}$, and passes through the two extra roots.
Figure 11: Our choice of cycles $\alpha_{1},\alpha_{2}$ and $\beta_{1},\beta_{2}$ near the first coincidence point (Figure 10). | | ${\rm U}(1)_{1}$ | $\times$ | ${\rm U}(1)_{2}$
---|---|---|---|---
${\bf 2}\times D$ | | $\phantom{-}1_{\rm ele}$ | | $0$
$B$ | | $-1_{\rm mag}$ | | $1$
Table 6: Particles and charges in the coincidence point.
This is the choice of basis more convenient for the $r=1$ vacuum. A monodromy
around the (104) singularity gives the charge of the dual quark
$\widetilde{D}$, $D$: a doublet of the flavor group and electrically charged
under the ${\rm U}(1)_{1}$. To get the charge of the other massless particle,
we compute the monodromy around the singularity (103). Choosing the vector of
cycles like $(\alpha_{1},\alpha_{2}|\beta_{1},\beta_{2})$, the monodromy is
$M=\left(\begin{array}[]{c|c}\mathbf{1}-h\otimes
q&\phantom{\mathbf{1}}-h\otimes h\\\ \hline\cr\phantom{\mathbf{1}\,}q\otimes
q&\mathbf{1}+q\otimes
h\end{array}\right)=\left(\begin{array}[]{cc|cc}\phantom{-}1&\phantom{-}1&-1&\\\
&\phantom{-}1&&\\\ \hline\cr&&\phantom{-}1&\\\
&\phantom{-}1&-1&\phantom{-}1\end{array}\right)$ (113)
This particle, which we call $B$, is a singlet with respect to the color,
magnetically charged with respect to ${\rm U}(1)_{1}$, and electrically
charged with respect to ${\rm U}(1)_{2}$ (see Table 6). This is nothing but
one of the two particles of Table 5 that are massless in the $r=0$ vacuum. But
due to the change of basis, it is now a magnetic object.
This is a particular kind of Argyres-Douglas singularity [12, 13, 14]. The
low-energy dynamics consists of a non-local, strongly interacting
superconformal field theory. The superpotentials (107) select, respectively,
these two coincidence points breaking ${\cal N}=2$ down to ${\cal N}=1$. No
condensate is developed, and this means that the theory in the infrared still
remains superconformal.101010The fact that the condensate vanishes in this
particular point has also been previously noted in [15]. Our claim is that in
these points there is an interpolation between ${\cal N}=2$ super-QCD and pure
${\cal N}=1$ super-QCD. We shall come back at the end of the next section,
after the generalization to arbitrary $n_{c},n_{f}$, for the discussion and
interpretation of this important issue. For the moment, let us keep in mind
the lessons that we learn from this example:
* •
We have $2n_{c}-n_{f}$ ($2$ in this case) coincidence points symmetric under
the $\mathbb{Z}_{2n_{c}-n_{f}}$ remnant of the ${\rm U}(1)_{R}$ symmetry.
* •
The coincidence vacua lie at the intersections between the $r$ vacua
singularities.
* •
Mutual non-local particles become massless in these vacua. They are particular
cases of Argyres-Douglas singularities.
Let us now consider a few more specific examples.
* * *
Another example is still $n_{c}=2$ but now with $n_{f}=3$. This belongs to a
special class $n_{f}=2n_{c}-1$. The number $2n_{c}-n_{f}$ is particularly
important due to the discrete $R$ symmetry. The number of coincidence points
is in fact equal to $2n_{c}-n_{f}$ since they spontaneously break this
discrete symmetry. The previous example had two coincidence points. This
example has only one coincidence point, and it lies in the $u_{1}=0$ section
of the moduli space. The SW curve is
${y}^{2}=\frac{1}{4}\left(z^{2}-u_{1}z+\frac{u_{1}^{2}}{2}-u_{2}\right)^{2}-\Lambda
z^{3}\ .$ (114)
Now we cannot use the last passage of (96) and there is no easy expression for
the roots. But the maximal singularity is nevertheless easy to detect. The
singularity is only one and is in the center of the moduli space at $u_{1}=0$,
$u_{2}=0$. The curve factorizes as
${y}^{2}=\frac{1}{4}z^{3}(z-4\Lambda)\ .$ (115)
So in this case the classical notion of the origin of the moduli space
persists. It can be seen that of three singularities of co-dimension two meet
at this point. The superpotential that leaves this point is simply
$W(\Phi)=\mu\hbox{\rm Tr}\,\Phi^{2}/2$.
Another fully computable example, with different $2n_{c}-n_{f}$ from the
previous ones, is $n_{c}=4$, and $n_{f}=4$. Now $2n_{c}-n_{f}=4$ and we expect
four coincidence points. The non-baryonic branch with $r=2$ has the following
curve
$\displaystyle{y}^{2}$ $\displaystyle=$ $\displaystyle
z^{4}\left(\frac{1}{4}\left(z^{2}-u_{1}z+\frac{u_{1}^{2}}{2}-u_{2}\right)^{2}-\Lambda^{4}\right)$
(116) $\displaystyle=$
$\displaystyle\frac{1}{4}z^{4}\left(z^{2}-u_{1}z+\frac{u_{1}^{2}}{2}-u_{2}-2\Lambda^{2}\right)\left(z^{2}-u_{1}z+\frac{u_{1}^{2}}{2}-u_{2}+2\Lambda^{2}\right)\
.$
The four extra-roots are
$\displaystyle\frac{u_{1}\pm\sqrt{-u_{1}^{2}+4u_{2}+8\Lambda^{2}}}{2}\ ,$
$\displaystyle\frac{u_{1}\pm\sqrt{-u_{1}^{2}+4u_{2}-8\Lambda^{2}}}{2}\ .$
(117)
Two must collide, and one of the other two, the split ones, must be at the
mass value zero. One choice is
$\displaystyle-u_{1}^{2}+4u_{2}-8\Lambda^{2}=0\ ,$ $\displaystyle
u_{1}=\pm\sqrt{-u_{1}^{2}+4u_{2}+8\Lambda^{2}}\ ,$ (118)
whose solution is $u_{1}=\pm 4\Lambda$ and $u_{2}=6\Lambda^{2}$. The other is
$\displaystyle-u_{1}^{2}+4u_{2}+8\Lambda^{2}=0\ ,$ $\displaystyle
u_{1}=\pm\sqrt{-u_{1}^{2}+4u_{2}-8\Lambda^{2}}\ ,$ (119)
whose solution is $u_{1}=\pm i4\Lambda$ and $u_{2}=-6\Lambda^{2}$. We thus
find exactly four points as expected. They are related by the
$\mathbb{Z}_{2n_{c}-n_{f}}$ symmetry, which in this case is $\mathbb{Z}_{4}$.
Again, with these four points, there is a collision with the four singularity
($r=0,1,2$) that departs from $u_{1}=0$. The superpotentials
$W(\Phi)=\mu\hbox{\rm Tr}\,\left(\frac{\Phi^{2}}{2}\mp 2\Lambda\Phi\right)\
,\qquad W(\Phi)=\mu\hbox{\rm Tr}\,\left(\frac{\Phi^{2}}{2}\mp
2i\Lambda\Phi\right)$ (120)
select the four coincidence vacua.111111Consider the singularity
$u_{2}=u_{1}^{2}/4-2\Lambda$. If we want to select a generic point on this
singularity, then $W=\mu(u_{2}-\alpha u_{1})=\mu(u_{1}^{2}/4-2\Lambda-\alpha
u_{1})$. The vacuum selected will be thus at $u_{1}=2\alpha$ and
$u_{2}=\alpha^{2}-2\Lambda$.
For the case $u_{1}=4\Lambda$, and $u_{2}=6\Lambda^{2}$, the roots are all
real, and the factorization of the curve is
$y^{2}=\frac{1}{4}z^{5}(z-2\Lambda)^{2}(z-4\Lambda)$ (121)
Note that $2=4\left(\cos{\frac{\pi}{4}}\right)^{2}$. We keep in mind this for
the generalization we are going to do in the coming section.
## 5 General Case
We now consider the case of generic $n_{c}$ and $n_{f}$. The Seiberg-Witten
curve is
$\displaystyle{y}^{2}$ $\displaystyle=$ $\displaystyle{\cal
P}_{(n_{c},n_{f})}(z)$ (122) $\displaystyle=$
$\displaystyle\frac{1}{4}\det(z-\phi)^{2}-\Lambda^{2n_{c}-n_{f}}z^{n_{f}}\ ,$
where we have defined for convenience the polynomial ${\cal
P}_{(n_{c},n_{f})}(z)$. The quark singularities are labeled by an integer $r$
that runs from $\widetilde{n}_{c}$ to $[n_{f}/2]$. Along these singular sub-
manifolds, the curve has a $2r$ zero at the hypermultiplet mass. We consider
$m=0$ here for simplicity. The curve is thus factorized as follows
${y}^{2}=z^{2r}{\cal P}_{n_{c}-r,n_{f}-2r}(z)\ ,$ (123)
where what remains is the curve for a gauge group $n_{c}-r$ and $n_{f}-2r$
flavors. The adjoint scalar is
$\phi={\rm diag}(0,\dots,0,\phi_{r+1},\dots,\phi_{n_{c}}),$ (124)
where the first $r$ diagonal elements are locked to the mass $m=0$ and the
other $n_{c}-r$ coordinates span the singular sub-manifold of the Coulomb
branch. This manifold is a root of a $r$ non-baryonic Higgs branch.
Let’s consider, to begin with, the maximal case $r=[n_{f}/2]$. Let us restrict
also for simplicity to the case $n_{f}$ even. Along this $n_{f}/2$ sub-
manifold, the SW curve is factorized as follows
$y^{2}=z^{n_{f}}{\cal P}_{(n_{c}-n_{f}/2,0)}(z)\ .$ (125)
What remains after extracting the $z^{n_{f}}$ factor is the curve of pure
${\rm U}(n_{c}-n_{f}/2)$ without matter fields. The maximal singularity points
for ${\cal P}_{(n_{c}-n_{f}/2,0)}(z)$ are given by the solution of Douglas and
Shenker [16]. There are $n_{c}-n_{f}/2$ of these maximal singularity points.
They arise when the $n_{c}-n_{f}/2$ cuts are lined up, and all the roots,
apart from two of them, are doubled. The simplest solution is when all the
roots are on the real axis. The others are related by a $\exp{\frac{2\pi
ik}{2n_{c}-n_{f}}}$ transformation. In the real, case we have
$\phi=\mathrm{diag}(\phi_{1},\dots,\phi_{n_{c}-n_{f}/2})$ and
$\phi_{j}=2\Lambda\cos{\frac{\pi(j-1/2)}{n_{c}-n_{f}/2}}$. The curve
factorization is obtained by using properties of the Chebyshev polynomials:
$\displaystyle{\cal P}_{(n_{c}-n_{f}/2,0)}(z)$ $\displaystyle=$
$\displaystyle\frac{1}{4}\prod_{j=1}^{n_{c}-n_{f}/2}\left(z-2\Lambda\cos{\frac{\pi(j-1/2)}{n_{c}-n_{f}/2}}\right)^{2}-\Lambda^{n_{c}-n_{f}/2}$
(126) $\displaystyle=$
$\displaystyle\left(\frac{1}{4}T_{n_{c}-n_{f}/2}\left(\frac{z}{2\Lambda}\right)^{2}-1\right)\Lambda^{n_{c}-n_{f}/2}$
$\displaystyle=$
$\displaystyle\left(\frac{z^{2}}{4}-\Lambda^{2}\right)U_{n_{c}-n_{f}/2-1}\left(\frac{z}{2\Lambda}\right)^{2}\Lambda^{n_{c}-n_{f}/2-2}\
,$
where
$U_{n_{c}-n_{f}/2-1}\left(\frac{z}{2\Lambda}\right)^{2}=\prod_{j=1}^{n_{c}-n_{f}/2-1}\left(\frac{z}{2\Lambda}-\cos{\frac{\pi
j}{n_{c}-n_{f}/2}}\right)$. In order to factorize the curve, we have used the
important identity
$T_{N}^{2}(z)-(z^{2}-1)U_{N-1}(z)=1\ .$ (127)
The Douglas-Shenker solution provides the exact position of the maximal
singularities where all the ${\rm U}(1)$ low-energy factors have their own
monopole (or dyon) massless. These discrete vacua are the ones that are
selected by the mass perturbation in the superpotential. The case of
$r=n_{f}/2$ is particularly simple due to the existence of this exact analytic
solution. The Douglas-Shenker solution will become particularly useful at the
end of the section when we shall describe the exact location of the
coincidence points.
Now let us describe the low-energy dynamics in a generic $r$ vacuum. We
already said that the curve factorizes like Eq.(123) in a sub-manifold of the
Coulomb branch of dimension $n_{c}-r$. An ${\rm SU}(r)\times{\rm U}(1)_{0}$
gauge group with $n_{f}$ flavors in the fundamental representation survives in
the low-energy spectrum. The non-Abelian gauge group is infrared free if the
condition $r<n_{f}/2$ is satisfied. It becomes superconformal for the maximal
case $r=n_{f}/2$. The other $n_{c}-r$ dimensions of the Coulomb branch
represent ${\rm U}(1)_{j}$ vector multiplets with $j$ that runs from $1$ to
$n_{c}-r$ . We are then interested in the points of maximal singularity where
all the ${\rm U}(1)$ gauge groups, except one of them, have their own massless
hypermultiplet. In these discrete points, the low-energy physics can thus be
summarized in Table 7 containing the gauge groups and corresponding charged
hypermultiplets where we have chosen a convenient basis for the cycles in the
SW curve so that the charges are all diagonal.
| | ${\rm SU}(r)$ | $\times$ | ${\rm U}(1)_{0}$ | $\times$ | ${\rm U}(1)_{1}$ | $\times$ | $\cdots$ | ${\rm U}(1)_{n_{c}-r-1}$ | $\times$ | ${\rm U}(1)_{n_{c}-r}$
---|---|---|---|---|---|---|---|---|---|---|---
${\bf n_{f}}\times D\phantom{....}$ | | $\bf r$ | | $1$ | | | | | | |
$E_{1\phantom{n_{c}-r-}}$ | | | | | | $1$ | | | | |
$\vdots\phantom{.....}$ | | | | | | | | $\ddots$ | | |
$E_{n_{c}-r-1}$ | | | | | | | | | $1$ | |
Table 7: Low-energy particles and charges in a generic $r$ vacuum with
$\widetilde{n}_{c}<r\leq[n_{f}/2]$.
It is easy to check that the Higgs branch emanating from this special vacuum
is identical to the baryonic Higgs branch determined in the classical theory.
These $r$ vacua, where all the particles can be put in a diagonal and local
form, are the complete list of the critical points located at $u_{1}=0$ that
survive after the perturbation by a mass term $\mu\hbox{\rm Tr}\,\Phi^{2}/2$.
Let us call $A_{\bf r}$ the adjoint chiral superfield of the ${\cal N}=2$
${\rm SU}(r)$ vector multiplet, $A_{j}$ with $j=0,\dots,n_{c}-r$ the real
chiral superfield for the ${\rm U}(1)_{j}$ vector multiplets, and
$D,\widetilde{D}$, $E_{j},\widetilde{E}_{j}$ the chiral superfields of the
matter hypermultiplets. The ${\cal N}=2$ low-energy Lagrangian simply follows
from the information provided in Table 7. After breaking to ${\cal N}=1$ with
the mass term $\mu\hbox{\rm Tr}\,\Phi^{2}/2$ in the microscopic theory, the
low-energy effective superpotential is
${\cal W}_{r}^{\prime}=\sqrt{2}\left(\widetilde{D}_{i}A_{\bf
r}D^{i}+\widetilde{D}_{i}A_{0}D^{i}+\sum_{j=1}^{n_{c}-r-1}\widetilde{E}_{j}A_{j}E_{j}-\mu
u_{2}(A_{\bf r},A_{0},\dots,A_{n_{c}-r})\right)\ .$ (128)
All the matter fields $\widetilde{D}D$ and $\widetilde{E}_{j}E_{j}$ acquire a
condensate due to the vanishing condition for the corresponding $F_{A_{\bf
r}}$ and $F_{A_{j}}$ terms. All the gauge groups, except from the last one
${\rm U}(1)_{n_{c}-r}$, are then Higgsed at an energy scale
$\sim\sqrt{\mu\Lambda}$. The theory in the IR thus loses completely the
information about the non-Abelian nature of the microscopic theory.
We now come to the minimal case when $r=\widetilde{n}_{c}$. It is better to
consider this case separately due to some peculiarities that shall soon be
evident. The curve in the $r=\widetilde{n}_{c}$ sub-manifold factorizes like
${y}^{2}=z^{2\widetilde{n}_{c}}{\cal P}_{(2n_{c}-n_{f},2n_{c}-n_{f})}(z)\ .$
(129)
What emerges is the curve for gauge group ${\rm U}(2n_{c}-n_{f})$ with a
number of flavors $2n_{c}-n_{f}$. There is now a particularly nice solution
for the points of maximal singularity. They actually consist of a single
vacuum, that is invariant under the $\mathbb{Z}_{2n_{c}-n_{f}}$ symmetry of
the theory. The location of this point is given by
$\phi=(0,\dots,0,\Lambda\omega_{2n_{c}-n_{f}},\Lambda\omega_{2n_{c}-n_{f}}^{2},\dots,\Lambda\omega_{2n_{c}-n_{f}}^{-1},\Lambda)\
,$ (130)
where $\omega_{2n_{c}-n_{f}}$ is the $(2n_{c}-n_{f})$’th root of unity. The
factorization of the curve is given by the following algebraic steps
$\displaystyle{\cal P}_{(2n_{c}-n_{f},2n_{c}-n_{f})}(z)$ $\displaystyle=$
$\displaystyle\frac{1}{4}\prod_{j=1}^{2n_{c}-n_{f}}(z-\Lambda\omega_{2n_{c}-n_{f}}^{j})^{2}-\Lambda^{2n_{c}-n_{f}}z^{2n_{c}-n_{f}}$
(131) $\displaystyle=$
$\displaystyle\frac{1}{4}\left(z^{2n_{c}-n_{f}}+\Lambda^{2n_{c}-n_{f}}\right)^{2}-\Lambda^{2n_{c}-n_{f}}z^{2n_{c}-n_{f}}$
$\displaystyle=$
$\displaystyle\frac{1}{4}\left(z^{2n_{c}-n_{f}}-\Lambda^{2n_{c}-n_{f}}\right)^{2}\
.$
Note the peculiarity that all the roots are now doubled. That is what makes
the $\widetilde{n}_{c}$ vacuum different from the other generic $r$ vacua.
Away from $u_{1}=0$, the singularity splints into $2n_{c}-n_{f}$ different
branches. As an example, consider $2n_{c}-n_{f}=2$ where two lines depart from
the $\widetilde{n}_{c}$ vacuum as in Figure 8.
The $r=\widetilde{n}_{c}$ maximal critical point, called the root of the
baryonic branch in [2]121212Since we are now working in ${\rm U}(n_{c})$ and
not ${\rm SU}(n_{c})$, there is no baryonic branch, only a $\widetilde{n}_{c}$
non-baryonic Higgs branch., is a single point, invariant under the discrete
global $\mathbb{Z}_{2n_{c}-n_{f}}$ symmetry of the theory (the anomaly-free
part of the classical ${\rm U}(1)$ R-symmetry). The peculiarity with respect
to the previously discussed $r$ vacua is that now there is an extra
degeneracy. The curve given by (129) and (131) has in fact no unpaired roots.
This means that now every ${\rm U}(1)$ factor in the low energy has its own
low-energy massless hypermultiplet. Table 7 must now be supplemented with an
additional hypermultiplet. By an appropriate choice of basis for the ${\rm
U}(1)$’s, the charges can be taken to be as in Table 8.131313Since we are in
${\rm U}(n_{c})$ and not ${\rm SU}(n_{c})$, we can choose a basis so that all
the charges are diagonal, even in this maximal singularity case.
| ${\rm SU}(\widetilde{n}_{c})$ | $\times$ | ${\rm U}(1)_{0}$ | $\times$ | ${\rm U}(1)_{1}$ | $\times$ | $\cdots$ | ${\rm U}(1)_{2n_{c}-n_{f}-1}$ | $\times$ | ${\rm U}(1)_{2n_{c}-n_{f}}$
---|---|---|---|---|---|---|---|---|---|---
${\bf n_{f}}\times D\phantom{....}$ | $\bf\widetilde{n}_{c}$ | | $\phantom{-}1$ | | | | | | |
$E_{1\phantom{n_{c}-n_{f}-1}}$ | | | | | $\phantom{-}1$ | | | | |
$\vdots\phantom{....}$ | | | | | | | $\ddots$ | | |
$E_{2n_{c}-n_{f}-1}$ | | | | | | | | $\phantom{-}1$ | |
$E_{2n_{c}-n_{f}\phantom{-1}}$ | | | | | | | | | | $\phantom{-}1$
Table 8: Low-energy particles and charges in the $r=\widetilde{n}_{c}$ vacuum.
Now let us examine the breaking of the effective theory at the
$\widetilde{n}_{c}$ singular point. In this case, the superpotential is
$\displaystyle{\cal W}_{r=\widetilde{n}_{c}}^{\prime}$ $\displaystyle=$
$\displaystyle\sqrt{2}\Big{(}\widetilde{D}_{i}A_{\mathbf{r}}D^{i}+\widetilde{D}_{i}A_{0}D^{i}+\Big{.}$
(132)
$\displaystyle\Big{.}+\sum_{j=1}^{2n_{c}-n_{f}}\widetilde{E}_{j}A_{j}E_{i}-\mu
u_{2}(A_{\mathbf{r}},A_{0},\dots,A_{n_{c}-r})\Big{)}\ .$
The important difference with the generic $r$ vacua is that now there is an
hypermultiplet $E_{j}$ for each ${\rm U}(1)$ and the dual-quarks
$\widetilde{D}$, $D$ are now relieved from the duty of condensation. The role
of $W^{\prime 2}+f$ is now played by two roots in zero. All the ${\rm U}(1)$
factors are then Higgsed and can be integrated out. The low-energy theory is
thus a non-Abelian ${\rm SU}(\widetilde{n}_{c})$ gauge theory with the
effective superpotential
${\cal
W}_{r=\widetilde{n}_{c}}^{\prime}=\sqrt{2}\left(\widetilde{D}_{i}A_{\mathbf{r}}D^{i}-\frac{\mu}{2}\hbox{\rm
Tr}{A_{\mathbf{r}}}^{2}\right)\ .$ (133)
$A_{\mathbf{r}}$ is also massive and can be integrated out. We are thus left
with ${\rm SU}(\widetilde{n}_{c})$ ${\cal N}=1$ super–QCD with $n_{f}$
flavors.
This cannot be the infrared of pure ${\cal N}=1$ SQCD. According to the
Seiberg duality, the IR of pure ${\cal N}=1$ SQCD is described by ${\rm
SU}(\widetilde{n}_{c})$ gauge theory with $n_{f}$ flavors plus a meson
$M_{i}^{j}$ and an opportune superpotential $\widetilde{D}MD$ [17]. We have
the right gauge group and the right dual-quark, but the meson $M_{i}^{j}$ is
missing. Note that $A_{\mathbf{r}}$ cannot be identified with the meson of
Seiberg duality. First of all, $A_{\mathbf{r}}$ has no flavor charge. Second,
from (133) we can see that it becomes massive after the $\mu$ breaking. Thus,
it should be integrated out to get the infrared conformal fixed point.
The cases $n_{f}=n_{c}+1$ and $n_{f}=n_{c}$ require special attention. In
these cases, there is no non-Abelian group ${\rm SU}(\widetilde{n}_{c})$. For
$n_{f}=n_{c}+1$, we can still use Table 8 and just delete the ${\rm
SU}(\widetilde{n}_{c})$ column. For $n_{f}=n_{f}$, there is no more a flavor
charged particle, and we should also delete the ${\rm U}(1)_{0}$ column and
the flavored particle ${\bf n_{f}}\times D$.
* * *
Now we are going to find the general solution for the coincidence vacua. We
need to generalize the findings of Section 4. The task seems apparently
difficult, but using a trick, and the help of the Doulgas-Shenker solution, we
shall find quite easily the general solution. First of all, we have to take a
look at the example we already found in the previous section and guess from
the particular case. Then we shall prove that the guess is right.
Consider, for the moment, another theory with double the number of colors and
flavors
$N_{c}=2n_{c}\ ,\qquad N_{f}=2n_{f}\ .$ (134)
Then take the maximal $r=N_{f}/2$ vacua for this theory. We already discussed
at the beginning of the section the exact solution for maximal $r$ vacua, when
the number of flavors is even. The curve is given by
$\displaystyle Y^{2}$ $\displaystyle=$ $\displaystyle\frac{1}{4}{\rm
det}(Z-\Phi)-\Delta Z^{N_{f}}$ (135) $\displaystyle=$ $\displaystyle
Z^{N_{f}}\,{\cal P}_{(N_{c}-N_{f}/2,0)}(Z)\ ,$
and for ${\cal P}_{(N_{c}-N_{f}/2,0)}(Z)$ we have the $N_{c}-N_{f}/2$ Douglas-
Shenker solutions, as in (126).
$Y^{2}=Z^{N_{f}}\left(\frac{Z^{2}}{4}-\Delta^{2}\right)\prod_{j=1}^{N_{c}-N_{f}/2-1}\left(\frac{Z}{2\Delta}-\cos{\frac{\pi
j}{N_{c}-N_{f}/2}}\right)^{2}\Delta^{N_{c}-N_{f}/2-2}\ .$ (136)
We called $\Phi$ the adjoint scalar field for the theory $(N_{c},N_{f})$. The
Coulomb moduli space has dimension $N_{c}$ and is parameterized by the
coordinates
$U_{k}=\frac{1}{k}\hbox{\rm Tr}\,\Phi^{k}\ .$ (137)
Now note that the roots of the curve (136) are symmetric if we exchange $j$
with $N_{c}-N_{f}/2-j$ and simultaneously the sign of the roots. We can thus
combine the roots with the same modulus and opposite sign and rewrite the
curve in the following way, for $N_{f}/2$ odd
$Y^{2}=Z^{N_{f}}\left(\frac{Z^{2}}{4}-\Delta^{2}\right)\prod_{j=1}^{N_{c}/2-[N_{f}/4]}\left(\frac{Z^{2}}{4\Delta^{2}}-\left(\cos{\frac{\pi
j}{N_{c}-N_{f}/2}}\right)^{2}\right)^{2}\Delta^{N_{c}-N_{f}/2-2}\ ,$ (138)
and in the following for $N_{f}/2$ even
$Y^{2}=Z^{N_{f}}\left(\frac{Z^{2}}{4}-\Delta^{2}\right)\prod_{j=1}^{N_{c}/2-N_{f}/4-1}\left(\frac{Z^{2}}{4\Delta^{2}}-\left(\cos{\frac{\pi
j}{N_{c}-N_{f}/2}}\right)^{2}\right)^{2}Z^{2}\Delta^{N_{c}-N_{f}/2-2}\ .$
(139)
Note that this looks quite similar to the coincidence points we found in
Section 4.
To consolidate this guess, note that for the maximal $r=N_{f}/2$ vacua we can
certainly say that the odd part of the moduli space coordinates vanishes
$U_{{\rm odd}}=0\ .$ (140)
We can thus focus our attention on the sub-moduli space of $U_{\rm even}$ that
has exactly the same dimension of the moduli space for the $(n_{c},n_{f})$
theory. We finally make the following mapping between the $(N_{c},N_{f})$
theory and the original $(n_{c},n_{f})$:
$Z^{2}=z\ ,\qquad\Delta^{2}=\Lambda\ .$ (141)
And between the moduli spaces:
$U_{2k}=u_{k}\ .$ (142)
Now is just a matter of rewriting (138) and (139) with the new coordinates,
and we get the following factorization of the SW curve
$y^{2}=\frac{1}{4}z^{2[n_{f}/2]+1}(z-4\Lambda)\prod_{j=1}^{n_{c}-[n_{f}/2]-1}\left(z-4\Lambda\left(\cos{\frac{\pi
j}{2n_{c}-n_{f}}}\right)^{2}\right)^{2}\ .$ (143)
This is the coincidence vacuum we were looking for. This is valid for both
$n_{f}$ even or odd. At the end, everything still follows from the identity
(127).141414The use of the theory $(N_{c},N_{f})$ has been only a mathematical
trick to get the solution for the coincidence vacua passing through Douglas-
Shenker. But maybe there is something physical behind this bigger theory. To
pursue this idea, certainly a cubic superpotential should be used. We thus
have the expected solution with $2[n_{f}/2]+1$ roots in zeros and all the
others, except one, doubled. The coordinate $\hbox{\rm Tr}\,\phi$ of this
coincidence vacuum is given by
$\displaystyle u_{1}$ $\displaystyle=$ $\displaystyle U_{2}$ (144)
$\displaystyle=$
$\displaystyle\sum_{j=1}^{2n_{c}-n_{f}}2\Lambda\left(\cos{\frac{\pi(j-1/2)}{2n_{c}-n_{f}}}\right)^{2}=(2n_{c}-n_{f})\Lambda\
,$
with the exception $u_{1}=0$ for $2n_{c}-n_{f}=1$. The coordinate $\hbox{\rm
Tr}\,\phi^{2}/2$ of this coincidence vacuum is given by
$\displaystyle u_{2}$ $\displaystyle=$ $\displaystyle U_{4}$ (145)
$\displaystyle=$
$\displaystyle\sum_{j=1}^{2n_{c}-n_{f}}4\Lambda^{2}\left(\cos{\frac{\pi(j-1/2)}{2n_{c}-n_{f}}}\right)^{4}=\frac{3}{2}(2n_{c}-n_{f})\Lambda^{2}\
,$
with the exceptions $u_{2}=0,2$ for $2n_{c}-n_{f}=1,2$. The other coincidence
vacua are just obtained with a $\mathbb{Z}_{2n_{c}-n_{f}}\in{\rm U}(1)_{R}$
transformation. To select these vacua, we need then to use the following
superpotential
$W(\Phi)=\mu\hbox{\rm Tr}\,\left(\frac{\Phi^{2}}{2}-2\Lambda e^{\frac{i\,2\pi
k}{2n_{c}-n_{f}}}\Phi\right)\ ,\qquad k=1,\dots,2n_{c}-n_{f}\ ,$ (146)
where the index $k$ corresponds to the various coincidence vacua. As in the
$n_{c}=2$, $n_{f}=2$ example, the macroscopic dynamic in these coincidence
points is a non-local superconformal field theory.
We are also now ready to understand the multiple collision anticipated in
Figure 4. At the level $\hbox{\rm Tr}\,\phi=0$ of the moduli space, there are
multiple discrete $r$ vacua where the curve is factorized similar to Eq.
(123), and the residual part ${\cal P}_{(n_{c}-r,n_{f}-2r)}(z)$ is further
factorized so that all roots are doubled with the exception of two of them.
These two unpaired roots correspond the roots of $W^{\prime}(z)^{2}$, split by
the quantum effect of the polynomial $f(z)$. We want to start from any one of
these $r$ vacua and leave the $u_{1}=0$ plane, but remaining in this maximal
singularity sub-manifold (that is, all the roots paired except two of them).
One complex parameter, $u_{1}$, can thus be adjusted in order to bring one of
the unpaired roots anywhere we want in the complex plane. To have a
coincidence, we need one of these unpaired roots to collide with the bunch of
zeros $z^{2r}$ representing the quark singularity. So it seems that starting
from any one of the $r$ vacua, and tuning the coordinate $u_{1}$, we can
create a different coincidence vacuum. The fact is that they all end up in the
same coincidence point, as described by Figure 4. The reason is the following:
an odd number of zeros $r^{2h+1}$ is not possible unless $h=[n_{f}/2]$. That
means that if we start from any $r$ vacuum and we tune $u_{1}$ to bring an
extra root in zero, that automatically brings other $[n_{f}/2]-r$ couples of
roots also in zero. This is exactly what happens in Figure 8 for
$(n_{c},n_{f})=(2,2)$.
Let us check the previous claim. In the generic $r$ singularity, the curve
factorizes as
$y^{2}=z^{2r}\left(\frac{1}{4}P_{n_{c}-r}(z)^{2}-\Lambda^{2n_{c}-2r}z^{n_{f}-2r}\right)\
.$ (147)
It is clear that, if $n_{f}-2r>1$, is not possible to bring another unpaired
root in zero. Since $P_{n_{c}-r}(z)^{2}$ should also vanish, they always come
in pairs. When $n_{f}-2r=0$ or $n_{f}-2r=1$, it then is possible to bring a
single unpaired root in zero.
We can understand better the intersections of the $r$ branches by considering
only the minimal $\widetilde{n}_{c}$ and the maximal $[n_{f}/2]$. Consider
also for simplicity $n_{f}$ even. We have one $\widetilde{n}_{c}$ singularity,
that is, the intersection of $2n_{c}-n_{f}$ different singularity co-dimension
two surfaces at the level $u_{1}=0$. Then the maximal $r=n_{f}/2$ non-baryonic
roots, which are $n_{c}-n_{f}/2$ vacua previously described. Each one of the
$n_{c}-n_{f}/2$ roots collide with two of the surfaces emanating from the
baryonic branch. For example in the $n_{c},n_{f}=2,2$ case the singularity
that departs from the $r=1$ vacuum intersects with the two singularities that
depart from the $r=0$ vacuum (Figure 8). The same thing happens in the general
case. Every singularity that departs from an $n_{f}/2$ vacuum (there are
$n_{c}-n_{f}/2$ of them) intersects with two singularities of the
$2n_{c}-n_{f}$ that depart from the $\widetilde{n}_{c}$ vacuum.
So we can finally try to summarize and conjecture how the pure ${\cal N}=1$
super-QCD emerges from these coincidence points. We can consider the simplest
example where $n_{f}=2n_{c}-2$. In this case, we have only two vacua at
$u_{1}=0$: the $\widetilde{n}_{c}$ vacuum and the $[n_{f}/2]$ vacuum. The
situation is completely analogous to the example $n_{c}=n_{f}=2$ discussed in
Section 4. We can thus refer to Figure 8 to understand what is happening. The
$\widetilde{n}_{c}$ vacuum has low-energy dynamics summarized in Table 8 while
the $[n_{f}/2]$ vacuum is summarized in Table 7. The two non-Abelian
multiplets are non-local between them. Let’s use $D,\widetilde{D}$ to denote
the $\widetilde{n}_{c}$ dual-quark and $Q,{\widetilde{Q}}$ to denote the
$[n_{f}/2]$ low-energy hypermultiplet. We can suspect that in the
$\mu\to\infty$ limit the $Q$ and ${\widetilde{Q}}$ condense, forming a mesonic
bound state $M={\widetilde{Q}}Q$. This is now local with the dual quarks $D$
and $\widetilde{D}$. This is probably the way Seiberg duality is obtained in
the $\mu\to\infty$ limit.
It is good to add more comments about this last, crucial point. The infrared
theory at the coincidence point is a superconformal fixed point where many
non-local degrees of freedom are entangled together. Since it is not possible
to write an effective local Lagrangian, the study of this theory becomes
difficult. Our approach has been to consider the theory at the various $r$
vacua, where a low-energy description is available, and progressively move
toward this coincidence point. In these $r$ vacua, there is a ${\cal N}=2$,
${\rm SU}(r)\times{\rm U}(1)_{0}$ gauge theory with a dual-quark in the
fundamental of flavor ${\bf n_{f}}$ and in the fundamental of gauge ${\bf
n_{f}}$. The beta function for the non-Abelian gauge coupling is proportional
to $2r-n_{f}$. This is always infrared free (except for the case the maximal
branch $r=n_{f}/2$ when $n_{f}$ even). The lowest case $r=\widetilde{n}_{c}$
and the maximal case $r=[n_{f}/2]$ are particularly important. As we already
mentioned, the first one provides the ${\rm SU}(\widetilde{n}_{c})$ and the
dual-quarks $\widetilde{D}$, $D$, which are essential degrees of freedom in
the Seiberg duality. The last $r=[n_{f}/2]$ is also important because it gives
the maximal non-baryonic branch $r=[n_{f}/2]$, which is crucial if we want to
recover ${\cal N}=1$ SQCD in the $\mu\to\infty$ limit. Note that the beta
functions for the ${\rm SU}(\widetilde{n}_{c})$ and ${\rm SU}([n_{f}/2])$ are,
respectively, the lowest and the highest.
When the various $r$ vacua collide in the coincidence point, all the degrees
of freedom are merged in this superconformal field theory. We cannot write an
explicit theory since they are mutually non-local. But we can nevertheless say
that these degrees of freedom are there. The weakest ${\rm
SU}(\widetilde{n}_{c})$ and the strongest ${\rm SU}([n_{f}/2])$ play an
essential role. As we break ${\cal N}=2$ with the opportune superpotential
(146), we select this coincidence point, and we give mass to all the adjoint
scalar fields $A_{{\bf r}}$. The dual-quarks do not condense, and so there is
no Higgs effect. In the $\mu\to\infty$, we have to compute again the beta
functions that are now $3r-n_{f}$. In the maximal case, the beta function is
proportional to $3[n_{f}/2]-n_{f}$ and is now strong in the infrared. It is
the strongest one among the various $r$. This supports the previous claim that
the dual-quarks of the maximal branch condense and provide the meson
$M_{i}^{j}={\widetilde{Q}}_{i}Q^{j}$, essential in the Seiberg dual theory.
For $n_{f}=n_{c}+1$ and $n_{f}=n_{c}$, there is no non-Abelian group ${\rm
SU}(\widetilde{n}_{c})$. But still the minimal $r$ vacuum ($r=1$ for the first
case and $r=0$ for the second) provides an essential ingredient for the phase
of low-energy ${\cal N}=1$ SQCD: the baryon. In the $n_{f}=n_{c}+1$, it is
fundamental in flavor (see Table 8 without the ${\rm SU}(\widetilde{n}_{c})$
column) while for the $n_{c}=n_{f}$ case is flavor neutral (see Table 8
without the ${\rm SU}(\widetilde{n}_{c})$ and ${\rm U}(1)_{0}$ columns). So
the minimal $r=1,0$ vacuum provides the baryon while the maximal $r=[n_{f}/2]$
vacuum provides the meson.
## 6 MQCD
Figure 12: Classical brane configuration whose low energy spectrum in
$\mathbb{R}^{3,1}$ coincide with super-QCD.
Super-QCD can be implemented as a low-energy theory of certain brane setup
configurations. A well-known approach consists of taking type IIA string
theory and a set of orthogonal NS$5$ and D$4$ branes. Using the common
convections, we have the displacement of branes summarized in Table 9 and
Figure 12.
| | $\mathbb{R}^{3,1=}x^{0},\dots,x^{3}$ | | $v=x^{4}+ix^{5}$ | | $x^{6}$ | | $w=x^{7}+ix^{8}$ | | $x^{9}$ |
---|---|---|---|---|---|---|---|---|---|---|---
$\phantom{n_{c}\,\,\,}\mathrm{NS}5$ | | $*$ | | $*$ | | $0$ | | $0$ | | $0$ |
$n_{c}\,\,\,\mathrm{D}4$ | | $*$ | | $0$ | | $*$ | | $0$ | | $0$ |
$\phantom{n_{f}\,\,\,}\mathrm{NS}5^{\prime}$ | | $*$ | | $*$ | | $L_{6}$ | | $\mu v$ | | $0$ |
$n_{f}\,\,\,\mathrm{D}4$ | | $*$ | | $0$ | | $*$ | | $0$ | | $0$ |
Table 9: Brane setup in type IIA string theory whose low energy on
$\mathbb{R}^{3,1}$ coincide with ${\cal N}=2$ and ${\cal N}=1$ SQCD.
This is just the classical configuration. Taking into account the string
coupling effects, the NS$5$ branes are logarithmically bended due to the
pulling of the D4 branes. This log-bending corresponds to the running of the
coupling constant in the four-dimensional, low-energy action. Other strong
coupling effects are related to the junctions between the D$4$ branes and the
NS5 branes. A way to resolve these singularities provides a window into the
strong coupling of the four-dimensional theory.
One way to study non-perturbative effects is to lift type IIA string theory to
M-theory [19, 20]. Now the D4 and NS5 branes are all described by the same
object, an M$5$ brane. The embedding of the M5 brane is related to the
Seiberg-Witten curve and the factorization equation in the ${\cal N}=1$ case.
We thus obtain a beautiful geometric interpretation of many field theoretical
quantities, mostly the chiral and topological ones. The goal for this section
is to describe the $r$ vacua and the coincidence vacua in the MQCD framework.
The MQCD curve shall provide further evidence for why the coincidence points
are so special.
After the M-theory lifting, the internal space is now
$Y=\mathbb{R}^{6}\times\mathbb{S}^{1}$, which we parameterize with three
complex coordinates $v,\ w$ and $s$, and one real coordinate $x^{7}$. $v$, $w$
and $x^{7}$ parameterize the
$\mathbb{R}^{5}\subset\mathbb{R}^{6}\times\mathbb{S}^{1}$ while the complex
coordinate $s=x^{6}+ix^{10}$ parameterizes the remaining
$\mathbb{R}\times\mathbb{S}^{1}$. $x^{10}$ is a periodic coordinate
parameterizing the M-theory circle. Weakly coupled type IIA string is
recovered when the compactification radius is very small. We define the
exponential mapping $t=\exp(s)$, which is valued in $\mathbb{C}^{*}$.
The brane setup of Figure 12, now becomes a unique M$5$ brane extended along
$\mathbb{R}^{3,1}$ times a Riemann surface $\Sigma$ embedded in $Y$. This
embedding $\Sigma\subset Y$ contains some crucial information about the
quantum field theory of interest. The surface $\Sigma$ is what finally is
related to the Seiberg-Witten data of the original QFT (we have to use the
change of coordinates $t=y+P_{n_{c}}(v)$). In the ${\cal N}=2$ case, the M$5$
brane embedding is the algebraic surface by the equations
$t^{2}-2P_{n_{c}}(v)t+\Lambda^{2n_{c}-n_{f}}v^{n_{f}}=0\ ,$ (148)
combined with $w=0$. The solutions to this equation are
$t_{1,2}=P_{n_{c}}(v)\pm\sqrt{P_{n_{c}}(v)^{2}-4\Lambda^{2n_{c}-n_{f}}v^{n_{f}}}\
.$ (149)
In the square root, we have exactly the Seiberg-Witten curve. These two
branches correspond, asymptotically, to the two NS$5$ branes. The Riemann
surface $\Sigma$ is thus a double cover of the $v$ plane. The solutions at
$v\to\infty$ are
$t_{1}\sim v^{n_{c}}\ ,\qquad t_{2}\sim v^{n_{f}-n_{c}}\ .$ (150)
and correspond to the NS$5$ and NS$5^{\prime}$ branes. The flavor branes
correspond to an expansion around zero. In a generic point of the moduli
space, where $P_{n_{c}}(0)\neq 0$, the flavor brane is attached to the second
sheet
$t_{1}\sim P_{n_{c}}(0)\ ,\qquad t_{2}\sim v^{n_{f}}\ .$ (151)
These correspond to the $n_{f}$ flavor D4 branes attached to the
NS$5^{\prime}$ brane.
So far for the ${\cal N}=2$ theory. The breaking to ${\cal N}=1$ by mean of
the superpotential $\hbox{\rm Tr}\,W(\Phi)$, correspond in this setting to a
deformation of the NS$5^{\prime}$ brane into the $w$ plane [21]. From the MQCD
perspective, we have to supplement Eq. (148) with a second equation that
provides the information about the embedding in the $w$ plane
$w^{2}-2W^{\prime}(v)w-f(v)=0\ ,$ (152)
where the polynomial $f(v)$ captures the quantum corrections to the
superpotential. The two branches of the solutions are
$w_{1,2}=W^{\prime}(v)\pm\sqrt{W^{\prime}(v)^{2}+f(v)}\ .$ (153)
Classically, without $f(v)$, we have $w=0$ that corresponds to the NS$5$ brane
and $w=2W^{\prime}(v)$ that corresponds to the NS$5^{\prime}$ brane
deformation in the $w$ plane.
Without the superpotential the ${\cal N}=2$, curve (148) enjoys a moduli space
of the solutions parameterized by the $n_{c}$ coefficients contained in the
polynomial $P_{n_{c}}(v)$. But things are more complicated when a
superpotential is introduced and we have to deal also with Eq. (152). The
equations (149) and (153) are both two branches that cover the $v$ plane.
Passages from one branch to another are determined by what is inside the
square root: the Seiberg-Witten curve in the first case and the ${\cal N}=1$
curve in the second. For a generic value of $P_{n_{c}}$, we have a total of
four branches connected together. So nothing that resembles the classical
picture. Only for a particular discrete set of solutions, exactly when the
factorization (2) is satisfied, can we separate the four branches into two
disconnected parts: $t_{1},w_{1}$ together with $t_{2},w_{2}$, and
$t_{1},w_{2}$ together with $t_{2},w_{1}$. And this is the MQCD explanation
for the ${\cal N}=1$ factorization of the Seiberg-Witten curve.
For the generic $r$ vacuum, the curve is given by
$t^{2}-2T_{n_{c}-n_{f}/2}\left(\frac{v^{2}}{2}\right)^{2}v^{n_{f}/2}t+\Lambda^{2n_{c}-n_{f}}v^{n_{f}}=0\
,$ $w^{2}-2\mu vw-1=0\ .$ (154)
Flavor branes are separated into two groups of $r$ and $n_{f}-r$ units (see
Figure 14). These two correspond to two different spikes of the M5 brane
separated in the $w$ plane by a distance $\sqrt{W^{\prime}(m)^{2}+f(m)}$
(remember the two solutions in Eq. (153)). This in fact corresponds to the
dual-quark condensate (95).
Figure 13: MQCD curve corresponding to a generic $r$ vacuum.
Figure 14: MQCD curve for the specific case $r=\widetilde{n}_{c}$.
The $r=\widetilde{n}_{c}$ branch is an exception. Now the curve is given by
$t^{2}-2\left(z^{2n_{c}-n_{f}}+\Lambda^{2n_{c}-n_{f}}\right)v^{\widetilde{n}_{c}}t+\Lambda^{2n_{c}-n_{f}}v^{n_{f}}=0\
,$ $w^{2}-2\mu vw=0\ .$ (155)
In this case, there is an extra massless particle, and the MQCD becomes
divided into two distinct pieces. The two planes meet only at
$x_{6}\to\infty$. The reason can be seen from the factorization of the SW
curve (131). All the roots are now paired and positioned at
$\Lambda\omega_{2n_{c}-n_{f}}^{j}$. The cuts are in a closed polygonal shape
with $2n_{c}-n_{f}$ sides; they topologically separate the interior and the
exterior of the polygon in the complex $v$ plane. A spike corresponding to
$n_{f}-n_{c}$ D$4$ branes departs from the NS$5^{\prime}$ branes and joins at
$x_{6}\to\infty$ with a spike of $n_{c}$ D$4$ branes departed from the left
NS$5$ branes (see Figure 14). There is no asymptotic separation in the $w$
plane and that means no condensation of the dual quark. Note also that the
topology of the curve is different. It consists of two disconnected pieces
that join only at $x_{6}\to\infty$ (see Figure 14).
Figure 15: MQCD curve for a coincidence point.
We have thus seen that none of the singularities at $\hbox{\rm Tr}\,\phi=0$
have the right MQCD curve to describe ${\cal N}=1$ SQCD at $\mu\to\infty$. In
all cases, there is a mass gap and nothing in the low energy. In the baryonic
root, the quarks have zero expectation value, but the topology of the curve is
changed. The coincidence points are instead very special. The condensate of
the quark is zero, and there is no change of topology. The curve is given by
$t^{2}-2v^{[n_{f}/2]}\prod_{j=1}^{n_{c}-[n_{f}/2]}\left(z-4\Lambda\left(\cos{\frac{\pi(j-1/2)}{2n_{c}-n_{f}}}\right)^{2}\right)^{2}t+\Lambda^{2n_{c}-n_{f}}v^{n_{f}}=0\
,$ $w^{2}-2\mu vw-v=0\ .$ (156)
Good things happen in this circumstance. First of all, the flavor branes
correspond to a unique spike of the M$5$ brane. There is no asymptotic
separation in the $w$ plane and that means no condensation. Furthermore,
despite what happened for the $\widetilde{n}_{c}$ vacuum, the Riemann surface
preserves its topology. The curve is still a double cover of the $v$ plane
(Figure 15).
## 7 Conclusion
In the paper, we addressed a problem of extended supersymmetry breaking, from
${\cal N}=2$ to ${\cal N}=1$. Generically, this is achieved by giving a mass
term to the adjoint scalar field $\phi$ of the ${\cal N}=2$ gauge
supermultiplet. Classically, this works perfectly fine, with or without matter
hypermultiplets. When quarks fields are present, some of the flat directions
of the ${\cal N}=1$ theory are already present in the ${\cal N}=2$ theory.
Others are recovered as pseudo-moduli with mass proportional to $\propto
1/\mu$, that become massless in the $\mu\to\infty$ limit.
In quantum theory, there are new subtleties that spoil this breaking pattern.
First of all, the notion of the origin of the moduli space does not hold
anymore. What was the classical origin of the moduli space is in some sense
split into various vacua labeled by an integer $r$. A mass term for the
adjoint scalar field $\phi$ selects any one of these vacua, but none of them,
in the $\mu\to\infty$ limit, flows exactly to pure ${\cal N}=1$ super-QCD.
Figure 16: RG flow for various values of $\mu$. The five corners are fixed
points of the RG flow.
The reason is quantum mechanical and must be traced back to the operator that
is generated after integrating out the field $\phi$. This operator
$W=\frac{1}{{\mu\sqrt{2}}}\ {\widetilde{Q}}Q\ {\widetilde{Q}}Q\ ,$ (157)
looks apparently harmless. Since it is proportional to $1/\mu$, we would be
tempted to conclude that its effects become negligible in the $\mu\to\infty$
limit. But quantum mechanically it leads to an operator of dimension $d$
generally smaller than three. In ${\cal N}=1$ SQCD, the dimension of the meson
operator is
$D({\widetilde{Q}}Q)=3R^{\prime}({\widetilde{Q}}Q)/2=3(n_{f}-n_{c})/n_{f}$. So
(157) leads generally to an operator of dimension
$D({\widetilde{Q}}Q\ {\widetilde{Q}}Q)=6\frac{n_{f}-n_{c}}{n_{c}}\ ,$ (158)
for very large $\mu$. For $n_{f}=2n_{c}$ this is exactly marginal.151515Note
the non-triviality of that. The ${\cal N}=1$ is in general not sensitive to
the equality $n_{f}=2n_{c}$ which is instead very important for the ${\cal
N}=2$ theory. It is thus a relevant operator, and it grows as
$(\epsilon/\epsilon_{0})^{d-3}$ as the energy scale of the renormalization
group flow $\epsilon$ goes to zero. No matter how small the coefficient in
front of it is, it will always blow up for sufficiently small energies of the
RG flow and force the theory to flow to another IR fixed point. Most of the
time, like in the cases $\widetilde{n}_{c}<r\leq[n_{f}/2]$, the theory flows
to nothing, and there is a mass gap for the gauge degrees of freedom. There is
a special case, the minimal value $r=\widetilde{n}_{c}$, where the theory
flows to ${\rm SU}(\widetilde{n}_{c})$ with $n_{f}$ dual-quarks
$\widetilde{D},D$. Although very near to what we should expect in pure ${\cal
N}=1$ SQCD, still there is a missing piece: the meson $M_{i}^{j}$ and its
superpotential interaction with $\widetilde{D},D$. We can interpret this
missing piece as damage still caused by the relevant operator (157). We can
thus conclude that for generic ${\rm SU}(n_{c})$ SQCD there is no possibility
to flow exactly from ${\cal N}=2$ to ${\cal N}=1$ (Figure 16).
But the classical analysis comes back as a source of inspiration. As we saw
from the prototype model, super-QED with hypermultiplets of mass $m$ and
superpotential $W(\phi)$, and the classical theories have a large spectrum of
vacua, and a few of them, in general, flow to pure ${\cal N}=1$ super-QED as
the superpotential goes to infinity. There are Coulomb vacua where $\phi$ is
equal to the root of the superpotential and the quarks are massive. There are
gauge-flavor locked vacua where $\phi$ is locked to some quark mass and the
gauge group is broken by the quark condensate $W^{\prime}(m)$. Finally, there
are the “coincidence” vacua that arise under the very particular circumstance
in which a root $a$ of $W^{\prime}$ precisely coincides with a hypermultiplet
mass $m$. Choosing $\phi=m=a$, we get a particular vacuum in which the quarks
are massless, but they do not condense. These are the right vacua in which to
flow from ${\cal N}=2$ to ${\cal N}=1$ super-QED with massless quarks.
Returning to the non-Abelian case, we thus decided to extend the space of
interest from ${\rm SU}(n_{c})$ to ${\rm U}(n_{c})$. ${\rm U}(n_{c})$ ${\cal
N}=2$ super-QCD has, with respect to ${\rm SU}(n_{c})$, one dimension more in
the Coulomb moduli space given by the coordinate $u_{1}=\hbox{\rm Tr}\,\phi$.
This extra dimension is crucial. The quantum effects, as already said, erase
the notion of the origin of the moduli space and split it into various $r$
vacua (in the $\hbox{\rm Tr}\,\phi=0$ section). None of these vacua can be
considered as the right generalization of the classical notion of coincidence
vacua. None flows to ${\cal N}=1$ super-QCD after the mass breaking term is
sent to infinity. But extending the search to the $u_{1}$ dimension we find a
nice surprise. We find $2n_{c}-n_{f}$ points that, by all rights, can be
considered as the quantum generalization of the classical notion of
coincidence vacua. With a suitable superpotential
$W(\Phi)=\mu\hbox{\rm Tr}\,\left(\frac{\Phi^{2}}{2}-2\Lambda e^{\frac{i\,2\pi
k}{2n_{c}-n_{f}}}\Phi\right)\ ,\qquad k=1,\dots,2n_{c}-n_{f}\ ,$ (159)
we can select any one of these vacua. And each one is a good coincidence
point.
The statement is thus that these coincidence points are the right ones to flow
from ${\cal N}=2$ super-QCD to ${\cal N}=1$ super-QCD. They are the points in
which the effects of the relevant operator (157) are minimized and the theory
flows as close as possible to ${\cal N}=1$ super-QCD. Let us summarize the
arguments we presented in the paper.
* •
The formula for the quark condensate,
$\widetilde{q}q\propto\sqrt{W^{\prime}(m)^{2}+f(m)}$, is the starting point in
the search for the generalization of the classical notion of coincidence.
Quantum mechanically, the roots of $W^{\prime}(z)^{2}$ are in general split by
the polynomial $f(z)$. For the quark condensate to vanish, we need to send one
of these roots near the bunch of zeros at the mass $m$. This guiding principle
gives us the location of the coincidence points and the factorization of the
Seiberg-Witten curve in these vacua.
* •
Having done this, we achieve also another, a priori not required, result.
These coincidence points are located at the intersection between singularities
that depart from all the $r$ vacua, from $\widetilde{n}_{c}$ to $[n_{f}/2]$
(the collision of Figure 4). We thus recover the notion of the origin of the
moduli space that in the $\hbox{\rm Tr}\,\phi=0$ section of the moduli space
was lost. The Higgs branch emanating from these vacua is equal to the maximal
non-baryonic branch for $r=[n_{f}/2]$. But the theory at the root of the
branch is much richer than the original $r=[n_{f}/2]$ vacuum at $\hbox{\rm
Tr}\,\phi=0$. It is a non-local strongly interacting theory, a particular kind
of Argyres-Douglas singularity.
* •
The MQCD approach provides an interesting point of view, giving some geometric
intuition about what is going on. The classical brane setup consists of NS$5$
branes and perpendicular D$4$ branes. Some quantum effects, such as the
Seiberg-Witten curve and the factorization due to the superpotential, can be
analyzed by lifting to M-theory where the branes are described by a single
M$5$-brane with a proper embedding in the $v,t,w$ space. In ordinary $r$
vacua, the flavor D$4$-branes are divided into two sets, $r$ and $n_{f}-r$,
separated by a $w$ distance of $\sqrt{W^{\prime}(m)^{2}+f(m)}$. This is the
signal of quark condensation. For the particular case $r=\widetilde{n}_{c}$,
the flavor branes are not separated in the $w$ plane (the quarks do not
condense). But in this particular case the M$5$ brane gets divided into two
separated curves that meet only at $v\to+\infty$. Only for the coincidence
vacua we have are the quarks D$4$-branes non-separated and the topology of the
curve does not change. This is certainly what is closer to the classical
realization of a rotation of the NS$5^{\prime}$ brane while keeping the D$4$
gauge and flavor branes all allineated.
* •
We said that the vacua at $u_{1}=0$ are not the right coincidence ones because
of the relevant operator (157). In $\mu\to\infty$, the coefficient in front of
the operator goes to zero, but the operator itself always goes to infinity for
sufficiently small energies of the RG flow. This is a relevant perturbation
that makes the theory flow down from the pure ${\cal N}=1$ super-QCD infrared
fixed point. In general, we cannot predict where the theory will end, but one
thing we can certainly say: it is something smaller than the pure ${\cal N}=1$
vacuum. We should not underestimate this piece of information. The generic $r$
vacua are trivial examples, since the theory flows to nothing. The
$r=\widetilde{n}_{c}$ is an interesting case. From ${\cal N}=1$ ${\rm
SU}(\widetilde{n}_{c})$ with $n_{f}$ dual-quarks $\widetilde{D}$, $D$ and the
meson $M_{i}^{j}$ (the IR of pure ${\cal N}=1$ SQCD), we flow to the same
theory, just without the meson.
The coincidence points, as part of the $r=\widetilde{n}_{c}$ singularity,
contain the degrees of freedom of the gauge group ${\rm
SU}(\widetilde{n}_{c})$ and the dual-quark $\widetilde{D}$, $D$. Since they
are also part of the other $r$ singularities, they contain many more degrees
of freedom. It is hard to imagine that there is something different where the
theory could flow, lower than pure ${\cal N}=1$ and higher than the one at the
$r=\widetilde{n}_{c}$ vacuum. What is missing is just the meson, and, as we
saw, the coincidence vacua are certainly capable of providing it.
| $n_{f}<n_{c}$ | | $\mu\Lambda_{\mbox{\tiny${\cal N}=2$ }}\longrightarrow\infty\phantom{0\Lambda_{\mbox{\tiny${\cal N}=1$ }}^{2}}$
---|---|---|---
| $n_{f}=n_{c}$ | | $\mu\Lambda_{\mbox{\tiny${\cal N}=2$ }}\longrightarrow\Lambda_{\mbox{\tiny${\cal N}=1$ }}^{2}\phantom{0\infty}$
| $n_{f}>n_{c}$ | | $\mu\Lambda_{\mbox{\tiny${\cal N}=2$ }}\longrightarrow 0\phantom{\infty\Lambda_{\mbox{\tiny${\cal N}=1$ }}^{2}}$
Table 10: Scaling of $\mu\Lambda_{\mbox{\tiny${\cal N}=2$ }}$ as
$\mu\to\infty$.
The conclusion is that these particular coincidence points should be
considered as the quantum analog of coincidence between the hypermultiplet
mass and the root of the $W^{\prime}$. Although the quark condensate vanishes,
it is plausible that the Higgs moduli space will be modified by the
$\mu\Lambda_{\mbox{\tiny${\cal N}=2$ }}$ term. In Table 10, we have the
$\mu\Lambda_{\mbox{\tiny${\cal N}=2$ }}$ condensate scales as $\mu$ is sent to
infinity (remember the relationship (32) between the two scales). This is
probably related to the fact that the moduli space for ${\cal N}=1$ SQCD is
quantum modified for $n_{f}=n_{c}$ and not modified from the classical one for
$n_{f}>n_{c}$.
Note that, at the contrary of color-flavor locked vacua, coincidence vacua can
also exist for $n_{f}<n_{c}$. Since the color branes are locked to a root of
$W^{\prime}$, there is no lower bound on the number of flavors we can attach
to the color branes. Although we have focused our attention on $n_{f}\geq
n_{c}$, many things go unchanged for $n_{f}$ smaller. In particular, there are
still $2n_{c}-n_{f}$ coincidence points in the moduli space and their position
and curve are still described by the findings of Section 5. It is known that
${\cal N}=1$ SQCD for $n_{f}<n_{c}$ has an instanton generated run-away
potential. This could be probably related to the scaling of
$\mu\Lambda_{\mbox{\tiny${\cal N}=2$ }}$ of Table 10.
We are left with challenging questions for the future. One question is what
happens to the operator (157) in the coincidence vacua. Another question
regards the $\mu$ transition from $0$ to $\infty$. The superconformal field
theory at the coincidence vacua is like a boiling soup, with many non-local
degrees of freedom. We have conjectured how the dual-quark and the meson of
the Seiberg duality emerge out of it in the $\mu\to\infty$ limit. Certainly, a
more detailed understanding of this transition is needed. In particular, we do
not know if, in the IR fixed point, this transition is sharp or if it is a
marginal deformation from the superconformal ${\cal N}=2$ to the infrared of
the Seiberg duality. The completion of this program should eventually be
considered the field theoretical proof of the Seiberg duality that was
initiated in [2].
###### Acknowledgments.
I want to thank M. Shifman and A. Yung for useful discussions about susy-QCD,
and in particular the heterotic vortex problem. I am grateful to M. Shifman
for his help and support. I want to thank K. Konishi for discussions in the
past regarding the susy-QCD moduli space and the APS paper. I want also to
thank various collaborators who helped me in the past in the study of related
problems: R. Auzzi, J. Evslin and M. Matone. I want to thank Ki-Myeong Lee and
people at KIAS for their hospitality in early June 2008. I want also to thank
A. Vainshtein for recent discussions. This work is supported by DOE grant DE-
FG02-94ER40823.
## References
* [1]
* [2] P. C. Argyres, M. R. Plesser and N. Seiberg, “The Moduli Space of N=2 SUSY QCD and Duality in N=1 SUSY QCD,” Nucl. Phys. B 471, 159 (1996) [arXiv:hep-th/9603042].
* [3] S. Bolognesi, “The holomorphic tension of vortices,” JHEP 0501, 044 (2005) [arXiv:hep-th/0411075]. S. Bolognesi, “The holomorphic tension of nonabelian vortices,” Nucl. Phys. B 719, 67 (2005) [arXiv:hep-th/0412241].
* [4] M. Shifman and A. Yung, “Non-abelian flux tubes in SQCD: Supersizing world-sheet supersymmetry,” Phys. Rev. D 72, 085017 (2005) [arXiv:hep-th/0501211].
* [5] M. Edalati and D. Tong, “Heterotic vortex strings,” JHEP 0705, 005 (2007) [arXiv:hep-th/0703045]. D. Tong, “The quantum dynamics of heterotic vortex strings,” JHEP 0709, 022 (2007) [arXiv:hep-th/0703235].
* [6] M. Shifman and A. Yung, “Heterotic Flux Tubes in N=2 SQCD with N=1 Preserving Deformations,” arXiv:0803.0158 [hep-th]. M. Shifman and A. Yung, “Large-N Solution of the Heterotic N=(0,2) Two-Dimensional CP(N-1) Model,” arXiv:0803.0698 [hep-th].
* [7] R. Auzzi, S. Bolognesi and J. Evslin, “Monopoles can be confined by 0, 1 or 2 vortices,” JHEP 0502, 046 (2005) [arXiv:hep-th/0411074]. S. Bolognesi and J. Evslin, “Stable vs unstable vortices in SQCD,” JHEP 0603, 023 (2006) [arXiv:hep-th/0506174].
* [8] K. A. Intriligator and N. Seiberg, “Lectures on supersymmetric gauge theories and electric-magnetic duality,” Nucl. Phys. Proc. Suppl. 45BC, 1 (1996) [arXiv:hep-th/9509066].
* [9] N. Seiberg and E. Witten, “Electric - magnetic duality, monopole condensation, and confinement in N=2 supersymmetric Yang-Mills theory,” Nucl. Phys. B 426, 19 (1994) [Erratum-ibid. B 430, 485 (1994)] [arXiv:hep-th/9407087]. N. Seiberg and E. Witten, “Monopoles, duality and chiral symmetry breaking in N=2 supersymmetric QCD,” Nucl. Phys. B 431, 484 (1994) [arXiv:hep-th/9408099].
* [10] F. Cachazo, M. R. Douglas, N. Seiberg and E. Witten, “Chiral rings and anomalies in supersymmetric gauge theory,” JHEP 0212, 071 (2002) [arXiv:hep-th/0211170]. F. Cachazo, N. Seiberg and E. Witten, “Phases of N = 1 supersymmetric gauge theories and matrices,” JHEP 0302, 042 (2003) [arXiv:hep-th/0301006].
* [11] F. Cachazo, N. Seiberg and E. Witten, “Chiral rings and phases of supersymmetric gauge theories,” JHEP 0304, 018 (2003) [arXiv:hep-th/0303207].
* [12] P. C. Argyres and M. R. Douglas, “New phenomena in SU(3) supersymmetric gauge theory,” Nucl. Phys. B 448, 93 (1995) [arXiv:hep-th/9505062].
* [13] P. C. Argyres, M. Ronen Plesser, N. Seiberg and E. Witten, “New N=2 Superconformal Field Theories in Four Dimensions,” Nucl. Phys. B 461, 71 (1996) [arXiv:hep-th/9511154].
* [14] T. Eguchi, K. Hori, K. Ito and S. K. Yang, “Study of $N=2$ Superconformal Field Theories in $4$ Dimensions,” Nucl. Phys. B 471, 430 (1996) [arXiv:hep-th/9603002].
* [15] A. Gorsky, A. I. Vainshtein and A. Yung, “Deconfinement at the Argyres-Douglas point in SU(2) gauge theory with broken N = 2 supersymmetry,” Nucl. Phys. B 584, 197 (2000) [arXiv:hep-th/0004087].
* [16] M. R. Douglas and S. H. Shenker, “Dynamics of SU(N) supersymmetric gauge theory,” Nucl. Phys. B 447, 271 (1995) [arXiv:hep-th/9503163].
* [17] N. Seiberg, “Electric - magnetic duality in supersymmetric nonAbelian gauge theories,” Nucl. Phys. B 435, 129 (1995) [arXiv:hep-th/9411149]. N. Seiberg, “Exact Results On The Space Of Vacua Of Four-Dimensional Susy Gauge Theories,” Phys. Rev. D 49, 6857 (1994) [arXiv:hep-th/9402044].
* [18] R. Auzzi, S. Bolognesi, J. Evslin, K. Konishi and A. Yung, “Nonabelian superconductors: Vortices and confinement in N = 2 SQCD,” Nucl. Phys. B 673, 187 (2003) [arXiv:hep-th/0307287]. S. Bolognesi and K. Konishi, “Non-Abelian magnetic monopoles and dynamics of confinement,” Nucl. Phys. B 645, 337 (2002) [arXiv:hep-th/0207161]. G. Carlino, K. Konishi and H. Murayama, “Dynamical symmetry breaking in supersymmetric SU(n(c)) and USp(2n(c)) gauge theories,” Nucl. Phys. B 590, 37 (2000) [arXiv:hep-th/0005076].
* [19] E. Witten, “Solutions of four-dimensional field theories via M-theory,” Nucl. Phys. B 500, 3 (1997) [arXiv:hep-th/9703166]. E. Witten, “Branes and the dynamics of QCD,” Nucl. Phys. B 507, 658 (1997) [arXiv:hep-th/9706109].
* [20] A. Giveon and D. Kutasov, “Brane dynamics and gauge theory,” Rev. Mod. Phys. 71, 983 (1999) [arXiv:hep-th/9802067].
* [21] J. de Boer and S. de Haro, “The off-shell M5-brane and non-perturbative gauge theory,” Nucl. Phys. B 696, 174 (2004) [arXiv:hep-th/0403035]. J. de Boer and Y. Oz, “Monopole condensation and confining phase of N = 1 gauge theories via M-theory fivebrane,” Nucl. Phys. B 511, 155 (1998) [arXiv:hep-th/9708044].
| arxiv-papers | 2008-07-16T19:27:56 | 2024-09-04T02:48:56.809623 | {
"license": "Public Domain",
"authors": "Stefano Bolognesi",
"submitter": "Stefano Bolognesi",
"url": "https://arxiv.org/abs/0807.2456"
} |
0807.2466 | 11institutetext: T-7, Center for Non-Linear Studies
Los Alamos National Laboratory
Los Alamos, New Mexico 87545
marko@lanl.gov, vadasg@lanl.gov
22institutetext: Center for Embedded Networked Sensing
University of California at Los Angeles
Los Angeles, California 90095
apepe@ucla.edu
# A Grateful Dead Analysis: The Relationship Between Concert and Listening
Behavior
Marko A. Rodriguez1 Vadas Gintautas1 Alberto Pepe2
###### Abstract
The Grateful Dead were an American band that was born out of the San
Francisco, California psychedelic movement of the 1960s. The band played music
together from 1965 to 1995 and is well known for concert performances
containing extended improvisations and long and unique set lists. This article
presents a comparative analysis between 1,590 of the Grateful Dead’s concert
set lists from 1972 to 1995 and 2,616,990 last.fm Grateful Dead listening
events from August 2005 to October 2007. While there is a strong correlation
between how songs were played in concert and how they are listened to by
last.fm members, the outlying songs in this trend identify interesting aspects
of the band and their fans 10 years after the band’s dissolution.
## 1 Introduction
The Grateful Dead were an American band which, despite relatively little
popular radio airtime, enjoyed a cult-like following from a fan base that
numbered in the millions trip:mcnally2002 . The Grateful Dead originated in
San Francisco, California in the early 1960s and toured the world playing
concerts until the untimely death of the foreman and lead guitarist Jerry
Garcia in 1995. The primary source of revenue and exposure for the band came
through their concert tours. They played over 37,000 songs live, in some 2,300
concerts over their 30 years as a band setlists:1996 . Throughout their years
together, the Grateful Dead accumulated a large repertoire that included over
450 unique songs setlists:1996 . The Grateful Dead’s success and continuity
across multiple generations of music listeners is perhaps due in part to their
fundamentally eclectic nature. The band utilized many song writers, composers
and singers, and this resulted in a broad diversity in sound. Robert Hunter
and John Barlow were the primary lyricists for scores written by Jerry Garcia
and Bob Weir, respectively annotatedead:dobb2007 . While Jerry Garcia and Bob
Weir were the primary singers as well, other singers included Ron McKernan,
Brent Mydland, and Phil Lesh. Moreover, their eclectic nature can be seen in
the large number of graphic icons they used to represent themselves. These
icons include skeletons, roses, dancing bears, terrapins, etc. Perhaps their
most famous and recognizable image is the “Steal Your Face” icon in Figure 1
that was released as the album cover art to the live 1976 Steal Your Face
album stealface:dead1976 .
Figure 1: The Grateful Dead “Steal Your Face” icon.
The history of the Grateful Dead’s album releases (13 studio albums and 77
live albums) further reinforces the band’s emphasis on concerts. More live
albums are released regularly as high quality recordings of good performances
are discovered in the Grateful Dead concert archive. For the band and for the
fans, the performances of the Grateful Dead were all about diversity in the
live music experience. In any given show, the concert set list, the
improvisations, and the mood of the band all varied. In concert, all of these
factors came together to create a unique experience for their fans each and
every time.
Perhaps even more astounding than their prolific concert performances is the
dedication that their fans (known as “deadheads”) had to their music
deadheads:grushkin1983 ; adams:deadhead1998 ; sardiello:deadhead ;
deadhead:pattacini . The typical deadhead was not a passive consumer of
recorded studio albums, but an active concert goer that traveled with the band
from concert to concert, city to city, and country to country. Some 10 years
after the Grateful Dead disbanded, the band’s music is still heavily listened
to as evinced by statistics gathered from the popular online music service
known as last.fm.111last.fm is available at: http://www.last.fm/ The last.fm
“audioscrobbler” plug-in is recommendation software that works with popular
computer music players such as iTunes or Winamp. Whenever a song is played
using, say, iTunes, the plug-in reports this activity to the last.fm server
where it is aggregated. From August 2005 to October 2007, there were over 2.5
million Grateful Dead song usages recorded by last.fm. With 72% of the users
of last.fm under the age of 35222Source: last.fm internal web statistics,
courtesy Anil Bawa-Cavia., the popularity of the Grateful Dead, a generation
of fans later and 10 years after the band’s dissolution, is still very strong.
This article presents an analysis of the Grateful Dead’s concert behavior and
exposes a relationship between the concert song patterns from 1972 to 1995 and
the last.fm listening statistics of the band’s songs from August 2005 to
October 2007. First the available set list data is summarized and presented
with an analysis of the concert behavior of the band. Next the usage data from
last.fm is presented with an analysis of the listening behavior of last.fm
members. Finally, a comparative analysis of the concert and listening behavior
of the Grateful Dead is presented.
## 2 The Grateful Dead Concert Behavior
Concert set lists provide the raw data from which to study the concert
behavior of the Grateful Dead.333Set list data obtained from
http://www.cs.cmu.edu/People/gdead/setlists.html. The data were cleaned to
remedy various typographical alterations (e.g. “Trucking” and “Truckin”’ are
the same song), to fix various spelling errors (e.g. “Warf Rat” and “Wharf
Rat” are the same song), and to fix various abbreviations (e.g. “China Cat”
and “China Cat Sunflower” are the same song). The data gathered include 1,590
set lists for concerts from 1972 to 1995, with 28,904 individual song plays. A
typical, unmodified set list is presented below:
Winterland Arena, San Francisco, CA (12/31/77)
Music Never Stopped
Tennessee Jed
Funiculi Funicula
Me and My Uncle
Loser
Jack Straw
Friend of the Devil
Lazy Lightnin’
Supplication
Sugar Magnolia
Scarlet Begonias
Fire on the Mountain
Truckin’
Wharf Rat
drums
Not Fade Away
Around and Around
One More Saturday Night
Casey Jones
Blank lines divide the set list into 4 components. The first component is the
concert venue and location along with the date that the concert was played.
The second component is the first set list in the sequence in which the songs
were played. For example, “Friend of the Devil” was played immediately after
“Jack Straw” in the above example. The third component is the second set list
of the concert. The second set of the Grateful Dead is known for fewer songs
and extended improvisational sessions. Furthermore, the second set of a
Grateful Dead concert is known for its “blending” of songs in which there
exist fewer pauses between the end of one song and the beginning of the next;
that is, the second set was often a large medley of sorts. Second set medleys
often featured pairs of songs that were almost always played together. For
example, “China Cat Sunflower” almost always preceded “I Know You Rider”, but
never in the opposite order. Also, “China Cat Sunflower” was very rarely
played with a different song following.444A similar pattern exists for other
song pairs such as “Scarlet Begonias” and “Fire on the Mountain”, “Saint of
Circumstance” and “Lost Sailor”, and “Cryptic Envelopment” and “The Other
One”. The fourth and last component, which is usually the shortest, is the
encore set list. The Grateful Dead were known to typically play their concerts
in this 3 set form.
A basic measure, calculated using many concert set lists, is to simply count
the number of times a given song is played over all concerts. This ranked list
of songs is a rudimentary “greatest hits” list of sorts, but also a histogram
of concert plays sheds light on the distribution of these counts. Did most
songs get approximately the same number of concert plays, or did the band play
a small set of favorite songs interspersed with less popular songs to provide
variety? Table 3 shows the raw counts for the 15 most played songs. Note that
of the 1,590 concerts analyzed, 1,386 of those concerts included the “Drums”
improvisational rhythm sequence, which typically appeared in the second set of
most concerts. The Grateful Dead very often used “Drums” in the second set to
bridge two songs that would not otherwise be simple to link as a medley.
Figure 3 presents a histogram denoting the number of songs that were played a
given number of times. In summary, many songs were played only a few times and
few songs were played many times.
song | times played
---|---
“Drums” | 1386
“Playing in the Band” | 651
“Sugar Magnolia” | 494
“Not Fade Away” | 486
“The Other One” | 438
“Jack Straw” | 437
“Trucking” | 427
“Me and My Uncle” | 412
“Looks Like Rain” | 407
“Promised Land” | 407
“I Know You Rider” | 406
“China Cat Sunflower” | 403
“New Minglewood Blues” | 398
“Around and Around” | 395
“Tennessee Jed” | 390
Figure 2: The top 15 Grateful Dead songs played in concert from 1972 to 1995.
Figure 3: A histogram of the number of times a song was played in concert.
Most songs were played only a few times and very few songs were played many
times. (The vertical axis was trimmed from its maximum of 150 in order to
preserve diagram clarity).
## 3 The Grateful Dead Usage Statistics
The online music service last.fm tracks how registered members enjoy music by
what songs they play on Internet radio or through a computer music player
plug-in. The last.fm service maintains a database of the listening behavior of
its registered members. From this database, the last.fm service is able to
recommend songs and artists to its members based on the listening behavior of
similar members. This service is analogous to Amazon.com using historical
purchasing behavior to recommend products to customers. Table 5 lists the top
$15$ Grateful Dead songs listened to by last.fm members. These data were
gathered from August 2005 to October 2007 and include 2,616,990 unique
listening events. Figure 5 presents a histogram of the listening counts of
songs. In summary, similar to the concert behavior of the Grateful Dead, many
songs were listened to a few times and a few songs were listened to many
times.
song | times used
---|---
“Friend of the Devil” | 143988
“Sugar Magnolia” | 124736
“Trucking” | 122877
“Casey Jones” | 102449
“Box of Rain” | 88340
“Uncle John’s Band” | 82431
“Ripple” | 80629
“Touch of Grey” | 71270
“Brokedown Palace” | 54675
“Candyman” | 54344
“Fire on the Mountain” | 48516
“Franklin’s Tower” | 45404
“Scarlet Begonias” | 42137
“Dark Star” | 39953
“China Cat Sunflower” | 36479
Figure 4: The top $15$ downloads of Grateful Dead songs on last.fm from August
2005 to October 2007.
Figure 5: A histogram of times a song was listened to, from last.fm data. (The
vertical axis was trimmed from its maximum of 200 in order to preserve diagram
clarity.)
## 4 The Relationship Between Concert and Usage Behavior
The last.fm music service records listening behavior of songs that have
appeared in some published form, such as studio albums, singles, and
recordings of live concerts. The Grateful Dead deviated from the standard
model of releasing studio albums, in that their primary revenue stream was
through concert performances, even from the start. The Grateful Dead produced
13 studio albums and 77 live albums. The band’s first live album, Live/Dead,
was released in 1969 livedead:dead1969 . With respect to the influence of live
recordings on the present generation of listeners, nearly all live albums were
a direct reflection of a particular live concert performance and as such,
respected that concert set list’s song sequence. Thus, if a last.fm listener
were to listen to any one of the many live albums, he or she is, in fact,
replaying concert history and contributing proportionately to the number of
songs listened to as times played in concert. This notion is further
accentuated by fans that created digital renditions of their favorite concert
tapes.555Prior to the advent of the Internet and the easy distribution of
digital audio files, Grateful Dead concert tape trading was an extremely
popular way of disseminating the Grateful Dead’s live experience. In fact, the
band encouraged this tapers:dwork1998 . Many tapes have now been digitized and
shared online by fans. Given that there are currently 5 times more live albums
than studio albums, one may expect that last.fm users would primarily listen
to recordings of concerts and that the usage data would be directly correlated
with that of the set list data. However, as shown in this section, there are
significant deviations from a perfect correlation. Reasons for these
deviations may include the fact that listeners can replay only their favorite
songs from live albums, may prefer studio albums to recordings of live shows,
and are able to make compilations of tracks (“playlists”) that differ from the
live and studio productions.
Figure 6 plots each song in a two-dimensional space. Each song is provided a
coordinate in this space, where the horizontal coordinate is the number of
times the song was played in concert and the vertical coordinate is the number
of times the song was listened to by last.fm members.
Figure 6: Grateful Dead concert plays vs. last.fm usage. Unfortunately, not
all song names could be displayed due to severe overlapping of the labels. In
order to remedy this situation, in the more dense regions, song labels were
randomly removed.
Given that each song has two associated values (horizontal and vertical
coordinates), it is possible to form two vectors of these numbers and to
measure the correlation of these vectors to determine how strongly their
values are related. The analysis reveals that the listening behavior of
last.fm members is strongly correlated with the concert behavior of the
Grateful Dead (a measured correlation of $0.763$ where $0$ is uncorrelated and
$1$ is perfectly correlated).666A Spearman $\rho$ rank-order correlation
reveals a correlation of $\rho=0.763$ with a $p$-value $<2.2\times 10^{-16}$,
where $\rho=-1$ is inversely related, $\rho=0$ is unrelated, and $\rho=1$ is
correlated. The $p$-value indicates the probability of such a correlation
occurring randomly; the $p$ value for this correlation indicates that, at
random, this correlation would occur 5 times out of $10^{15}$
parame:sheskin2004 . Since the data are not perfectly correlated, it is
valuable to examine songs which are exceptions to this trend.
The solid curved line in Figure 6 shows regions for which songs display a
strong correlation.777Specifically, the line represents a linear regression
model that predicts usage in terms of concert plays. The data were fit to a
line; the line appears curved because the values on both axes are plotted on
logarithmic (base-$10$) scale to preserve diagram clarity. The two dashed
lines represent the $40^{\text{th}}$-quantile and $75^{\text{th}}$-quantile of
concert plays. These quantiles were subjectively chosen in order to provide a
discussion of songs that are noteworthy outliers from the trend. Finally, the
two closed polygons labeled A and B represent the interesting outlying songs.
Polygon A encapsulates songs that were heavily listened to by last.fm members,
but played few times in concert by the Grateful Dead. On the other hand, B
encapsulates songs that were heavily played by the Grateful Dead, but less
listened to (and possibly overlooked) by last.fm members.888A random sample of
songs were removed from this diagram to preserve clarity as to ensure that
song labels did not overlap. Furthermore, while A and B occupy regions outside
of their quantile delimitations, this is to ensure that song labels were
encapsulated.
Polygon A encapsulates those songs that were heavily listened to by last.fm
users but not heavily played by the Grateful Dead in concert. Many of the
songs in polygon A are old classics that did not persist due to either changes
in the band or to the band’s extremely prolific period of songwriting in the
early 1970s. For example, although the song “Mister Charlie” was played for
the last time in concert on May 26, 1972 at the Strand Lyceum in London,
England, it was featured on the Europe ‘72 album europe:dead1972 . This album
reached a peak spot of 24 on the Billboard pop albums chart in 1973. Thus,
while not being played much in concert during the band’s lifetime, “Mister
Charlie” remains a fan treasure by virtue of securing a place on a much
celebrated album. Similar arguments can be made for “Sage and Spirit” and
“Blues for Allah” which, while technically difficult and therefore generally
avoided in concert, are songs that were released on the Blues For Allah LP
allah:dead1975 which reached spot 12 on the Billboard charts for pop album in
1975. “Good Morning Little School Girl” was primarily sung by Ron McKernan,
whose untimely death in 1973 caused the song to be removed from the Grateful
Dead concert play list except for a few special appearances in the late 80s
and early 90s.999It is worth noting that “Mister Charlie” was also primarily
sung by Ron McKernan and thus, didn’t last with the Grateful Dead past his
lifetime. However, “Good Morning Little School Girl” is the opening track of
the Two from the Vault album twovalut:dead1992 that contains a live recording
of the August 24, 1968 Shrine Exhibit Hall performance of the Grateful Dead.
Although Jerry Garcia was the primary singer of “Death Don’t Have No Mercy”,
the song had a similar concert fate as “Good Morning Little School Girl” and
was also released on the Two from the Vault live album. Two from the Vault
reached position 119 in 1992 on the Billboard charts.101010“The Eleven” and
“Cryptic Envelopment” are also on Two from the Vault. Likewise, “Hard to
Handle”, like “Good Morning Little School Girl” was a Ron McKernan specialty
that was dropped after his death, except for being played during the New
Year’s Eve show of 1982 in Oakland, California. Finally, “Cryptic Envelopment”
provided a medley prelude to the popular “The Other One” and only later in the
bands life was “The Other One” separated from “Cryptic Envelopment” and
preceded by “Drums”. It is worth noting that 28 of the 77 live albums of the
Grateful Dead include “The Other One” while only 7 include “Cryptic
Envelopment”.
Polygon B encapsulates those songs that were heavily played in concert
throughout the Grateful Dead’s career, but for various reasons, were less
frequently listened to by last.fm members. “Drums” is perhaps the most salient
of this collection of songs at the extreme of the boundary. “Drums” is an all-
drum improvisational piece that usually appeared in the second set of a
Grateful Dead concert. Usually appearing with “Drums” was the full sonic
spectrum improvisation of “Space” which included all band members. Both
“Drums” and “Space” found a stable home on the Infrared Roses album
infrared:dead1991 , but unfortunately, due to the esoteric nature of these
improvisations, Infrared Roses has been less well received by the general
public and thus received no popular awards and did not make it on any music
charts. Furthermore, to compound the situation, the Grateful Dead provided
unique names for Infrared Roses tracks and thus, when played by last.fm users,
are not associated with the typical “Drums” and “Space” songs of the concert
set lists. It is interesting to note the songs “Saint of Circumstance,” “When
I Paint My Masterpiece,” “Victim or the Crime,” “Lost Sailor,” and “Greatest
Story” in the bottom left of polygon B. All of these songs were created by the
song writing duo of Barlow and Weir and sung in concert often by Bob Weir.
While these songs were played extensively in concert, they received relatively
little attention from last.fm users.
Finally, the extreme upper right of this plot is important as “Trucking” and
“Sugar Magnolia” represent not only the most popular songs in terms of times
played in concert, but in terms of times listened to on last.fm. “Trucking” is
on 25 of the 90 released Grateful Dead albums and “Sugar Magnolia” is on 32 of
those albums. Both “Trucking” and “Sugar Magnolia” were also well received
publicly. “Trucking” reached position 64 in 1971 and “Sugar Magnolia” reached
position 91 in 1973 on the Billboard pop singles charts. Also in this area is
“Touch of Grey”. “Touch of Grey” was the only Grateful Dead song with an
accompanying music video and in 1987, reached the top 10 Billboard single’s
chart. By comparison to produced greatest hits albums, Table 2 lists the songs
that were released on the 2003 Very Best of the Grateful Dead compilation. Of
these songs, 12 out of the 17 songs are in the top right quadrant (these songs
are marked with an ∗ in Table 2), meaning that they were both played and
listened to heavily by the Grateful Dead and their fans,
respectively.111111“Eyes of the World” and “Estimated Prophet” are not
displayed as they were randomly removed to preserve diagram clarity. A similar
situation exists with the Skeletons from the Closet greatest hits album for
which 8 out of 11 songs are in the top right quadrant skelcloset:dead1974
(Table 2 presents the songs on the album).121212“Mexicali Blues” is not
displayed as it was randomly removed to preserve diagram clarity. Of
particular importance is “Box of Rain” (on Very Best of the Grateful Dead) by
Phil Lesh and Robert Hunter. This song, written after the death of Phil Lesh’s
father, is not only the last song ever played by the Grateful Dead in
concert131313The final Grateful Dead performance took place at Soldier Field
on July 9, 1995 in Chicago. Jerry Garcia died exactly one month later on
August 9, 1995., but also unique in that it is one of the few songs for which
Phil Lesh was the primary singer. In summary, the upper right hand quadrant of
this diagram is ripe for creating compilation and greatest hits albums as it
reflects both what the band as well as present day fans appreciate.
# | Track name
---|---
1 | “Trucking” ∗
2 | “Touch of Grey” ∗
3 | “Sugar Magnolia” ∗
4 | “Casey Jones” ∗
5 | “Uncle John’s Band” ∗
6 | “Friend of the Devil” ∗
7 | “Franklin’s Tower” ∗
8 | “Estimated Prophet” ∗
9 | “Eyes of the World” ∗
10 | “Box of Rain” ∗
11 | “U.S. Blues”
12 | “The Golden Road to Ultimate Devotion”
13 | “One More Saturday Night” ∗
14 | “Fire on the Mountain” ∗
15 | “The Music Never Stopped”
16 | “Hell in a Bucket”
17 | “Ripple”
Table 1: The tracks of the Very Best of the Grateful Dead greatest hits album
hits:dead2003 . Upper right quadrant songs in Figure 6 are marked with ∗.
# | Track name
---|---
1 | “The Golden Road to Ultimate Devotion”
2 | “Trucking” ∗
3 | “Rosemary”
4 | “Sugar Magnolia” ∗
5 | “St. Stephen”
6 | “Uncle John’s Band” ∗
7 | “Casey Jones” ∗
8 | “Mexicali Blues” ∗
9 | “Turn on Your Love Light” ∗
10 | “One More Saturday Night” ∗
11 | “Friend of the Devil” ∗
Table 2: The tracks of the Skeletons in the Closet greatest hits album
skelcloset:dead1974 . Upper right quadrant songs in Figure 6 are marked with
∗.
## 5 Conclusion
The Grateful Dead were an American music phenomenon that influenced multiple
generations of music lovers. For 30 years, the Grateful Dead made a career out
of an unrelenting tour schedule that took them around the world, and, unlike
typical bands, took their fans with them. It is now 43 years since the band
started and with the use of online music providers and services, it is
possible to track the listening behavior of the many Grateful Dead fans in the
world today. This article presented an analysis comparing the popularity of
Grateful Dead songs as identified by both how many times they were played in
concert and how many times they were listened to by members of the last.fm
online music service. The correlation between concert plays and fan listens is
strong, but not perfect. Those songs that existed as outliers to a perfect
correlation were analyzed to understand what made these songs deviate from the
model. These deviations can be understood by changes in the band, by live
performance album releases, and by the very nature of the songs themselves.
There is much to be learned about American concert tour culture and the bands
that bring this culture to fruition concert:black2007 . Perhaps more than any
other band, there exists large amounts of Grateful Dead data that go beyond
set lists to include lyrics, chord progressions, concert reviews, and history.
Many books have been published about the Grateful Dead and albums continue to
be released 13 years after their final concert in 1995. Without a doubt, the
Grateful Dead have made a profound impact on that which is American rock
music.
## Acknowledgements
The authors would like to thank Anil Bawa-Cavia of last.fm for providing the
Grateful Dead raw usage data, Jerry Stratton for providing the set list data,
and the Grateful Dead for doing what they do best.
## References
* (1) Rebecca G. Adams. Inciting sociological thought by studying the deadhead community: Engaging publics in dialogue. Social Forces, 77(1):1–25, 1998.
* (2) Grant C. Black, Mark A. Fox, and Paul Kochanowski. Concert tour success in North America: An examination of the top 100 tours from 1997 to 2005. Popular Music and Society, 30(2):149–172, 2007.
* (3) Grateful Dead. Two from the Vault. Music Album, May 1992.
* (4) David Dobb. The Complete Annotated Grateful Dead Lyrics. Free Press, 2007.
* (5) John R. Dwork and Michael Getz, editors. The Deadhead’s Taping Compendium: An In-Depth Guide to the Music of the Grateful Dead on Tape. Holt, New York, NY, 1998.
* (6) Grateful Dead. Live/Dead. Music Album, November 1969.
* (7) Grateful Dead. Europe ‘72. Music Album, November 1972.
* (8) Grateful Dead. Skeletons from the Closet: The Best of Grateful Dead. Music Album, February 1974.
* (9) Grateful Dead. Blues for Allah. Music Album, September 1975.
* (10) Grateful Dead. Steal Your Face. Music Album, June 1976.
* (11) Grateful Dead. Infrared Roses. Music Album, November 1991.
* (12) Grateful Dead. The Very Best of the Grateful Dead. Music Album, September 2003.
* (13) Paul Grushkin, Cynthia Bassett, and Jonas Grushkin. Grateful Dead: The Official Book of the Dead Heads. Harper Paperbacks, 1983.
* (14) Madhu Lundquist. The SetList Program: Grateful Dead Setlists and Listener Experiences. Webpage, 1996.
* (15) Dennis McNally. A Long Strange Trip: The Inside History of the Grateful Dead. Broadway Books, 2002.
* (16) Melissa McCray Pattacini. Deadheads yesterday and today: An audience study. Popular Music and Society, 24, 2000.
* (17) Robert Sardiello. Youth Culture: Identity in a Postmodern World, chapter Identity and Status Stratification in Deadhead Subculture. Blackwell Publishers, 1998.
* (18) David J. Sheskin. Parametric and nonparametric statistical procedures. Chapman and HallCRC, New York, NY, 2004.
| arxiv-papers | 2008-07-15T21:31:36 | 2024-09-04T02:48:56.819410 | {
"license": "Public Domain",
"authors": "Marko A. Rodriguez and Vadas Gintautas and Alberto Pepe",
"submitter": "Marko A. Rodriguez",
"url": "https://arxiv.org/abs/0807.2466"
} |
0807.2506 | # Nonlinear molecular excitations in a completely inhomogeneous DNA chain
M. Daniel daniel@cnld.bdu.ac.in V. Vasumathi Corresponding Author.
Telephone:+91-431-2407057, Fax:+91-431-2407093 Centre for Nonlinear Dynamics,
School of Physics, Bharathidasan University, Tiruchirappalli - 620 024, India.
###### Abstract
We study the nonlinear dynamics of a completely inhomogeneous DNA chain which
is governed by a perturbed sine-Gordon equation. A multiple scale perturbation
analysis provides perturbed kink-antikink solitons to represent open state
configuration with small fluctuation. The perturbation due to inhomogeneities
changes the velocity of the soliton. However, the width of the soliton remains
constant.
###### keywords:
DNA , Soliton , Multiple Scale Perturbation
###### PACS:
87.15.He, 66.90.+r, 63.20.Ry
,
## 1 Introduction
A number of theoretical models have been proposed in the recent times to study
the nonlinear dynamics of Deoxyribonucleic acid (DNA) molecule to understand
the conservation and transformation of genetic information (see for e.g [1,
2]). These models are based on longitudinal, transverse, rotational and
stretching motions of bases. Among these different possible motions,
rotational motion of bases is found to contribute more towards the opening of
base pairs and to the nonlinear dynamics of DNA. The first contribution
towards nonlinear dynamics of DNA was made by Englander and his co-workers [3]
and they studied the base pair opening in DNA by taking into account the
rotational motion. Yomosa [4, 5] proposed a plane base rotator model by taking
into account the rotational motion of bases in a plane normal to the helical
axis, and Takeno and Homma generalized the same [6, 7, 8]. Later using this
model, several authors found solitons to govern the fluctuation of DNA double
helix between an open state and its equilibrium states [9, 10, 11, 12, 13,
14]. Peyrard and Bishop [15, 16] and Christiansen and his collegues [17]
studied the process of base pair opening by taking into account the transverse
and longitudinal motions of bases in DNA. Very recently, there have been
extensions of the radial model of Bishop and Peyrard [18, 19], composite
models for DNA torsion dynamics [20] and models interplaying between radial
and torsional dynamics [21, 22, 23, 24]. In all the above studies, homogeneous
strands and hydrogen bonds have been considered for the analysis.
However, in nature the presence of different sites along the strands such as
promotor, coding, terminator, etc., each of which has a specific sequence of
bases is related to a particular function and thus making the strands site-
dependent or inhomogeneous [25, 26]. Also, the presence of abasic sites leads
to inhomogeneity in stacking [27]. In this context, in a recent paper the
present authors [28] studied the nonlinear molecular excitations in DNA with
site-dependent stacking energy along the strands based on the plane base
rotator model. The nonlinear dynamics of DNA in this case was found to be
governed by a perturbed sine-Gordon (s-G) equation. The perturbed kink and
antikink soliton solutions of the perturbed s-G equation represented an open
state configuration of base pairs with small fluctuation. The perturbation in
this case introduces small fluctuations in the localized region of the soliton
retaining the overall shape of the soliton. However, the width of the soliton
remains constant and the velocity changes for different inhomogeneities. The
results indicate that the presence of inhomogeneity in stacking changes the
number of base pairs that participate in the open state configuration and
modifies the speed with which the open state configuration travels along the
double helical chain. In reality, the presence of site-dependent strands in
DNA changes the nature of hydrogen bonds between adjacent base pairs and the
presence of abasic sites leads to absence of hydrogen bonds. Thus, when the
strands are site-dependent in stacking, naturally the hydrogen bonds that
connect the bases between the strands are also site-dependent. Hence, it has
become necessary to consider inhomogeneity in hydrogen bonds also in the study
of nonlinear dynamics of DNA. In the present paper, we study the dynamics of
DNA with inhomogeneity both in stacking and in hydrogen bonds using the plane
base rotator model. The paper is organized as follows. In section 2, we
present the Hamiltonian of our model and derive the associated dynamical
equation for the inhomogeneous DNA. The effect of stacking and hydrogen bond
inhomogeneity on base-pair opening is studied by solving the dynamical
equations using a multiple scale soliton perturbation theory in section 3. The
results are concluded in section 4.
## 2 Hamiltonian and the dynamical equation
We consider the B-form of a DNA double helix with site-dependent strands as
well as base-pair sequence and study the nonlinear molecular excitations by
considering a plane-base rotator model. In Fig. (1a) we have presented a
sketch of the DNA double helix. Here, $S$ and $S^{\prime}$ represent the two
complementary strands in the DNA double helix and each arrow represents the
direction of the base attached to the strand and the dots between arrows
represent the net hydrogen bonding effect between the complementary bases. The
z-axis is chosen along the helical axis of the DNA. Fig. (1b) represents a
horizontal projection of the $n^{th}$ base pair in the xy-plane. In this
figure $Q_{n}$ and $Q^{\prime}_{n}$ denote the tips of the $n^{th}$ bases
belonging to the strands $S$ and $S^{\prime}$. $P_{n}$ and $P^{\prime}_{n}$
represent the points where the bases in the $n^{th}$ base pair are attached to
the strands $S$ and $S^{\prime}$ respectively.
As we are looking for opening of base pairs in DNA which is related to
important DNA functions such as replication and transcription, we consider the
rotational motion of bases (due to its importance among other motions) in a
plane normal to the helical axis (z-direction) represented by the angles
$\phi_{n}$ and $\phi^{\prime}_{n}$ at the $n^{th}$ site of the base pair. The
stacking and hydrogen bonding energies are the major components of the energy
in a DNA double helix. In the case of a homogeneous DNA system, Yomosa [4, 5]
expressed the Hamiltonian involving these energies in terms of the rotational
angles $\phi_{n}$ and $\phi^{\prime}_{n}$ under the plane base rotator model
which was later modified by the present authors [28] in the case of site-
dependent stacking. When both the stacking and hydrogen bonds are site-
dependent, the Hamiltonian for our plane base rotator model of DNA double
helix is written in terms of the rotational angles as
$\displaystyle H$ $\displaystyle=$
$\displaystyle\sum_{n}\left[\frac{I}{2}({\dot{\phi}_{n}}^{2}+{\dot{\phi}_{n}}^{{}^{\prime}2})+Jf_{n}\left[2-\cos(\phi_{n+1}-\phi_{n})-\cos(\phi^{\prime}_{n+1}-\phi^{\prime}_{n})\right]\right.$
(1) $\displaystyle\left.-\eta
g_{n}\left[1-\cos(\phi_{n}-\phi^{\prime}_{n})\right]\right].$
The first two terms in the Hamiltonian (1) represent the kinetic energies of
the rotational motion of the $n^{th}$ nucleotide bases with $I$ their moments
of inertia and the remaining terms represent the potential energy due to
stacking and hydrogen bonds. While $J$ and $\eta$ represent a measure of
stacking and hydrogen bonding energies respectively, $f_{n}$ and $g_{n}$
indicate the site-dependent(inhomogeneous) character of stacking and hydrogen
bonds respectively.
The Hamilton’s equations of motion corresponding to Hamiltonian (1) is written
as
$\displaystyle I\frac{\partial^{2}\phi_{n}}{\partial t^{2}}$ $\displaystyle=$
$\displaystyle
J\left[f_{n}\sin(\phi_{n+1}-\phi_{n})-f_{n-1}\sin(\phi_{n}-\phi_{n-1})\right]$
(2a) $\displaystyle+\eta g_{n}\sin(\phi_{n}-\phi^{\prime}_{n}),$
$\displaystyle I\frac{\partial^{2}\phi^{\prime}_{n}}{\partial t^{2}}$
$\displaystyle=$ $\displaystyle
J\left[f_{n}\sin(\phi^{\prime}_{n+1}-\phi^{\prime}_{n})-f_{n-1}\sin(\phi^{\prime}_{n}-\phi^{\prime}_{n-1})\right]$
(2b) $\displaystyle+\eta g_{n}\sin(\phi^{\prime}_{n}-\phi_{n}).$
It is expected that the difference in the angular rotation of neighbouring
bases along the two strands in the case of B-form of DNA double helix is small
[6, 7]. Therefore under small angle approximation, in the continuum limit Eqs.
(2a) and (2b) are written as
$\displaystyle\phi_{\hat{t}\hat{t}}$ $\displaystyle=$ $\displaystyle
f(z)\phi_{zz}+f_{z}\phi_{z}-\frac{1}{2}g(z)\sin(\phi-\phi^{\prime}),$ (3a)
$\displaystyle\phi^{\prime}_{\hat{t}\hat{t}}$ $\displaystyle=$ $\displaystyle
f(z)\phi^{\prime}_{zz}+f_{z}\phi^{\prime}_{z}-\frac{1}{2}g(z)\sin(\phi^{\prime}-\phi).$
(3b)
While writing Eqs. (3a) and (3b) we have chosen $\eta=-\frac{Ja^{2}}{2}$ and
rescaled the time variable as $\hat{t}=\sqrt{\frac{Ja^{2}}{I}}t$. Now, adding
and subtracting Eqs.(3a) and (3b) and by choosing $\phi=-\phi^{\prime}$, we
obtain
$\displaystyle\Psi_{\hat{t}\hat{t}}-\Psi_{zz}+\sin\Psi=\epsilon\left[A\left(s(z)\Psi_{z}\right)_{z}-B~{}h(z)\sin\Psi\right],$
(4)
where $\Psi=2\phi$. Further, as the contribution due to inhomogeneity is small
compared to homogeneous stacking and hydrogen bonding energies while writing
Eq.(4), the inhomogeneity in stacking and hydrogen bonding energies are
expressed in terms of a small parameter $\epsilon$ as
$f(z)=1+\epsilon~{}As(z)$ and $g(z)=1+\epsilon~{}Bh(z)$, where $A$ and $B$ are
arbitrary constants. When $\epsilon=0,$ Eq. (4) reduces to the completely
integrable sine-Gordon(s-G) equation which admits N-soliton solutions in the
form of kink and antikink [29]. Hence, we call Eq. (4) as a perturbed sine-
Gordon equation. For instance, the one soliton solution of the integrable s-G
equation obtained through Inverse Scattering Transform (IST) method is written
as,
$\displaystyle\Psi(z,\hat{t})=4arctan{exp[\pm m_{0}(z-v_{0}\hat{t})]},\quad
m_{0}=\frac{1}{\sqrt{1-v_{0}^{2}}}.$ (5)
Here $v_{0}$ and $m^{-1}_{0}$ are real parameters that represent the velocity
and width of the soliton respectively. In Eq.(5) while the upper sign
corresponds to the kink soliton, the lower sign represents the antikink
soliton. In Figs. (2a) and (2b) the one soliton kink-antikink solutions
(Eq.(5)) are plotted by choosing $v_{0}=0.4$. The kink-antikink soliton
solution of the integrable sine-Gordon equation describes an open state
configuration in the DNA double helix. In Figure (2c), we present a sketch of
how the base pairs open locally in the form of kink-antikink structure in each
strand and propagate along the direction of the helical axis.
The base pair opening will help in the process of replication that duplicates
DNA and in transcription which helps to synthesize messenger RNA. However,
when $\epsilon\neq 0$, the inhomogeneity in stacking and hydrogen bonds may
affect the base-pair opening through a perturbation on the kink-antikink
solitons of the s-G equation. Therefore in the next section, we solve the
perturbed s-G equation (4) using a multiple-scale soliton perturbation theory
[30, 31, 32] (as has been carried out in the case corresponding to $B=0$ [28])
to understand the effect of stacking and hydrogen bond inhomogeneities on the
base pair opening.
## 3 Effect of stacking and hydrogen bonding inhomogeneities on the open
state configuration
### 3.1 Multiple-scale soliton perturbation theory
In order to study the effect of inhomogeneity in stacking and hydrogen bonds
on the base pair opening in the form of kink-antikink soliton by treating them
as a perturbation, the time variable $\hat{t}$ is transformed into several
variables as $t_{n}=\epsilon^{n}\hat{t}$ where $n=0,1,2,...$ and $\epsilon$ is
a very small parameter [30, 31, 32]. In view of this, the time derivative and
$\Psi$ in Eq. (4) are replaced by the expansions
$\frac{\partial}{\partial\hat{t}}=\frac{\partial}{\partial
t_{0}}+\epsilon~{}\frac{\partial}{\partial
t_{1}}+\epsilon^{2}\frac{\partial}{\partial t_{2}}+...$ and
$\Psi=\Psi^{(0)}+\epsilon\Psi^{(1)}+\epsilon^{2}\Psi^{(2)}+....$ We then
equate the coefficients of different powers of $\epsilon$ and obtain the
following equations. At $O(\epsilon^{(0)})$ we obtain the integrable sine-
Gordon equation
$\displaystyle\Psi^{(0)}_{t_{0}t_{0}}-\Psi^{(0)}_{zz}+sin\Psi^{(0)}=0,$ (6)
for which the one soliton solution takes the form as given in Eq.(5) with
$\hat{t}$ replaced by $t_{0}$. Due to perturbation, the soliton parameters
namely $m$ and $\xi(\xi=v\hat{t})$ are now treated as functions of the slow
time variables $t_{0},t_{1},t_{2},....$ However, $m$ is treated as independent
of $t_{0}$. The equation at $O(\epsilon^{(1)})$ takes the form
$\displaystyle{\Psi^{(1)}_{\tau\zeta}}-{\Psi_{\zeta\zeta}^{(1)}}+(1-2{sech^{2}}\zeta)\Psi^{(1)}=F^{(1)}(\zeta,\tau),$
(7) $\displaystyle
where\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad$
$\displaystyle
F^{(1)}(\zeta,t_{0})=2\left[As(\zeta)sech\zeta\right]_{\zeta}-2aBh(\zeta)\tanh\zeta\mbox{sech}\zeta\qquad\qquad$
$\displaystyle+4v_{0}sech\zeta\left[m_{t_{1}}+(m^{2}\xi_{t_{1}}-\zeta
m_{t_{1}})\tanh\zeta\right].$ (8)
While writing the above equation we have used the transformation
$\hat{\zeta}=m(z-vt_{0})$ and $\hat{t}_{0}=t_{0}$ to represent everything in a
co-ordinate frame moving with the soliton. Then, we have used another set of
transformations given by
$\tau=\frac{\hat{t}_{0}}{2m}-\frac{(1+v)\hat{\zeta}}{2}$ and
$\zeta=\hat{\zeta}$ for our later convenience. We have also replaced
$sin\Psi^{(0)}$ by $2a\tanh\zeta$ $sech\zeta$, where $a=\pm 1$. The solution
of Eq. (7) is searched by assuming $\Psi^{(1)}(\zeta,\tau)=X(\zeta)T(\tau)$
and $F^{(1)}(\zeta,\tau)=X_{\zeta}(\zeta)H(\tau).$ Substituting the above in
Eq. (7) and simplifying we obtain
$\displaystyle X_{\zeta\zeta}+(2sech^{2}\zeta-1)X=\lambda_{0}X_{\zeta},$ (9)
$\displaystyle T_{\tau}-\lambda_{0}T=H(\tau),$ (10)
where $\lambda_{0}$ is a constant. Thus, the problem of constructing the
perturbed soliton at this moment turns out to be solving Eqs. (9) and (10) by
constructing the eigenfunctions and finding the eigen values [28, 32].
It may be noted that Eq.(9) differs from the normal eigen value problem, with
$X_{\zeta}$ in the right hand side instead of $X$. Hence, before actually
solving the eigen value equation (9), we first consider it in a more general
form given by
$\displaystyle L_{1}X=\lambda\tilde{X},\qquad
L_{1}=\partial_{\zeta\zeta}+2sech^{2}\zeta-1,$ (11)
where $\lambda$ is the eigen value. To find the adjoint eigen function to $X$,
we consider another eigen value problem
$\displaystyle L_{2}\tilde{X}=\lambda X,$ (12)
where $L_{2}$ is to be determined. Combining the two eigen value problems we
get
$\displaystyle
L_{2}L_{1}X=\lambda^{2}X,~{}~{}L_{1}L_{2}\tilde{X}=\lambda^{2}\tilde{X}.$ (13)
From the above equations we conclude that the operator $L_{1}L_{2}$ is the
adjoint of $L_{2}L_{1}$ and also $X$ and $\tilde{X}$ are expected to be
adjoint eigen functions. Hence, we can find the eigenfunction by solving
Eqs.(11) and (12). However, eventhough $L_{1}$ is known as given in Eq.(11),
the operator $L_{2}$ is still unknown. So, by experience we choose
$L_{2}=\partial_{\zeta\zeta}+6sech^{2}\zeta-1$, and solve the eigen value
equations by choosing the eigen functions as
$\displaystyle
X(\zeta,k)=p(\zeta,k)e^{ik\zeta},~{}~{}\tilde{X}(\zeta,k)=q(\zeta,k)e^{ik\zeta},$
(14)
where $k$ is the propagation constant. Further, as the operator $L_{2}L_{1}$
is self-adjoint, non-negative and satisfying a regular eigenvalue problem, the
sine-Gordon soliton is expected to be stable. On substituting Eqs. (14) in
Eqs.(11) and (12) in the asymptotic limit, we obtain the eigen value as
$\lambda=-(1+k^{2})$. Now, in order to find the eigen functions we expand
$p(\zeta,k)$ and $q(\zeta,k)$ [30, 32] as
$\displaystyle p(\zeta,k)$ $\displaystyle=$ $\displaystyle
p_{0}+p_{1}\frac{\sinh\zeta}{\cosh\zeta}+p_{2}\frac{1}{\cosh^{2}\zeta}+p_{3}\frac{\sinh\zeta}{\cosh^{3}\zeta}+p_{4}\frac{1}{\cosh^{4}\zeta}+...,$
(15a) $\displaystyle q(\zeta,k)$ $\displaystyle=$ $\displaystyle
q_{0}+q_{1}\frac{\sinh\zeta}{\cosh\zeta}+q_{2}\frac{1}{\cosh^{2}\zeta}+q_{3}\frac{\sinh\zeta}{\cosh^{3}\zeta}+q_{4}\frac{1}{\cosh^{4}\zeta}+...,$
(15b)
where $p_{j}$ and $q_{j}$, j=0,1,2,… are functions of $k$ to be determined. On
substituting Eqs. (14), (15a) and (15b) in Eqs. (11) and (12) and collecting
the coefficients of $1,~{}\frac{\sinh\zeta}{\cosh\zeta}$,
$\frac{1}{\cosh^{2}\zeta}$,… we get a set of simultaneous equations. On
solving those equations we obtain the following eigen functions [28, 32].
$\displaystyle X(\zeta,k)$ $\displaystyle=$
$\displaystyle\frac{(1-k^{2}-2ik\tanh\zeta)}{\sqrt{2\pi}(1+k^{2})}e^{ik\zeta},$
(16) $\displaystyle\tilde{X}(\zeta,k)$ $\displaystyle=$
$\displaystyle\frac{(1-k^{2}-2ik\tanh\zeta-2sech^{2}\zeta)}{\sqrt{2\pi}(1+k^{2})}e^{ik\zeta}.$
(17)
The higher order coefficients $p_{3},p_{4},...$ and $q_{3},q_{4},...$ vanish.
It may be noted that Eq.(10) is a linear inhomogeneous differential equation
and can be solved using known procedures. The solution reads
$\displaystyle
T(\tau,k)=\frac{1}{i\lambda_{0}k(1+k^{2})}\int_{-\infty}^{\infty}d\zeta^{\prime}F^{(1)}(\zeta^{\prime},\tau){X}^{\ast}(\zeta^{\prime},k)(e^{\lambda_{0}[\tau-\frac{(1+v)}{2}\zeta^{\prime}]}-1),$
(18)
where $\lambda_{0}=\frac{i(1+k^{2})}{k}$. The first order correction to the
soliton can be computed using the following expression.
$\displaystyle\Psi^{(1)}(\zeta,\tau)=\int_{-\infty}^{\infty}X(\zeta,k)T(\tau,k)dk+\sum_{j=0,1}X_{j}(\zeta)T_{j}(\tau).$
(19)
Here the continuous eigenfunctions $X(\zeta,k)$ and $T(\tau,k)$ are already
known as given in Eqs. (16) and (18). However, the discrete eigen states
$X_{0},X_{1}$ and $T_{0},T_{1}$ are unknown. The two discrete eigenstates
$X_{0},X_{1}$ corresponding to the discrete eigen value $\lambda=0$ can be
found using the completeness of the continuous eigenfunctions as
$\displaystyle X_{0}(\zeta)$ $\displaystyle=$ $\displaystyle
sech\zeta,~{}X_{1}(\zeta)=\zeta sech\zeta.$ (20)
In order to find $T_{0}$ and $T_{1}$, we substitute Eq.(19) in Eq.(7) and
multiply by $X_{0}(\zeta)$ and $X_{1}(\zeta)$ separately and use the
orthonormal relations to get
$\displaystyle{T_{1}}_{\tau}(\tau)$ $\displaystyle=$
$\displaystyle\int_{-\infty}^{\infty}F^{(1)}(\zeta,\tau)X_{0}(\zeta)d\zeta,$
(21a) $\displaystyle{T_{0}}_{\tau}(\tau)-2T_{1}(\tau)$ $\displaystyle=$
$\displaystyle-\int_{-\infty}^{\infty}F^{(1)}(\zeta,\tau)X_{1}(\zeta)d\zeta.\qquad$
(21b)
As $F^{(1)}(\zeta,\tau)$ given in Eq.(8) does not contain the time variable
$\tau$ explicitly, the right hand side of Eqs. (21a) and (21b) also should be
independent of time, and hence we write the nonsecularity conditions as
$\displaystyle\int_{-\infty}^{\infty}F^{(1)}(\zeta,\tau)X_{0}(\zeta)d\zeta=0,$
(22a)
$\displaystyle\int_{-\infty}^{\infty}F^{(1)}(\zeta,\tau)X_{1}(\zeta)d\zeta=0.$
(22b)
It may be indicated that, the nonsecularity conditions give way for the
stability of soliton [33] under perturbation through an understanding of the
evolution of soliton parameters such as width and velocity. Substituting the
above equations in Eqs. (21a) and (21b), we choose $T_{1}(\tau)=0$ and find
$T_{0}(\tau)=C$ which has to be determined. For this, we substitute
$T_{1}(\tau)=0$ in Eq. (21b) and integrate to obtain
$\displaystyle T_{0}(\tau)$ $\displaystyle=$
$\displaystyle\frac{(1+v)}{2}\int_{-\infty}^{\infty}d\zeta~{}\zeta
F^{(1)}(\zeta,\tau)X_{1}(\zeta).$ (23)
In order to find the first order correction to soliton $\Psi^{(1)}$ we need to
evaluate the eigen states $T(\tau,k)$ and $T_{0}(\tau)$ explicitly for which
we have to find the explicit form of $F^{(1)}(\zeta,\tau)$ which contains
unknown quantities like $m_{t_{1}},~{}\xi_{t_{1}},~{}s(\zeta)$ and $h(\zeta)$.
Therefore, we substitute Eqs. (8) and (20) in the nonsecularity conditions
given in Eqs. (22a) and (22b) and evaluate the integrals to obtain
$\displaystyle m_{t_{1}}$ $\displaystyle=$
$\displaystyle-\frac{1}{2v_{0}}\int_{-\infty}^{\infty}\left(A\left[{s(\zeta)sech\zeta}\right]_{\zeta}-aBh(\zeta)sech\zeta\tanh\zeta\right)sech\zeta
d\zeta,$ (24a) $\displaystyle\xi_{t_{1}}$ $\displaystyle=$
$\displaystyle-\frac{1}{2m^{2}v_{0}}\int_{-\infty}^{\infty}\left(A\left[{s(\zeta)sech\zeta}\right]_{\zeta}-aBh(\zeta)sech\zeta\tanh\zeta\right)\zeta
sech\zeta d\zeta.$ (24b)
Here $m_{t_{1}}$ and $\xi_{t_{1}}$ represent the time variation of the inverse
of the width and velocity of the soliton respectively.
### 3.2 Variation of soliton parameters
In order to find the variation of the width and velocity of the soliton while
propagating along the inhomogeneous DNA chain, we assume that the soliton with
a width of $m_{0}^{-1}$ is travelling with a speed of $v_{0}(\xi_{t_{0}})$
when the perturbation is switched on. In other words, the problem now boils
down to understand the propagation of the kink-antikink soliton in an
inhomogeneous DNA chain and to measure the change of the soliton parameters
due to inhomogeneity. To find the variation of the soliton parameters
explicitly and to construct the perturbed soliton solution we have to evaluate
the integrals found in the right hand sides of Eqs. (24a) and (24b) which can
be carried out only on supplying specific forms of $s(\zeta)$ and $h(\zeta)$.
Hence, we consider inhomogeneity in the form of localized and periodic
functions separately and further assume that the inhomogeneity in the stacking
and in the hydrogen bonds are equal. The localized inhomogeneity represents
the intercalation of a compound between neighbouring base pairs or the
presence of defect or the presence of abasic site in the DNA double helical
chain. The periodic nature of inhomogeneity may represent a periodic
repetition of similar base pairs along the helical chain.
#### 3.2.1 Localized inhomogeneity
To understand the effect of localized inhomogeneity in stacking and hydrogen
bonds on the width and velocity of the soliton during propagation, we
substitute $s(\zeta)=h(\zeta)=sech\zeta$ in Eqs.(24a) and (24b). On evaluating
the integrals, we obtain
$m_{t_{1}}=0,~{}\xi_{t_{1}}=\frac{\pi}{12m^{2}v_{0}}(2A+aB)$ which can be
written in terms of the original time variable $\hat{t}$ by using the
expansions $m_{\hat{t}}=m_{t_{0}}+\epsilon m_{t_{1}}$ and
$\xi_{\hat{t}}=\xi_{t_{0}}+\epsilon\xi_{t_{1}}$ as
$\displaystyle m=m_{0},\quad\xi_{\hat{t}}\equiv
v=v_{0}+\frac{\epsilon\pi(2A+aB)}{12m^{2}v_{0}},$ (25)
where $1/m_{0}$ is the width and $v_{0}$ is the velocity of the soliton in the
absence of inhomogeneity. The first of Eq.(25) says that when the
inhomogeneities in stacking and in hydrogen bonds are in the form
$s(\zeta)=h(\zeta)=sech\zeta$, the width of the soliton remains constant,
thereby showing that the number of base pairs participating in the opening
process remain constant during propagation. However, from the second of
Eq.(25), we find that the velocity of the soliton gets a correction. The
nature of correction in the velocity depends on the value of ‘$a$’ which takes
$+1$ or $-1$ and also on the nature of $A$ and $B$ which can be either
positive or negative. When $A$ and $B$ are greater than zero, the
inhomogeneity will correspond to an energetic barrier and on the other hand
when $A$ and $B$ are less than zero, the inhomogeneity will correspond to $a$
potential well. First, we consider the case corresponding to $a=+1$. In this
case when $(2A+B)>0$, the velocity of the soliton gets a positive correction
and it may propagate along the chain without formation of a bound state. On
the other hand, when $(2A+B)<0$, the inhomogeneities slow down the soliton.
Ofcourse, when $(2A+B)=0$, that is when the inhomogeneities in the stacking
and in the hydrogen bonding suitably balance each other, the velocity of the
soliton remains unaltered. Finally, the soliton stops when the original
velocity satisfies the condition
$v_{0}^{2}=\frac{1}{1-\frac{12}{\epsilon\pi(2A+B)}}$. In all the above cases,
the stability of the soliton is guaranteed. A similar argument can be made in
the case when $a=-1$ with $(2A+B)$ replaced by $(2A-B)$. It may also be noted
that similar results have been obtained in the case of resonant kink-impurity
interaction and kink scattering in a perturbed sine-Gordon model by Zhang Fei
et al [34], say that the kink will pass the impurity and escape to the
positive infinity when the initial velocity of kink soliton is larger than the
critical value. In a similar context Yakushevich et al [35] while studying the
interaction between soliton and the point defect in DNA chain showed
numerically, that the solitons are stable. At this point it is also worth
mentioning that Dandoloff and Saxena [36] realized that in the case of an XY-
coupled spin chain model which is identifiable with our plane-base rotator
model of DNA, the ansatz $sech\zeta$ energetically favours the deformation of
spin chain.
#### 3.2.2 Periodic inhomogeneity
We then choose the periodic inhomogeneity in the form
$s(\zeta)=h(\zeta)=\cos\zeta$ and substitute the same in Eqs. (24a) and (24b).
On evaluating the integrals, we get $m_{t_{1}}=0,~{}\xi_{t_{1}}=\frac{\pi[\pi
A+a(4-\pi)B]}{16m^{2}v_{0}}$ which can be written in terms of the original
time variable $\hat{t}$ as
$\displaystyle m=m_{0},\quad\xi_{\hat{t}}\equiv v=v_{0}+\frac{\epsilon\pi[\pi
A+a(4-\pi)B]}{16m^{2}v_{0}}.$ (26)
From Eq.(26), we observe that the width of the soliton remains constant and
the velocity gets a correction. Here also, one can project an argument similar
to the case of localized inhomogeneity. The only difference between the two
cases is the quantum of correction added to soliton velocity. Normally in
DNAs, inhomogeneity in hydrogen bonds is expected to be dominant and therefore
one would expect that $B>A$. One can verify that it is possible to obtain the
above condition from the velocity corrections in Eqs. (25) and (26) by writing
$\frac{\epsilon\pi[\pi
A+(4-\pi)B]}{16m^{2}v_{0}}>\frac{\epsilon\pi(2A+B)}{12m^{2}v_{0}}$ when
$a=+1$. The above condition indicates that the correction in velocity in the
case of periodic inhomogeneity is larger than that in the case of localized
inhomogeneity. This is because in this case, the inhomogeneity occurs
periodically in the entire length of the DNA chain. In a recent paper,
Yakushevich et al [35] studied numerically the dynamics of topological
solitons describing open states in an inhomogeneous DNA and investigated
interaction of soliton with the inhomogeneity and the results, have very close
analogy with the results of our perturbation analysis. It was shown that the
soliton can easily propagate along the DNA chain without forming a bound state
thus showing that soliton moving with sufficiently large velocity along the
DNA chain is stable with respect to defect or inhomogeneity.
### 3.3 First order perturbed soliton
Having understood the variation of the width and velocity of the soliton in a
slow time scale due to perturbation, we now construct the first order
perturbed soliton by substituting the values of the basis functions
$\\{X\\}\equiv\\{X(\zeta,k),X_{0}(\zeta),X_{1}(\zeta)\\}$ and
$\\{T\\}\equiv\\{T(\tau,k),T_{0}(\tau),T_{1}(\tau)\\}$ given in Eqs. (16),
(20), (18), (23) with $T_{1}(\tau)=0$ in Eq.(19), we get
$\displaystyle\Psi^{(1)}(\zeta,t_{0})$ $\displaystyle=$
$\displaystyle-\frac{1}{\pi}\int_{-\infty}^{\infty}dk\frac{1}{(1+k^{2})^{3}}(1-k^{2}-2ik\tanh\zeta)e^{ik\zeta}$
(27)
$\displaystyle\times\int_{-\infty}^{\infty}d\zeta^{\prime}\left(\left[As(\zeta^{\prime})sech\zeta^{\prime}\right]_{\zeta^{\prime}}-aBh(\zeta^{\prime})sech\zeta^{\prime}\tanh\zeta^{\prime}\right.+2v_{0}~{}sech\zeta^{\prime}$
$\displaystyle\times\left.\left[m_{t_{1}}+(m^{2}\xi_{t_{1}}-\zeta^{\prime}m_{t_{1}})\tanh\zeta^{\prime}\right]\right)(1-k^{2}+2ik\tanh\zeta^{\prime})e^{-ik\zeta^{\prime}}$
$\displaystyle\times\left(e^{\lambda_{0}[\tau+\frac{(1+v_{0})}{2}\zeta^{\prime}]}-1\right)+(1+v)sech\zeta\int_{-\infty}^{\infty}d\zeta^{\prime}\left(\left[As(\zeta^{\prime})sech\zeta^{\prime}\right]_{\zeta^{\prime}}\right.$
$\displaystyle-
aBh(\zeta^{\prime})sech\zeta^{\prime}\tanh\zeta^{\prime}+2v_{0}~{}sech\zeta^{\prime}\left[m_{t_{1}}\right.$
$\displaystyle\left.+(m^{2}\xi_{t_{1}}-\zeta^{\prime}\left.m_{t_{1}})\tanh\zeta^{\prime}\right]\right)\zeta^{\prime
2}sech\zeta^{\prime}.$
While writing the above, we have also used
$\tau=\frac{1}{2m}[t_{0}-m(1+v)\zeta]$. It may be noted that majority of the
poles that lie within the contour in Eq. (27), are purely imaginary giving
rise to exponentially localized residues and hence do not give rise to any
radiation thus will lead to stable form of soliton [37].
#### 3.3.1 Localized inhomogeneity
Now, we explicitly construct the first order perturbation correction to the
one soliton in the case of the localized inhomogeneity (i.e) when
$s(\zeta)=h(\zeta)=sech\zeta$, by substituting the corresponding values of
$F^{(1)},m_{t_{1}}$ and $\xi_{t_{1}}$ in Eq. (27). The resultant integrals are
then evaluated using standard residue theorem [38] which involves very lengthy
algebra and at the end we obtain
$\displaystyle\Psi^{(1)}(\zeta,t_{0})$ $\displaystyle\approx$
$\displaystyle(2A+aB)\left[\frac{40}{27\sqrt{2}}\sqrt{sech\zeta}~{}e^{-\frac{3\zeta}{2}}+\frac{32}{27\sqrt[3]{2}}\tanh\zeta\sqrt[3]{sech\zeta}~{}e^{-\frac{5}{3}\zeta}\right.$
(28)
$\displaystyle+\frac{\pi}{12v^{2}}\left.[2vt_{0}+m^{2}(v^{2}-1)]sech\zeta\right].$
Finally, the perturbed one soliton solution, that is
$\Psi(z,t_{0})=\Psi^{(0)}(z,t_{0})+\Psi^{(1)}(z,t_{0})$ (choosing
$\epsilon=1$) is written in terms of the original variables as
$\displaystyle\Psi(z,t_{0})$ $\displaystyle\approx$ $\displaystyle
4arc\tan\exp[\pm
m_{0}(z-v_{0}t_{0})]+(2A+aB)\left[\frac{40}{27\sqrt{2}}e^{\mp\frac{3(m(z-vt_{0})}{2}}\right.$
(29) $\displaystyle\times\sqrt{sech[\pm
m(z-vt_{0})]}+\frac{32}{27\sqrt[3]{2}}\tanh[\pm m(z-vt_{0})]$
$\displaystyle\times\sqrt[3]{sech[\pm
m(z-vt_{0})]}e^{\mp\frac{5}{3}m(z-vt_{0})}+\frac{\pi}{12mv^{2}}$
$\displaystyle\left.\times\left[m(v^{2}-1)+2t_{0}v\right]sech[\pm
m(z-vt_{0})]\right].$
Knowing $\Psi$, the angle of rotation of bases $\phi(z,t_{0})$ can be
immediately written down by using the relation $\phi=\frac{\Psi}{2}$. In Figs.
3(a,b) we have plotted $\phi(z,t_{0})$, the rotation of bases under the
perturbation $g(z)=h(z)=sechz$ by choosing $a=A=B=1$ and $v_{0}=0.4$. From the
figures we observe that there appears fluctuation in the form of a train of
pulses closely resembling the shape of the inhomogeneity profile in the width
of the soliton as time progresses. But, there is no change in the topological
character and no fluctuations appear in the asymptotic region of the soliton.
In Figs. (4a) and (4b), we have plotted the perturbed kink and antikink
solitons respectively corresponding to the case when the hydrogen bond
inhomogeneity is absent (B=0) for comparison. From Figs. (3) and (4), we
observe that the fluctuation in the form of a train of pulses appear in both
the cases. Eventhough the perturbed solitons in both the cases appear
qualitatively the same, the inhomogeneity in hydrogen bonds adds more
fluctuation in the width of the soliton.
#### 3.3.2 Periodic inhomogeneity
We then repeat the procedure for constructing the perturbed one soliton
solution $\Psi(\zeta,t_{0})$ in the case of periodic inhomogeneity by choosing
$s(\zeta)=h(\zeta)=\cos\zeta$, and obtain the perturbed soliton solution as
$\displaystyle\Psi(z,t_{0})$ $\displaystyle\approx$ $\displaystyle
4~{}arctan\exp[\pm m_{0}(z-v_{0}t_{0})]+[A\pi+aB(4-\pi)][b+dt_{0}]$ (30)
$\displaystyle sech[m_{0}(z-vt_{0})],$
where $b=\frac{\pi^{2}}{16v^{2}}(v^{2}-1)$ and $d=\frac{\pi^{2}}{8m_{0}v}$. In
Figs. 5(a,b), we plot the angle of rotation of bases
$\phi$($=\frac{\Psi}{2}$), using the perturbed soliton given in Eq.(30) for
the same parametric choices as in the case of localized inhomogeneity. We
observe from the figures that, the fluctuation appears in the width of the
soliton without any change in its topological character asymptotically. In
order to compare the results with that of the case when the hydrogen bonding
inhomogeneity is absent as also found in [28](see Figs. (6a,b)), we plot the
perturbed soliton found in (30) when $B=0$. As in the previous case, here also
we observe that fluctuation appears in the width of the soliton in both the
cases and the inhomogeneity in hydrogen bonds introduces more fluctuation.
## 4 Conclusions
In this paper, we studied the nonlinear dynamics of a completely inhomogeneous
(inhomogeneity in both stacking and hydrogen bonding) DNA double helix by
considering the dynamical plane-base rotator model. The dynamics of this model
in the continuum limit gives rise to a perturbed sine-Gordon equation, which
was derived from the Hamiltonian consisting of site-dependent stacking and
hydrogen bonding energies. In the unperturbed limit, the dynamics is governed
by the kink-antikink soliton of the integrable sine-Gordon equation which
represents the opening of base pairs in a homogeneous DNA. In order to
understand the effect of inhomogeneity in stacking and hydrogen bonds on the
base pair opening, we carried out a perturbation analysis using multiple-scale
soliton perturbation theory. The perturbation not only modifies the shape of
the soliton but also introduces change in the velocity of the soliton. From
the results, we observe that when the inhomogeneity is in a localized or
periodic form, the width of the soliton remains constant. However, the
velocity of the soliton increases, decreases or remains uniform and even the
soliton stops, depending on the values of inhomogeneity represented by $A$ and
$B$. The soliton in all the cases are found to be stable. From the results of
the perturbed soliton we observe that the inhomogeneity in stacking and
hydrogen bonds in both the cases (localized and periodic forms) introduce
fluctuation in the form of pulses in the width of the soliton. However, there
is no change in the topological character of the soliton in the asymptotic
region(see Figs. 3 and 5). The fluctuation is more when both the
inhomogeneities (site-dependent stacking and hydrogen bonds) are present,
whereas it is less in the case of homogeneous hydrogen bonds and site-
dependent stacking. Hence, we conclude that, the addition of inhomogeneity in
hydrogen bonds does not introduce big changes in the soliton parameters and
shape except a correction in the velocity of the soliton and fluctuation. The
above dynamical behaviour may act as energetic activators of the RNA-
polymerase transport process during transcription in DNA. In the case of short
DNA chains the discreteness effect assumes importance and hence we will
analyse the discrete dynamical equations (2) and the results will be published
elsewhere.
## 5 Acknowledgments
The work of M. D forms part of a major DST project. V. V thanks SBI for
financial support.
## References
* [1] L. V. Yakushevich, Physica D 79 (1994) 77.
* [2] L. V. Yakushevich, Nonlinear Physics of DNA ( Wiley-VCH, Berlin, 2004).
* [3] S. W. Englander, N. R. Kallenbanch, A. J. Heeger, J. A. Krumhansl and S. Litwin, Proc. Natl. Acad. Sci. U.S.A 77 (1980) 7222.
* [4] S. Yomosa, Phys. Rev. A 27 (1983) 2120.
* [5] S. Yomosa, Phys. Rev. A 30 (1984) 474.
* [6] S. Takeno and S. Homma, Prog. Theor. Phys. 70 (1983) 308.
* [7] S. Takeno and S. Homma, Prog. Theor. Phys. 72 (1984) 679.
* [8] S. Takeno, Phys. Letts. A 339 (2005) 352.
* [9] P. Jensen, M. V. Jaric and K. H. Bannenmann, Phys. Letts. A 95 (1983) 204.
* [10] A. Khan, D. Bhaumik and B. Dutta-Roy, Bull. Math. Biol. 47 (1985) 783.
* [11] V. K. Fedyanin and V. Lisy, Studia Biophys. 116 (1986) 65.
* [12] R. V. Polozov and L. V. Yakushevich, J. Theor. Biol. 130 (1988) 423.
* [13] J. A. Gonzalez and M. M. Landrove, Phys. Letts. A 292 (2002) 256.
* [14] L. V. Yakushevich, Nanobiology 1 (1992) 343 .
* [15] M. Peyrard and A. R. Bishop, Phys. Rev. Lett. 62 (1989) 2755.
* [16] G. Kalosakas, K. Q. Rasmussen and A. R. Bishop, Synthetic Metals, 141 (2004) 93.
* [17] P. L. Christiansen, P. C. Lomdahl and V. Muto, Nonlinearity 4 (1991) 477.
* [18] T. Dauxios , Phys. Lett. A, 159 (1991) 390.
* [19] C. B. Tabi, A. Mohamdou and T. C. Kofane, Phys. Scr. 77 (2008) 045002.
* [20] M. Cadoni and R. De Leo, Phys . Rev. E, 021919 (2007).
* [21] M. Barbi, S. Cocco and M. Peyrard, Phys. Letts. A, 253 (1999) 358.
* [22] M. Barbi, S. Cocco , M. Peyrard and N. Theodorakopoulos, Phys. Rev. E, 68 (2003) 061909.
* [23] S. Cocco and R. Manasson, Phys. Rev. Lett. 83 (1999) 5178.
* [24] A. Campa, Phys. Rev. E , 63 (2001) 021901.
* [25] J. Ladik and J. Cizek, Int. J. Quantum Chem. 26 (1984) 955.
* [26] E. Cubero, E. C. Sherer, F. J. Luque, M. Orozco and C. A. Laughton, J. Am. Chem. Soc. 121 (1999) 8653 .
* [27] M. Hisakado and M. Wadati, J. Phys. Soc. Jpn. 64 (1995) 1098.
* [28] M. Daniel and V. Vasumathi, Physica D 231 (2007) 10.
* [29] M. J. Ablowitz, D. J. Kaup, A. C. Newell, and H. Segur, Stud. Appl. Math. 53 (1974) 249.
* [30] J. Yan and Y. Tang, Phys. Rev. E 54 (1996) 6816.
* [31] Y. Tang and W. Wang, Phys. Rev. E 62 (2000) 8842.
* [32] J. Yan, Y. Tang, G. Zhou and Z. Chen, Phys. Rev. E 58 (1998) 1064.
* [33] Y. Matsuno, J. Math. Phys. 41 (2000) 7061.
* [34] F. Zhang, Y. S. Kivshar and L. Vazquez, Phys. Rev. A 45 (1992) 6019.
* [35] L. V. Yakushevich, A. V. Savin and L. I. Manevitch, Phys. Rev. E 66 (2002) 016614.
* [36] R. Dandoloff and A. Saxena, J. Phys.: Condens. Matter 9 (1997) L667.
* [37] A. Sanchez, A. R. Bishop and F. Dominguez-Adame, Phys. Rev. E 49 (1994) 4603.
* [38] E. Kreyszig, Advanced Engineering Mathematics (John-Wiley, New York, 2002).
Fig.1. (a) A schematic structure of the B-form DNA double helix. (b) A
horizontal projection of the $n^{th}$ base pair in the xy-plane.
Fig.2. (a) Kink and (b) antikink soliton solutions of the sine-Gordon equation
(Eq.(4) when $\epsilon=0$ ) with $v_{0}=0.4$. (c) A sketch of the formation of
open state configuration in terms of kink-antikink soliton in DNA double
helix.
Fig.3.(a) The perturbed kink-soliton and (b) the perturbed antikink-soliton
for the inhomogeneity $g(z)=h(z)=sechz$ with $A=B=1$ and $v_{0}=0.4$.
Fig.4. (a) The perturbed kink-soliton and (b) the perturbed antikink-soliton
for the inhomogeneity $s(z)=sechz,h(z)=0$ with $A=1$ and $v_{0}=0.4$.
Fig.5.(a) The perturbed kink-soliton and (b) the perturbed antikink-soliton
for the inhomogeneity $g(z)=h(z)=\cos z$ with $A=B=1$ and $v_{0}=0.4$.
Fig.6.(a) The perturbed kink-soliton and (b) the perturbed antikink-soliton
for the inhomogeneity $g(z)=\cos z,h(z)=0$ with $A=1$ and $v_{0}=0.4$.
Figure 1: (a) A schematic structure of the B-form DNA double helix. (b) A
horizontal projection of the $n^{th}$ base pair in the xy-plane. Figure 2: (a)
Kink and (b) antikink soliton solutions of the sine-Gordon equation (Eq.(4)
when $\epsilon=0$ ) with $v_{0}=0.4$. (c) A sketch of the formation of open
state configuration in terms of kink-antikink soliton in DNA double helix.
Figure 3: (a) The perturbed kink-soliton and (b) the perturbed antikink-
soliton for the inhomogeneity $g(z)=h(z)=sechz$ with $A=B=1$ and $v_{0}=0.4$.
Figure 4: (a) The perturbed kink-soliton and (b) the perturbed antikink-
soliton for the inhomogeneity $s(z)=sechz,h(z)=0$ with $A=1$ and $v_{0}=0.4$.
Figure 5: (a) The perturbed kink-soliton and (b) the perturbed antikink-
soliton for the inhomogeneity $g(z)=h(z)=\cos z$ with $A=B=1$ and $v_{0}=0.4$.
Figure 6: (a) The perturbed kink-soliton and (b) the perturbed antikink-
soliton for the inhomogeneity $g(z)=\cos z,h(z)=0$ with $A=1$ and $v_{0}=0.4$.
| arxiv-papers | 2008-07-16T07:22:43 | 2024-09-04T02:48:56.825681 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "M. Daniel and V. Vasumathi",
"submitter": "Vasumathi Velachi",
"url": "https://arxiv.org/abs/0807.2506"
} |
0807.2509 | # Challenge on the Astrophysical R-process Calculation with Nuclear Mass
Models 111Partly supported by Major State Basic Research Developing Program
2007CB815000, the National Natural Science Foundation of China under Grant
Nos. 10435010, 10775004 and 10221003.
SUN Baohua1222010-62767013(O), 13810098634, Meng Jie 1,2,3,4
333mengj@pku.edu.cn, 010-62765620(O) 1School of Physics and SK Lab. of Nucl.
Phys. & Tech., Peking University, Beijing 100871, China
2 Department of Physics, University of Stellenbosch, Stellenbosch, South
Africa
3Institute of Theoretical Physics, Chinese Academy of Sciences, Beijing, China
4Center of Theoretical Nuclear Physics, National Laboratory of Heavy Ion
Accelerator, 730000 Lanzhou, China
###### Abstract
Our understanding of the rapid neutron capture nucleosynthesis process in
universe depends on the reliability of nuclear mass predictions. Initiated by
the newly developed mass table in the relativistic mean field theory (RMF), in
this paper the influence of mass models on the $r$-process calculations is
investigated assuming the same astrophysical conditions. The different model
predictions on the so far unreachable nuclei lead to significant deviations in
the calculated $r$-process abundances.
###### pacs:
21.10.-k, 21.10.Dr, 21.60.-n, 21.60.Jz, 23.40.-s, 26.30.Hj
The rapid neutron capture ($r$-) process is introduced more than 50 years ago
to explain the solar abundances not creating from the slow-neutron capture
($s$-) process and the proton-capture ($p$-) process BBFH . It is responsible
for the synthesis of about half of the nuclei beyond the iron group. The
recent observations Cowan-nature06 of extremely metal-poor ([Fe/H] $\approx$
-3) $r$-process enriched ([Ba/Eu] $<$ 0) stars, show a consistent elemental
abundances from Ba to the third $r$-process peak with the scaled solar system
$r$-process abundance distribution. This consistency may indicate that the
$r$-process abundance patterns are most possibly produced only by a single or
at most a few $r$-process events for the heavier elements with $Z\geq 56$,
i.e., there may be only one or a few $r$-process sites in the early Galaxy.
However, the exact astrophysical site where the $r$-process proceeds has not
been unambiguously identified, despite decades of work. This research is
complicated by the required knowledge of both the astrophysical environments
and the nuclear properties of very neutron-rich nuclei. Previous
phenomenological studies indicate that the $r$-process occurs at temperatures
around T $\sim 10^{9}$ K and at extreme neutron fluxes with neutron number
densities $n_{n}>10^{20}$ cm-3 Cowan-PR91 ; Kartz-APJ93 ; SMG07 . Moreover,
the $r$-process should be a dynamical process with changing conditions and
paths Kartz-APJ93 .
On the other hand, the determined astrophysical condition for the $r$-process
site relies on the extrapolation of theoretical nuclear models for the “terra
incognita”. Among the required nuclear properties, the key one is the nuclear
mass, from which one can directly determine the one-neutron separation energy,
shell gap and also the beta-decay energy. As discussed in Ref. SMG07 , though
various nuclear mass models agree quite well with the known data, they
disagree among each other towards the very neutron-rich side, where the
$r$-process runs along. As a result, the required astrophysical condition for
the $r$-process nucleosynthesis can vary for different model predictions SMG07
.
In order to investigate the impact of nuclear mass models on the $r$-process
nucleosynthesis, one should distinguish the astrophysical uncertainty from the
nuclear physics uncertainty. In this letter, adopting the newly constructed
mass table RMFBCS in the relativistic mean field (RMF) theory, and assuming
the same astrophysical conditions, the impact of different mass models on the
$r$-process calculation will be investigated. Other mass models used include
the finite-range droplet model (FRDM) FRDM , extended Thomas-Fermi plus
Strutinsky integral with quenched shell (ETFSI-Q) ETFSIQ and the recent
Hartree-Fock-Bogolyubov (HFB-13) model HFB-13 .
The RMF approach has made lots of successes in describing the nuclear
properties far away from the $\beta$-stability line as reviewed, for example
in Ref. Meng06 . The first systematic calculation of the nuclear ground state
properties including nuclear masses, radii and deformations has been done
recently for all the nuclei lying between the proton drip line and the neutron
drip line RMFBCS . A good agreement with the available data is found. The
detailed analysis shows that the predictions of nuclear masses in the RMF are
generally underestimated with respect to the experimental data. However,
considering the factor that with less than 10 parameters obtained from fitting
several doubly magic nuclei, the RMF Lagrangian achieves almost the same
prediction power ($\sim$ 0.65 MeV) for one neutron separation energy $S_{n}$
AME03 as those highly parameterized mass models FRDM ; ETFSIQ ; HFB-13 , thus
it is very interesting to see to what extent the solar $r$-abundances can be
reproduced using this new table.
We adopt a site-independent $r$-process calculation as in Ref. SMG07 , where a
configuration of 16 $r$-process components is chosen as a reasonable
approximation to the real $r$-process site. The seed-nuclei iron are
irradiated by neutron sources of high and continuous neutron densities $n_{n}$
ranging from $10^{20}$ to $10^{28}$ cm-3 over a timescale $\tau$ at a high
temperature ($T\sim 10^{9}$ K). The neutron captures proceed in (n,
$\gamma$)$\leftrightarrow$($\gamma$, n) equilibrium, and the abundance flow
from one isotopic chain to the next is governed by $\beta$-decays. Roughly,
the $r$-process runs along the contour lines between 2 and 5 MeV of one
neutron separation energies as illustrated in Fig. 1. Similar to the classical
$s$-process kappeler89 , these different $r$-process components are assumed to
satisfy a simply exponential formula, i.e.,
$\omega(n_{n})=n_{n}^{a},\tau(n_{n})=b\times n_{n}^{c}$, where $\omega(n_{n})$
and $\tau(n_{n})$ are the corresponding weighting factor and neutron
irradiation time for the component with a neutron density $n_{n}$. The
parameters a, b and c are determined from the least-square fit to the solar
$r$-process abundances.
The best simulation using the new RMF mass table is presented in Fig. 1. In
this simulation, the $\beta$-decay properties are taken from the recent
calculation Moller03 . Furthermore, the available experimental results AME03 ;
NNDC have been included. The corresponding $r$-process path is indicated by
the dark grey squares. The $r$-process abundance distribution, in general, can
be well reproduced. However, the predicted large shell towards the neutron
drip line in the RMF model leads to a large gap before the second and third
abundance peaks. Previous investigations Kartz-APJ93 ; pfeiffer-ZPA97 showed
that a quenched shell could avoid the jump in the $r$-process path and thus
result in a better simulation to the observation. Nevertheless, it is still
unclear whether there is a shell quenching effect or to what extent it is
towards the neutron drip line, since the present experimental results are
still in debate. One example is the shell gap at the critical waiting point
130Cd. It is firstly suggested to be a quenched shell in Ref. dillmann03 .
However, a recent experiment result rising07 , interpreted by the state-of-
the-art nuclear shell-model calculations, shows no evidence of shell
quenching.
In order to disentangle the nuclear physics uncertainty from the astrophysical
environment uncertainty, we have done the $r$-process calculations using
respectively FRDM FRDM , ETFSI-Q ETFSIQ and HFB-13 HFB-13 mass inputs while
keeping the same $\beta$-decay properties. All the calculations have used the
same astrophysical condition, i.e., the best case obtained using the RMF
masses. Along with the observation kappeler89 , the abundance distributions
around the third peak are compared with the RMF result in Fig. 2. It shows
that the abundances around the neutron shells $N=126$ strongly depends on the
mass models applied. Before the abundance peak, the RMF simulation shows a
broad dip around $A=170$, while the ETFSI-Q and HFB-13 simulations have a dip
towards a large mass number ($\sim$180). Differences also exist after the
abundance peak.
The different deficiencies in reproducing the solar r-process abundances can
be mainly traced back to two aspects. First, different nuclear mass models
predict quite different shell evolution towards the neutron drip line. The
shell gap energy at $N=126$ in the RMF model is about 2 MeV larger than that
in other models. A stronger shell structure will result in more nuclear matter
accumulated in nuclides with a neutron magic number and less in nuclides
around, which is reflected by dips on both sides of the abundance peak in the
calculated r-process abundance. Second, different models assign different
locations of the shape transition before the shell closures. The predicted
r-process paths pass through the shape-transition ranges before going to the
magic numbers $N=82$ and 126. The first oblate nuclides for $Z=60$-63 are in
the mass number of 172-177 for the RMF predictions, while in the HFB-13 model
the corresponding mass numbers are 178-179. One neutron separation energies
for these nuclides at the shape-transition point will deviate from the
approximately linear dependence of the mass number as predicted by a classical
liquid droplet mass model, and eventually affect the r-process path. Together,
both aspects result in a direct jump in the r-process path from 169Pr to 185Pr
(15 mass units) in the RMF simulation (see Fig. 3). Different from the RMF
simulation, abundances calculated in the HFB-13 simulation are also
accumulated in the nuclides with the mass number 170-177, therefore a better
reproduction of the solar r-process abundances at A$\sim$170, however a worse
reproduction at A$\sim$180\. The gap around the $N=126$ shell also exist in
other simulations though varying in magnitude and mass number. In the same
astrophysical environment, the $r$-process path in the RMF simulation runs
about 1-2 mass units towards the neutron drip line than those of other
simulations.
In the present investigation, it is shown that the nature of the $r$-process
is complicated due to the interplay between the nuclear physics and the
astrophysical environment. Since it is still not accessible to measure most of
the nuclei masses along the r-process path in the near future, further
theoretical development aiming at the description of the know and unknown
masses simultaneously is badly needed. If the astrophysical conditions for the
$r$-process are identified precisely, this may serve as a constraint for the
nuclear mass models. Equivalently, if nuclear masses are known in a good
precision, it can be used to constraint the potential site for the $r$-process
as well.
In the present letter, we have shown that the $r$-process calculation is quit
sensitive to the nuclear mass inputs. Precise mass measurement of neutron-rich
nuclides with an accuracy less than 100 keV are needed to decisively determine
the shell evolution at $N=82$ and 126 towards the neutron drip line, as well
as the locations of the shape transition before these shell closures. These
experimental results will offer a primary constraint to the existing mass
models and a strong motivation for further exploration of theoretical mass
models, and furthermore, a better understanding the nature of the r-process
nucleosynthesis.. Meanwhile, as there is no clear evidence to accept or reject
any mass model mentioned above, it is necessary to take different mass models
into account in the $r$-process study.
## References
* (1) Burbidge E M, Burbidge G R, Fowler W A and Hoyle F 1957 Rev. Mod. Phys. 29 547
* (2) Cowan J J and Snden C 2006 Nature 440 1151
* (3) Cowan J J, Thieleman F K and Truran J W 1991 Phys. Rep. 208 267
* (4) Kratz K L, Bitouzet J P, Thielemann F K, Möller P and Pfeiffer B 1993 ASTROPHYS. J. 403 216
* (5) Sun B, Montes F, Geng L S, Gessel H, Litvinov Yu A and Meng J, submitted to Phys. Rev. C
* (6) Geng L S, Toki H and Meng J 2005 Prog. Theor. Phys. 113 785
* (7) Möller P, Nix J R, Myers W D and Swiatecki W J 1995 At. Data Nucl. Data Tables 59 185
* (8) Pearson J M, Nayak R C and Goriely S 1996 Phys. Lett. B 387 455
* (9) Goriely S, Samyn M and Pearson J M 2006 Nucl. Phys. A 773 279
* (10) Meng J, Toki H, Zhou S G, Zhang S Q, Long W H and Geng L S 2006 Prog. Part. Nucl. Phys. 57 470
* (11) Audi G, Wapstra A H and Thibault C 2003 Nucl. Phys. A 729 337
* (12) Käppeler F, Berr H and Wisshak K 1989 Rep. Prog. Phys. 52 945
* (13) Möller P, Nix J R and Kratz K L 1997 At. Data Nucl. Data Tables 66 131; Möller P, Pfeiffer B and Kratz K L 2003 Phys. Rev. C 67 055802
* (14) National Nuclear Data Center, http://www.nndc.bnl.gov
* (15) Pfeiffer B, Kratz K L and Thielemann F K 1997 Z. Phys. A 357 235
* (16) Dillmann I et al 2003 Phys. Rev. Lett. 91 162503
* (17) Jungclaus A et al 2007 Phys. Rev. Lett. 99 132501
Figure 1: Features of the $r$-process calculated using the new RMF mass table.
Black squares denote $\beta$-stable nuclei, and magic proton and neutron
numbers are indicated by pairs of parallel lines. The region in the main graph
shows the calculated average one neutron separation energy ($S_{2n}/2$). The
solid line denotes the border of nuclides with known masses in the neutron-
rich side. The dark grey squares show the $r$-process path when using the RMF
mass predictions and the FRDM half-lives. The observed and calculated solar
$r$-process abundance curves are plotted versus the mass number A in the
inset, whose x-axis is curved slightly to follow the $r$-process path. Figure
2: Comparison of observed solar $r$-process abundances (filled circles) with
theoretical abundance after $\beta$-decays calculated using RMF, FRDM, ETFSI-Q
and HFB-13 mass models. The calculated abundances have been scaled to the
solar $r$-process abundance at $A=130$. Figure 3: The corresponding
$r$-process pathes of Fig. 2 for the RMF and HFB-13 mass models. Shown are
those isotopes with more than 10% population of each isotopic chain. For
comparison the stable nuclei are labeled by black squares.
| arxiv-papers | 2008-07-16T07:42:07 | 2024-09-04T02:48:56.830633 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Sun Baohua and Meng Jie",
"submitter": "Baohua Sun",
"url": "https://arxiv.org/abs/0807.2509"
} |
0807.2521 | ## 0.0.1 Fundamentals and Applications of Isotope Effect in Modern
Technology.
V.G. Plekhanov.
Fonoriton Science Lab., Garon Ltd., P.O. Box 2632, Tallinn, 13802, ESTONIA
$<$e-mail$>$ vgplekhanov@hotmail.com
Different crystals (semiconductors and insulators) with varying isotopic
composition have been recently grown. I discuss here the effect of isotopic
mass and isotopic disorder on the properties (vibrational, elastic, thermal
and optical) of different crystals. The main applications of the stable
isotopes are included self-diffusion, neutron transmutative doping (NTD) of
different semiconductors, optical fibers, isotope-based quantum computers,
etc. Because of space limitations this discussion will not exhaustive. I hope
however, to give sufficient references to published work so that the interest
reader can easily find the primary literature sources to this rapidly
expanding field of solid state physics.
Phonons, excitons, isotope-mixed crystals, laser materials, quantum
information, isotope-based quantum computers.
It is well-known that the presence of randomly distributed impurities in a
crystal can give rise to significant variations of its mechanical, electrical,
thermal, and optical properties with respect to those of the pure solid. All
these properties are, more or less, directly related to the structure of the
manifold of phonon states and any variation induced in this structure by the
presence of the impurities, will produce a corresponding alteration of the
physical properties of the material. Of particular interest is the case in
which the impurity species is of the same chemical nature, but with a
different mass, i.e. the case of isotopic impurities. The mechanisms by which
the impurities (isotopes) perturb the phonon distribution will depend on the
mass difference between the host and guest species [1-3]. Phonons are the
crystal excitations most directly related to the isotopic masses. In monatomic
crystals (like C, Ge, Si., etc.), and within the harmonic approximation, all
phonon frequencies scale like the square root of the average isotopic mass.
Namely, this feature can be used for the nondestructive isotopic
characterization investigated materials. The isotopic effect can be classified
into two categories: 1) The first type is caused by the variation of the
phonon frequencies with the average isotopic mass. To this type belongs the
isotope effect in superconductors, which plays an important role in the search
for the mechanism of high Tc superconductivity (see, e.g. [4]). The effect of
changing the atomic mass M is to change the phonon frequencies $\omega$
according to:
$\omega$ = $\sqrt{\frac{\alpha}{\text{M}}}$, (1)
where $\alpha$ is a force constant characteristic of the phonon under
consideration. The change in atomic mass implies, at low temperatures (see
below), a change in the average atomic displacement for each phonon mode. In
the case of one atom per primitive cell the mean squared phonon amplitude
$\langle$u${}^{2}\rangle$ is given by [1;2]:
$\langle$u${}^{2}\rangle$ =
$\langle\frac{\hbar^{2}}{\text{4M}\omega}\left[1\text{ +
2n}_{B}\text{(}\omega\text{)}\right]\rangle$ =
$\langle\frac{\hbar}{\text{4M}^{1/2}\alpha^{1/2}}\left[1\text{ +
2n}_{B}\text{(}\omega\text{)}\right]\rangle$, (2)
where nB($\omega$) is the Bose - Einstein statistical factor, $\omega$ is the
frequency of a given phonon and $\langle$…$\rangle$ represents an average over
all phonon modes. The average in r.h.s. of (2) is often simplified by taking
the value inside $\langle$…$\rangle$ at an average frequency $\omega_{D}$
which usually turns out to be close to the Debye frequency. We should
distinguish between the low temperature ($\hbar\omega$ $>$$>$ kBT) and the
high temperature ($\hbar\omega$ $<$$<$ kBT) limits and see:
($\hbar\omega$ $>$$>$ kBT), $\langle$u${}^{2}\rangle$ =
$\frac{\hbar}{\text{4M}\omega_{D}}$ $\sim$ M-1/2 independent of T and
($\hbar\omega$ $<$$<$ kBT), $\langle$u${}^{2}\rangle$ =
$\frac{\text{k}_{B}\text{T}}{\text{2M}\omega^{2}}\sim$ T independent of M (3).
Using Eq. (1) we can find from last equations that $\langle$u${}^{2}\rangle$,
the zero-point vibrational amplitude, is proportional to M-1/2 at low
temperatures: it thus decrease with increasing M and vanishes for
M$\longrightarrow\infty$. For high T, however, we find that
$\langle$u${}^{2}\rangle$ is independent of M and linear in T (details see [3]
and references therein).
Another type of isotope effects is produced by the isotopic mass fluctuations
about the average mass $\langle$M$\rangle$. These fluctuations perturb the
translational invariance of a crystal and lift, at least in part, k \- vector
conservation. The most striking effect of this type is observed in the thermal
conductivity which has a maximum at a temperature TM $<$$<$ $\Theta_{D}$(here
$\Theta_{D}$ is Debye temperature, TM = 80 K for diamond, TM = 20 K for
silicon (see, also Figs. 64 - 66 in [3]). Reduction of the concentration of
13C from the standard 1% (against 99% of 12C) by a factor of ten increases the
thermal conductivity of diamond by about a factor of two, a fact that leads to
amplifications in situations where a large amount of generated heat has to be
driven away (e.g. as substrates for high power electronic devices [5]). As is
well-known this maximum represents the transition from boundary scattering to
the phonon unklapp scattering regime and its value Km is determined by the
isotopic fluctuation parameter g (mass variance):
g = $\frac{\left\langle\text{M}^{2}\right\rangle}{\left\langle
M\right\rangle^{2}}$ \- 1, (4)
the larger g - the smaller Km [6].
It is known that materials having a diamond structure are characterized by the
triply degenerate phonon states in the $\Gamma$ \- point of the Brillouin zone
(k = 0). These phonons are active in the Raman scattering (RS) spectra, but
not in the IR absorption ones (see, e.g. [7]). First - order Raman light -
scattering spectrum in diamond crystals includes one line with the maximum at
$\omega_{LTO}$($\Gamma$) = 1332.5 cm-1. In Fig. 1a, the first-order scattering
spectrum in diamond crystals with different isotope concentration is shown
[8]. As was shown, the maximum and the width of the first-order scattering
line in isotopically-mixed diamond crystals are nonlinearly dependent on the
concentration of isotopes x (see also [7]). The maximum shift of this line is
52.3 cm-1, corresponding to the limiting values of x = 0 and x = 1.
Fig. 1b demonstrates the dependence of the shape and position of the first-
order line of optical phonons in germanium crystal on the isotope composition
at liquid nitrogen temperatures [9]. The coordinate of the center of the
scattering line is proportional to the square root of the reduced mass of the
unit cell, i.e. M-1/2. It is precisely this dependence that is expected in the
harmonic approximation (details see [3]). An additional frequency shift of the
line is observed for the natural and enriched germanium specimens and is
equal, as shown in Refs. [7, 9] to 0.34$\pm$0.04 and 1.06$\pm$0.04 cm-1,
respectively (see also Fig. 7 in Chap. 4 of Ref. [10]). Detailed calculation
of the shape of the lines in RS of semiconductors have been performed by
Spitzer et al. [11]. In their paper a quantitative agreement with the
experimental data on diamond and germanium has been obtained. Comparing the
half-widths of the scattering lines in first-order RS in diamond and germanium
(see Fig. 1), it is easy to see that the observed line broadening due to
isotopic disorder in diamond is much greater than that in germanium. The
reason for this is that the k = 0 point is not the highest point in the
diamond dispersion curve (see Fig. 10b in Ref. [7]), whereas in the case of
germanium it is the highest point [12]. This shift of the maximum from the
$\Gamma$ \- point (k = 0) leads to a much larger density of states in the
vicinity of $\omega_{LTO}$ in comparison with the normal one calculated by the
formula:
Nd $\sim$ Re($\omega_{LTO}$ \- $\omega$ \+
i$\left[\frac{\Delta\omega_{LTO}}{\text{2}}\right]$)1/2 (5).
(for more details see Ref. [12]). The density of states in diamond is
asymmetric with respect to $\omega_{LTO}$, causing asymmetry in the shape of
the scattering line [7]. This asymmetry also leads to the asymmetric
concentration dependence of the half-width of the scattering line. As was
shown early (see, e.g. [3] and references therein), in the case of a weak
potential of isotopic scattering of phonons, their self-energy
$\varepsilon\left(\omega\right)$ does not depend on q (- phonon quasiimpuls).
This is precisely the situation observed for C and Ge. Indeed, if we express
the mass fluctuation $\Delta$M/$\overline{\text{M}}$ ($\overline{\text{M}}$ is
the mean mass of all isotopes) in the form of the variation of the phonon band
width $\Delta\omega_{0}$ = 12 cm-1 at q = 0 and compare it with the width of
the band of optical phonons in Ge equals to $\approx$ 100 cm-1, we will see
that the variations very small. Under this conditions the localization of
optical phonons in Ge is naturally, absent, and as observed in experiment,
they stay delocalized (see below, however opposite case in LiHxD1-x crystals).
Moreover, direct measurements of the phonon lifetime in Ge show that, in the
case of anharmonic decay, it is two orders of magnitude shorter than the
lifetime that is due to the additional scattering by isotopes, i.e.
$\tau_{anharm}$ = $\tau_{disord}\cdot$ 10-2[13]. Therefore, the contribution
of anharmonicity to the half-width of the first-order light scattering line in
Ge is two orders of magnitude greater than that caused by the isotopic
disorder in crystal lattice. In conclusion of this part of our report we
should mention that analogous structure of first-order RS and their dependence
on isotope composition has by now been observed many times, not only in
elementary Si and $\alpha$-Sn, but also in compound CuCl, CuBr, ZnSe, GaN
semiconductors (details see Ref. [3]).
In Fig. 2 (curve 1) the spectrum of second-order RS of light in pure LiD
crystal is shown [7]. In spite of the fact, according to the nomogram of
exciton states [14], the crystal studied should be considered to be pure, its
RS spectrum contains a clear high-frequency peak around 1850 cm-1. The
observed peak does not have an analogue in RS of pure LiH (Fig. 2, curve 4)
and has already been observed earlier in the second-order RS and has been
interpreted (see [7] and references therein) as a local vibration of the
hydrogen in LiD crystals. Further we note that as the concentration grows
further (x $>$ 0.15) one observes in the spectra a decreasing intensity in the
maximum of 2LO($\Gamma$) phonons in LiD crystal with a simultaneous growth in
intensity of the highest frequency peak in mixed LiHxD1-xcrystals (Fig. 2,
curve 3). The origin of the last one is in the renormalization of LO($\Gamma$)
vibrations in mixed crystals [7]. Comparison of the structure of RS spectra
(curves 1 and 2 in Fig. 2) allows us, therefore, to conclude that in the
concentration range of 0.1 $<$ x $<$ 0.45 the RS spectra simultaneously
contain peaks of the LO($\Gamma$) phonon of pure LiD and the LO($\Gamma$)
phonon of the mixed LiHxD1-xcrystal. Thus, the second-order RS spectra of
LiHxD1-xcrystals have one- and two-mode character for LO($\Gamma$) phonons,
and also contain a contribution from the local excitation at small values of
x. Moreover, we should add that an additional structure in RS spectra on the
short-side of the 2LO($\Gamma$) peak (see Fig. 21 in Ref. [7]) was observed
relatively ago in mixed LiHxD1-xcrystals and, very recently, in isotopically
mixed crystals of diamond, germanium and $\alpha$-Sn (details see [3, 11]).
These effects caused by isotopic disorder in the crystal lattice of
isotopically mixed crystals [3]. The observation of two-mode behavior of the
LO($\Gamma$) phonons in RS spectra of LiHxD1-xcrystals contradicts the
prediction of the CPA [15], according to which the width W of optical
vibration band should be smaller than the frequency shift ($\Delta$) of
transverse optical phonon. However, as was shown early (see, e.g. [7] and
references therein) in LiHxD1-x mixed crystals, the reverse inequality is
valid, i.e. W $>$ $\left|\Delta\right|$. According [16], this discrepancy
between experimental results and theory based on CPA [15] is mainly explained
by the strong potential of scattering of phonons, caused by a large change in
the mass upon substitution of deuterium for hydrogen. Once more reason of the
discrepancy between theory and results of the experiment may be connected with
not taking into account in theory the change of the force-constant at the
isotope substitution of the smaller in size D by H ion. We should stress once
more that among the various possible isotope substitution, by far the most
important in vibrational spectroscopy is the substitution of hydrogen by
deuterium. As is well-known, in the limit of the Born-Oppenheimer
approximation the force-constant calculated at the minimum of the total energy
depends upon the electronic structure and not upon the mass of the atoms. It
is usually assumed that the theoretical values of the phonon frequencies
depend upon the force-constants determined at the minimum of the adiabatic
potential energy surface. This leads to a theoretical ratio
$\omega\left(\text{H}\right)$/$\omega\left(\text{D }\right)$of the phonon
frequencies that always exceed the experimental data. Very often anharmonicity
has been proposed to be responsible for lower value of this ratio. In isotope
effect two different species of the same atom will have different vibrational
frequencies only because of the difference in isotopic masses. The ratio p of
the optical phonon frequencies for LiH and LiD crystals is given in harmonic
approximation by:
p = $\frac{\omega\left(\text{H}\right)}{\omega\left(\text{D }\right)}$ =
$\sqrt{\frac{\text{M}\left(\text{LiD}\right)}{\text{M}\left(\text{LiH}\right)}}\simeq$
$\sqrt{\text{2 }}$ (6)
while the experimental value (which includes anharmonic effects) is 1.396
$\div$ 1.288 (see Table in Ref. [17]). In this Table there are the
experimental and theoretical values of p according to formula (6), as well as
the deviation $\delta$ = $\frac{\text{P}_{Theory}\text{ -
p}_{\exp}}{\text{p}_{theory}}$ of these values from theoretical ones. Using
the least squares method it was found the empirical formula of ln($\delta$%)
$\sim$ f(ln[$\frac{\partial\text{E}}{\partial\text{M}}]$) which is depicted on
Fig.3. As can be seen the indicated dependence has in the first approximation
a linear character:
ln($\delta$%) = -7.5 + 2ln($\frac{\partial\text{E}}{\partial\text{M}}$). (7)
From the results of Fig. 3, it can be concluded that only hydrogen compounds
(and its isotope analog - deuterium) need to take into account the force-
constant changes in isotope effect. It is also seen that for semiconductor
compounds (on Fig. 3 - points, which is below of Ox line) the isotope effect
has only the changes of the isotope mass (details see [3, 7]).
The dependence of the band gap energy on isotopic composition (via mechanism
of electron-phonon interaction) has already been observed for insulators (Fig.
4) and lowest (indirect - direct) gap of different semiconductors ([3] and
references therein). It has been shown to result primarily from the effect of
the average isotopic mass on the electron-phonon interaction, with a smaller
contribution from the change in lattice constant. It was the first paper [19]
where the exciton binding energy EB was found to depend on the isotopic
composition. It was shown further that this change in EB was attributed to the
exciton-phonon interaction (originally with LO phonons) (see, also [3]). At
present time such dependence of EB $\sim$ f(x) (x- isotope concentration) was
found for different bound excitons in semiconductors [20 - 21]. The simplest
approximation, in which crystals of mixed isotopic composition are treated as
crystals of identical atoms having the average isotopic mass is referred to as
virtual crystal approximation (VCA) [15]. Going beyond the VCA, in
isotopically mixed crystals one would also expect local fluctuations in the
band-gap energy from statistical fluctuations in local isotopic composition
within some effective volume, such as that of an exciton (see, e.g. Fig. 2 of
Ref. [18]). Using the least-squares method it was found the empirical
dependence of ln$\left(\frac{\partial\text{E}_{g}}{\partial\text{M}}\right)$
$\sim$ f$\left(\text{lnE}_{g}\right),$ which is presented on Fig. 5. As can be
seen the mentioned dependence has a parabolic character:
ln$\left(\frac{\partial\text{E}_{g}}{\partial\text{M}}\right)$ =
6.105$\left(\text{lnE}_{g}\right)^{2}$ \- 7.870$\left(\text{lnE}_{g}\right)$
\+ 0.565. (8)
From this figure it can be concluded also that the small variation of the
nuclear mass (semiconductors) causes the small changes in Eg also. When the
nuclear mass increases it causes the large changes in Eg (C, LiH, CsH, etc.)
(details, see [18, 3]).
Detail analyze the process of self-diffusion in isotope pure materials and
hetero-structures was done in [5]. Interest in diffusion in solids is as old
as metallurgy or ceramics, but the scientific study of the phenomenon may
probably be dated some sixth-seven decades ago. As is well-known, the measured
diffusion coefficients depends on the chemistry and structure of the sample on
which it is measured. In cited paper [5] it was shown to use the stable
isotopes for the study of diffusion process in different semiconducting
structures (bulk, hetero-structures etc.).
Chapter 6 indicated book [5] describes the new reactor technology - neutron
transmutative doping (NTD). Capture of thermal neutrons by isotope nuclei
followed by nuclear decay produces new elements, resulting in a very number of
possibilities for isotope selective doping of solids. The importance of NTD
technology for studies of the semiconductor doping as well as metal-insulator
transitions and neutral impurity scattering process is underlined. The low-
temperature mobility of free carriers in semiconductors is mainly determined
by ionized- and neutral-impurity scattering. The ionized-impurity scattering
mechanism has been extensively studied (see e.g. [5] and references therein),
and various aspects of this process are now quite well understood. Scattering
by neutral impurities is much less than by ionized centers, i.e., its
contribution is significant only in crystals with low compensation and at very
low temperatures where most of the free carriers are frozen on the impurity
sites. The availability of highly enriched isotopes of Ge which can be
purified to residual dopant levels $<$ 1012 cm-3 has provided the first
opportunity to measure neutral impurity scattering over a wide temperature
range. In paper [22] three Ge isotopes transmute into shallow acceptors (Ga),
shallow donors (As) and double donors (Se) (see also above):
${}_{32}^{70}$Ge + n $\rightarrow$
${}_{32}^{71}$Ge${}_{EC(t_{1/2}=11.2days)}\rightarrow$ ${}_{32}^{71}$Ga +
$\nu_{e},$
${}_{32}^{74}$Ge + n $\rightarrow_{32}^{75}$Ge ${}_{\beta^{-}(t_{1/2}\text{ =
}82.2min)}$ $\rightarrow$ ${}_{32}^{75}$As + $\beta^{-}$ \+ $\bar{\nu}_{e}$,
${}_{32}^{76}$Ge + n $\rightarrow$
${}_{32}^{77}$Ge${}_{\beta^{-}\text{(t}_{1/2}}$ = 11.3 ${}_{h\text{)
}}\rightarrow$ $\beta^{-}$ \+ $\bar{\nu}_{e}$ \+ ${}_{32}^{77}$As
${}_{\beta^{-}\text{(t}_{1/2}}$ = 38.8 ${}_{h\text{)}}$ $\rightarrow$
${}_{32}^{77}$Se + $\beta^{-}$ \+ $\bar{\nu}_{e}$. (9)
The isotopes 72Ge and 73Ge are transmuted into the stable 73Ge and 74Ge
respectively. Controlling the ratio of 70Ge and 74Ge in bulk Ge crystals
allows fine tuning of the majority- as well as the minority carrier
concentration. Currently, this is the best method to vary the free-carrier
concentration independently from compensation ratio. As opposed to other
doping methods, NTD yields a very homogeneous, perfectly random distribution
of the dopants down to the atomic levels [5]. Thus isotopically controlled
crystals offer a unique possibility to study systematically the scattering
mechanism of the charge carriers in semiconductors. Extensive Hall-effect and
resistivity measurements from room temperature down to 4.2K yielded very
accurate free-carrier concentrations and mobilities as a function of
temperature and doping level were done in paper [5]. Itoh et al. [22] have
performed temperature-dependent Hall measurements on four different p-type and
two-different n-type Ge crystals (Fig. 6). Fig. 6 shows the relative strength
of the scattering from the ionized and the neutral impurities. There is only a
relatively small temperature region in which the scattering from the neutral
impurities dominates. This range extends to higher temperatures as the free-
carrier concentration is increased. The calculated ”transition temperatures”
above which the ionized impurities are the main scattering centres compare
very well with experimental results of Itoh et al [22] (see also Fig. 6.31 in
Ref. [5]). In order to demonstrate the importance of the homogeneous dopant
distribution, Itoh et al. have performed the same study on samples cut from Ge
: Ga crystals grown by the conventional Czochralski method, where Ga
impurities were introduced to Ge melt during the crystal growth. These authors
observed deviations of the measured mobility from the theoretical
calculations, which are most likely due to inhomogeneous Ga impurity
distributions in melt-doped Ge. Only the use of NTD semiconductors with
randomly distributed dopants allows for an accurate test of the neutral
impurity-scattering models (details, see [5]).
Another application of isotope pure and isotope mixed crystals that will be
discussed here is related to the possibility of using an isotopically mixed
medium (e.g. LiHxD1-x or 12Cx 13C1-x) as an oscillator of coherent radiation
in the ultraviolet spectral range. To achieve this, the use of indirect
electron transitions involving, say, LO phonons was planned [23]. The
detection of LO phonon replicas of free - exciton luminescence in wide - gap
insulators attracted considerable attention to these crystals (see e.g. [10;
23]). At the same time it is allowed one to pose a question about the
possibility of obtaining stimulated emission in UV (VUV) region (4 - 6 eV) of
the spectrum, where no solid state sources for coherent radiation exist yet.
In the first place this related to the emitters working on the transitions of
the intrinsic electronic excitation (exciton). The last one provides the high
energetical yield of the coherent emission per unit volume of the substance.
In this part we will discuss the investigation results of the influence of the
excitation light density on the resonant secondary emission spectra of the
free - exciton in the wide - gap insulator LiHxD1-x (LiH1-xFx) crystals. The
cubic LiH crystals are typical wide - gap ionic insulator with Eg = 4.992 eV
[10] with relatively weak exciton - phonon interaction however:
EB/$\hbar\omega_{LO}$ = 0.29 where E${}_{B\text{ }}$and $\hbar\omega_{LO}$ are
exciton binding energy and longitudinal optical phonon’s energy ,
respectively. Besides it might be pointed out that the analogous relation for
CdS, diamond and NaI is 0.73; 0.45 and 12.7, respectively . In the insert of
Fig. 7 depicts the luminescence of 1LO and 2LO phonon replicas in LiH
crystals. An increase in the density of the exciting light causes a burst of
the radiation energy in the long-wave wing of the emission of the 1LO and 2LO
repetitions (see Fig. 7) at a rate is higher for the 1LO replica line [23]. A
detailed dependence of the luminescence intensity and the shape of the 2LO
phonon replica line are presented in Fig. 7. The further investigations have
shown [5] that with the increase of the excitation light intensity at the
beginning a certain narrowing can be observed, followed by widening of the
line of 2LO phonon replica with a simultaneous appearance of a
characteristics, probably mode structure (see Fig. 8.11 in Ref. [5]). From
this Fig. it can be seen that the coupling between longwavelength luminescence
intensity and excitation light intensity is not only linear, but, in fact, of
a threshold character as in case of other crystals . A proximity of the
exciton parameters of LiH and CdS (ZnO) crystals allowed to carry out the
interpretation of the density effects in LiH on the analogy with these
semiconducting compounds. Coming from this in the paper [23] it was shown that
for the observed experimental picture on LiH crystals to suppose the exciton-
phonon mechanism of light generation [5] is enough the excitons density about
1015 cm-3. This is reasonable value, if the high quality of the resonator
mirrow - the crystal cleavage ”in situ” and relatively large exciton radius (r
= 40 Å [10] )is taken into account. To this light mechanism generation must be
also promoting a large value of the LO phonon energy
$\left(\hbar\omega_{LO}\text{ = 140 meV }\right).$ Owing to this the radiative
transition is being realized in the spectral region with a small value of the
absorption coefficient, and thus with a small losses in resonator (details see
[5]).
In conclusion of this section we should underlined that if the observable mode
structure is really caused by the laser generation it may be smoothly tuned in
the region of energies 4.5 $\pm$ 5.1 eV owing to smooth transition of the line
emission energy in the LiHxD1-x (LiHxF1-x; LiDxF1-x) mixed crystals as well as
in the range 5.35 - 5.10 eV in 12Cx 13C1-x mixed crystals (see also [10]).
Concluding our report we should be paid your attention to the reports of
Professors Schoven, Weston, Wendt as well as Dr. Chai of our conference which
are devoted in the first step of radioactive isotope applications.
Figure Captions.
Fig. 1. a) First-order Raman spectra of 12C${}_{x}^{13}$C1-x diamonds with
different isotope compositions. The labels A,B, C, D, E and F correspond to x
= 0.989; 0.90; 0.60; 0.50; 0.30 and 0.01 respectively. The intensity is
normalized at each peak (after [8]); b) First-order Raman scattering spectra
in Ge with different isotope contents (after [13]).
Fig. 2. Second-order Raman spectra of LiHxD1-x crystals at room temperature:
(1); (2); (3) and (4) x = 0; 0.42; 0.76 and 1, respectively (after [7]).
Fig. 3. The dependence of ln($\delta$%) $\sim$
f(ln[$\frac{\partial\text{E}}{\partial\text{M}}]$): points are experimental
values and continuous line - calculation on the formula (7) (after [17]).
Fig. 4. Mirror reflection spectra of crystals: LiH, curve 1; LiHxD1-x, curve 2
and LiD, curve 3 at 4.2 K. Light source without crystals, curve 4 (after
[18]).
Fig. 5. The dependence of
ln$\left(\frac{\partial\text{E}_{g}}{\partial\text{M}}\right)$ $\sim$
f$\left(\text{lnE}_{g}\right)$: points are experimental date and continuous
line - calculation on the formula (8) (after [18]).
Fig. 6. Temperature dependence of the carrier mobility of a) p - type and b) n
- type NTD Ge crystals. c) Temperature dependence of relative contributions to
the mobility. Note that the mobility is dominated by neutral impurity
scattering below 20 K (70Ge:Ga $\sharp$ 2 crystal) (after [22]).
Fig. 7. The dependence of the intensity in the maximum (1) and on the long-
wavelength side (2) of 2LO replica emission line of LiH crystals on the
excitation light intensity. In insert: luminescence spectra of free excitons
in LiH crystals in the region of the emission lines of 1LO and 2LO phonon
repetitions at 4.2 K for low (1) and high (2) density of excitations of 4.99
eV photons (after [23]).
References.
1\. I.M. Lifshitz, Physics of Real Crystals and Disordered Systems, Selected
Works (Eds. M.I. Kaganov, A.M. Kosevich, Science, Moscow, 1987) (in Russian).
2\. A.A. Maradudin, E.W. Montroll, G.H. Weiss and I.P. Ipatova, Theory of
Lattice Dynamics in the Harmonic Approximation, Solid State Physics, Vol.3,
(Eds. F. Seitz, D. Turnbull and H. Ehrenreich, Academic, New York, 1971).
3\. V.G. Plekhanov, Elementary Excitations in Isotope-Mixed Crystals, Physics
Reports, 410 [1-3] 1 (2005).
4\. J.P. Franck, in: Physical Properties of High Tc Superconductors (ed. D.M.
Ginsberg, Vol. 4., World Scientific, Singapore, 1984) p. 189.
5\. For a review, see, V.G. Plekhanov, Applications of the Isotopic Effect in
Solids, Springer, Berlin - Heidelberg, 2004.
6\. See, for example, R. Berman, Thermal Conduction of Solids (Clarendon
Press, Oxford, 1976); T.M. Tritt, Thermal Conductivity (Springer, Berlin -
Heidelberg, 2005).
7\. V.G. Plekhanov, Isotopic Effects in Lattice Dynamics, Physics - Uspekhi
(Moscow) 46 [7] 689 (2003).
8\. H. Hanzawa, N. Umemura, Y. Nisida et al., Disorder Effects of Nitrogen
Impurities, Irradiation - Induced Defects, and 13C Isotope Composition on the
Raman Spectrum in Synthetic Ib Diamond, Phys. Rev. B54 [6] 3793 (1996).
9.M. Cardona, Semiconductor Crystals with Tailor - Made Isotopic Compositions,
in: Festkorperprobleme/Advances in Solid State Physics (ed. R. Helbig, Vieweg,
Braunschweig, Wiesbaden, Vol. 34, 1994) p. 35.
10\. V.G. Plekhanov, Isotope Effects in Solid State Physics (Academic, New
York, 2001).
11\. J. Spitzer, P. Etchegoin, M. Cardona et al., Isotopic - Disorder Induced
Raman Scattering in Diamond, Solid State Commun. 88 [6] 509 (1993).
12\. M. Cardona, Isotopic Effects in the Phonon and Electron Dispersion
Relations of Crystals, Phys. Stat. Sol. (b) 220 [1] 5 (2000).
13\. M. Cardona, P. Etchegoin, H.D. Fuchs et al., Effect of Isotopic Disorder
and Mass on the Electronic and Vibronic Properties of Three-, Two- and One -
Dimensional Solids, J. Phys.: Condens. Matter 5 [1] A61 (1993).
14\. V.G. Plekhanov, Phonon renormalization of Interband Transition Energy in
the Mixed Crystals, Solid State Commun. 76 [1] 51 (1990).
15\. R.J. Elliott, J.A. Krumhansl and P.L. Leath, The Theory and Properties of
Randomly Disordered Crystals and Related Physical Systems, Rev. Mod. Phys. 46
[3] 465 (1974).
16\. V.G. Plekhanov, Experimental Evidence of Strong Phonon Scattering in
Isotope Disordered Systems: The Case of LiHxD1-x, Phys. Rev. B51 [14] 8874
(1995).
17\. V.G. Plekhanov (unpublished results, 2004)
18\. V.G. Plekhanov, N.V. Plekhanov, Isotope Dependence of Band - Gap Energy,
Phys. Lett., A313 [3] 231 (2003).
19\. V.G. Plekhanov, T.A. Betenekova, V.A. Pustovarov, Excitons and
Peculiarities of Exciton-Phonon Interaction in LiH and LiD, Sov. Phys. Solid
State 18 [8] 1422 (1976).
20\. M. Cardona, Dependence of the Excitation Energies of Boron in Diamond on
Isotopic Mass, Solid State Commun. 121 [1] 7 (2002).
21\. D. Karaiskaj, T.A. Meyer, M.L.W. Thewalt et al., Dependence of the
Ionization Energy of Shallow Donors and Acceptors in Silicon on the Host
Isotopic Mass, Phys. Rev. B68 [2] 121201 (2003).
22\. H.D. Fuchs, K.M. Itoh and E.E. Haller, Isotopically Controlled Germanium:
A New Medium for the Study of Carrier Scattering by Neutral Impurities,
Philos. Mag. B70 [2] 662 (1994); K.M. Itoh, E.E. Haller, V.I. Ozogin, Neutral
- Impurity Scattering in Isotopically Engineering Ge, Phys. Rev. B50 [23]
16995 (1994).
23\. V.G. Plekhanov, Wide-Gap Insulators Excitonic Nonlinearity and Its
Potential Applications in Solid State Lasers, in: Proc. Int. Conf. Advances
Solid State Lasers, USA, SOQUE, 1990.
Table. Values of the coefficients dE/dM (mev, cm-1) for the optical phonons
and the experimental and theoretical values of p as well as deviation
$\delta$% of these values from theoretical ones.
Substances | Frequencies | pexp | p${}_{\text{theory}}$ | $\delta$% = $\frac{\text{p}_{\text{theory}}\text{ - p}_{\text{exp}}}{\text{p}_{\text{theory}}}$
---|---|---|---|---
LiH/LiD | 140(meV)/104(meV)[10,24] | 1.288-1.346 | $\sqrt{\text{2}}$ = 1.414 | 4.8 - 8.9
SiH4/SiD${}_{\text{4}}$ | 2186.87/1563.3(cm-1)[10] | 1.399 | $\sqrt{\text{2}}$ = 1.414 | 1.5
12C/13C | 1332,5/1280(cm-1)[10;25] | 1.041 | $\sqrt{\frac{\text{13}}{\text{12}}}$ = 1.041 | 0.001
70Ge/76Ge | 309.8/297.7(cm-1)[10,26,27] | 1.041 | $\sqrt{\frac{\text{76}}{\text{70}}}$ = 1.042 | 0.096
28Si/30Si | 524.8/509.8(cm-1)[28] | 1.029 | $\sqrt{\frac{\text{30}}{\text{28}}}$ = 1.035 | 0.58
64Zn76Se/68Zn80Se | 213.2/207.4(cm-1)[10] | 1.028 | $\sqrt{\frac{\mu_{1}}{\mu_{2}}}$ = 1.029 | 0.097
$\alpha-^{112}$Sn/$\alpha$\- 124Sn | 206.5/196.8(cm-1)[10,30] | 1.049 | $\sqrt{\frac{\text{124}}{\text{112}}}$ = 1.052 | 0.30
Ga14N/Ga15N | 535/518(cm-1)[10,31] | 1.033 | $\sqrt{\frac{\text{15}}{\text{14}}}$ = 1.035 | 0.19
63Cu35Cl/65Cu37Cl | 174.4/171.6(cm-1)[10,32] | 1.016 | $\sqrt{\frac{\mu_{1}}{\mu_{2}}}$ = 1.022 | 0.59
| arxiv-papers | 2008-07-16T09:15:25 | 2024-09-04T02:48:56.835265 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "V.G. Plekhanov",
"submitter": "Vladimir Plekhanov",
"url": "https://arxiv.org/abs/0807.2521"
} |
0807.2619 |
# Escherlike quasiperiodic heterostructures
Alberto G. Barriuso Departamento de Óptica, Facultad de Física, Universidad
Complutense, 28040 Madrid, Spain Juan J. Monzón Departamento de Óptica,
Facultad de Física, Universidad Complutense, 28040 Madrid, Spain Luis L.
Sánchez-Soto Departamento de Óptica, Facultad de Física, Universidad
Complutense, 28040 Madrid, Spain Antonio F. Costa Departamento de
Matemáticas Fundamentales, Facultad de Ciencias, Universidad Nacional de
Educación a Distancia, Senda del Rey 9, 28040 Madrid, Spain
###### Abstract
We propose quasiperiodic heterostructures associated with the tessellations of
the unit disk by regular hyperbolic triangles. We present explicit
construction rules and explore some of the properties exhibited by these
geometric-based systems.
###### pacs:
61.44.Br, 68.65.Cd, 71.55.Jv, 78.67.Pt
Quasiperiodic (QP) systems have been receiving a lot of attention over the
last years Maciá (2006). The interest was originally motivated by the
theoretical predictions that they should manifest peculiar electron and phonon
critical states Ostlund and Pandit (1984); Kohmoto et al. (1987), associated
with highly fragmented fractal energy spectra Kohmoto et al. (1983); Sütö
(1989); Bellissard et al. (1989). On the other hand, the practical fabrication
of Fibonacci Merlin et al. (1985) and Thue-Morse Merlin et al. (1987)
superlattices has triggered a number of experimental achievements that have
provided new insights into the capabilities of QP structures Velasco and
García-Moliner (2003). In particular, possible optical applications have
deserved major attention and some intriguing properties have been demonstrated
Tamura and Nori (1989); Vasconcelos and Albuquerque (1999); Lusk et al.
(2001); Barriuso et al. (2005). Underlying all these theoretical and
experimental efforts a crucial fundamental question remains concerning whether
QP devices would achieve better performance than usual periodic ones for some
specific applications Maciá (2001).
The QP systems considered thus far rely for their explicit construction in
substitutional rules among the elements of a basic alphabet. In the common
case of a two-letter alphabet $\\{A,B\\}$, the algorithm takes the form
$A\mapsto\sigma_{A}(A,B)$, $B\mapsto\sigma_{B}(A,B)$, where $\sigma_{A}$ and
$\sigma_{B}$ can be any string of the letters. The sequences generated after
$n$ applications of the algorithm are of significance in fields as diverse as
cryptography, time-series analysis, and cellular automata Cheng and Savit
(1990). In addition, they have interesting algebraic properties, which are
usually characterized by the nature of their Fourier or multifractal spectra
Spinadel (1999).
We wish to approach the problem from an alternative geometrical perspective.
To this end, we first observe that in many problems of physical interest Maciá
and Rodríguez-Oliveros (2006) the letters of the alphabet can be identified
with one-dimensional linear lossless systems (i.e., with two input and two
output channels). Under these general conditions, it turns out that the
associated transfer matrix belongs to the group SU(1,1), which is also the
basic symmetry group of the hyperbolic geometry Coxeter (1968). In
consequence, the unit disk appears as the natural arena to discuss their
performance Yonte et al. (2002); Monzón et al. (2002); Barriuso et al. (2003).
Since in the Euclidean plane, QP behavior is intimately linked with
tessellations, one is unfailingly led to consider the role of hyperbolic
tessellations in the unit disk, much in the spirit of Escher’s masterpiece
woodcut Circle Limit III Coxeter (1996). The answer we propose is promising:
the tessellations by different regular polygons provide new sequences with
properties that may open avenues of research in this field.
A QP system can thus be seen as a word generated by stacking different letters
of the basic alphabet. To be specific, we focus our attention on the optical
response. Let us consider one of these letters (which in practice is made of
several plane-parallel layers), which we assume to be sandwiched between two
semi-infinite identical ambient ($a$) and substrate ($s$). We suppose
monochromatic plane waves incident, in general, from both the ambient and the
substrate. As a result of multiple reflections in all the interfaces, the
total electric field can be decomposed in terms of forward- and backward-
traveling plane waves, denoted by $E^{(+)}$ and $E^{(-)}$, respectively. If we
take these components as a vector
$\mathbf{E}=\left(\begin{array}[]{c}E^{(+)}\\\ E^{(-)}\end{array}\right)\,,$
(1)
then the amplitudes at both the ambient and the substrate sides are related by
the transfer matrix ${\bm{\mathsf{M}}}$
$\mathbf{E}_{a}={\bm{\mathsf{M}}}\,\mathbf{E}_{s}\,.$ (2)
It can be shown that ${\bm{\mathsf{M}}}$ is of the form
${\bm{\mathsf{M}}}=\left(\begin{array}[]{cc}1/T&R^{\ast}/T^{\ast}\\\
R/T&1/T^{\ast}\end{array}\right)\,,$ (3)
where the complex numbers $R$ and $T$ are, respectively, the overall
reflection and transmission coefficients for a wave incident from the ambient.
The condition $\det{\bm{\mathsf{M}}}=+1$ is equivalent to $|R|^{2}+|T|^{2}=1$,
and then the set of transfer matrices reduces to the group SU(1,1). Obviously,
the matrix of a word obtained by putting together letters of the alphabet is
the product of the matrices representing each one of them, taken in the
appropriate order.
In many instances we are interested in the transformation properties of field
quotients rather than the fields themselves. Therefore, it seems natural to
consider the complex numbers
$z=\frac{E^{(-)}}{E^{(+)}}\,,$ (4)
for both ambient and substrate. The action of the matrix given in Eq. (2) can
be then seen as a function $z_{a}=f(z_{s})$ that can be appropriately called
the transfer function. From a geometrical viewpoint, this function defines a
transformation of the complex plane $\mathbb{C}$, mapping the point $z_{s}$
into the point $z_{a}$ according to
$z_{a}=\frac{\beta^{\ast}+\alpha^{\ast}z_{s}}{\alpha+\beta z_{s}}\,,$ (5)
where $\alpha=1/T$ and $\beta=R^{\ast}/T^{\ast}$. When no light is incident
from the substrate, $z_{s}=0$ and then $z_{a}=R$. Equation (5) is a bilinear
(or Möbius) transformation. One can check that the unit disk, the external
region and the unit circle remain invariant under (5). This unit disk is then
a model for hyperbolic geometry in which a line is represented as an arc of a
circle that meets the boundary of the disk at right angles to it (and
diameters are also permitted). In this model, we have three different kinds of
lines: intersecting, parallel (they intersect at infinity, which is precisely
the boundary of the disk) and ultraparallel (they are neither intersecting nor
parallel).
To classify the possible actions it proves convenient to work out the fixed
points of the transfer function; that is, the field configurations such that
$z_{a}=z_{s}\equiv z_{f}$ in Eq. (5), whose solutions are
$z_{f}=\frac{1}{2\beta}\left\\{-2i\,\mathrm{Im}(\alpha)\pm\sqrt{[\mathop{\mathrm{Tr}}\nolimits({\bm{\mathsf{M}}})]^{2}-4}\right\\}.$
(6)
When $[\mathop{\mathrm{Tr}}\nolimits({\bm{\mathsf{M}}})]^{2}<4$ the action is
elliptic and it has only one fixed point inside the unit disk. Since in the
Euclidean geometry a rotation is characterized for having only one invariant
point, this action can be appropriately called a hyperbolic rotation.
When $[\mathop{\mathrm{Tr}}\nolimits({\bm{\mathsf{M}}})]^{2}>4$ the action is
hyperbolic and it has two fixed points, both on the boundary of the unit disk.
The geodesic line joining these two fixed points remains invariant and thus,
by analogy with the Euclidean case, this action is called a hyperbolic
translation.
Finally, when $[\mathop{\mathrm{Tr}}\nolimits({\bm{\mathsf{M}}})]^{2}=4$ the
action is parabolic and it has only one (double) fixed point on the boundary
of the unit disk.
As it is well known, SU(1, 1) is isomorphic to the group of real unimodular
matrices SL(2, $\mathbb{R}$), which allows us to translate the geometrical
structure defined in the unit disk to the complex upper semiplane, recovering
in this way an alternative model of the hyperbolic geometry that is useful in
some applications.
The notion of periodicity is intimately connected with tessellations, i.e.,
tilings by identical replicas of a unit cell (or fundamental domain) that fill
the plane with no overlaps and no gaps. Of special interest is the case when
the primitive cell is a regular polygon with a finite area Zieschang et al.
(1980). In the Euclidean plane, the associated regular tessellation is
generically noted $\\{p,q\\}$, where $p$ is the number of polygon edges and
$q$ is the number of polygons that meet at a vertex. We recall that
geometrical constraints limit the possible regular tilings $\\{p,q\\}$ to
those verifying $(p-2)(q-2)=4$. This includes the classical tilings
$\\{4,4\\}$ (tiling by squares) and $\\{6,3\\}$ (tiling by hexagons), plus a
third one, the tiling $\\{3,6\\}$ by triangles (which is dual to the
$\\{6,3\\}$).
On the contrary, in the hyperbolic disk regular tilings exist provided
$(p-2)(q-2)>4$, which now leads to an infinite number of possibilities. An
essential ingredient is the way to obtain fundamental polygons. These polygons
are directly connected to the discrete subgroups of isometries (or congruent
mappings). Such groups are called Fuchsian groups Ford (1972) and play for the
hyperbolic geometry a role similar to that of crystallographic groups for the
Euclidean geometry Beardon (1983).
A tessellation of the hyperbolic plane by regular polygons has a symmetry
group that is generated by reflections in geodesics, which are inversions
across circles in the unit disk. These geodesics correspond to edges or axes
of symmetry of the polygons. Therefore, to construct a tessellation of the
unit disk one just has to built one tile and to duplicate it by using
reflections in the edges.
Figure 1: (color online). A realistic implementation of the generators
${\bm{\mathsf{A}}}$ and ${\bm{\mathsf{B}}}$ (together with
${\bm{\mathsf{A}}}^{-1}$) of the tessellations by hyperbolic triangles in (9).
The external layers (in blue) are made of cryolite (Na3AlF6), while the
central medium (in brown) is zinc selenide (ZnSe). The wavelength in vacuum is
$\lambda=610$ nm and normal incidence (from left to right) has been assumed.
The corresponding thicknesses are expressed in nanometers.
In this Letter, we consider only the simple example of a tessellation by
triangles with vertices in the unit circle, although the treatment can be
extended to other polygons. The key idea is to consider the Fuchsian group
generated by an elliptic transformation whose fixed point is the middle of an
edge of the triangle and a parabolic one with its fixed point in the opposite
vertex. Proceeding in this way we get
$\displaystyle{\bm{\mathsf{A}}}$ $\displaystyle=$
$\displaystyle\left(\begin{array}[]{cc}1+i/\sqrt{3}&1/\sqrt{3}\\\
1/\sqrt{3}&1-i/\sqrt{3}\end{array}\right)\,,$ (9)
$\displaystyle{\bm{\mathsf{B}}}$ $\displaystyle=$
$\displaystyle\left(\begin{array}[]{cc}2i/\sqrt{3}&-1/\sqrt{3}\\\
-1/\sqrt{3}&-2i/\sqrt{3}\end{array}\right)\,.$ (13)
The fixed point of ${\bm{\mathsf{A}}}$ is $-i$; while for ${\bm{\mathsf{B}}}$
the fixed point in the disk is $i(2-\sqrt{3})$. In Fig. 1 we show a possible
way in which these matrices can be implemented in terms of two commonly
employed materials in optics. Note that, in physical terms, the inverses must
be constructed as independent systems, although in our case only
${\bm{\mathsf{A}}}^{-1}$ must be considered, since the action of
${\bm{\mathsf{B}}}^{-1}$ coincides with that of ${\bm{\mathsf{B}}}$.
Figure 2: (color online). Tiling of the unit disk with the matrices (9). The marked points are the barycenters of the triangles in the tessellation and all of them are the transformed of the origin by a matrix that have as reflection coefficient the complex number that links the origin with the center of the triangle. Table 1: Explicit rules to obtain the barycenter $z_{n+1}$ from the $z_{n}$. We have indicated the corresponding transformations, which depend on the color jumps and the sides crossed by going from $z_{n}$ to $z_{n+1}$. | red $\rightarrow$ yellow | yellow $\rightarrow$ red
---|---|---
Side | Trans | ${\bm{\mathsf{A}}}_{n+1}$ | ${\bm{\mathsf{B}}}_{n+1}$ | Trans | ${\bm{\mathsf{A}}}_{n+1}$ | ${\bm{\mathsf{B}}}_{n+1}$
0 | ${\bm{\mathsf{B}}}_{n}$ | ${\bm{\mathsf{B}}}_{n}{\bm{\mathsf{A}}}_{n}{\bm{\mathsf{B}}}_{n}$ | ${\bm{\mathsf{B}}}_{n}$ | ${\bm{\mathsf{B}}}_{n}$ | ${\bm{\mathsf{B}}}_{n}{\bm{\mathsf{A}}}_{n}^{-1}{\bm{\mathsf{B}}}_{n}$ | ${\bm{\mathsf{B}}}_{n}$
1 | ${\bm{\mathsf{A}}}_{n}$ | ${\bm{\mathsf{A}}}_{n}$ | ${\bm{\mathsf{A}}}_{n}{\bm{\mathsf{B}}}_{n}{\bm{\mathsf{A}}}_{n}^{-1}$ | ${\bm{\mathsf{A}}}_{n}^{-1}$ | ${\bm{\mathsf{A}}}_{n}$ | ${\bm{\mathsf{A}}}_{n}^{-1}{\bm{\mathsf{B}}}_{n}{\bm{\mathsf{A}}}_{n}$
2 | ${\bm{\mathsf{A}}}_{n}^{-1}$ | ${\bm{\mathsf{A}}}_{n}^{-1}$ | ${\bm{\mathsf{A}}}_{n}^{-1}{\bm{\mathsf{B}}}_{n}{\bm{\mathsf{A}}}_{n}$ | ${\bm{\mathsf{A}}}_{n}$ | ${\bm{\mathsf{A}}}_{n}^{-1}$ | ${\bm{\mathsf{A}}}_{n}{\bm{\mathsf{B}}}_{n}{\bm{\mathsf{A}}}_{n}^{-1}$
In Fig. 2 we have shown the tessellation obtained by transforming the
fundamental triangle with the Fuchsian group generated by the powers of
$\\{{\bm{\mathsf{A}}},{\bm{\mathsf{B}}}\\}$ (and the inverses). This triangle
is equilateral with vertices at the points $-i$, $\exp(i\pi/6)$ and
$\exp(i5\pi/6)$ (which are the fixed points of ${\bm{\mathsf{A}}}$,
${\bm{\mathsf{A}}}{\bm{\mathsf{B}}}$ and ${\bm{\mathsf{B}}}{\bm{\mathsf{A}}}$,
respectively). Moreover, all the other triangles are equal, with an area
$\pi$. In the figure we have plotted also the barycenters of each triangle
together with the resulting tree (that is called the dual graph of the
tessellation), which turns out to be a Farey tree Schröder (2006). In fact,
each line connecting two of these barycenters represent the action of a word
(with alphabet $\\{{\bm{\mathsf{A}}},{\bm{\mathsf{B}}}\\}$ and the inverses).
To give an explicit construction rule for the possible words, we proceed as
follows. First, we arbitrarily assign the number 0 to the upper side of the
fundamental triangle, while the other two sides are clockwise numbered as 1
and 2. It is easy to convince oneself that this assignment fixes once for all
the numbering for the sides of the other triangles in the tessellation.
However, these triangles can be distinguished by their orientation (as seen
from the corresponding barycenter): the clockwise oriented are filled in red,
while the counterclockwise are filled in yellow. In short, we have determined
a fundamental coloring of the tessellation Grünbaum and Shepard (1987).
To obtain one barycenter $z_{n+1}$ from the previous one $z_{n}$, one looks
first at the corresponding color jump. Next, the matrix that take $z_{n}$ into
$z_{n+1}$ depends on the numbering of the side (0, 1, or 2) one must cross,
and appear in the appropriate column “Trans” in Table 1. The next generation
is obtained much in the same way, except for the fact that
${\bm{\mathsf{A}}}_{n}$ and ${\bm{\mathsf{B}}}_{n}$ must be replaced by
${\bm{\mathsf{A}}}_{n+1}$ and ${\bm{\mathsf{B}}}_{n+1}$, respectively, as
indicated in the Table. In obtaining recursively any word, the origin is
denoted as $z_{0}$ and the matrices ${\bm{\mathsf{A}}}_{0}$ and
${\bm{\mathsf{B}}}_{0}$ coincide with ${\bm{\mathsf{A}}}$ and
${\bm{\mathsf{B}}}$.
With this rule, one can construct any word proceeding step by step. For
example, the word that transform $z_{0}$ into $z_{5}$ in the zig-zag path
sketched in Fig. 2 results
$\begin{array}[]{lcl}z_{0}\rightarrow z_{1}\quad:\quad{\bm{\mathsf{B}}}\,,\\\
z_{1}\rightarrow
z_{2}\quad:\quad{\bm{\mathsf{B}}}{\bm{\mathsf{A}}}^{-1}{\bm{\mathsf{B}}}\,,\\\
z_{2}\rightarrow
z_{3}\quad:\quad{\bm{\mathsf{B}}}{\bm{\mathsf{A}}}^{-1}{\bm{\mathsf{B}}}\,,\\\
z_{3}\rightarrow
z_{4}\quad:\quad{\bm{\mathsf{B}}}{\bm{\mathsf{A}}}^{-1}{\bm{\mathsf{A}}}^{-1}{\bm{\mathsf{B}}}{\bm{\mathsf{A}}}{\bm{\mathsf{A}}}{\bm{\mathsf{B}}}\,,\\\
z_{4}\rightarrow
z_{5}\quad:\quad{\bm{\mathsf{B}}}{\bm{\mathsf{A}}}^{-1}{\bm{\mathsf{A}}}^{-1}{\bm{\mathsf{B}}}{\bm{\mathsf{A}}}{\bm{\mathsf{B}}}{\bm{\mathsf{A}}}{\bm{\mathsf{A}}}{\bm{\mathsf{B}}}\,.\end{array}$
(14)
Obviously, the total word is obtained by composing these partial words.
In fact, one can show that given an arbitrary sequence of nonzero integers
$\\{k_{1},k_{2},\ldots,k_{r}\\}$, the word represented by the transfer matrix
${\bm{\mathsf{M}}}(s,k_{1},\ldots,k_{r})={\bm{\mathsf{B}}}^{s_{1}}{\bm{\mathsf{A}}}^{k_{1}}{\bm{\mathsf{B}}}{\bm{\mathsf{A}}}^{k_{2}}\ldots{\bm{\mathsf{B}}}{\bm{\mathsf{A}}}^{k_{r}}{\bm{\mathsf{B}}}^{s_{2}}\,,$
(15)
where $\\{s_{1},s_{2}\\}\subset\\{0,1\\}$, transforms the origin in a
barycenter of the tessellation.
Figure 3: (color online). Normalized structure factor for the word (made of 60
letters) connecting the origin with the point $z_{8}$ in the zig-zag path
shown in Fig. 2.
Given the geometric regularity of the construction sketched in this Letter,
the sequences obtained must play a key role in the theory and practice of QP
systems. Of course, to put forward the relevant physical features of these
sequences, there are a number of quantities one can look at. Perhaps one of
the most appropriate ones to assess the performance of these systems is the
structure factor Cheng and Savit (1990). For a word ${\bm{\mathsf{M}}}$
(composed of $L$ letters) we define a numerical sequence $f_{n}$ by assigning
1, $e^{2\pi i/3}$, and $e^{-2\pi i/3}$ to the letters ${\bm{\mathsf{B}}}$,
${\bm{\mathsf{A}}}$, and ${\bm{\mathsf{A}}}^{-1}$, respectively. Next, we
calculate the discrete Fourier transform of the sequence $f_{n}$
$F_{k}=\sum_{n=0}^{L-1}f_{n}\exp\left(-\frac{2\pi ikn}{L}\right)\,,$ (16)
where $k=0,1,..L-1$. The structure factor (or power spectrum) is just
$|F_{k}|^{2}$. In Fig. 3 we have plotted this structure factor in terms of $k$
for the word connecting the origin with the point $z_{8}$ in the zig-zag path
of Fig. 2. The peaks reveal a rich behavior: a full analysis of these
questions is outside the scope of this Letter and will be presented elsewhere.
As a final and rather technical remark, we note that the quotient of the
hyperbolic disk by the Fuchsian group generated by ${\bm{\mathsf{A}}}$ and
${\bm{\mathsf{B}}}$ is a 2-orbifold of genus 0, with a conical point of order
two and a cusp. Each word as given in Eq. (15) represents a hyperbolic
transformation of the disk, and the axis of the transformation is projected
onto a closed geodesic of such an orbifold. This provides an orbifold
interpretation of our QP sequences.
In summary, we expect to have presented new schemes to generate QP sequences
based on hyperbolic tessellations of the unit disk. Apart from the intrinsic
beauty of the formalism, our preliminary results seem to be quite encouraging
for future applications of these systems.
The authors wish to express their warmest gratitude to E. Maciá and J. M.
Montesinos for their help and interest in the present work.
## References
* Maciá (2006) E. Maciá, Rep. Prog. Phys. 69, 397 (2006).
* Ostlund and Pandit (1984) S. Ostlund and R. Pandit, Phys. Rev. B 29, 1394 (1984).
* Kohmoto et al. (1987) M. Kohmoto, B. Sutherland, and K. Iguchi, Phys. Rev. Lett. 58, 2436 (1987).
* Kohmoto et al. (1983) M. Kohmoto, L. P. Kadanoff, and C. Tang, Phys. Rev. Lett. 50, 1870 (1983).
* Sütö (1989) A. Sütö, J. Stat. Phys. 56, 525 (1989).
* Bellissard et al. (1989) J. Bellissard, B. Iochum, E. Scoppola, and D. Testard, Comm. Math. Phys. 125, 527 (1989).
* Merlin et al. (1985) R. Merlin, K. Bajema, R. Clarke, F. Y. Juang, and P. K. Bhattacharya, Phys. Rev. Lett. 55, 1768 (1985).
* Merlin et al. (1987) R. Merlin, K. Bajema, J. Nagle, and K. Ploog, J. Phys. Colloq. 48, C5 503 (1987).
* Velasco and García-Moliner (2003) V. R. Velasco and F. García-Moliner, Prog. Surf. Sci. 74, 343 (2003).
* Tamura and Nori (1989) S. Tamura and F. Nori, Phys. Rev. B 40, 9790 (1989).
* Vasconcelos and Albuquerque (1999) M. S. Vasconcelos and E. L. Albuquerque, Phys. Rev. B 59, 11128 (1999).
* Lusk et al. (2001) D. Lusk, I. Abdulhalim, and F. Placido, Opt. Commun. 198, 273 (2001).
* Barriuso et al. (2005) A. G. Barriuso, J. J. Monzón, L. L. Sánchez-Soto, and A. Felipe, Opt. Express 13, 3913 (2005).
* Maciá (2001) E. Maciá, Phys. Rev. B 63, 205421 (2001).
* Cheng and Savit (1990) Z. Cheng and R. Savit, J. Stat. Phys. 60, 383 (1990).
* Spinadel (1999) V. W. Spinadel, Nonlinear Anal. 36, 721 (1999).
* Maciá and Rodríguez-Oliveros (2006) E. Maciá and R. Rodríguez-Oliveros, Phys. Rev. B 74, 144202 (2006).
* Coxeter (1968) H. S. M. Coxeter, _Non-Euclidean Geometry_ (University of Toronto Press, Toronto, 1968).
* Yonte et al. (2002) T. Yonte, J. J. Monzón, L. L. Sánchez-Soto, J. F. Cariñena, and C. López-Lacasta, J. Opt. Soc. Am. A 19, 603 (2002).
* Monzón et al. (2002) J. J. Monzón, T. Yonte, L. L. Sánchez-Soto, and J. F. Cariñena, J. Opt. Soc. Am. A 19, 985 (2002).
* Barriuso et al. (2003) A. G. Barriuso, J. J. Monzón, and L. L. Sánchez-Soto, Opt. Lett. 28, 1501 (2003).
* Coxeter (1996) H. S. M. Coxeter, Math. Intell. 18, 42 (1996).
* Zieschang et al. (1980) H. Zieschang, E. Vogt, and H. D. Coldeway, _Surfaces and planar discontinuous groups_ , vol. 835 of _Lect. Not. Math._ (Springer, Berlin, 1980).
* Ford (1972) L. R. Ford, _Automorphic functions_ (Chelsea Publising Company, New York, 1972).
* Beardon (1983) A. F. Beardon, _The geometry of discrete groups_ (Springer, Berlin, 1983).
* Schröder (2006) M. R. Schröder, _Number Theory in Science and Communication_ (Springer, New York, 2006), 4th ed.
* Grünbaum and Shepard (1987) B. Grünbaum and G. C. Shepard, _Tilings and Patterns_ (Freeman, New York, 1987).
| arxiv-papers | 2008-07-16T17:42:04 | 2024-09-04T02:48:56.842516 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "A. G. Barriuso, J. J. Monzon, L. L. Sanchez-Soto, and A. F. Costa",
"submitter": "Luis L. Sanchez. Soto",
"url": "https://arxiv.org/abs/0807.2619"
} |
0807.2656 | Through a Glass Darkly
Steven G. Krantz111It is a pleasure to thank David H. Bailey, Jonathan
Borwein, Robert Burckel, David Collins, Marvin Greenberg, Reece Harris,
Deborah K. Nelson, and James S. Walker for many useful remarks and suggestions
about different drafts of this essay. Certainly their insights have
contributed a number of significant improvements.
## 1 Prolegomena
> Education is a repetition of civilization in little.
> — Herbert Spencer
Being a mathematician is like being a manic depressive. One experiences
occasional moments of giddy elation, interwoven with protracted periods of
black despair. Yet this is the life path that we choose for ourselves. And we
wonder why nobody understands us.
The budding mathematician spends an extraordinarily long period of study and
backbreaking hard work in order to attain the Ph.D. And that is only an entry
card into the profession. It hardly makes one a mathematician.
To be able to call oneself a mathematician, one must have proved some good
theorems and written some good papers thereon. One must have given a number of
talks on his work, and (ideally) one should have either an academic job or a
job in the research infrastructure. Then, and only then, can one hold one’s
head up in the community and call oneself a peer of the realm. Often one is
thirty years old before this comes about. It is a protracted period of
apprenticeship, and there are many fallen and discouraged and indeed lost
along the way.
The professional mathematician spends his life thinking about problems that he
cannot solve, and learning from his (repeated and often maddening) mistakes.
That he can very occasionally pull the fat out of the fire and make something
worthwhile of it is in fact a small miracle. And even when he can pull off
such a feat, what are the chances that his peers in the community will toss
their hats in the air and proclaim him a hail fellow well met? Slim to none at
best.
In the end we learn to do mathematics because of its intrinsic beauty, and its
enduring value, and for the personal satisfaction it gives us. It is an
important, worthwhile, dignified way to spend one’s time, and it beats almost
any other avocation that I can think of. But it has its frustrations.
There are few outside of the mathematical community who have even the vaguest
notion of what we do, or how we spend our time. Surely they have no sense of
what a theorem is, or how one proves a theorem, or why one would want
to.222From my solipsistic perspective as a mathematician, this is truly
tragic. For mathematical thinking is at the very basis of human thought. It is
the key to an examined life. How could one spend a year or two studying other
people’s work, only so that one can spend yet several more years to develop
one’s own work? Were it not for tenure, how could any mathematics ever get
done?
We in the mathematics community expect (as we should) the state legislature to
provide funds for the universities (to pay our salaries, for instance). We
expect the members of Congress to allocate funds for the National Science
Foundation and other agencies to subvent our research. We expect the White
House Science Advisor to speak well of academics, and of mathematicians in
particular, so that we can live our lives and enjoy the fruits of our labors.
But what do these people know of our values and our goals? How can we hope
that, when they do the obvious and necessary ranking of priorities that must
be a part of their jobs, we will somehow get sorted near the top of the list?
This last paragraph explains in part why we as a profession can be aggravated
and demoralized, and why we endure periods of frustration and hopelessness. We
are not by nature articulate—especially at presenting our case to those who do
not speak our language—and we pay a price for that incoherence. We tend to be
solipsistic and focused on our scientific activities, and trust that the value
of our results will speak for themselves. When competing with the Wii and the
iPod, we are bound therefore to be daunted.
## 2 Life in the Big City
> The most savage controversies are about those matters as to which there is
> no good evidence either way.
>
> — Bertrand Russell
If you have ever been Chair of your department, put in the position of
explaining to the Dean what the department’s needs are, you know how hard it
is to explain our mission to the great unwashed. You waltz into the Dean’s
office and start telling him how we must have someone in Ricci flows, we
certainly need a worker in mirror symmetry, and what about that hot new stuff
about the distribution of primes using additive combinatorics? The Dean,
probably a chemist, has no idea what you are talking about.
Of course the person who had the previous appointment with the Dean was the
Chair of Chemistry, and he glibly told the Dean how they are woefully shy of
people in radiochemistry and organic chemistry. And an extra physical chemist
or two would be nice as well. The Dean said “sure”, he understood immediately.
It was a real shift of gears then for the Dean to have to figure out what in
the world you (from the Mathematics Department) are talking about. How do you
put your case in words that the Dean will understand? How do you sell yourself
(and your department) to him?333It is arguable that a mathematics department
is better off with a Dean who is a musicologist or perhaps a philologist. Such
a scholar is not hampered by the Realpolitik of lab science dynamics, and can
perhaps think imaginatively about what our goals are.
Certainly we have the same problem with society at large. People understand,
just because of their social milieu, why medicine is important and useful.
Computers and their offspring make good sense; we all encounter computers
every day and have at least a heuristic sense of what they are good for. Even
certain parts of engineering resonate with the average citizen (aeronautics,
biomedical engineering, civil engineering). But, after getting out of school,
most people have little or no use for mathematics. Most financial transactions
are handled by machines. Most of us bring our taxes to professionals for
preparation. Most of us farm out construction projects around the house to
contractors. If any mathematics, or even arithmetic, is required in the
workplace it is probably handled by software.
One of my wife’s uncles, a farmer, once said to me—thinking himself to be in a
puckish mood—that we obviously no longer need mathematicians because we have
computers. I gave him a patient look and said yes, and we obviously no longer
need farmers because we have vending machines. He was not amused. But the
analogy is a good one. Computers are great for manipulating data, but not for
thinking. Vending machines are great for handing you a morsel of food that
someone else has produced in the traditional fashion.
People had a hard time understanding what Picasso’s art was about—or even Andy
Warhol’s art—but they had a visceral sense that it was interesting and
important. The fact that people would spend millions of dollars for the
paintings gave the activity a certain gravitas, but there is something in the
nature of art that makes it resonate with our collective unconscious. With
mathematics, people spend their lives coming to grips with what was likely a
negative experience in school, reinforced by uninspiring teachers and dreadful
textbooks. If you are at a cocktail party and announce that you don’t like
art, or don’t like music, people are liable to conclude that you are some kind
of philistine. If instead you announce that you don’t like mathematics, people
conclude that you are a regular guy. [If you choose to announce that you do
like mathematics, people are liable to get up and walk away.] To the
uninitiated, mathematics is cold and austere and unforgiving. It is difficult
to get even an intuitive sense of what the typical mathematician is up to.
Unlike physicists and biologists (who have been successfully communicating
with the press and the public for more than fifty years), we are not good at
telling half-truths so that we can paint a picture of our meaning and get our
point across. We are too wedded to the mathematical method. We think in terms
of definitions and axioms and theorems.
## 3 Living the Good Life
> One normally thinks that everything that is true is true for a reason. I’ve
> found mathematical truths that are true for no reason at all. These
> mathematical truths are beyond the power of mathematical reasoning because
> they are accidental and random.
>
> — G. J. Chaitin
The life of a mathematician is a wonderful experience. It is an exhilarating,
blissful existence for those who are prone to enjoy it. One gets to spend
one’s time with like-minded people who are in pursuit of a holy grail that is
part of an important and valuable larger picture that we are all bound to. One
gets to travel, and spend time with friends all over the world, and hang out
in hotels, and eat exotic foods, and drink lovely drinks. One gets to teach
bright students and engage in the marketplace of ideas, and actually to
develop new ones. What could be better? There is hardly a more rewarding way
to be professionally engaged.
It is a special privilege to be able to spend one’s time—and be paid for
it—thinking original (and occasionally profound) thoughts and developing new
programs and ideas. One actually feels that he is changing the fabric of the
cosmos, helping people to see things that they have not seen before, affecting
people’s lives.444I have long been inspired by Freeman Dyson’s book [DYS]. It
describes both poignantly and passionately the life of the scientist, and how
he can feel that he is altering and influencing the world around him. Teaching
can and probably should be a part of this process. For surely bringing along
the next generation, training a new flank of scholars, is one of the more
enlightened and certainly important pursuits. Also interacting with young
minds is a beautiful way to stay vibrant and plugged in, and to keep in touch
with the development of new ideas.
Of course there are different types of teaching. The teaching of rudimentary
calculus to freshmen has different rewards from teaching your latest research
ideas to graduate students. But both are important, and both yield palpable
results. What is more, this is an activity that others understand and
appreciate. If the public does not think of us in any other way, surely they
think of us as teachers. And better that we should have to do it. After all,
it is our bailiwick.
The hard fact of the matter is that the powers that be in the university also
appreciate our teaching rather more than they do our many other activities.
After all, mathematics is a key part of the core curriculum. A university
could hardly survive without mathematics. Other majors could not function,
could not advance their students, could not build their curricula, without a
basis in mathematics. So our teaching role at the institution is both
fundamental and essential. Our research role is less well understood,
especially because we do not by instinct communicate naturally with scholars
in other departments.
This is actually a key point. We all recall the crisis at the University of
Rochester thirteen years ago, when the Dean shut down the graduate program in
mathematics. His reasoning, quite simply, was that he felt that the
mathematics department was isolated, did not interact productively with other
units on campus, did not carry its own weight. The event at Rochester rang a
knell throughout the profession, for we all knew that similar allegations
could be leveled at any of us. Institutions like Princeton or Harvard are
truly ivory towers, and unlikely to suffer the sort of indignity being
described here. But if you work at a public institution then look out. I work
at a very private university, and I can tell you that, in my negotiations as
Chair with our Dean, he sometimes brought up Rochester. And he did not do so
in an effort to be friendly. He was in fact threatening me.
Some departments, like Earth & Planetary Science or Biomedical Engineering,
interact very naturally with other subjects. Their material is intrinsically
interdisciplinary. It makes perfect sense for these people to develop cross-
disciplinary curricula and joint majors with other departments. It is very
obvious and sensible for them to apply for grants with people from departments
even outside of their School. A faculty member of such a department will speak
several languages fluently.
It is different for mathematics. It is a challenge just to speak the one
language of mathematics, and to speak it well. Most of us do a pretty good job
at it, and those outside of mathematics cannot do it at all. So there is a
natural barrier to communication and collaboration. In meetings with other
faculty—even from physics and engineering—we find difficulty identifying a
common vocabulary. We find that we have widely disparate goals, and very
different means of achieving them.
Also our value systems are different. Our methods for gauging success vary
dramatically. Our reward systems deviate markedly. Once you become a full
Professor you will serve on tenure and promotion committees for other
departments. This experience is a real eye-opener, for you will find that the
criteria used in English and History and Geography are quite different from
what we are accustomed to.555I still recall serving on the committee for
promotion to Professor of a candidate in Geography. One of his published
writings was called A Walk Through China Town. It described the experience of
walking down Grant Avenue in San Francisco and smelling the wonton soup. What
would be the analogue of this in a case for promotion in Mathematics? Even our
views of truth can be markedly different.
## 4 The Why and the Wherefore
> The lofty light of the a priori outshines the dim light of the world and
> makes for us incontrovertible truths because of their “clearness and
> distinctness.”
> — René Descartes
A mathematician typically goes through most of his early life as a flaming
success at everything he does. One excels in grade school, one excels in high
school, one excels in college. Even in graduate school one can do quite well
if one is willing to put forth the effort.
Put in slightly different terms: One can get a long way in the basic material
just by being smart. Not so much effort or discipline is required. And this
may explain why so many truly brilliant people get left in the dust. They
reach a point where some real Sitzfleisch and true effort are required, and
they are simply not up to it. They have never had to expend such disciplined
study before, so why start now?
While there is no question that being smart can take one a long way, there
comes a point—for all of us—where it becomes clear that a capacity for hard
work can really make a difference. Most professional mathematicians put in at
least ten hours per day, at least six days per week. There are many who do
much more. And we tend to enjoy it. The great thing about mathematics is that
it does not fight you. It will not sneak behind your back and bite you. It is
always satisfying and always rewarding.
Doing mathematics is not like laying bricks or mowing the grass. The quantity
of end product is not a linear function of the time expended. Far from it. As
Charles Fefferman, Fields Medalist, once said, a good mathematician throws 90%
of his work in the trash. Of course one learns from all that work, and it
makes one stronger for the next sortie. But one often, at the end of six
months or a year, does not have much to show.
On the other hand, one can be blessed with extraordinary periods of
productivity. The accumulated skills and insights of many years of study
suddenly begin to pay off, and one finds that he has plenty to say. And it is
quite worthwhile. Certainly worth writing up and sharing with others and
publishing. This is what makes life rewarding, and this is what we live for.
Economists like to use professors as a model, because they run contrary to
many of the truisms of elementary economic theory. For example, if you pay a
Professor of Mathematics twice as much, that does not mean that he will be
able to prove twice as many theorems, or produce twice as many graduate
students. The truth is that he is probably already working to his capacity.
There are only so many hours in the day. What more could he do? It is
difficult to say what a Professor of Mathematics should be compensated,
because we do not fit the classical economic model.
Flipped on its head, we could also note that if you give a Professor of
Mathematics twice as much to do, it does not follow that he will have a
nervous breakdown, or quit, or go into open rebellion. Many of us now have a
teaching load of two courses per semester. But sixty years ago the norm—even
at the very best universities in the United States—was three courses (or
more!) per semester. Also, in those days, there was very little secretarial
help. Professors did a lot of the drudgery themselves. There were also no NSF
grants, and very little discretionary departmental money, so travel was often
subvented from one’s own pocket. Today life is much better for everyone.
The fact is that a Professor of Mathematics has a good deal of slack built
into his schedule. If you double his teaching load, it means that he has less
time to go to seminars, or to talk to his colleagues, or just to sit and
think. But he will still get through the day. Just with considerably less
enthusiasm. And notably less creativity. Universities are holding faculty much
more accountable for their time these days. Total Quality Management is one of
many insidious ideas from the business world that is starting to get a grip at
our institutions of higher learning. In twenty years we may find that we are
much more like teachers (in the way that we spend our time) and much less like
scholars.
Sad to say, the Dean or the Provost has only the vaguest sense of what our
scholarly activities are. When they think of the math department at all, they
think of us as “those guys who teach calculus.” They certainly do not think of
us as “those guys who proved the Bieberbach conjecture.” Such a statement
would have little meaning for the typical university administrator. Of course
they are pleased when the faculty garners kudos and awards, but the awards
that Louis de Branges received for his achievement were fairly low key.666When
I was Chair of the Mathematics Department, the Dean was constantly reminding
me that he thought of us as a gang of incompetent, fairly uncooperative boobs.
One of his very favorite Chairs at that time was the Head of Earth & Planetary
Sciences. This man was in fact the leader of the Mars space probe team, and he
actually designed the vehicle that was being used to explore Mars. Well, you
can imagine the kind of presentations that this guy could give—lots of
animated graphics, lots of panoramic vistas, lots of dreamy speculation, lots
of stories about other-worldly adventures. His talks were given in the biggest
auditoriums on campus, and they were always packed. The Dean was front and
center, with his tongue hanging out, every time; he fairly glowed in the dark
because he was so pleased and excited. How can a mathematician compete with
that sort of showmanship? Even if I were to prove the Riemann Hypothesis, it
would pale by comparison. They probably would not even raise an eyebrow among
the Board of Trustees.
## 5 Such is Life
> There is no religious denomination in which the misuse of metaphysical
> expressions has been responsible for so much sin as it has in mathematics.
>
> — Ludwig Wittgenstein
Mathematicians are very much like oboe players. They do something quite
difficult that nobody else understands. That is fine, but it comes with a
price.
We take it for granted that we work in a rarified stratum of the universe that
nobody else will understand. We do not expect to be able to communicate with
others. When we meet someone at a cocktail party and say, “I am a
mathematician,” we expect to be snubbed, or perhaps greeted with a witty
rejoinder like, “I was never any good in math.” Or, “I was good at math until
we got to that stuff with the letters—like algebra.”
When I meet a brain surgeon I never say, “I was never any good at brain
surgery. Those lobotomies always got me down.” When I meet a proctologist, I
am never tempted to say, “I was never any good at ….” Why do we mathematicians
elicit such foolish behavior from people?
One friend of mine suggested that what people are really saying to us, when
they make a statement of the sort just indicated, is that they spent their
college years screwing around. They never buckled down and studied anything
serious. So now they are apologizing for it. This is perhaps too simplistic.
For taxi drivers say these foolish things too. And so do mailmen and butchers.
Perhaps what people are telling us is that they know that they should
understand and appreciate mathematics, but they do not. So instead they are
resentful.
There is a real disconnect when it comes to mathematics. Most people, by the
time that they get to college, have had enough mathematics so that they can be
pretty sure they do not like it. They certainly do not want to major in the
subject, and their preference is to avoid it as much as possible.
Unfortunately, for many of these folks, their major may require a nontrivial
amount of math (not so much because the subject area actually uses
mathematics, but rather because the people who run the department seem to want
to use mathematics as a filter). And also unfortunately it happens, much more
often than it should, that people end up changing their majors (from
engineering to psychology or physics to media studies) simply because they
cannot hack the math.
In recent years I have been collaborating with plastic surgeons, and I find
that this is a wonderful device for cutting through the sort of conversational
impasse that we have been describing. Everyone, at least everyone past a
certain age, is quite interested in plastic surgery. People want to understand
it, they want to know what it entails, they want to know what are the
guarantees of success. When they learn that there are connections between
plastic surgery and mathematics then that is a hint of a human side of math.
It gives me an entree that I never enjoyed in the past.
I also once wrote a paper with a picture of the space shuttle in it. That did
not prove to be quite so salubrious for casual conversation; after all,
engineering piled on top of mathematics does not make the mathematics any more
palatable. But at least it was an indication that I could speak several
tongues.
And that is certainly a point worth pondering if we want to fit into a social
milieu. Speaking many tongues is a distinct advantage, and gives one a wedge
for making real contact with people. It provides another way of looking at
things, a new point of contact. Trying to talk to people about mathematics, in
the language of mathematics, using the logic of mathematics is not going to
get you very far. It will not work with newspaper reporters and it also will
not work with ordinary folks that you are going to meet in the course of your
life.
## 6 Mathematics and Art
> It takes a long time to understand nothing.
>
> — Edward Dahlberg
Even in the times of ancient Greece there was an understanding that
mathematics and art were related. Both disciplines entail symmetry, order,
perspective, and intricate relationships among the components. The golden mean
is but one of many artifacts of this putative symbiosis.
M. C. Escher spent a good deal of time at the Moorish castle the Alhambra,
studying the very mathematical artwork displayed there. This served to inspire
his later studies (which are considered to be a very remarkable synthesis of
mathematics and art).
Today there is more formal recognition of the interrelationship of mathematics
and art. No less an eminence than Louis Vuitton offers a substantial prize
each year for innovative work on the interface of mathematics and art. Benoit
Mandelbrot has received this prize (for his work on fractals—see [MAN]), and
so has David Hoffman for his work with Jim Hoffman and Bill Meeks on embedded
minimal surfaces (see [HOF]).
Mathematics and art make a wonderful and fecund pairing for, as we have
discussed here, mathematics is perceived in general to be austere,
unforgiving, cold, and perhaps even lifeless. By contrast, art is warm, human,
inspiring, even divine. If I had to give an after-dinner talk about what I do,
I would not get very far trying to discuss the automorphism groups of
pseudoconvex domains. I would probably have much better luck discussing the
mathematics in the art of M. C. Escher, or the art that led to the
mathematical work of Celso Costa on minimal surfaces.
Of course we as mathematicians perceive our craft to be an art form. Those
among us who can see—and actually prove!—profound new theorems are held in the
greatest reverence, much as artists. We see the process of divining a new
result and then determining how to verify it much like the process of eking
out a new artwork. It would be in our best interest to convey this view of
what we do to the world at large. Whatever the merits of fractal geometry may
be, Benoit Mandelbrot has done a wonderful job of conveying both the art and
the excitement of mathematics to the public.
Those who wish to do so may seek mathematics exhibited in art throughout the
ages. Examples are
* •
A marble mosaic featuring the small stellated dodecahedron, attributed to
Paolo Uccello, in the floor of the San Marco Basilica in Venice.
* •
Leonardo da Vinci’s outstanding diagrams of regular polyhedra drawn as
illustrations for Luca Pacioli’s book The Divine Proportion.
* •
A glass rhombicuboctahedron in Jacopo de’ Barbari’s portrait of Pacioli,
painted in 1495.
* •
A truncated polyhedron (and various other mathematical objects) which feature
in Albrecht Dürer’s engraving Melancholia I.
* •
Salvador Dalí’s painting The Last Supper in which Christ and his disciples are
pictured inside a giant dodecahedron.
Sculptor Helaman Ferguson [FER] has made sculptures in various materials of a
wide range of complex surfaces and other topological objects. His work is
motivated specifically by the desire to create visual representations of
mathematical objects. There are many artists today who conceive of themselves,
and indeed advertise themselves, as mathematical artists. There are probably
rather fewer mathematicians who conceive of themselves as artistic
mathematicians.
Mathematics and music have a longstanding and deeply developed relationship.
Abstract algebra and number theory can be used to understand musical
structure. There is even a well-defined subject of musical set theory
(although it is used primarily to describe atonal pieces). Pythagorean tuning
is based on the perfect consonances. Many mathematicians are musicians, and
take great comfort and joy from musical pastimes. Music can be an opportunity
for mathematicians to interact meaningfully with a broad cross section of our
world. Mathematicians Noam Elkies and David Wright have developed wonderful
presentations—even full courses—about the symbiosis between mathematics and
music.
Mathematics can learn a lot from art, especially from the way that art reaches
out to humanity. Part of art is the interface between the artist and the
observer. Mathematics is like that too, but typically the observer is another
mathematician. We would do well, as a profession, to think about how to expand
our pool of observers.
## 7 Mathematics vs. Physics
> I do still believe that rigor is a relative notion, not an absolute one. It
> depends on the background readers have and are expected to use in their
> judgment.
>
> — René Thom
Certainly “versus” is the wrong word here. Ever since the time of Isaac
Newton, mathematics and physics have been closely allied. After all, Isaac
Newton virtually invented physics as we know it today. And mathematics in his
day was a free-for-all. So the field was open for Newton to create any
synthesis that he chose.
But mathematics and physics are divided by a common goal, which is to
understand the world around us. Physicists perceive that “world” by observing
and recording and thinking. Mathematicians perceive that “world” by looking
within themselves (but see the next section on Platonism vs. Kantianism).
And thus arises a difference in styles. The physicist thinks of himself as an
observer, and is often content to describe what he sees. The mathematician is
never so content. Even when he “sees” with utmost clarity, the mathematician
wants to confirm that vision with a proof. This fact makes us precise and
austere and exacting, but it also sets us apart and makes us mysterious and
difficult to deal with.
I once heard Fields Medalist Charles Fefferman give a lecture (to a mixed
audience of mathematicians and physicists) about the existence of matter. In
those days Fefferman’s goal was to prove the existence of matter from first
principles—in an axiomatic fashion. I thought that this was a fascinating
quest, and I think that some of the other mathematicians in the audience
agreed with me. But at some point during the talk a frustrated physicist
raised his hand and shouted, “Why do you need to do this? All you have to do
is look out the window to see that matter exists!”
Isn’t it wonderful? Different people have different value systems and
different ways to view the very same scientific facts. If there is a schism
between the way that mathematicians view themselves and the way that
physicists see us, then there is little surprise that there is such a schism
between our view of ourselves and the way that non-scientists see us. Most
laymen are content to accept the world phenomenologically—it is what it is.
Certainly it is not the average person’s job to try to dope out why things are
the way they are, or who made them that way. This all borders on theology, and
that is a distinctly painful topic. Better to go have a beer and watch a
sporting event on the large-screen TV. This is not the view that a
mathematician takes.
The world of the mathematician is a world that we have built for ourselves.
And it makes good sense that we have done so, for we need this infrastructure
in order to pursue the truths that we care about. But the nature of our
subject also sets us apart from others—even from close allies like the
physicists. We not only have a divergence of points of view, but also an
impasse in communication. We often cannot find the words to enunciate what we
are seeing, or what we are thinking.
In fact it has taken more than 2500 years for the modern mathematical mode of
discourse to evolve. Although the history of proof is rather obscure, we know
that the efforts of Thales and Protagoras and Hippocrates and Theaetetus and
Plato and Pythagoras and Aristotle, culminating in Euclid’s magnificent
Elements, have given us the axiomatic method and the language of proof. In
modern times, the work of David Hilbert and Nicolas Bourbaki have helped us to
sharpen our focus and nail down a universal language and methodology for
mathematics (see [KRA] for a detailed history of these matters and for many
relevant references). The idea of mathematical proof is still changing and
evolving, but it is definitely part of who we are and what we believe.
The discussion of Platonism and Kantianism in the next section sheds further
light on these issues.
## 8 Plato vs. Kant
> It is by logic we prove, it is by intuition that we invent.
>
> — Henri Poincaré
A debate has been festering in the mathematics profession for a good time now,
and it seems to have heated up in the past few years (see, for instance
[DAV]). And the debate says quite a lot about who we are and how we endeavor
to think of ourselves. It is the question of whether our subject is Platonic
or Kantian.
The Platonic view of the world is that mathematical facts have an independent
existence—very much like classical Platonic ideals—and the research
mathematician discovers those facts—very much like Amerigo Vespucci discovered
America, or Jonas Salk discovered his polio vaccine. But it should be clearly
understood that, in the Platonic view, mathematical ideas exist in some higher
realm that is independent of the physical world, and certainly independent of
any particular person. Also independent of time. The Platonic view poses the
notion that a theorem can be “true” before it is proved.
The Kantian view of the world is that the mathematician creates the subject
from within himself. The idea of set, the idea of group, the idea of
pseudoconvexity, are all products of the human mind. They do not exist out
there in nature. We (the mathematical community) have created them.
My own view is that both these paradigms are valid, and both play a role in
the life of any mathematician. On a typical day, the mathematician goes to his
office and sits down and thinks. He will certainly examine mathematical ideas
that already exist, and can be found in some paper penned by some other
mathematician. But he will also cook things up from whole cloth. Maybe create
a new axiom system, or define a new concept, or formulate a new hypothesis.
These two activities are by no means mutually exclusive, and they both
contribute to the rich broth that is mathematics.
Of course the Kantian position raises interesting epistemological questions.
Do we think of mathematics as being created by each individual? If that is so,
then there are hundreds if not thousands of distinct individuals creating
mathematics from within. How can they communicate and share their ideas? Or
perhaps the Kantian position is that mathematics is created by some shared
consciousness of the aggregate humanity of mathematicians. And then is it up
to each individual to “discover” what the aggregate consciousness has been
creating? Which is starting to sound awfully Platonic. Saunders Mac Lane [MAC]
argues cogently that mathematical ideas are elicited or abstracted from the
world around is. This is perhaps a middle path between the two points of view.
The Platonic view of reality seems to border on theism. For if mathematical
truths have an independent existence—floating out there in the ether
somewhere—then who created those truths? And by what means? Is it some
higher power, with whom we would be well-advised to become better acquainted?
The Platonic view makes us more like physicists. It would not make much sense
for a physicist to study his subject by simply making things up. Or cooking
them up through pure cogitation. For the physicist is supposed to be
describing the world around him. A physicist like Stephen Hawking, who is very
creative and filled with imagination, is certainly capable of cooking up ideas
like “black hole” and “supergravity” and “wormholes”, but these are all
intended to help explain how the universe works. They are not like
manufacturing a fairy tale.
There are philosophical consequences for the thoughts expressed in the last
paragraph. Physicists do not feel honor-bound to prove the claims made in
their research papers. They frequently use other modes of discourse, ranging
from description to analogy to experiment to calculation. If we mathematicians
are Platonists, describing a world that is “already out there”, then why
cannot we use the same discourse that the physicists use? Why do we need to be
so wedded to proofs?
One can hardly imagine an English Professor trying to decide whether his
discipline is Platonic or Kantian. Nor would a physicist ever waste his time
on such a quest. People in those disciplines know where the grist of their
mill lives, and what they are about. The questions do not really make sense
for them. We are somewhat alone in this quandary, and it is our job to take
possession of it. If we can.
It appears that literary critics and physicists are certainly Platonists. What
else could they be?777Although a physicist may put a finer point on it and
assert that he has no care for a Platonic realm of ideas. Rather, he wishes to
run experiments and “ask questions of nature.” It is unimaginable that they
would cook up their subject from within themselves. Certainly philosophers can
and do engage in this discussion, and they would also be well-equipped (from a
strictly intellectual perspective) to engage in the Platonic vs. Kantian
debate. But they have other concerns. This does not seem to be their primary
beat.
The article [MAZ] sheds new and profound light on the questions being
considered here. This is a discussion that will last a long time, and probably
will never come to any clear resolution.
Once again the Platonic vs. Kantian debate illustrates the remove that
mathematicians have from the ordinary current of social discourse. How can the
layman identify with these questions? How can the layman even care about them?
If I were a real estate salesman or a dental technician, what would these
questions mean to me?
## 9 Seeking the Truth
> In what we really understand, we reason but little.
>
> — William Hazlitt
Mathematicians are good at solving problems. But we have recognized for a long
time that we have a problem with communicating with laymen, with the public at
large, with the press, and with government agencies. We have made little
progress in solving this particular problem. What is the difficulty?
Part of the problem is that we are not well-motivated. It is not entirely
clear what the rewards would be for solving this problem. But it is also not
clear what the methodology should be. Standard mathematical argot will not
turn the trick. Proceeding from definitions to axioms to theorems will, in
this context, fall on deaf ears. We must learn a new modus operandi, and we
must learn how to implement it.
This is not something that anyone is particularly good at, and we
mathematicians have little practice in the matter. We have all concentrated
our lives in learning how to communicate with each other. And such activity
certainly has its own rewards. But it tends to make us blind to broader
issues. It tends to make us not listen, and not perceive, and not process the
information that we are given. Even when useful information trickles through,
we are not sure what to do with it. It does not fit into the usual
infrastructure of our ideas. We are not comfortable processing the data.
This is our own fault. This is how we have trained ourselves, and it is how we
train our students. We are not by nature open and outreaching. We are rather
parochial and closed. We are more comfortable sticking close to home. And, to
repeat a tired adage, we pay a price for this isolation.
## 10 Brave New World
> For most wearers of white coats, philosophy is to science as pornography is
> to sex: it is cheaper, easier, and some people seem, bafflingly, to prefer
> it. Outside of psychology it plays almost no part in the functions of the
> research machine.
> — Steve Jones
For the past 2,000 years, mathematicians have enjoyed a sense of keeping to
themselves, and playing their own tune.888Although it would be remiss not to
note that Archimedes, Newton, and Gauss were public figures, and very much a
part of society. It has given us the freedom to think our own thoughts and to
pursue our own truths. By not being answerable to anyone except ourselves, we
have been able to keep our subject pure and insulated from untoward
influences.
But the world has changed around us. Because of the rise of computers, because
of the infusion of engineering ideas into all aspects of life, because of the
changing nature of research funding, we find ourselves not only isolated but
actually cut off from many of the things that we need in order to prosper and
grow.
So it may be time to re-assess our goals, and our milieu, and indeed our very
lingua franca, and think about how to fit in more naturally with the flow of
life. Every medical student takes a course on medical ethics. Perhaps every
mathematics graduate student should take a course on communication. This would
include not only good language skills, but how to use electronic media, how to
talk to people with varying (non-mathematical) backgrounds, how to seek the
right level for a presentation, how to select a topic, and many of the other
details that make for effective verbal and visual skills. Doing so would
strengthen us as individuals, and it would strengthen our profession. We would
be able to get along more effectively as members of the university, and also
as members of society at large. Surely the benefits would outweigh the
inconvenience and aggravation, and we would likely learn something from the
process. But we must train ourselves (in some instances re-train ourselves) to
be welcoming to new points of view, to new perspectives, to new value systems.
These different value systems need not be perceived as inimical to our own.
Rather they are complementary, and we can grow by internalizing them.
Mathematics is one of the oldest avenues of human intellectual endeavor and
discourse. It has a long and glorious history, and in many ways it represents
the best of what we as a species are capable of doing. We, the mathematics
profession, are the vessels in which the subject lives. It is up to us to
nurture it and to ensure that it grows and prospers. We can no longer do this
in isolation. We must become part of the growing and diversifying process that
is human development, and we must learn to communicate with all parts of our
culture. It is in our best interest, and it is in everyone else’s best
interest as well.
References
1. [DAV]
E. B. Davies, Let Platonism die, Newsletter of the European Mathematical
Society 64(1007), 24–25.
2. [DYS]
F. Dyson, Disturbing the Universe, Basic Books, New York, 2001.
3. [FER]
H. Ferguson, Sculpture Gallery,
`http://www.helasculpt.com/gallery/index.html`.
4. [HOF]
D. Hoffman, The computer-aided discovery of new embedded minimal surfaces,
Math. Intelligencer 9(1987), 8–21.
5. [KRA]
S. Krantz, The Proof is in the Pudding: A Look at the Changing Nature of
Mathematical Proof, Springer Publishing, to appear.
6. [MAC]
S. Mac Lane, Mathematical models: a sketch for the philosophy of mathematics,
American Mathematical Monthly 88(1981), 462–472.
7. [MAN]
B. Mandelbrot, The Fractal Geometry of Nature, Freeman, New York, 1977.
8. [MAZ]
B. Mazur, Mathematical Platonism and its opposites,
`http://www.math.harvard.edu/~mazur/`.
Department of Mathematics, Washington University in St. Louis, St. Louis,
Missouri 63130
sk@math.wustl.edu
| arxiv-papers | 2008-07-16T20:39:55 | 2024-09-04T02:48:56.847943 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Steven G. Krantz",
"submitter": "Steven G. Krantz",
"url": "https://arxiv.org/abs/0807.2656"
} |
0807.2685 | # Cardy-Verlinde formula in Taub-NUT/Bolt-(A)dS space
Chong Oh Lee Department of Physics, Chonbuk National University, Jeonju
561-756, Republic of Korea cohlee@chonbuk.ac.kr
###### Abstract
We consider a finite action for a higher dimensional Taub-NUT/Bolt-(A)dS space
via the so-called counter term subtraction method. In the limit of high
temperature, we show that the Cardy-Verlinde formula holds for the Taub-Bolt-
AdS metric and for the specific dimensional Taub-NUT-(A)dS metric, except for
the Taub-Bolt-dS metric.
###### pacs:
11.25.Hf, 11.25.Tq
## I Introduction
The AdS/CFT duality was first conjectured by 't Hooft:1973jz in his search
for relationship between gauge theories and strings. The AdS/CFT
correspondence Maldacena:1997re ; Gubser:1998bc ; Witten:1998qj ;
Aharony:1999ti ; D'Hoker:2002aw asserts there is an equivalence between a
gravitational theory in the bulk and a conformal field theory in the boundary.
According to AdS/CFT, a $(d+1)$-dimensional S-(A)dS action $A$ is given by
$\displaystyle A=A_{B}+A_{\partial B}+A_{ct}$ (1)
where the bulk action $A_{B}$, action boundary $A_{\partial B}$, and
counterterm action $A_{ct}$ are given as
$\displaystyle A_{B}$ $\displaystyle=$ $\displaystyle\frac{1}{16\pi
G_{d+1}}\int_{\cal M}d^{d+1}x\sqrt{-g}({\cal R}-2\Lambda),$ $\displaystyle
A_{\partial B}$ $\displaystyle=$ $\displaystyle-\frac{1}{8\pi
G_{d+1}}\int_{\partial{\cal M}}d^{d}x\sqrt{-\gamma}\Theta,$ $\displaystyle
A_{ct}$ $\displaystyle=$ $\displaystyle-\frac{1}{8\pi
G_{d+1}}\int_{\partial{\cal
M}}d^{d}x\sqrt{-\gamma}\left\\{-\frac{d-1}{l}\right.$ (2)
$\displaystyle-\frac{lR}{2(d-2)}{\cal F}(d-3)$
$\displaystyle-\frac{l^{3}}{2(d-2)^{2}(d-4)}$ $\displaystyle\hskip
28.45274pt\times\left(R_{ab}R^{ab}-\frac{d}{4(d-1)}R^{2}\right){\cal F}(d-5)$
$\displaystyle+\frac{l^{5}}{(d-2)^{3}(d-4)(d-6)}$ $\displaystyle\hskip
28.45274pt\times\left(\frac{3d+2}{4(d-1)}RR_{ab}R^{ab}-\frac{d(d+2)}{16(d-1)^{2}}R^{3}\right.$
$\displaystyle\left.\left.+\frac{d-2}{2(d-1)}R^{ab}\nabla_{a}\nabla_{b}R-R^{ab}\Box
R_{ab}\right.\right.$ $\displaystyle\left.\left.+\frac{1}{2(d-1)}R\Box
R\right){\cal F}(d-7)+\cdots\right\\},$
where a negative cosmological constant $\Lambda$ is
$\Lambda=-{d(d-1)}/{2l^{2}}$, $\Theta$ is the trace of extrinsic curvature.
Here, ${\cal F}(d)$ is a step function, 1 when $d\geq 0$, 0 otherwise. The
boundary action $A_{\partial B}$ is added to the action $A$ to obtain
equations of motion well behaved at the boundary. Then the boundary energy-
momentum tensor is expressed in Brown:1992br
$\displaystyle\frac{2}{\sqrt{-\gamma}}\frac{\delta A_{\partial
B}}{\delta\gamma^{ab}}=\Theta_{ab}-\gamma_{ab}\Theta.$ (3)
The counterterm action $A_{ct}$ is added to the action $A$ to remove the
divergence appearing as the boundary goes to infinity Balasubramanian:1999re .
For low dimensional S-AdS, a few terms in the counterterm action $A_{ct}$ were
explicitly evaluated in Balasubramanian:1999re ; Emparan:1999pm . Using the
universality of the structure of divergences, the counterterm action $A_{ct}$
for arbitrary dimension is suggested in Kraus:1999di . This action $A$ (1)
leads to the entropy $S$ via the Gibbs-Duhem relation
$\displaystyle S=\frac{E}{T}-A$ (4)
where $T$ denotes the temperature and $E$ is the total energy.
The entropy of the (1+1)-dimensional CFT is expressed in terms of the Virasoro
operator $L_{0}$ and the central charge $c$, the so-called the Cardy formula
Cardy:1986ie . Using conformal invariance, the generalized Cardy formula in
arbitrary dimension is shown to be given universal form as Verlinde:2000wg
(for the review articles of the issue, see, e.g., Nojiri:2001fa ;
Nojiri:2002hz ; Lidsey:2002ah )
$S_{\rm CFT}=\frac{2\pi R}{\sqrt{ab}}\sqrt{E_{c}(2E-E_{c})}\,,$ (5)
where $a$ and $b$ are certain constants. $R$ denotes the radius of the
universe at a given time and $E_{c}$ is the Casimir energy defined by
$E_{c}=d\,E-(d-1)TS\,.$ (6)
Employing AdS/CFT dual picture, $\sqrt{ab}$ is fixed to $(d-1)$ exactly, in
particular, for a $d$-dimensional CFT on ${\rm\mathbf{R}}$$\times S^{d-1}$
Verlinde:2000wg . Then, the entropy is given as
$S_{\rm CFT}=\frac{2\pi R}{d-1}\sqrt{E_{c}(2E-E_{c})}\,,$ (7)
which is shown to hold for Schwarzschild (A)dS (S-(A)dS) Verlinde:2000wg ;
Cai:2001sn , charged (A)dS Cai:2001jc ; Cai:2001tv , Kerr-(A)dS Klemm:2001db ;
Cai:2001tv , and Taub-Bolt-AdS4 Birmingham:2001vd . There are many other
relevant papers on the subject Wang:2001bf ; Wang:2001bv ; Setare:2002ss ;
Setare:2002qa ; Setare:2003fg . Thus, one may naively expect that the entropy
of all CFTs that have an AdS-dual description is given as the form (7).
However, AdS black holes do not always satisfy the Cardy-Verlinde formula
(see, e.g., Cai:2001jc ; Gibbons:2005vp ). Therefore, one intriguing question
is whether this formula is valid for higher dimensional Taub-NUT-(A)dS at high
temperature. In this Letter, we will endeavor to do this.
## II Taub-NUT/Bolt-AdS black hole
When the total number of dimension of the spacetime is even, $(d+1)=2u+2$, for
some integer $u$, the Euclidean section of the arbitrary (d+1)-dimensional-
Taub-NUT-AdS metric, for a $U(1)$ fibration over a series of the space ${\cal
M}^{2}$ as the base space $\bigotimes_{i=1}^{u}{\cal M}^{2}$, is given by
Chamblin:1998pz ; Hawking:1998ct ; Mann:1999bt ; Awad:2000gg ; Clarkson:2002uj
; Astefanesei:2004ji ; Astefanesei:2004kn (for the generalized versions of
the issue, see, e.g., Mann:2003zh ; Mann:2005ra )
$\displaystyle ds^{2}_{\rm AdS}$ $\displaystyle=$ $\displaystyle
f(r)\left[dt_{E}+2N\sum_{i=1}^{u}\cos(\theta_{i})d\phi_{i}\right]^{2}$
$\displaystyle+$
$\displaystyle\frac{dr^{2}}{f(r)}+(r^{2}-N^{2})\sum_{i=1}^{u}\Biggr{[}d\theta_{i}^{2}+\sin^{2}(\theta_{i})d\phi_{i}^{2}\Biggr{]},$
where $N$ represents a NUT charge for the Euclidean section, and the metric
function $f(r)$ has the general form
$\displaystyle f(r)$ $\displaystyle=$
$\displaystyle\frac{r}{(r^{2}-N^{2})^{u}}\int^{r}\left[\frac{(p^{2}-N^{2})^{u}}{p^{2}}\right.$
(9) $\displaystyle+$
$\displaystyle\left.\frac{(2u+1)(p^{2}-N^{2})^{u+1}}{l^{2}p^{2}}\right]dp-\frac{2mr}{(r^{2}-N^{2})^{u}},$
with a cosmological parameter $l$ and a geometric mass $m$.
Requiring $f(r)|_{r=N}$ the NUT solution occurs. Then for AdS spacetime the
inverse of the temperature $\beta$ arises from imposed condition in order to
ensure regularity in the Euclidean time $t_{E}$ and radial coordinate $r$
Chamblin:1998pz ; Hawking:1998ct ; Mann:1999bt ; Awad:2000gg ; Clarkson:2002uj
; Astefanesei:2004ji ; Astefanesei:2004kn
$\displaystyle\beta=\left.\frac{4\pi}{f^{\prime}(r)}\right|_{r=N}=\frac{2(d+1)\pi
N}{q},$ (10)
where $\beta$ is the period of $t_{E}$. Here $q$ is a positive integer, which
originates from removing Misner string singularities. Using counter term
subtraction method the regularized action is given as Chamblin:1998pz ;
Hawking:1998ct ; Mann:1999bt ; Awad:2000gg ; Clarkson:2002uj ;
Astefanesei:2004ji ; Astefanesei:2004kn
$\displaystyle I_{\rm NUT}$ $\displaystyle=$
$\displaystyle\frac{(4\pi)^{\frac{d}{2}}N^{d-2}\biggr{(}(d-1)N^{2}-l^{2}\biggr{)}}{32\pi^{2}l^{2}}$
(11) $\displaystyle\hskip
8.53581pt\times\Gamma(\frac{2-d}{2})\Gamma(\frac{d+1}{2})\beta.$
Employing the thermal relation $E=\partial_{\beta}I$ the total energy can also
be written by
$\displaystyle E$ $\displaystyle=$
$\displaystyle\frac{(4\pi)^{\frac{d}{2}}(d-1)N^{d-2}\biggr{(}(d+1)N^{2}-l^{2}\biggr{)}}{32\pi^{2}ql^{2}}$
(12) $\displaystyle\hskip
8.53581pt\times\Gamma(\frac{2-d}{2})\Gamma(\frac{d+1}{2}),$
and the entropy is given as Chamblin:1998pz ; Hawking:1998ct ; Mann:1999bt ;
Awad:2000gg ; Clarkson:2002uj ; Astefanesei:2004ji ; Astefanesei:2004kn
$\displaystyle S_{\rm NUT,AdS}$ $\displaystyle=$
$\displaystyle\frac{(4\pi)^{\frac{d}{2}}N^{d-2}\biggr{(}d(d-1)N^{2}-(d-2)l^{2}\biggr{)}}{32\pi^{2}l^{2}}$
(13) $\displaystyle\hskip
8.53581pt\times\Gamma(\frac{2-d}{2})\Gamma(\frac{d+1}{2})\beta,$
by the Gibbs-Duhem relation $S=\beta M-I$ where $M$ denotes the conserved mass
$\displaystyle M=\frac{(d-1)(4\pi)^{\frac{d}{2}}}{16\pi^{\frac{3}{2}}}m.$ (14)
Substituting (10), (12), and (13) into (6), one gets the Casimir energy
Verlinde:2000wg
$\displaystyle E_{c}$ $\displaystyle=$
$\displaystyle\frac{(4\pi)^{\frac{d}{2}}(d-1)N^{d-2}\biggr{(}dN^{2}-l^{2}\biggr{)}}{16\pi^{2}ql^{2}}$
(15) $\displaystyle\hskip
8.53581pt\times\Gamma(\frac{2-d}{2})\Gamma(\frac{d+1}{2}).$
From now on, for convenience we use $l/z$ instead of the universe radius $R$
in (7) since the AdS metric is always asymptotically taken to be
Fefferman:1984aa
$\displaystyle
ds^{2}=\frac{l^{2}}{z^{2}}dz^{2}+\frac{l^{2}}{z^{2}}g_{ab}(x,z)dx^{a}dx^{b},$
(16)
where the $r=\infty$ is put to $z=0$, and the roman indexes $a$ and $b$ refer
to boundary coordinates. When $1/\sqrt{ab}$ in the formula (7) is taken to be
$2/(d+1)(d-1)(d-2)$, the CFT entropy is given as
$\displaystyle S_{\rm CFT}$ $\displaystyle=$ $\displaystyle\frac{4\pi
l\sqrt{|E_{c}(2E-E_{c})|}}{(d+1)(d-1)(d-2)}\,,$ (17) $\displaystyle=$
$\displaystyle\frac{(4\pi)^{\frac{d}{2}}|dN^{2}-l^{2}|(-1)^{[\frac{d}{2}]}\Gamma(\frac{d+1}{2})\Gamma(\frac{2-d}{2})}{4\pi(d+1)(d-2)q},$
where $[x]$ is the Gauss number (greatest integer less than or equal to x).
Here it seems that the difference from the standard Cardy-Verlinde formula (7)
is due to the distinctive nature of NUT solution in AdS space like
asymptotically locally AdS (ALAdS) metric. In the limit of high temperature,
$N\rightarrow 0$, leading term in the entropy of CFT can be expressed as
$\displaystyle S_{\rm CFT}$ $\displaystyle=$
$\displaystyle\frac{(4\pi)^{\frac{d}{2}}(d+1)(d-2)N^{d-1}}{16\pi q}$ (18)
$\displaystyle\hskip
8.53581pt\times(-1)^{[\frac{d}{2}]}\Gamma(\frac{d+1}{2})\Gamma(\frac{2-d}{2})$
$\displaystyle=$ $\displaystyle(-1)^{[\frac{d}{2}]}S_{\rm NUT,AdS}.$
This result shows that the entropy of the Taub-NUT-AdS space suffices to be
the generalized Cardy-Verlinde formula (5) for all even $u$ $(d+1=2u+2)$. This
is reasonable because the Taub-NUT-AdS metric has the thermodynamically stable
range depending on the magnitude of the NUT charge i.e. any NUT solution in
AdS space for all odd $u$ is thermodynamically unstable in the lime
$N\rightarrow 0$.
Requiring $f(r)|_{r=r_{\rm B}>N}$ and $f^{\prime}(r)|_{r=r_{\rm
B}}=\frac{1}{(u+1)N}$, the Bolt solution occurs. In Taub-Bolt-AdS metric, the
inverse of the temperature, the total energy, and the entropy are respectively
$\displaystyle\beta=\left.\frac{4\pi}{f^{\prime}(r)}\right|_{r=r_{\rm
B}}=\frac{4\pi l^{2}r_{\rm B}}{l^{2}+(2u+1)(r_{\rm B}^{2}-N^{2})},$ (19)
$\displaystyle E$ $\displaystyle=$
$\displaystyle\frac{(4\pi)^{u}\,u}{8\pi}\left(\sum_{k=0}^{u}\left(\begin{array}[]{l}u\\\
k\end{array}\right)\frac{(-1)^{k}N^{2k}r_{\rm B}^{2u-2k-1}}{2u-2k-1}\right.$
(25) $\displaystyle\hskip
14.22636pt+\left.\sum_{k=0}^{u+1}\left(\begin{array}[]{l}u+1\\\
k\end{array}\right)\frac{(-1)^{k}N^{2k}r_{\rm B}^{2u-2k-1}}{2u-2k-1}\right),$
$\displaystyle S_{\rm Bolt,AdS}$ $\displaystyle=$
$\displaystyle\frac{(4\pi)^{u}\beta}{16\pi
l^{2}}\left[\frac{(2u-1)(2u+1)(-1)^{u}N^{2u+2}}{r_{\rm B}}\right.$
$\displaystyle\hskip 28.45274pt+\sum_{k=0}^{u}\left(\begin{array}[]{l}u\\\
k\end{array}\right)(-1)^{k}N^{2k}r_{\rm B}^{2u-2k}$ $\displaystyle\hskip
48.36967pt\times\left(\frac{(2u-1)l^{2}}{(2u-2k-1)r_{\rm B}}\right.$
$\displaystyle+\left.\left.\frac{(2u+1)(2u^{2}+3u-2k+1)r_{\rm
B}}{(2u-2k+1)(u-k+1)}\right)\right],$
where $r_{\rm
B}=\frac{ql^{2}+\sqrt{q^{2}l^{4}+(2u+1)(2u+2)^{2}N^{2}[(2u+1)N^{2}-l^{2}]}}{(2u+1)(2u+2)N}$.
The CFT entropy is written as
$\displaystyle S_{\rm CFT}=\frac{2\pi l\sqrt{E_{\rm c}(2E-E_{\rm
c})}}{2u\sqrt{2u-1}},$ (30)
where $1/\sqrt{ab}$ is fixed to $1/2u\sqrt{2u-1}$. In the high temperature
limit, the CFT entropy well suffices to be Cardy-Verlinde formula as the
following
$\displaystyle S_{\rm
CFT}=\frac{(4\pi)^{u}}{4N^{2u}}\left(\frac{ql^{2}}{2u^{2}+3u+1}\right)^{2u}=S_{\rm
Bolt,AdS}.$ (31)
Note that the higher dimensional Taub-Bolt-AdS space follows the generalized
Cardy-Verlinde formula (5) even if Taub-Bolt-AdS4 space $(u=1)$ exactly
satisfies the Cardy-Verlinde formula (7) Birmingham:2001vd .
## III Taub-NUT/Bolt-dS black hole
Taub-NUT-dS metric is obtained from the Taub-NUT-AdS metric by replacing
$l^{2}\rightarrow-l^{2}$, and one has Clarkson:2003wa ; Clarkson:2003kt ;
Mann:2004mi
$\displaystyle ds^{2}_{\rm dS}$ $\displaystyle=$
$\displaystyle-g(r)\left[dt_{E}+2N\sum_{i=1}^{u}\cos(\theta_{i})d\phi_{i}\right]^{2}$
$\displaystyle-$
$\displaystyle\frac{dr^{2}}{g(r)}+(r^{2}-N^{2})\sum_{i=1}^{u}\Biggr{[}d\theta_{i}^{2}+\sin^{2}(\theta_{i})d\phi_{i}^{2}\Biggr{]},$
where $g(r)$ is given as
$\displaystyle g(r)$ $\displaystyle=$
$\displaystyle-\frac{r}{(r^{2}-N^{2})^{u}}\int^{r}\left[\frac{(p^{2}-N^{2})^{u}}{p^{2}}\right.$
(33) $\displaystyle-$
$\displaystyle\left.\frac{(2u+1)(p^{2}-N^{2})^{u+1}}{l^{2}p^{2}}\right]dp+\frac{2mr}{(r^{2}-N^{2})^{u}}.$
Using parallel way as in the previous case, the inverse of the temperature,
the total energy, the entropy, and the Casimir energy are obtained
$\displaystyle\beta=\left.\frac{4\pi}{g^{\prime}(r)}\right|_{r=N}=\frac{2(d+1)\pi|N|}{q},$
(34) $\displaystyle E$ $\displaystyle=$
$\displaystyle\frac{(4\pi)^{\frac{d}{2}}(d-1)N^{d-2}\biggr{(}(d+1)N^{2}+l^{2}\biggr{)}}{32\pi^{2}ql^{2}}$
(35) $\displaystyle\hskip
8.53581pt\times\Gamma(\frac{2-d}{2})\Gamma(\frac{d+1}{2}),$ $\displaystyle
S_{\rm NUT,AdS}$ $\displaystyle=$
$\displaystyle\frac{(4\pi)^{\frac{d}{2}}N^{d-2}\biggr{(}d(d-1)N^{2}+(d-2)l^{2}\biggr{)}}{32\pi^{2}l^{2}}$
(36) $\displaystyle\hskip
8.53581pt\times\Gamma(\frac{2-d}{2})\Gamma(\frac{d-1}{2})\beta,$
$\displaystyle E_{c}$ $\displaystyle=$
$\displaystyle\frac{(4\pi)^{\frac{d}{2}}(d-1)N^{d-2}\biggr{(}dN^{2}+l^{2}\biggr{)}}{16\pi^{2}ql^{2}}$
(37) $\displaystyle\hskip
8.53581pt\times\Gamma(\frac{2-d}{2})\Gamma(\frac{d+1}{2}).$
In this case, $1/\sqrt{ab}$ in (7) is fixed to $2/(d+1)(d-1)(d-2)$. Then, the
entropy in boundary CFT is expressed as
$\displaystyle S_{\rm CFT}$ $\displaystyle=$ $\displaystyle\frac{4\pi
l\sqrt{E_{c}(2E-E_{c})}}{(d+1)(d-1)(d-2)}\,,$ (38) $\displaystyle=$
$\displaystyle\frac{(4\pi)^{\frac{d}{2}}(dN^{2}+l^{2})(-1)^{[\frac{d}{2}]}\Gamma(\frac{d+1}{2})\Gamma(\frac{2-d}{2})}{4\pi(d+1)(d-2)q}.$
In high temperature limit, leading term in the entropy of the CFT for all even
$u$ is precisely matched with that in the entropy of the Taub-NUT-dS space as
the following
$\displaystyle S_{\rm CFT}$ $\displaystyle=$
$\displaystyle\frac{(4\pi)^{\frac{d}{2}}(d+1)(d-2)N^{d-1}\Gamma(\frac{4-d}{2})\Gamma(\frac{3+d}{2})}{16\pi^{\frac{3}{2}}q}$
(39) $\displaystyle=$ $\displaystyle S_{\rm NUT,dS}.$
For the Bolt solution in dS space, the inverse of the temperature, the total
energy, and the entropy are respectively
$\displaystyle\beta=\left.\frac{4\pi}{f^{\prime}(r)}\right|_{r=r_{\rm
B}}=-\frac{4\pi l^{2}r_{\rm B}}{l^{2}+(2u+1)(r_{\rm B}^{2}-N^{2})},$ (40)
$\displaystyle E$ $\displaystyle=$
$\displaystyle-\frac{(4\pi)^{u}\,u}{8\pi}\left(\sum_{k=0}^{u}\left(\begin{array}[]{l}u\\\
k\end{array}\right)\frac{(-1)^{k}N^{2k}r_{\rm B}^{2u-2k-1}}{2u-2k-1}\right.$
(46) $\displaystyle\hskip
14.22636pt+\left.\sum_{k=0}^{u+1}\left(\begin{array}[]{l}u+1\\\
k\end{array}\right)\frac{(-1)^{k}N^{2k}r_{\rm B}^{2u-2k-1}}{2u-2k-1}\right),$
$\displaystyle S_{\rm Bolt,dS}$ $\displaystyle=$
$\displaystyle\frac{(4\pi)^{u}\beta}{16\pi
l^{2}}\left[-\frac{(2u-1)(2u+1)(-1)^{u}N^{2u+2}}{r_{\rm B}}\right.$
$\displaystyle\hskip 28.45274pt+\sum_{k=0}^{u}\left(\begin{array}[]{l}u\\\
k\end{array}\right)(-1)^{k}N^{2k}r_{\rm B}^{2u-2k}$ $\displaystyle\hskip
48.36967pt\times\left(-\frac{(2u-1)l^{2}}{(2u-2k-1)r_{\rm B}}\right.$
$\displaystyle+\left.\left.\frac{(2u+1)(2u^{2}+3u-2k+1)r_{\rm
B}}{(2u-2k+1)(u-k+1)}\right)\right],$
where $r_{\rm
B}=\frac{ql^{4}+\sqrt{q^{2}l^{2}+(2u+1)(2u+2)^{2}N^{2}[(2u+1)N^{2}+l^{2}]}}{(2u+1)(2u+2)N}$.
The CFT entropy is given as
$\displaystyle\frac{2\pi l\sqrt{|E_{\rm c}|(2E-E_{\rm c})}}{2u\sqrt{2u-1}},$
(51)
where $1/\sqrt{ab}$ is fixed to $1/2u\sqrt{2u-1}$. As the NUT charge goes to
0, the CFT entropy becomes
$\displaystyle S_{\rm
CFT}=\frac{(4\pi)^{u}}{4N^{2u}}\left(\frac{ql^{2}}{2u^{2}+3u+1}\right)^{2u}=-S_{\rm
Bolt,dS},$ (52)
which shows that no entropy of the Taub-Bolt-dS metric satisfies the Cardy-
Verlinde formula. This means that any Bolt solution in dS space is
thermodynamically unstable at high temperature limit.
## IV Conclusion
We have considered that the Taub-NUT/Bolt-(A)dS metric in general even
dimension, and have checked that its metric suffices to be the the Cardy-
Verlinde formula. In the limit of high temperature, we showed that the Taub-
Bolt-AdS space well follows the generalized Cardy-Verlinde formula (7) rather
than the Cardy-Verlinde formula (5). It seems that the modification of the
standard Cardy-Verlinde formula (5) is due to the distinctive property of the
Taub-NUT solution such as the ALAdS metric. It was proven that the leading
term of the CFT entropy at the boundary for all even $u$ is exactly matched
with that of the entropy in the Taub-NUT-(A)dS space by using the generalized
Cardy-Verlinde formula at high temperature. Thermal stability of the Taub-
NUT-(A)dS solution for all odd $u$ and Taub-Bolt-dS solution for all $u$ is
determined by the magnitude of the NUT charge so that the negative entropy
occurs as the NUT charge goes to 0. Finally, the the breaking of Cardy-
Verlinde formula in the Taub-Bolt-dS metric seems to reflect the fact that
there is no Bolt solution in dS space due to the absence of hyperbolic NUT in
AdS space Mann:2004mi .
## Acknowledgements
We are grateful to Cristian Stelea for useful comments. This work was
supported by the BK 21 project of the Ministry of Education and Human
Resources Development, Korea (C.O.L.).
## References
* (1) G. ’t Hooft, Nucl. Phys. B 72, 461 (1974).
* (2) J. M. Maldacena, Adv. Theor. Math. Phys. 2, 231 (1998) [Int. J. Theor. Phys. 38, 1113 (1999)] [arXiv:hep-th/9711200];
* (3) S. S. Gubser, I. R. Klebanov and A. M. Polyakov, Phys. Lett. B 428, 105 (1998) [arXiv:hep-th/9802109];
* (4) E. Witten, Adv. Theor. Math. Phys. 2, 253 (1998) [arXiv:hep-th/9802150];
* (5) O. Aharony, S. S. Gubser, J. M. Maldacena, H. Ooguri and Y. Oz, Phys. Rept. 323, 183 (2000) [arXiv:hep-th/9905111].
* (6) E. D’Hoker and D. Z. Freedman, [arXiv:hep-th/0201253].
* (7) J. D. Brown and J. W. York, Phys. Rev. D 47, 1407 (1993).
* (8) V. Balasubramanian and P. Kraus, Commun. Math. Phys. 208, 413 (1999) [arXiv:hep-th/9902121].
* (9) R. Emparan, C. V. Johnson and R. C. Myers, Phys. Rev. D 60, 104001 (1999) [arXiv:hep-th/9903238].
* (10) P. Kraus, F. Larsen and R. Siebelink, Nucl. Phys. B 563, 259 (1999) [arXiv:hep-th/9906127].
* (11) J. L. Cardy, Nucl. Phys. B 270, 186 (1986).
* (12) E. P. Verlinde, arXiv:hep-th/0008140.
* (13) S. Nojiri, S. D. Odintsov and S. Ogushi, Int. J. Mod. Phys. A 16, 5085 (2001) [arXiv:hep-th/0105117].
* (14) S. Nojiri, S. D. Odintsov and S. Ogushi, Int. J. Mod. Phys. A 17, 4809 (2002) [arXiv:hep-th/0205187].
* (15) J. E. Lidsey, S. Nojiri, S. D. Odintsov and S. Ogushi, Phys. Lett. B 544, 337 (2002) [arXiv:hep-th/0207009].
* (16) R. G. Cai, Phys. Lett. B 525, 331 (2002) [arXiv:hep-th/0111093].
* (17) R. G. Cai, Phys. Rev. D 63, 124018 (2001) [arXiv:hep-th/0102113].
* (18) R. G. Cai, Nucl. Phys. B 628, 375 (2002) [arXiv:hep-th/0112253].
* (19) D. Klemm, A. C. Petkou and G. Siopsis, Nucl. Phys. B 601, 380 (2001) [arXiv:hep-th/0101076].
* (20) D. Birmingham and S. Mokhtari, Phys. Lett. B 508, 365 (2001) [arXiv:hep-th/0103108].
* (21) B. Wang, E. Abdalla and R. K. Su, Phys. Lett. B 503, 394 (2001) [arXiv:hep-th/0101073].
* (22) B. Wang, E. Abdalla and R. K. Su, Mod. Phys. Lett. A 17, 23 (2002) [arXiv:hep-th/0106086].
* (23) M. R. Setare, Mod. Phys. Lett. A 17, 2089 (2002) [arXiv:hep-th/0210187].
* (24) M. R. Setare and R. Mansouri, Int. J. Mod. Phys. A 18, 4443 (2003) [arXiv:hep-th/0210252].
* (25) M. R. Setare and E. C. Vagenas, Phys. Rev. D 68, 064014 (2003) [arXiv:hep-th/0304060].
* (26) G. W. Gibbons, M. J. Perry and C. N. Pope, Phys. Rev. D 72, 084028 (2005) [arXiv:hep-th/0506233].
* (27) A. Chamblin, R. Emparan, C. V. Johnson and R. C. Myers, Phys. Rev. D 59, 064010 (1999) [arXiv:hep-th/9808177].
* (28) S. W. Hawking, C. J. Hunter and D. N. Page, Phys. Rev. D 59, 044033 (1999) [arXiv:hep-th/9809035].
* (29) R. B. Mann, Phys. Rev. D 61, 084013 (2000) [arXiv:hep-th/9904148].
* (30) A. Awad and A. Chamblin, Class. Quant. Grav. 19, 2051 (2002) [arXiv:hep-th/0012240].
* (31) R. Clarkson, L. Fatibene and R. B. Mann, Nucl. Phys. B 652, 348 (2003) [arXiv:hep-th/0210280].
* (32) D. Astefanesei, R. B. Mann and E. Radu, Phys. Lett. B 620, 1 (2005) [arXiv:hep-th/0406050].
* (33) D. Astefanesei, R. B. Mann and E. Radu, JHEP 0501, 049 (2005) [arXiv:hep-th/0407110].
* (34) R. B. Mann and C. Stelea, Class. Quant. Grav. 21, 2937 (2004) [arXiv:hep-th/0312285].
* (35) R. B. Mann and C. Stelea, Phys. Lett. B 634, 448 (2006) [arXiv:hep-th/0508203].
* (36) C. Fefferman and C. Robin Graham, The mathematical heritage of Elie Cartan (Lyon, 1984). Asterisque 1985, Numero Hors Serie, 95-116.
* (37) R. Clarkson, A. M. Ghezelbash and R. B. Mann, Phys. Rev. Lett. 91, 061301 (2003) [arXiv:hep-th/0304097].
* (38) R. Clarkson, A. M. Ghezelbash and R. B. Mann, Nucl. Phys. B 674, 329 (2003) [arXiv:hep-th/0307059].
* (39) R. B. Mann and C. Stelea, Phys. Rev. D 72, 084032 (2005) [arXiv:hep-th/0408234].
| arxiv-papers | 2008-07-17T02:28:59 | 2024-09-04T02:48:56.854094 | {
"license": "Public Domain",
"authors": "Chong OH Lee",
"submitter": "Chongoh Lee",
"url": "https://arxiv.org/abs/0807.2685"
} |
0807.2746 | # Suppression of Weiss oscillations in the magnetoconductivity of modulated
graphene monolayer
M. Tahir∗ Department of Physics, University of Sargodha, Sargodha 40100,
Pakistan K. Sabeeh† Department of Physics, Quaid-i-Azam University, Islamabad
45320, Pakistan
([; date; date; date; date)
###### Abstract
We have investigated the electrical transport properties of Dirac electrons in
a monolayer graphene sheet in the presence of both electric and magnetic
modulations. The effects of the modulations on quantum transport when they are
in phase and out of phase are considered. We present the energy spectrum and
the bandwidth of the Dirac electrons in the presence of both the modulations.
We determine the $\sigma_{yy}$ component of the magnetoconductivity tensor for
this system which is shown to exhibit Weiss oscillations.Asymptotic
expressions for $\sigma_{yy}$ are also calculated to better illustrate the
effects of in-phase and out-of-phase modulations.We find that the position of
the oscillations in magnetoconductivity depends on the relative strength of
the two modulations. When the two modulations are out-of-phase there is
complete suppression of Weiss oscillations for particular relative strength of
the modulations.
one two three
###### pacs:
PACS number
††preprint:
year number number identifier Date text]date
LABEL:FirstPage1 LABEL:LastPage#12
## I Introduction
The recent successful preparation of monolayer graphene has generated a lot of
interest in the physics community and efforts are underway to study the
electronic properties of graphene 1 . The nature of quasiparticles called
Dirac electrons in these two-dimensional systems is very different from those
of the conventional two-dimensional electron gas (2DEG) realized in
semiconductor heterostructures. Graphene has a honeycomb lattice of carbon
atoms. The quasiparticles in graphene have a band structure in which electron
and hole bands touch at two points in the Brillouin zone. At these Dirac
points the quasiparticles obey the massless Dirac equation. In other words,
they behave as massless Dirac particles leading to a linear dispersion
relation $\epsilon_{k}=vk$ ( with the characteristic velocity $v\simeq
10^{6}m/s)$. This difference in the nature of the quasiparticles in graphene
from conventional 2DEG has given rise to a host of new and unusual phenomena
such as anomalous quantum Hall effects and a $\pi$ Berry phase1 2 . Among the
electronic properties of interest is the interaction of electrons with
artificially created periodic potentials. It has been observed that if
conventional 2DEG is subjected to artificially created periodic potentials in
the submicrometer range it leads to the appearance of Weiss oscillations in
the magnetoresistance. This type of electrical modulation of the 2D system can
be carried out by depositing an array of parallel metallic strips on the
surface or through two interfering laser beams 3 ; 4 ; 5 . Besides the
fundamental interest in understanding the electronic properties of graphene
there is also serious suggestions that it can serve as the building block for
nanoelectronic devices 6 .
In conventional 2DEG systems, electron transport in the presence of electric
and magnetic modulation has continued to be an active area of research 7 ; 8 ;
9 . Recently, electrical transport in graphene monolayer in the presence of
electrical modulation was considered and theoretical predictions made 10 . We
have also carried out a study of magnetoconductivity when graphene monolayer
is subjected to magnetic modulation alone11 . Along the same lines, in this
work we investigate low temperature magnetotransport of Dirac electrons in a
graphene monolayer subjected to both electric and magnetic modulations with
the same period. From a practical point of view, this is important to consider
as magnetic modulation of graphene can be realized by magnetic or
superconducting stripes placed on top of graphene which in turn act as
electrical gates that induce electric modulation. The relative phase of the
two modulations can have important consequences for magnetotransport in the
system as was seen in conventional 2DEG. Therefore in this work we investigate
Weiss oscillations in magnetoconductivity $\sigma_{yy}$ in a graphene
monolayer for both the cases when the electric and magnetic modulations are
in-phase and when they are out of phase.
## II Energy Spectrum and Bandwidth
We consider two-dimensional Dirac electrons in graphene moving in the
$x-y-$plane. The magnetic filed ($B$) is applied along the $z-$direction
perpendicular to the graphene plane. We consider the perpendicular magnetic
field $B$ modulated weakly and periodically along one direction such that
$\mathbf{B}=(B+B_{0}\cos(Kx))\mathbf{z}$. Here $B_{0}$ is the strength of the
magnetic modulation and $K=2\pi/a$ with $a$ being the modulation period. We
consider the modulation to be weak such that $B_{0}<<B$. We use the Landau
gauge for vector potential $\mathbf{A}=(0,Bx,0)$. In effective mass
approximation the one electron Hamiltonian is
$H=v\overleftrightarrow{\sigma}.(-i\hbar\mathbf{\nabla}+e\mathbf{A}).$ The low
energy excitations are described by the two-dimensional (2D) Dirac like
Hamiltonian 1 ; 2 ; 10
$H=v\overleftrightarrow{\sigma}.(-i\hbar\mathbf{\nabla}+e\mathbf{A}).$Here
$\overleftrightarrow{\sigma}=\\{\overleftrightarrow{\sigma}_{x},\overleftrightarrow{\sigma}_{y}\\}$
are the Pauli matrices and $v$ characterizes the electron velocity. We employ
the Landau gauge and write the vector potential as
$\mathbf{A}=(0,Bx+(B_{0}/K)\sin(Kx),0)$. The Hamiltonian can be written as
$H=H_{0}+H_{,}^{\prime}$where $H_{0}$ is the unmodulated Hamiltonian given as
$H_{0}=-i\hbar
v\overleftrightarrow{\sigma}.\mathbf{\nabla}+ev\overleftrightarrow{\sigma_{y}}Bx$
and $H^{\prime}=ev\overleftrightarrow{\sigma_{y}}\frac{B_{0}}{K}\sin(Kx).$The
Landau level energy eigenvalues without modulation are given by
$\varepsilon(n)=\hbar\omega_{g}\sqrt{n}$ (1)
where $n$ is an integer and $\omega_{g}=v\sqrt{2eB/\hbar}.$ As has been
pointed out 10 the Landau level spectrum for Dirac electrons is significantly
different from the spectrum for electrons in conventional 2DEG which is given
as $\varepsilon(n)=\hbar\omega_{c}(n+1/2)$, where $\omega_{c}=eB/m$ is the
cyclotron frequency.
The eigenfunctions without modulation are given by
$\Psi_{n,k_{y}}(r)=\frac{e^{ik_{y}y}}{\sqrt{2L_{y}}}\left(\begin{array}[c]{c}-i\varphi_{n-1}(x,x_{0})\\\
\varphi_{n}(x,x_{0})\end{array}\right)$ (2)
where
$\varphi_{n}(x,x_{0})=\frac{e^{-(x+x_{0})/2l}}{\sqrt{2^{n}n!l\sqrt{\pi}}}H_{n}(\frac{x+x_{0}}{l})$
are the normalized harmonic oscillator eigenfunctions$,l=\sqrt{\hbar/eB}$ is
the magnetic length, $x_{0}=l^{2}k_{y},$ $L_{y}$ is the $y$-dimension of the
graphene layer and $H_{n}(x)$ are the Hermite polynomials.
Since we are considering weak modulation $B_{0}<<$ $B$, we can apply standard
perturbation theory to determine the first order corrections to the
unmodulated energy eigenvalues in the presence of modulation with the result
$\Delta\varepsilon_{{}_{n,k_{y}}}=\omega_{0}\cos(Kx_{0})\left(\sqrt{\frac{2n}{u}}e^{-u/2}[L_{n-1}(u)-L_{n}(u)]\right)$where
$\omega_{0}=\frac{evB_{0}}{K}$, $u=K^{2}l^{2}/2$ and $L_{n}(u)$ are the
Laguerre polynomials. Hence the energy eigenvalues in the presence of the
periodic magnetic modulation are
$\varepsilon(n,x_{0})=\varepsilon(n)+\Delta\varepsilon_{{}_{n,k_{y}}}=\hbar\omega_{g}\sqrt{n}+\omega_{0}\cos(Kx_{0})G_{n}$
(3)
with $G_{n}(u)=\sqrt{\frac{2n}{u}}e^{-u/2}[L_{n-1}(u)-L_{n}(u)]$. We observe
that the degeneracy of the Landau level spectrum of the unmodulated system
with respect to $k_{y}$ is lifted in the presence of modulation with the
explicit presence of $k_{y}$ in $x_{0}.$ The $n=0$ landau level is different
from the rest as the energy of this level is zero and electrons in this level
do not contribute to diffusive conductivity calculated in the next section.
The rest of the Landau levels broaden into bands. The Landau bandwidths $\sim
G_{n}$ oscillates as a function of $n$ since $L_{n}(u)$ are oscillatory
functions of the index $n$.
Since we are interested in electron transport in the presence of both electric
and magnetic modulations, we consider an additional weak electric modulation
potential given as $V(x)=V_{0}\cos(Kx)$ on the system. Here $V_{0}$ is the
amplitude of modulation. We can determine the energy eigenvalues in the
presence of weak electric modulation where we take $V_{o}$ to be an order of
magnitude smaller than the Fermi Energy $\varepsilon_{F}=v_{F}\hslash k_{F}\
$with $k_{F}=\sqrt{2\pi n_{e}}$ is the magnitude of Fermi wave vector with
$n_{e}$ being the electron concentration. Hence we can apply standard first
order perturbation theory to determine the energy eigenvalues in the presence
of modulation. The first order correction in the energy eigenvalues when
electric modulation is present is given as
$\varepsilon(n,x_{0})=\varepsilon(n)+V_{0}F_{n}\cos(Kx_{0})$ (4)
Here, $\ F_{n}=\frac{1}{2}\exp(-\frac{u}{2})[L_{n}(u)+L_{n-1}(u)]$,
$u=\frac{K^{2}l^{2}}{2}$ and, $L_{n}(u)$ and $L_{n-1}(u)\ $are Laguerre
polynomials.
The energy eigenvalues of Dirac electrons in the presence of both modulations
can be expressed as
$\varepsilon(n,k_{y})=\hbar\omega_{g}\sqrt{n}+\omega_{0}\cos(Kx_{0})G_{n}+V_{0}F_{n}\cos(Kx_{0}).$
(5)
To better appreciate the modulation effects on the Landau levels we determine
the asymptotic expression for the bandwidth ($\Delta$) next. The width of the
$n$th Landau level in the presence of periodic electric and magnetic
modulation is given as $\Delta=\Delta_{B}+\Delta_{E},$ where $\Delta_{E}$ is
width of the electric modulation and $\Delta_{B}$ is the width of the magnetic
modulation:
$\Delta_{B}=2\left|G_{n}\right|=2\sqrt{\frac{2n}{u}}e^{-u/2}[L_{n-1}(u)-L_{n}(u)]$
(6)
The asymptotic expression of bandwidth can be obtained by using the following
asymptotic expression for the Laguerre polynomials by taking the large $n$
($n_{F}=\frac{\varepsilon_{F}^{2}}{\varepsilon^{2}(n)}=\frac{\hbar
K_{F}^{2}}{2eB}$) limit as
$exp^{-u/2}L_{n}(u)\rightarrow\frac{1}{\sqrt{\pi\sqrt{nu}}}\cos(2\sqrt{nu}-\frac{\pi}{4})$
with the result that the asymptotic expression for $\Delta_{B}$ is
$\Delta_{B}=\frac{8\omega_{0}}{Kl}\sqrt{\frac{\hbar
K_{F}^{2}}{2eB\pi^{2}R_{g}}}\sin\left(\frac{2eB\pi R_{g}}{2a\hbar
K_{F}^{2}}\right)\sin\left(\frac{2\pi R_{g}}{a}-\frac{\pi}{4}\right).$ (7)
Similarly, for electric modulation, the bandwidth $\Delta_{E}$ is given as
$\Delta_{E}=2\left|F_{N}\right|=V_{0}\exp^{-\frac{u}{2}}\left|L_{n}(u)+L_{n-1}(u)\right|\,.$
and the asymptotic expression for $\Delta_{E}$ is
$\Delta_{E}=V_{0}\left(\frac{a}{\pi^{2}R_{g}}\right)^{\frac{1}{2}}\cos\left(\frac{2eB\pi
R_{g}}{2a\hbar K_{F}^{2}}\right)\cos\left(\frac{2\pi
R_{g}}{a}-\frac{\pi}{4}\right).$ (8)
Therefore the bandwidth in the presence of both electric and magnetic
modulations can be expressed as
$\displaystyle\Delta$ $\displaystyle=\frac{8\omega_{0}}{Kl}\sqrt{\frac{\hbar
K_{F}^{2}}{2eB\pi^{2}R_{g}}}\sin\left(\frac{2eB\pi R_{g}}{2a\hbar
K_{F}^{2}}\right)\sin\left(\frac{2\pi R_{g}}{a}-\frac{\pi}{4}\right)+$
$\displaystyle
V_{0}\left(\frac{a}{\pi^{2}R_{g}}\right)^{\frac{1}{2}}\cos\left(\frac{2eB\pi
R_{g}}{2a\hbar K_{F}^{2}}\right)\cos\left(\frac{2\pi
R_{g}}{a}-\frac{\pi}{4}\right).$ (9)
In Fig.(1) we present the bandwidths as a function of the magnetic field for
temperature $T=6K$, electron density $n_{e}=3\times 10^{11}cm^{-2},$ the
period of modulation $a=350nm.$The strength of the electric modulation
$V_{0}=0.2meV$ where as $B_{0}=0.004T$ which corresponds to
$\omega_{0}=0.2meV$ with the result that both the modulations have equal
strengths. In the same figure we have also shown the bandwidths when either
the magnetic or electric modulation alone is present. We observe that electric
and magnetic bandwidths are out of phase while the positions of the extrema of
combined bandwidth are shifted with respect to electric and magnetic
bandwidths. The combined bandwidth when both modulations are present will
affect the conductivity and that is considered in the next section.
## III Magnetoconductivity with Periodic Electric and Magnetic Modulation:
In-phase
To calculate the electrical conductivity in the presence of weak electric and
magnetic modulations we use Kubo formula to calculate the linear response to
applied external fields. In a magnetic field, the main contribution to Weiss
oscillations comes from the scattering induced migration of the Larmor circle
center. This is diffusive conductivity and we shall determine it following the
approach in 7 ; 10 ; 11 ; 12 where it was shown that the diagonal component
of conductivity $\sigma_{yy}$ can be calculated by the following expression in
the case of quasielastic scattering of electrons
$\sigma_{yy}=\frac{\beta
e^{2}}{L_{x}L_{y}}\underset{\zeta}{{\displaystyle\sum}}f(E_{\zeta})[1-f(E_{\zeta})]\tau(E_{\zeta})(\upsilon_{y}^{\zeta})^{2}$
(10)
$L_{x}$, $L_{y}$, are the dimensions of the layer, $\beta=\frac{1}{k_{B}T}$ is
the inverse temperature with $k_{B}$ the Boltzmann constant, $f(E)$ is the
Fermi Dirac distribution function, $\tau(E)$ is the electron relaxation time
and $\zeta$ denotes the quantum numbers of the electron eigenstate. The
diagonal component of the conductivity $\sigma_{yy}$ is due to modulation
induced broadening of Landau bands and hence it carries the effects of
modulation in which we are primarily interested in this work. $\sigma_{xx}$
does not contribute as the component of velocity in the $x$-direction is zero
here. The collisional contribution due to impurities is not taken into account
in this work.
The summation in Eq.(10) over the quantum numbers $\zeta$ can be written as
$\frac{1}{L_{x}L_{y}}\underset{\zeta}{{\displaystyle\sum}}=\frac{1}{2\pi
L_{x}}{\displaystyle\int\limits_{0}^{\frac{L_{x}}{l^{2}}}}dk_{y}\underset{n=0}{\overset{\infty}{{\displaystyle\sum}}}=\frac{1}{2\pi
l^{2}}\underset{n=0}{\overset{\infty}{{\displaystyle\sum}}}$ (11)
The component of velocity required in Eq.(10) can be calculated from the
following expression
$\upsilon_{y}^{\zeta}=-\frac{\partial}{\hbar\partial
k_{y}}\varepsilon(n,k_{y}).$ (12)
Substituting the expression for $\varepsilon(n,k_{y})$ obtained in Eq.(5) into
Eq.(12) yields
$\upsilon_{y}^{\zeta}=\left[\frac{2\omega_{0}u}{\hbar
K}\sin(Kx_{0})G_{n}(u)+\frac{2V_{0}u}{\hbar K}\sin(Kx_{0})F_{n}(u)\right]$
(13)
As a result $\upsilon_{y}^{\zeta},$ the corresponding velocity given by
Eq.(13) contains the contribution from both the modulations (electric and
magnetic) obtained in Eq.(5) compared to one term in velocity component for
each10 ; 11 . This term ($\upsilon_{y}^{\zeta}$) has important consequences
for the quantum transport phenomena in modulated systems.
With the results obtained in Eqs.(11), (12) and (13) we can express the
diffusive contribution to the conductivity given by Eq.(10) as
$\sigma_{yy}=A_{0}\phi$ (14)
where
$A_{0}=\frac{e^{2}\tau\beta}{\pi\hbar^{2}}$ (15)
and $\phi$ is given as
$\phi={\displaystyle\sum_{n=0}^{\infty}}\frac{g(\varepsilon_{n})}{[g(\varepsilon_{n})+1)]^{2}}\left[\frac{\sqrt{u}e^{-u/2}V_{0}}{2}(L_{n}(u)+L_{n-1}(u))+\omega_{0}e^{-u/2}\sqrt{2n}(L_{n-1}(u)-L_{n}(u))\right]^{2}$
(16)
where $g(\varepsilon)=\exp[\beta(\varepsilon-\varepsilon_{F}]$ and
$\varepsilon_{F}$ is the Fermi energy.
In Fig.(2) we show the in-phase conductivity (the magnetic and the electric
modulations are in-phase) $\sigma_{yy}$ given by Eqs(14,15,16) as a function
of the inverse magnetic field for temperature $T=6K$, electron density
$n_{e}=3\times 10^{11}cm^{-2},$ the period of modulation $a=350nm.$ The
dimensionless magnetic field $\frac{B^{\prime}}{B}$ is introduced where
$B^{\prime}=\frac{\hslash}{ea^{2}}=0.0054T$ for $a=350nm.$ The strength of the
electric modulation $V_{0}=0.2meV$ where as $B_{0}=0.004T$ which corresponds
to $\omega_{0}=0.2meV$ with the result that both the modulations have equal
strengths. In the same figure we have also shown the conductivity when either
the magnetic or electric modulation alone is present. The $\frac{\pi}{2}$
phase difference in the bandwidths results in the same phase difference
appearing in the conductivity for electric and magnetic modulations as can be
seen in the figure. To better understand the effects of in-phase modulations
on the conductivity we consider the asymptotic expression of the quantity
$\phi$ given by Eq.(16) that appears in the magnetoconductivity $\sigma_{yy}$.
The asymptotic results are valid when applied magnetic field is weak such that
many Landau levels are filled. The asymptotic expression is obtained in the
next section.
## IV Asymptotic Expressions: In-Phase Modulations
To get a better understanding of the results of the previous section we will
consider the asymptotic expression of conductivity where analytic results in
terms of elementary functions can be obtained following 7 ; 10 ; 11 . The
asymptotic expression of $\phi$ can be obtained by employing the following
asymptotic expression for the Laguerre polynomials which is valid in the limit
of large $n$ when many landau levels are filled 13
$\exp^{-u/2}L_{n}(u)\rightarrow\frac{1}{\sqrt{\pi\sqrt{nu}}}\cos(2\sqrt{nu}-\frac{\pi}{4})$
(17)
with the result that the in-phase bandwidth can be written as
$\Delta(\text{in-
phase})=\frac{4\omega_{0}\times\sqrt{\frac{2n}{u}}\times\sin\left(1/2\sqrt{\frac{u}{n}}\right)\times\sqrt{1+\delta^{2}}}{\sqrt{\pi\sqrt{nu}}}\times\sin\left(2\sqrt{nu}-\frac{\pi}{4}+\Phi\right)$
where the ratio between the two modulation strengths
$\delta=\frac{V_{0}\cos\left(1/2\sqrt{\frac{u}{n}}\right)}{2\omega_{0}\sqrt{\frac{2n}{u}}\sin\left(1/2\sqrt{\frac{u}{n}}\right)}=\tan(\Phi).$
The flat band condition from the above equation is
$2\sqrt{nu}-\frac{\pi}{4}+\Phi=i\pi$ where $i$ is an integer. This condition
can also be expressed as
$\frac{\sqrt{2n}}{a}l=i+\frac{1}{4}-\frac{\Phi}{\pi}$, where
$n=n_{F}=\frac{\varepsilon_{F}^{2}}{\varepsilon^{2}(n)}=\frac{\varepsilon_{F}^{2}}{\hslash^{2}\omega_{g}^{2}}$
is the highest Fermi integer. We see that the flat band condition in this case
depends on the relative strength of the two modulations. We now take the
continuum limit:
$n-->\frac{1}{2}\left(\frac{l\varepsilon}{v\hbar}\right)^{2},\overset{\infty}{\underset{n=0}{{\displaystyle\sum}}}-->\left(\frac{l}{v\hbar}\right)^{2}{\displaystyle\int\limits_{0}^{\infty}}\varepsilon
d\varepsilon$ (18)
to express $\phi$ in Eq.(16) as the following integral
$\displaystyle\phi$
$\displaystyle=\frac{4\omega_{0}^{2}\times\frac{2\varepsilon_{F}^{2}}{u\hslash^{2}\omega_{g}^{2}}\times\sin^{2}\left(1/2\sqrt{\frac{u\hslash^{2}\omega_{g}^{2}}{\varepsilon_{F}^{2}}}\right)\times(1+\delta^{2})}{\pi\sqrt{u}}\times\left(\frac{l}{v\hbar}\right)^{2}\times$
$\displaystyle{\displaystyle\int\limits_{0}^{\infty}}\frac{\varepsilon}{\sqrt{n}}d\varepsilon\frac{g(\varepsilon)}{[g(\varepsilon)+1)]^{2}}\sin^{2}(1/2\sqrt{u/n})\sin^{2}(2\sqrt{nu}-\frac{\pi}{4}+\Phi)$
(19)
where $u=2\pi^{2}/b$ with $b=\frac{eBa^{2}}{\hbar}=\frac{B}{B^{\prime}}$ and
$B^{\prime}=\frac{ea^{2}}{\hslash}$.
Now assuming that the temperature is low such that
$\beta^{-1}\ll\varepsilon_{F}$ and replacing
$\varepsilon=\varepsilon_{F}+s\beta^{-1}$, we rewrite the above integral as
$\displaystyle\phi$
$\displaystyle=\frac{4\omega_{0}^{2}\times\frac{2\varepsilon_{F}^{2}}{u\hslash^{2}\omega_{g}^{2}}\times\sin^{2}\left(1/2\sqrt{\frac{u\hslash^{2}\omega_{g}^{2}}{\varepsilon_{F}^{2}}}\right)\times(1+\delta^{2})}{\pi\sqrt{u}\beta}\times\left(\frac{l}{v\hbar}\right)^{2}\times$
$\displaystyle{\displaystyle\int\limits_{-\infty}^{\infty}}\frac{dse^{s}}{(e^{s}+1)^{2}}\sin^{2}(\frac{2\pi
p}{b}-\frac{\pi}{4}+\Phi+\frac{\sqrt{2u}}{v\sqrt{B\hbar}\beta}s)$ (20)
where $p=\frac{\varepsilon_{F}a}{\hbar v}=k_{F}a=\sqrt{2\pi n_{s}}a$ is the
dimensionless Fermi momentum of the electron. To obtain an analytic solution
we have also replaced $\varepsilon$ by $\varepsilon_{F}$ in the above integral
except in the sine term in the integrand.
The above expression can be written as
$\displaystyle\phi$
$\displaystyle=\frac{4\omega_{0}^{2}\times\frac{2\varepsilon_{F}^{2}}{u\hslash^{2}\omega_{g}^{2}}\times\sin^{2}\left(1/2\sqrt{\frac{u\hslash^{2}\omega_{g}^{2}}{\varepsilon_{F}^{2}}}\right)\times(1+\delta^{2})}{\pi\sqrt{u}\beta}\times\left(\frac{l}{v\hbar}\right)^{2}\times$
$\displaystyle{\displaystyle\int\limits_{-\infty}^{\infty}}\frac{ds}{\cosh^{2}(s/2)}\sin^{2}(\frac{2\pi
p}{b}-\frac{\pi}{4}+\Phi+\frac{2\pi a}{vb\beta}s)$ (21)
The above integration can be performed by using the following identity 13 :
${\displaystyle\int\limits_{0}^{\infty}}dx\frac{\cos ax}{\cosh^{2}\beta
x}=\frac{a\pi}{2\beta^{2}\sinh(a\pi/2\beta)}$ (22)
with the result
$\displaystyle\phi$
$\displaystyle=\frac{2\omega_{0}^{2}\times\frac{2\varepsilon_{F}^{2}}{u\hslash^{2}\omega_{g}^{2}}}{4\pi^{2}}\times\frac{T}{T_{D}}\times\sin^{2}\left(\frac{\pi}{p}\right)\times(1+\delta^{2})\times$
$\displaystyle\left[1-A\left(\frac{T}{T_{D}}\right)+2A\left(\frac{T}{T_{D}}\right)\sin^{2}\left[2\pi\left(\frac{p}{b}-\frac{1}{8}+\frac{\Phi}{2\pi}\right)\right]\right]$
(23)
where $k_{B}T_{D}=\frac{\hbar vb}{4\pi^{2}a},$
$\frac{T}{T_{D}}=\frac{4\pi^{2}a}{\hbar vb\beta}$and
$A(x)=\frac{x}{\sinh(x)}-^{(x-->\infty)}->=2xe^{-x}.$
From the asymptotic expression of $\phi$ given by Eq.(23), we observe that the
effect of the in-phase electric and magnetic modulations is the appearance of
a phase factor $\Phi$ in the conductivity. The shift in the Weiss oscillations
when in-phase electric and magnetic modulations are present can be seen in
Fig.(3). The phase factor $\Phi$ depends on the relative strength of the two
modulations. How the Weiss oscillations are affected as $\Phi$ as well as the
magnetic field is varied can be seen in Fig.(3). The results shown are for a
fixed magnetic modulation of strength $\omega_{0}=0.2meV$ and the electric
modulation is varied. The change in $V_{0}$ results in a corresponding change
in both $\delta$ and $\Phi$. From Fig.(3), we observe that the position of the
extrema in magnetoconductivity as a function of the inverse magnetic field
depends on the relative strength of the modulations.
The effects of electric and magnetic modulations that are out-of-phase on the
conductivity can be better appreciated if we consider the asymptotic
expression for $\phi$ in this case. This is taken up in the next section.
## V Magnetoconductivity with Periodic Electric and Magnetic Modulation: Out-
of-phase
In this section, we calculate $\phi$ when electric and magnetic modulations
are out of phase by $\pi/2.$ We consider magnetic modulation out of phase with
the electric one: We take the electric modulation to have the same phase as
given in the previous section with the $\pi/2$ phase difference incorporated
in the magnetic field. The energy eigenvalues are
$\varepsilon(n,k_{y})=\hbar\omega_{g}\sqrt{n}+\omega_{0}\sin(Kx_{0})G_{n}+V_{0}F_{n}\cos(Kx_{o}),$
(24)
and the bandwidth is
$\Delta(\text{out of
phase})=\frac{4\omega_{0}\times\sqrt{\frac{2n}{u}}\times\sin\left(1/2\sqrt{\frac{u}{n}}\right)}{\sqrt{\pi\sqrt{nu}}}\times\sqrt{\delta^{2}+(1-\delta^{2})\sin\left(2\sqrt{nu}-\frac{\pi}{4}\right)}\,.$
(25)
The term responsible for Weiss oscillations is the
$\sin\left(2\sqrt{nu}-\frac{\pi}{4}\right)$ term under the square root which
can be readily seen by considering the large $n$ limit of the bandwidth.
Therefore for $\delta=\pm 1$ Weiss oscillations are no longer present in the
bandwidth. The velocity component $\upsilon_{y}$ is given as
$\upsilon_{y}=-\left[\frac{2\omega_{0}u}{\hbar
K}\cos(Kx_{0})G_{n}(u)-\frac{2V_{0}u}{\hbar K}\sin(Kx_{0})F_{n}(u)\right]$
(26)
and $\phi$ is given as
$\phi={\displaystyle\sum_{n=0}^{\infty}}\frac{g(\varepsilon_{n})}{[g(\varepsilon_{n})+1)]^{2}}\left[\frac{ue^{-u}V_{0}^{2}}{4}(L_{n}(u)+L_{n-1}(u))^{2}+\omega_{0}^{2}e^{-u}2n(L_{n-1}(u)-L_{n}(u))^{2}\right].$
(27)
The asymptotic expression for $\phi$ in the presence of both electric and
magnetic modulations that are out of phase is obtained by substituting the
asymptotic expressions for the Laguerre polynomials and converting the sum
into integration with the result
$\displaystyle\phi$
$\displaystyle=\frac{2\omega_{0}^{2}p^{2}}{\pi^{2}}\sin^{2}\left(\frac{\pi}{p}\right)\frac{T}{4\pi^{2}T_{D}}\times$
$\displaystyle\left[2\delta^{2}+(1-\delta^{2})\left(1-A\left(\frac{T}{T_{D}}\right)+2A\left(\frac{T}{T_{D}}\right)\sin^{2}\left[2\pi\left(\frac{p}{b}-\frac{1}{8}\right)\right]\right)\right].$
(28)
From the expression of the out-of-phase bandwidth given by Eq.(25) we find
that Weiss oscillations in the bandwidth are absent for relative modulation
strength $\delta=\pm 1$, the same is reflected in $\phi$ as the term
responsible for Weiss oscillations
($\sin^{2}\left[2\pi\left(\frac{p}{b}-\frac{1}{8}\right)\right]$) vanishes for
$\delta=\pm 1$ as can be seen from the above equation. Therefore the
magnetoconductivity $\sigma_{yy}$ does not exhibit Weiss oscillations when the
relative modulation strength $\delta=\pm 1.$The conductivity as a function of
magnetic field when the electric and magnetic modulations are out-of-phase is
shown in Fig.(4). The results shown are for a fixed magnetic modulation of
strength $\omega_{0}=0.2meV$ and the electric modulation $V_{0}$ is allowed to
vary between positive and negative values. The other parameters are the same
as in Figs.(1,2,3). As $V_{0}$ is varied there is a corresponding change in
$\delta.$ We find that the positions of the extrema of $\sigma_{yy}$ as a
function of the inverse magnetic field do not change as $\delta$ is varied
since the phase factor $\Phi$ does not appear in the expression of
conductivity when the two modulations are out of phase. It is also observed in
Fig.(4) that there is a $\frac{\pi}{2}$ phase difference between the curves
for $\delta\geq 1$ and $\delta<1.$ The same behavior is observed in the
bandwidth which is reflected in the magnetoconductivity. In this work we have
considered Weiss oscillations and have not taken pure Shubnikov de Hass (SdH)
oscillations into account but SdH oscillations do appear superimposed on Weiss
oscillations in the region of strong magnetic field as can been seen in all of
our figures. In addition, since this work was motivated by7 , it is important
to compare the results obtained here for graphene monolayer with those
presented for standard 2DEG. In contrast to the work presented here, in 7 SdH
are explicitly taken into account but we find that our principal expressions
Eq.(23), (28) reduce to the expressions there when the terms contributing to
SdH oscillations are ignored.
In the end, we would also like to mention relevance of this work in
experimental studies of transport in graphene. Experiments on graphene
monolayer in the presence of modulated electric and magnetic fields have not
been realized yet, this theoretical work is in anticipation of experimental
work. We expect that modulation effects predicted here can be observed in
graphene employing established techniques that were used for the two-
dimensional electron gas systems found in semiconductor heterostructures 14 .
In conventional semiconductor systems, modulation of the potential seen by
electrons can be produced by molecular beam epitaxy, chemical vapor deposition
as well as sputtering techniques 15 . In graphene, we expect that modulation
effects can be introduced by adsorbing adatoms on graphene surface using
similar techniques, by positioning and aligning impurities with scanning
tunneling microscopy or by applying top gates to graphene. Epitaxial growth of
graphene on a patterned substrate is also possible 16 .
In conclusion, we have determined the effects of both the electric and
magnetic modulations on the magnetoconductivity of a graphene monolayer.
Appearance of Weiss oscillations in the magnetoconductivity $\sigma_{yy}$ is
the main focus of this work. These oscillations are affected by the relative
phase of the two modulations and position of the oscillations depends on the
relative strength of the two modulations. We find complete suppression of
Weiss oscillations for particular relative strength of the modulations when
the modulations are out-of-phase.
## VI Acknowledgements
We gratefully acknowledge F. M. Peeters for suggesting the problem and for
comments on the manuscript. One of us (K.S.) would like to acknowledge the
support of the Pakistan Science Foundation (PSF) through project No. C-QU/Phys
(129). M. T. would like to acknowledge the support of the Pakistan Higher
Education Commission (HEC).
$\ast$Department of Physics, Blackett Laboratory, Imperial College London,
South Kensington Campus, London SW7 2AZ, United Kingdom.; Electronic address:
m.tahir06@imperial.ac.uk
$\dagger$Electronic address: ksabeeh@qau.edu.pk; kashifsabeeh@hotmail.com
## References
* (1) K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, M. I. Katsnelson, I. V. Grigorieva, S. V. Dubonos, and A. A. Firsov, Nature 438, 197 (2005); Y. Zhang, Y. W. Tan, H. L. Stormer, and P. Kim, 438, 201 (2005).
* (2) Y. Zheng and T. Ando, Phys. Rev. B 65, 245420 (2002); V. P. Gusynin and S. G. Sharapov, Phys. Rev. Lett. 95, 146801 (2005); N. M. R. Perez F. Guinea, and A. H. Castro Neto, Phys. Rev. B 73, 125411 (2006); M. I. Katsnelson , K. S. Novoselov, and A. K. Geim, Nat. Phys. 2, 620 (2006); K. S. Novoselov, E. McCann, S. V. Morozov, V. I. Fal’ko, M. I. Katsnelson, U. Zeitler, D. Jiang, F. Schedin, and A. K. Geim, Nat. Phys. 2, 177 (2006).
* (3) D. Weiss, K. v. Klitzing, K. Ploog, and G. Weimann, Europhys. Lett., 8, 179 (1989)
* (4) R. W. Winkler, J. P. Kotthaus, and K. Ploog, Phys. Rev. Lett. 62, 1177 (1989).
* (5) R. R. Gerhardts, D. Weiss, and K. v. Klitzing, Phys. Rev. Lett. 62, 1173 (1989).
* (6) C. Berger,et.al, Science 312, 1191 (2006).
* (7) F. M. Peeters and P. Vasilopoulos, Phys. Rev. B 47, 1466 (1993)
* (8) F. M. Peeters and P. Vasilopoulos, Phys. Rev. B 46, 4667 (1992); A. Manolescu, R. R. Gerhardts, Phys. Rev. B 56, 9707 (1997); R. R. Gerhardts, Phys. Rev. B 53, 11064 (1996); R. Menne, R. R. Gerhardts, Phys. Rev. B 57, 1707 (1998); U. J. Gossmann, A. Manolescu, R. R. Gerhardts, Phys. Rev. B 57, 1680 (1998); J. Shi and F. M. Peeters, K. W. Edmonds and B. L. Gallagher, Phys. Rev. B 66, 035328 (2002); P. De Ye, D. Weiss, R. R. Gerhardts, M. Seeger, K. v. Klitzing, K. Eberl and H. Nickel, Phys. Rev. Lett. 74, 3013 (1995); J. H. Ho, Y. H. Lai, Y. H. Chui and M. F. Lin, Nanotechnology 19, 035712 (2008);
* (9) P. Vasilopoulos, F. M. Peeters, Superlattices and Microstructures 7, 393 (1990); F. M. Peeters and A. Matulis, Phys. Rev. B 48, 15166 (1993); F. M. Peeters, P. Vasilopoulos and Jirong Shi, J. Phys. : Condens. Matter 14, 8803 (2002); X. F. Wang and P. Vasilopoulos, F. M. Peeters, Phys. Rev. B 71, 125301 (2005); X. F. Wang, P. Vasilopoulos and F. M. Peeters, Phys. Rev. B 69, 035331 (2004); D. P. Xue and G. Xiao, Phys. Rev. B 45, 5986 (1992).
* (10) A. Matulis and F. M. Peeters, Phys. Rev. B 75, 125429 (2007).
* (11) M. Tahir, K. Sabeeh, Phys. Rev. B 77, 195421 (2008).
* (12) M. Charbonneau, K. M. Van Vliet and P. Vasilopoulos, J. Math. Phys. 23, 318 (1982).
* (13) I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series and Products (Academic Press, New York, 1980).
* (14) H. A. Carmona,et.al, Phys. Rev. Lett. 74, 3009 (1995); K. W. Edmonds,et.al, Phys. Rev. B 64, 041303 (2001); P. De Ye, D. Weiss, R. R. Gerhardts, M. Seeger, K. v. Klitzing, K. Eberl and H. Nickel, Phys. Rev. Lett. 74, 3013 (1995).
* (15) R. Tsu, Superlattice to Nanoelectronics (Elsevier, Oxford, UK, 2005); M. G. Cottam, D. R. Tilley, Introduction to Surface and Superlattice Excitations (Cambridge Univ. Press, UK, 1989).
* (16) C- H Park,et.al, Nat. Phys. 4, 213 (2008).
| arxiv-papers | 2008-07-17T10:11:51 | 2024-09-04T02:48:56.859490 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "M. Tahir and K. Sabeeh",
"submitter": "Muhammad Tahir",
"url": "https://arxiv.org/abs/0807.2746"
} |
0807.2833 | # Anisotropic superconducting properties of aligned Sm0.95La0.05FeAsO0.85F0.15
microcrystalline powder
B. C. Chang1, C. H. Hsu1, Y. Y. Hsu2, Z. Wei3, K. Q. Ruan3, X. G. Li3, and H.
C. Ku1 hcku@phys.nthu.edu.tw 1Department of Physics, National Tsing Hua
University, Hsinchu 30013, Taiwan
2Department of Physics, National Taiwan Normal University, Taipei 10610,
Taiwan
3Hefei National Laboratory for Physical Sciences at Microscale and Department
of Physics, University of Science and Technology of China, Hefei 230026, China
###### Abstract
The Sm0.95La0.05FeAsO0.85F0.15 compound is a quasi-2D layered superconductor
with a superconducting transition temperature Tc = 52 K. Due to the Fe spin-
orbital related anisotropic exchange coupling (antiferromagnetic or
ferromagnetic fluctuation), the tetragonal microcrystalline powder can be
aligned at room temperature using the field-rotation method where the
tetragonal $\it{ab}$-plane is parallel to the aligned magnetic field Ba and
$\it{c}$-axis along the rotation axis. Anisotropic superconducting properties
with anisotropic diamagnetic ratio $\chi_{c}$/$\chi_{ab}\sim$ 2.4 + 0.6 was
observed from low field susceptibility $\chi$(T) and magnetization M(Ba). The
anisotropic low-field phase diagram with the variation of lower critical field
gives a zero-temperature penetration depth $\lambda_{c}$(0) = 280 nm and
$\lambda_{ab}$(0) = 120 nm. The magnetic fluctuation used for powder alignment
at 300 K may be related with the pairing mechanism of superconductivity at
lower temperature.
###### pacs:
74.72.-h, 74.25.Ha
High-Tc superconductivity with transition temperature Tc up to 55 K were
reported in the newly discovered iron-based RFeAsO1-xFx (rare earth R = La,
Ce, Pr, Nd or Sm) system p1 ; p2 ; p3 ; p4 ; p5 ; p6 ; p7 ; p8 ; p9 ; p10 ;
p11 ; p12 . The ZrCuAsSi-type tetragonal structure (space group P4/nmm) is a
layered structure where the metallic FeAs layer is separated by the insulating
RO1-xFx layer. The discovery of the iron-based superconductor has generated
enormous interest since these compounds are the first non-cuprate high-Tc
superconductors with Tc higher than 50 K. The parent compound LaFeAsO is a
normal semi-metal which shows a Fermi surface nesting or spin Peierls
instability below 150 K with a tetragonal to orthorhombic structural
transition, accompanied by a spin density wave (SDW) type antiferromagnetic
order p6 . Electron doping to the FeAs layer through F--substitution in the
O2- site or O2--deficiency suppresses both the magnetic order and the
structural distortion in favor of superconductivity p1 ; p3 . On the other
hand, hole doping through Sr2+-substitution in the R3+ site gives similar
effect p4 . Therefore, like the high-Tc cuprate systems, the superconductivity
in these iron-based compounds occurs in close proximity to a long range
antiferromagnetic ground state.
Since the FeAs layer is believed to be the superconducting layer of the
RFeAsO1-xFx system, studies on the anisotropic properties are crucial for
understanding this new iron-based system. High-quality single crystal is
essential for detailed in-depth studies p9 ; p12 . However, the anisotropic
superconducting properties can be easily obtained using a much simpler way. In
this report, we use the field-rotation alignment method to align the Tc = 52 K
Sm0.95La0.05FeAsO0.85F0.15 microcrystalline powder at room temperature.
Figure 1: Block diagram for field-rotation powder alignment method with
$\it{ab}$-plane along the aligned magnetic field Ba and $\it{c}$-axis along
the rotation axis.
The polycrystalline Sm0.95La0.05FeAsO0.85F0.15 was prepared by conventional
solid state reaction. First the SmAs powder was prepared by reacting Sm and As
powders at 650∘C for about 5 hours, and fine powders of SmAs, FeAs, Fe, LaF3,
Fe2O3 were mixed together according to the stoichiometric ratio of
Sm0.95La0.05FeAsO0.85F0.15, ground thoroughly, and then pressed into pellets.
The pellets were wrapped in Ta foil to be sealed in an evacuated quartz tube,
and annealed at 1160∘C for about 50 hours. The Sm0.95La0.05FeAsO0.85F0.15
powders with an average microcrystalline grain size of 1-10 $\mu$m were mixed
with epoxy (4-hour curing time) in a quartz tube ($\phi$ = 8 mm) with a
powder:epoxy ratio of 1:5. The quartz tube was immediately placed in a 0.9-T
electromagnet and rotated at a speed of 10 rpm with the rotation axis
perpendicular to the aligned magnetic field Ba (Fig. 1). Since the tetragonal
$\it{ab}$-plane of Sm0.95La0.05FeAsO0.85F0.15 microcrystalline is aligned
along Ba from X-ray diffraction study, the rotation perpendicular to Ba will
force the microcrystalline $\it{c}$-axis to align along the rotation axis.
Figure 2: Powder X-ray diffraction patterns for Sm0.95La0.05FeAsO0.85F0.15.
(a) random powder, (b) $\it{ab}$-plane aligned along Ba, (c) $\it{c}$-axis
aligned along rotation axis ($\perp$ Ba).
The X-ray diffraction patterns for Sm0.95La0.05FeAsO0.85F0.15 random powder,
partially $\it{ab}$-plane aligned along Ba, and partially $\it{c}$-axis
aligned along the rotation axis are shown collectively in Fig. 2. The
unaligned random powder diffraction pattern gives a slightly larger tetragonal
lattice parameters of $\it{a}$ = 0.3936(3) nm and $\it{c}$ = 0.8495(8) nm with
5$\%$ La doping than SmFeAsO0.85F0.15 p5 . At least 10$\%$ impurities were
observed as marked by asterisk in the diffraction pattern. For $\it{ab}$-plane
aligned along Ba, enhanced ($\it{hk}$0) diffraction lines ($\it{h}$ \+
$\it{k}$ = 2n) were observed. The lack of forbidden line (210) is consistent
with the space group P4/nmm. For $\it{c}$-axis aligned along the rotation
axis, enhanced (00$\it{l}$) diffraction lines were observed. The imperfect
alignment (80-90 $\%$) with the appearance of major (102) line may be due to
intrinsic weak magnetic anisotropy, low aligned field (0.9 T), imperfect
powder preparation or alignment procedure. The diffraction pattern for
nonsuperconducting SmFeAsO0.95F0.05 aligned powder gives similar result with
$\it{ab}$-plane aligned along Ba.
The field-rotation alignment method at room temperature is based on magnetic
anisotropy due to the spin-orbital related anisotropic exchange coupling.
There are two major contributions of magnetic moment at room temperature
$\chi\sim\chi$(Fe) + $\chi$(Sm). The large Sm3+ (4f55s25p6, S = 5/2, L = 5, J
= L - S = 5/2) localized moment in the insulating (Sm0.95La0.05)(O0.85F0.15)
layer has weak anisotropic exchange coupling. On the other hand, the Fe2+
(3d6) itinerant moment in the metallic FeAs layer is rather small from powder
neutron diffraction data and band structure calculation. The Fe2+ feels the
distorted FeAs4 tetrahedral crystal field with three low-lying manifold and
two up-lying manifold (dxz and dyz). The six 3$\it{d}$ electrons are
distributed in three quasi-2D hole-like bands and two electron-like bands with
Fermi surfaces from five 3$\it{d}$ orbitals. Anisotropic exchange coupling
(magnetic fluctuation) occurs from strong Fe-As 3dxz,yz-4p hybridization and
Fe-Fe 3dxy-3dxy direct electron hopping between neighboring Fe atoms in the
tetragonal basal $\it{ab}$-plane p6 ; p10 ; p11 . In the aligned magnetic
field, anisotropic Fe orbitals are tied to the spin direction, and a strong
spin-orbital related anisotropic exchange coupling at 300 K should dominate
the magnetic alignment. Due to the quasi-2D FeAs layer structure, anisotropic
Fe magnetic susceptibility $\chi_{ab}$(Fe) $>$ $\chi_{c}$(Fe) is expected to
be the dominant factor for $\it{ab}$-plane alignment along Ba at 300 K.
Indeed, anisotropic room temperature magnetization for
Sm0.95La0.05FeAsO0.85F0.15 aligned powder was observed with weak magnetic
anisotropy ratio $\chi_{ab}/\chi_{c}\sim$ 1.2 up to 7 T. In the low field
region, a cross-over from $\chi_{c}\geq\chi_{ab}$ to $\chi_{c}\leq\chi_{ab}$
at high field with nonlinear behavior may be caused by the magnetic
impurities.
Figure 3: Anisotropic low-field FC and ZFC susceptibility $\chi_{ab}$(T) and
$\chi_{c}$(T) for aligned Sm0.95La0.05FeAsO0.85F0.15.
Magnetization and magnetic susceptibility data were collected with a Quantum
Design 1-T $\mu$-metal shielded MPMS2 or a 7-T MPMS superconducting quantum
interference device (SQUID) magnetometer from 2 K to 300 K. The anisotropic
temperature dependence of molar magnetic susceptibility $\chi_{ab}$(T) and
$\chi_{c}$(T) for aligned powder Sm0.95La0.05FeAsO0.85F0.15 with applied field
along the $\it{ab}$-plane and $\it{c}$-axis are shown collectively in Fig. 3.
For aligned dispersed microcrystalline in low applied field of 10 G, both
zero-field-cooled (ZFC) and field-cooled (FC) data revealed a sharp
superconducting transition temperature Tc of 52 K, which is identical to the
measured Tc from bulk polycrystalline sample. Large ZFC intragrain Meissner
shielding signals were observed with an almost constant $\chi_{c}$ = -2.72
cm3/mol and $\chi_{ab}$ = -1.13 cm3/mol up to 20 K. The anisotropic
diamagnetic parameter $\gamma$ = $\chi_{c}$/$\chi_{ab}$ of 2.4 was deduced for
aligned microcrystalline. The FC flux-trapped signal gave the same anisotropic
parameter $\gamma$ of 2.4. Considering the imperfect alignment factor
(80-90$\%$), a larger anisotropic diamagnetic parameter $\gamma$ = 2.4 + 0.6
was derived. For temperature above Tc, normal state paramagnetic anisotropy
ratio $\chi_{c}$/$\chi_{ab}$ of 1.3 was observed in 10 G at 300 K, which is
consistent with low field magnetization at 300 K and may be caused by magnetic
impurities. No trace of Curie-Weiss-like behavior was observed in the normal
state.
Figure 4: Temperature variation of (a) $\it{c}$-axis superconducting initial
magnetization Mc(Ba) and (b) $\it{ab}$-plane superconducting initial
magnetization Mab(Ba) for aligned Sm0.95La0.05FeAsO0.85F0.15 powder.
The temperature variation of superconducting initial magnetization curves with
applied field Ba along $\it{c}$-axis, Mc(Ba,T), and $\it{ab}$-plane,
Mab(Ba,T), for aligned Sm0.95La0.05FeAsO0.85F0.15 powder from 2 K to 45 K are
shown in Fig. 4(a) and 4(b), respectively. For example, at T = 5 K, the
maximum diamagnetic signal was observed at magnetization peak field
$\mu_{0}$Hc(peak) of 850 G for Mc and a larger $\mu_{0}$Hab(peak) of 1250 G
for Mab, indicating anisotropic pinning for different applied field
directions. The peak field corresponds to the field that magnetic flux
penetrates into the center of aligned superconducting microcrystallines
forming a vortex glass where vortex are mainly pinned by impurities and
imperfections. At the low-field region, linear variation of magnetization as
an evidence of Meissner state was clearly observed for both directions p13 .
The low-field linear slopes of initial magnetization are consistent with the
low-field (Ba = 10 G) susceptibility as measured in Fig. 3. Since the lower
critical field $\mu_{0}$Hc1 is the field flux start to penetrate into the
superconductor forming mixed state , the magnetization M(Ba) deviates from the
linear response of Meissner state. Using 10-G/5-K linear magnetic
susceptibility $\chi_{c}$ = -2.7 cm3/mol and $\chi_{ab}$ = -1.1 cm3/mol in
Fig. 3, lower critical field $\mu_{0}$H${}_{c1}^{c}$ of 200 G and
$\mu_{0}$H${}_{c1}^{ab}$ of 85 G were deduced from the linear susceptibility
extrapolation lines.
The low temperature, low magnetic field anisotropic phase diagram for aligned
Sm0.95La0.05FeAsO0.85F0.15 powder is shown in Fig. 5(a). The zero temperature
lower critical field $\mu_{0}$H${}_{c1}^{c}$(0) of 230 G and
$\mu_{0}$H${}_{c1}^{ab}$(0) of 95 G were extrapolated. The temperature
dependence of anisotropic penetration depth $\lambda_{ab}$(T) and
$\lambda_{c}$(T) are shown in Fig. 5(b), using the anisotropic formula
$\mu_{0}$H${}_{c1}^{c}$ = $\Phi_{0}$/2$\pi\lambda_{ab}^{2}$ and
$\mu_{0}$H${}_{c1}^{ab}$ = $\Phi_{0}$/2$\pi\lambda_{ab}\lambda_{c}$, where
$\Phi_{0}$ is the flux quantum. The zero temperature penetration depths
$\lambda_{c}$(0) = 280 nm and $\lambda_{ab}$(0) = 120 nm were extrapolated.
These values reflect the quasi-2D FeAs layer structure and are smaller than
the microcrystalline grain size of $\sim$ 1-10 $\mu$m. The anisotropy
parameter $\gamma$ derived from the penetration depth ratio
$\lambda_{c}(0)/\lambda_{ab}(0)$ of 2.33 is smaller than the reported value
from single crystal by penetration depth ($\sim$ 4) p12 or $\mu_{0}$Hc2
($\sim$ 4.34-4.9) p9 . However, the differences of anisotropy parameter
$\gamma$ may be attributed to the different demagnetization field in slab-
shaped single crystal with different applied field directions when the
geometric shape is considered, i.e. the slab shape of single crystals and the
sphere-like shape of the aligned microcrystalline in this work.
Figure 5: (a) Low field anisotropic phase diagram $\mu_{0}$H(T) for aligned
Sm0.95La0.05FeAsO0.85F0.15 powder. (b) Temperature dependence of anisotropic
penetration depth $\lambda_{ab}$(T) and $\lambda_{c}$(T).
Since the band structure calculations suggested quasi-2D nature of the Fermi
surface p10 ; p11 in the layered Fe-based compounds, the $\it{ab}$-plane
penetration depth behavior was expected to provide more intrinsic information
than $\it{c}$-axis. The low temperature (2-25 K) penetration depth
$\lambda_{ab}$(T) can be roughly fitted by a BCS-type s-wave exponential law
p12
$\frac{\lambda_{ab}(T)-\lambda_{ab}(0)}{\lambda_{ab}(0)}=\sqrt{\frac{\pi\Delta_{0}}{2T}}exp(-\frac{\Delta_{0}}{k_{B}T}).$
(1)
but with a much smaller $\Delta_{0}$/kBTc of 0.4 than typical value of 1.76,
indicating that isotropic s-wave model may not be a good one for this new
high-Tc superconductor, although a BCS-like gap is reported for
SmFeAsO0.85F0.15 from Andreev spectra p7 . However, the low temperature
$\lambda_{ab}$(T) can also be well fitted with a simple linear-T dependence,
indicating existence of nodal quasiparticle excitation, possibly d-wave or
extended s-wave pairing in nature p11 . A triplet p-wave pairing model with
strong ferromagnetic fluctuation and degenerate electron Fermi surfaces was
also proposed from band structure consideration p10 . In the powder alignment
procedure, Fe spin-orbital-related anisotropic magnetic fluctuation is needed
at 300 K. However, we can not distinguish whether the magnetic fluctuation is
ferromagnetic or antiferromagnetic in origin.
The anisotropic high-field ($\pm$7 T) isothermal superconducting hysteresis
loops M-Ba at 5 K (Fig. 6) indicate that upper critical field $\mu_{0}$Hc2 in
both directions are much larger than maximum applied field of 7 T with very
short anisotropic coherence lengths $\xi_{ab}$ and $\xi_{c}$. Moreover,
hysteresis loops with large paramagnetic background indicate the large
magnetic/impurity contribution.
Figure 6: Anisotropic high-field superconducting hysteresis loop Mc(Ba) and
Mab(Ba) at 5 K for aligned Sm0.95La0.05FeAsO0.85F0.15 powder.
In conclusion, due to Fe spin-orbital-related short-range anisotropic exchange
interaction $\chi_{ab}$(Fe) $>\chi_{c}$(Fe) at 300 K,
Sm0.95La0.05FeAsO0.85F0.15 microcrystalline powder can be aligned using the
field-rotation alignment method where $\it{ab}$-plane is parallel to the
aligned field Ba and $\it{c}$-axis is parallel to the rotation axis. Rather
small superconducting anisotropy parameter $\gamma\sim$ 2.4 \+ 0.6 was
observed (from both susceptibility and lower critical field). The magnetic
fluctuation used for powder alignment at 300 K may be related with the pairing
mechanism of superconductivity at lower temperature.
This work was supported by NSC95-2112-M-007-056-MY3, NSC97-2112-M-003-001-MY3,
NSFC50421201, MSTC2006CB601003, and MSTC2006CB922005.
## References
* (1) Y. Kamihara, T. Watanabe, M. Hirano, and H. Hosono, J. Am. Chem. Soc. 130, 3296 (2008).
* (2) H. Takahashi, K. Igawa, K. Arii, Y. Kamihara, M. Hirano, and H. Hosono, Nature 453, 376 (2008).
* (3) T. A. Ren, G. C. Che, X. L. Dong, J. Yang, W. Lu, W. Yi, X. L. Shen, Z. C. Li, L. L. Sun, F. Zhou, and Z. X. Zhao, Europhys. Lett. 83, 17002 (2008).
* (4) H. H. Wen, G. Mu, L. Fang, H. Yang, and X. Zhu, Europhys. Lett. 83, 17009 (2008).
* (5) X. H. Chen, T. Wu, G. Wu, R. H. Liu, H. Chen, and D. F. Fang, Nature 453, 761 (2008).
* (6) C. de la Cruz, Q. Huang, J. W. Lynn, J. Y. Li, W. Ratcliff II, J. L. Zarestky, H. A. Mook, G. F. Chen, J. L. Luo, N. L. Wang, P. C. Dai, Nature 453, 899 (2008).
* (7) T. Y. Chen, Z. Tesanovic, R. H. Liu, X. H. Chen, and C. L. Chien, Nature 453, 1224 (2008).
* (8) G. F. Chen, Z. Li, D. Wu, G. Li, W. Z. Hu, J. Dong, P. Zheng, J. L. Luo, and N. L. Wang, Phys. Rev. Lett. 100, 247002 (2008).
* (9) Y. Jia, P. Cheng, L. Fang, H. Q. Luo, H. Yang, C. Ren, L. Shan, C. Z. Gu, and H. H. Wen, arXiv:cond-mat/0806.0532 (2008).
* (10) G. Xu, W. Ming, Y. Yao, X. Dai, S. C. Zhang, and Z. Fang, Europhys. Lett. 82, 67002 (2008).
* (11) X. Dai, Z. Fang, Y. Zhou, and F. C. Zhang, arXiv:cond-mat/0803.3982 (2008).
* (12) C. Martin, R. T. Gordon, M. A. Tanatar, M. D. Vannette, M. E. Tillman, E. D. Mun, P. C. Canfield, V. G. Kogan, G. D. Samolyuk, J. Schmalian, and R. Prozorov, arXiv:cond-mat/0807.0876v1 (2008).
* (13) X. G. Li, X. F. Sun, Y. H. Toh, Y. Y. Hsu, and H. C. Ku, Phys. Rev. B 58, 1000 (1998).
| arxiv-papers | 2008-07-17T16:57:13 | 2024-09-04T02:48:56.865909 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "B. C. Chang, C. H. Hsu, Y. Y. Hsu, Z. Wei, K. Q. Ruan, X. G. Li, and\n H. C. Ku",
"submitter": "Huan-Chiu Ku",
"url": "https://arxiv.org/abs/0807.2833"
} |
0807.2967 | # Local cohomology: Associated primes,
artinianness and asymptotic behaviour
Moharram Aghapournahr Moharram Aghapournahr
Faculty of Mathematical Sciences, Teacher Training University, 599 Taleghani
Avenue, Tehran 15614, Iran. m.aghapour@gmail.com and Leif Melkersson Leif
Melkersson
Department of Mathematics, Linköping University, S-581 83 Linköping, Sweden.
lemel@mai.liu.se
###### Abstract.
Let $R$ be a noetherian ring, $\mathfrak{a}$ an ideal of $R$, $M$ an
$R$–module and $n$ a non-negative integer. In this paper we first will study
the finiteness properties of the kernel and the cokernel of the natural map
$f:\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M)\longrightarrow\operatorname{Hom}_{R}(R/\mathfrak{a},\operatorname{H}^{n}_{\mathfrak{a}}(M))$.
Then we will get some corollaries about the associated primes and artinianness
of local cohomology modules. Finally we will study the asymptotic behaviour of
the kernel and the cokernel of this natural map in the graded case.
###### Key words and phrases:
Local cohomology, associated prime, asymptotic behaviors, tameness.
###### 2000 Mathematics Subject Classification:
13D45, 13D07
## 1\. Introduction
Throughout $R$ is a commutative noetherian ring. Our terminology follows the
book [5] on local cohomology. Huneke formulated and discussed several problems
in [12] about local cohomology modules
$\operatorname{H}^{n}_{\mathfrak{a}}(M)$ of $M$. One of them is when do they
have just finitely many associated prime ideals. Another one is about when
such modules are artinian. Furthermore the following conjecture was made by
Grothendieck in [10]:
Conjecture: For any ideal $\mathfrak{a}$ and finite $R$–module $M$, the module
$\operatorname{Hom}_{R}(R/\mathfrak{a},\operatorname{H}^{n}_{\mathfrak{a}}(M))$
is finite for all $n\geq 0$.
This was thought of as a substitute for the artinianness of
$\operatorname{H}^{n}_{\mathfrak{a}}(M)$, in the case when $\mathfrak{a}$ is
the maximal ideal of a local ring. Morover the finiteness of
$\operatorname{Hom}_{R}(R/\mathfrak{a},\operatorname{H}^{n}_{\mathfrak{a}}(M))$
implies the finiteness of the set of associated prime ideals of
$\operatorname{H}^{n}_{\mathfrak{a}}(M)$. Although this conjecture is not true
in general as shown by Hartshorne in [11], there are some attempts to show
that under some conditions, for some number $n$, the module
$\operatorname{Hom}_{R}(R/\mathfrak{a},\operatorname{H}^{n}_{\mathfrak{a}}(M))$
is finite, see [1, Theorem 3.3], [7, Theorem 6.3.9] and [8, Theorem 2.1].
In section 2, in view of Grothendieck’s conjecture and [7, Theorem 6.3.9], we
first study finiteness properties of the kernel and the cokernel of the
natural homomorphism
$f:\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M)\longrightarrow\operatorname{Hom}_{R}(R/\mathfrak{a},\operatorname{H}^{n}_{\mathfrak{a}}(M))$
under certain conditions, using the technique introduced by the second author
in [14]. We will find some relations between finiteness of the associated
primes of $\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M)$ and
$\operatorname{H}^{n}_{\mathfrak{a}}(M)$ that will give us a generalization of
[13, Proposition 1.1 part (b)], [2, Lemma 2.5 part (c)] and [4, Proposition
2.2]. Also we show that under certain conditions
$\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M)$ is artinian if and only if
$\operatorname{H}^{n}_{\mathfrak{a}}(M)$ is. One of our main results of this
paper is
Corollary 2.2.
(a)If
$\operatorname{Ext}^{n-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))=0$
for all $j<n$, then $f$ is injective.
(b)If
$\operatorname{Ext}^{n+1-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))=0$
for all $j<n$, then $f$ is surjective.
(c)If
$\operatorname{Ext}^{t-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))=0$
for $t=n,n+1$ and for all $j<n$, then
$f$ is an isomorphism.
In section 3 we will study the asymptotic behaviour of the kernel and the
cokernel of the above mentioned natural map in the graded case.
## 2\. Associated primes and artinianness
###### Proposition 2.1.
Let $R$ be a noetherian ring, $\mathfrak{a}$ an ideal of $R$ and $M$ an
$R$–module. Let $n$ be a non-negative integer and $\mathcal{S}$ be a Serre
subcategory of the category of $R$–modules i.e. it is closed under taking
submodules, quotients and extensions. Consider the natural homomorphism
$f:\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M)\longrightarrow\operatorname{Hom}_{R}(R/\mathfrak{a},\operatorname{H}^{n}_{\mathfrak{a}}(M))$
If
$\operatorname{Ext}^{n-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))$
belongs to $\mathcal{S}$ for all $j<n$, then $\operatorname{Ker}{f}$ belongs
to $\mathcal{S}$. (In particular if
$\operatorname{Ext}^{n-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))$
is finite for all $j<n$, then $\operatorname{Ker}{f}$ is finite.)
If
$\operatorname{Ext}^{n+1-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))$
belongs to $\mathcal{S}$ for all $j<n$, then $\operatorname{Coker}{f}$ belongs
to $\mathcal{S}$. ( In particular if
$\operatorname{Ext}^{n+1-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))$
is finite for all $j<n$, then $\operatorname{Coker}{f}$ is finite.)
If
$\operatorname{Ext}^{t-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))$
belongs to $\mathcal{S}$ for $t=n,n+1$ and for all $j<n$, then
$\operatorname{Ker}{f}$ and $\operatorname{Coker}{f}$ both belong to
$\mathcal{S}$. Thus $\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M)$ belongs to
$\mathcal{S}$ if and only if
$\operatorname{Hom}_{R}(R/\mathfrak{a},\operatorname{H}^{n}_{\mathfrak{a}}(M))$
belongs to $\mathcal{S}$. (In particular if
$\operatorname{Ext}^{t-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))$
is finite for $t=n,n+1$ and for all $j<n$ then $\operatorname{Ker}{f}$ and
$\operatorname{Coker}{f}$ both are finite, thus
$\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M)$ is finite if and only if
$\operatorname{Hom}_{R}(R/\mathfrak{a},\operatorname{H}^{n}_{\mathfrak{a}}(M))$
is finite. (See also [7, Theorem 6.3.9])).
###### Proof.
By the exact sequence $0\rightarrow\Gamma_{\mathfrak{a}}(M)\rightarrow
M\rightarrow M/\Gamma_{\mathfrak{a}}(M)\rightarrow 0$ we obtain the exact
sequence (*) below:
$\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},\Gamma_{\mathfrak{a}}(M))\rightarrow\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M)\overset{h}{\rightarrow}\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M/\Gamma_{\mathfrak{a}}(M))\rightarrow\operatorname{Ext}^{n+1}_{R}(R/\mathfrak{a},\Gamma_{\mathfrak{a}}(M)).$
We have the commutative diagram
$\begin{matrix}\\\
&&\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M)&\overset{h}{\longrightarrow}&\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M/\Gamma_{\mathfrak{a}}(M))\\\
&&f\big{\downarrow}&&\big{\downarrow}{\bar{f}}\\\
&&\operatorname{Hom}_{R}(R/\mathfrak{a},\operatorname{H}^{n}_{\mathfrak{a}}(M))&\overset{\bar{h}}{\longrightarrow}&\operatorname{Hom}_{R}(R/\mathfrak{a},\operatorname{H}^{n}_{\mathfrak{a}}(M/\Gamma_{\mathfrak{a}}(M)))\\\
\end{matrix}.$
Note that $\bar{h}$ is an isomorphism. Let $g=\bar{h}^{-1}\circ{\bar{f}}$.
Thus $f=g\circ{h}$ and we obtain the exact sequence
$0\rightarrow\operatorname{Ker}{h}\rightarrow\operatorname{Ker}{f}\rightarrow\operatorname{Ker}{g}\rightarrow\operatorname{Coker}{h}\rightarrow\operatorname{Coker}{f}\rightarrow\operatorname{Coker}{g}\rightarrow{0}$
Proof of (a):
For $n=0$, since
$\operatorname{Ext}^{0}_{R}(R/\mathfrak{a},M)\cong\operatorname{Hom}_{R}(R/\mathfrak{a},\Gamma_{\mathfrak{a}}(M))$
thus $\operatorname{Ker}{f}=0$ and the result is clear. For $n\geq 1$ by
hypothesis and the exact sequence (*), $\operatorname{Ker}{h}$ belongs to
$\mathcal{S}$. Therefore it is enough to show that $\operatorname{Ker}{g}$
belongs to $\mathcal{S}$. We prove this by induction on $n$. For $n=1$ since
$\bar{h}$ is an isomorphism, hence
$\operatorname{Ker}{g}=\operatorname{Ker}{\bar{f}}$ but $\bar{f}$ is an
isomorphism by [14, Lemma 7.9]. Assume $n>1$ and the case $n-1$ is settled.
Let $E$ be the injective hull of $M/\Gamma_{\mathfrak{a}}(M)$ and set
$N=E/(M/\Gamma_{\mathfrak{a}}(M))$. Since
$\Gamma_{\mathfrak{a}}(M/\Gamma_{\mathfrak{a}}(M))=0$, we have
$\operatorname{Hom}_{R}(R/\mathfrak{a},E)=0$ and $\Gamma_{\mathfrak{a}}(E)=0$.
Thus we get the isomorphisms
$\operatorname{Ext}^{i}_{R}(R/\mathfrak{a},N)\cong\operatorname{Ext}^{i+1}_{R}(R/\mathfrak{a},M/\Gamma_{\mathfrak{a}}(M))$
and
$\operatorname{H}^{i}_{\mathfrak{a}}(N)\cong\operatorname{H}^{i+1}_{\mathfrak{a}}(M/\Gamma_{\mathfrak{a}}(M))\cong\operatorname{H}^{i+1}_{\mathfrak{a}}(M)$
for all $i\geq 0$. Therefore
$\operatorname{Hom}_{R}(R/\mathfrak{a},\operatorname{H}^{n-1}_{\mathfrak{a}}(N))\cong\operatorname{Hom}_{R}(R/\mathfrak{a},\operatorname{H}^{n}_{\mathfrak{a}}(M))$.
In addition, for all $j<n-1$ the modules
$\operatorname{Ext}^{n-1-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(N))\cong\operatorname{Ext}^{n-1-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j+1}_{\mathfrak{a}}(M))$
belong to $\mathcal{S}$. Now, by the induction hypothesis, the kernel of the
natural homomorphism
$f^{\prime}:\operatorname{Ext}^{n-1}_{R}(R/\mathfrak{a},N)\longrightarrow\operatorname{Hom}_{R}(R/\mathfrak{a},\operatorname{H}^{n-1}_{\mathfrak{a}}(N))$
belongs to $\mathcal{S}$. But
$\operatorname{Ker}{f^{\prime}}=\operatorname{Ker}{g}$, so
$\operatorname{Ker}{g}$ belongs to $\mathcal{S}$.
Proof of (b):
The proof is similar to (a).
Proof of (c):
It is clear by (a) and (b). ∎
The following corollary is one of our main results of this paper. It is well-
known by using a spectral sequence argument as in [13, Proposition 1.1 (b)] or
another method as in [2, Lemma 2.5 (b)] to prove that, (when $M$ is a finite
module and $n=\operatorname{depth}_{\mathfrak{a}}M$) or more generally when
the local cohomology vanishes below $n$, for example use [15, Theorem 11.3],
then the natural map
$f:\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M)\longrightarrow\operatorname{Hom}_{R}(R/\mathfrak{a},\operatorname{H}^{n}_{\mathfrak{a}}(M))$
is an isomorphism. In the following corollary we show that without any
condition on $M$ it is not necessary that these local cohomology modules are
zero in order that the natural map is an isomorphism.
###### Corollary 2.2.
If
$\operatorname{Ext}^{n-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))=0$
for all $j<n$, then $f$ is injective.
If
$\operatorname{Ext}^{n+1-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))=0$
for all $j<n$, then $f$ is surjective.
If
$\operatorname{Ext}^{t-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))=0$
for $t=n,n+1$ and for all $j<n$, then
$f$ is an isomorphism.
###### Proof.
In proposition 2.1 set $\mathcal{S}=\\{0\\}$. ∎
###### Corollary 2.3.
If $\operatorname{H}^{i}_{\mathfrak{a}}(M)=0$ for all $i<n$ (in particular,
when $M$ is finite and $n=\operatorname{depth}_{\mathfrak{a}}M$) then
$\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M)\cong\operatorname{Hom}_{R}(R/\mathfrak{a},\operatorname{H}^{n}_{\mathfrak{a}}(M))$
###### Corollary 2.4.
If
$\operatorname{Ext}^{t-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))$
is artinian for $t=n,n+1$ and for all $j<n$, then
$\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M)$ is artinian if and only if
$\operatorname{H}^{n}_{\mathfrak{a}}(M)$ is artinian.
###### Proof.
It is immediate by using 2.1 (c) and [5, Theorem 7.1.2]. ∎
###### Corollary 2.5.
If
$\operatorname{Ext}^{n-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))=0$
for all $j<n$, then
$\operatorname{Ass}_{R}(\operatorname{H}^{n}_{\mathfrak{a}}(M))\subset\operatorname{Ass}_{R}(\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M))\cup\operatorname{Ass}_{R}(\operatorname{Coker}{f})$.
If
$\operatorname{Ext}^{n+1-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))=0$
for all $j<n$, then
$\operatorname{Ass}_{R}(\operatorname{H}^{n}_{\mathfrak{a}}(M))\subset\operatorname{Ass}_{R}(\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M))\cup\operatorname{Supp}_{R}(\operatorname{Ker}{f})$.
If
$\operatorname{Ext}^{t-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))=0$
for $t=n,n+1$ and for all $j<n$, then
$\operatorname{Ass}_{R}(\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M))=\operatorname{Ass}_{R}(\operatorname{H}^{n}_{\mathfrak{a}}(M))$.
For the proof of part (b) in corollary 2.5 we need the following lemma.
###### Lemma 2.6.
For any exact sequence $N\longrightarrow L\longrightarrow T\longrightarrow 0$
of $R$–modules and $R$–homomorphisms we have
$\operatorname{Ass}_{R}(T)\subset\operatorname{Ass}_{R}(L)\cup\operatorname{Supp}_{R}(N)$
###### Proof.
Let
$\mathfrak{p}\in\operatorname{Ass}_{R}(T)\setminus\operatorname{Supp}_{R}(N)$
then $N_{\mathfrak{p}}=0$, so $L_{\mathfrak{p}}\cong T_{\mathfrak{p}}$. Now
${\mathfrak{p}}R_{\mathfrak{p}}\in\operatorname{Ass}_{R_{\mathfrak{p}}}(T_{\mathfrak{p}})$,
hence
${\mathfrak{p}}R_{\mathfrak{p}}\in\operatorname{Ass}_{R_{\mathfrak{p}}}(L_{\mathfrak{p}})$,
so $\mathfrak{p}\in\operatorname{Ass}_{R}(L)$. ∎
Divaani-Aazar and Mafi introduced in [9] weakly Laskerian modules. An
$R$–module $M$ is weakly Laskerian if for any submodule $N$ of $M$ the
quotient $M/N$ has finitely many associated primes. The weakly Laskerian
modules form a Serre subcategory of the category of $R$–modules. In fact it is
the largest Serre subcategory such that each module in it has just finitely
many associated prime ideals. In the following corollary we give some
conditions for the finiteness of the sets of associated primes of
$\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M)$ and
$\operatorname{H}^{n}_{\mathfrak{a}}(M)$ respectively.
###### Corollary 2.7.
If $\operatorname{Ass}_{R}(\operatorname{H}^{n}_{\mathfrak{a}}(M))$ is a
finite set and
$\operatorname{Ext}^{n-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))$
is weakly Laskerian for all $j<n$ ( in particular if
$\operatorname{H}^{j}_{\mathfrak{a}}(M)$ is weakly Laskerian for all $j<n$),
then $\operatorname{Ass}_{R}(\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M))$ is
a finite set. (See also [7, Corollary 6.3.11])
If $\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M)$ and
$\operatorname{Ext}^{n+1-j}_{R}(R/\mathfrak{a},\operatorname{H}^{j}_{\mathfrak{a}}(M))$
for all $j<n$ are weakly Laskerian (in particular if $M$ and
$\operatorname{H}^{j}_{\mathfrak{a}}(M)$ for all $j<n$ are weakly Laskerian),
then $\operatorname{Ass}_{R}(\operatorname{H}^{n}_{\mathfrak{a}}(M))$ is a
finite set. (Compare with [9, Corollary 2.7])
###### Proof.
Note that
$\operatorname{Ass}_{R}(\operatorname{H}^{n}_{\mathfrak{a}}(M))=\operatorname{Ass}_{R}(\operatorname{Hom}_{R}(R/\mathfrak{a},\operatorname{H}^{n}_{\mathfrak{a}}(M)))$.
We use the exact sequence
$0\rightarrow\operatorname{Ker}f\rightarrow\operatorname{Ext}^{n}_{R}(R/\mathfrak{a},M)\overset{f}{\rightarrow}\operatorname{Hom}_{R}(R/\mathfrak{a},\operatorname{H}^{n}_{\mathfrak{a}}(M))\rightarrow\operatorname{Coker}f\rightarrow
0$.
(a) By 2.1 (a), $\operatorname{Ker}f$ is weakly Laskerian.
(b) By 2.1 (b), $\operatorname{Coker}f$ is weakly Laskerian.
∎
## 3\. Asymptotic behaviours
Assume that $R=\underset{i\geq{0}}{\bigoplus}{R_{i}}$ is a homogeneous graded
noetherian ring and $M=\underset{i\in\mathbb{Z}}{\bigoplus}{M_{i}}$ is a
graded $R$–module. The module $M$ is said to have
(a) the property of asymptotic stablity of associated primes if there exists
an integer $n_{0}$ such that
$\operatorname{Ass}_{R_{0}}(M_{n})=\operatorname{Ass}_{R_{0}}(M_{n_{0}})$ for
all $n\leq{n_{0}}$.
(b) the property of asymptotic stablity of supports if there exists an integer
$n_{0}$ such that
$\operatorname{Supp}_{R_{0}}(M_{n})=\operatorname{Supp}_{R_{0}}(M_{n_{0}})$
for all $n\leq{n_{0}}$.
(c) the property of tameness if $M_{i}=0$ for all $i<<0$ or else $M_{i}\neq 0$
for all $i<<0$.
Note that if $M$ has the property of (a) then $M$ has the property of (b) and
if $M$ has the property of (b) then $M$ is tame i.e. has the property of (c).
Let $s$ be a non-negative integer. In order to refine and complete [6, Theorem
4.4] we study the asymptotic behaviour of the kernel and cokernel of the
natural homomorphism between the graded modules
$\operatorname{Ext}_{R}^{s}(R/R_{+},M)$ and
$\operatorname{Hom}_{R}(R/R_{+},\operatorname{H}_{R_{+}}^{s}(M))$. For more
information see also [3].
###### Definition 3.1.
A graded module $M$ over a homogeneous graded ring $R$ is called
asymptotically zero if $M_{i}=0$ for all $i<<0$.
All finite graded $R$–modules are asymptotically zero.
###### Theorem 3.2.
Assume that $M$ is a graded $R$–module and let $s$ be a fixed non-negative
integer such that the modules
$\operatorname{Ext}_{R}^{s-j}(R/R_{+},\operatorname{H}_{R_{+}}^{j}(M))$ and
$\operatorname{Ext}_{R}^{s-j+1}(R/R_{+},\operatorname{H}_{R_{+}}^{j}(M))$ ,
$0\leq j<s$
are asymptotically zero (e.g. they might be finite). Then the kernel and the
cokernel of the natural homomorphism
$f:\operatorname{Ext}_{R}^{s}(R/R_{+},M)\longrightarrow\operatorname{Hom}_{R}(R/R_{+},\operatorname{H}_{R_{+}}^{s}(M))$
are asymptotically zero. Therefore $\operatorname{Ext}_{R}^{s}(R/R_{+},M)$ has
one of the properties of (a), (b) and (c) if and only if
$\operatorname{Hom}_{R}(R/R_{+},\operatorname{H}_{R_{+}}^{s}(M))$ has.
###### Proof.
Note that the class of asymptotically zero graded modules is Serre subcategory
of the category of graded $R$–modules and use 2.1 (c). ∎
## References
* [1] J. Asadollahi, K. Khashyarmanesh, Sh. Salarian, _A generalization of the cofiniteness problem in local cohomology modules_ , J. Aust. Math. Soc. 75 (2003), 313–324
* [2] J. Asadollahi and P. Schenzel, Some results on associated primes of local cohomology modules, Japan. J. Math. 29(2003), 285-296.
* [3] M.P. Brodmann, Asymptotic behavior of cohomology : tameness, supports and associated primes, to appear in Contemporary Mathematics: Proceedings of the international conference in Mathematics, Bangalore / India, December, 2003.
* [4] M.P. Brodmann, Ch. Rotthaus, R.Y. Sharp, On annihilators and associated primes of local cohomology modules. J. Pure Appl. Algebra 153 (2000), 197–227.
* [5] M.P. Brodmann, R.Y Sharp, Local cohomology: an algebraic introduction with geometric applications, Cambridge University Press, 1998.
* [6] M.T. Dibaei, A. Nazari, Graded local cohomology: attached and associated primes, asymptotic behaviors, Comm.Alg. 35 (2007), 1567- 1576.
* [7] M.T. Dibaei, S. Yassemi, Associated primes of the local cohomology modules. Abelian groups, rings, modules, and homological algebra, 49–56, Chapman and Hall/CRC, 2006.
* [8] M.T . Dibaei, S. Yassemi, Associated primes and cofiniteness of local cohomology modules, manuscripta math, 117(2005), 199-205.
* [9] K. Divaani-Aazar, A. Mafi, Associated primes of local cohomology modules, Proc. Amer. Math. Soc. 133 (2005), 655–660 117(2005), 199-205.
* [10] A. Grothendieck, Cohomologie locale des faisceaux coh$\acute{e}$rents et th$\acute{e}$or$\grave{e}$mes de Lefschetz locaux et globaux (SGA 2), North-Holland, Amsterdam, 1968.
* [11] R. Hartshorne, _Affine duality and cofiniteness_ , Invent. Math. 9 (1970), 145–164.
* [12] C. Huneke, Problems on local cohomology :Free resolutions in commutative algebra and algebraic geometry, (Sundance, UT, 1990), 93-108, Jones and Bartlett, 1992.
* [13] T. Marley, The associated primes of local cohomology modules over rings of small dimension, manuscripta math. 104(2001), 519-525
* [14] L. Melkersson, Modules cofinite with respect to an ideal, J. Algebra. 285(2005), 649-668.
* [15] J. Rotman, An introduction to homological algebra, Academic Press, 1979.
| arxiv-papers | 2008-07-18T13:01:16 | 2024-09-04T02:48:56.871081 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Moharram Aghapournahr and Leif Melkersson",
"submitter": "Moharram Aghapournahr",
"url": "https://arxiv.org/abs/0807.2967"
} |
0807.3035 | Vol.0 (200x) No.0, 000–000
11institutetext: North West University (Potchefstroom Campus), School of
Physics (Unit for Space Research), Private Bag $X6001$, Potchefstroom $2531$,
Republic of South Africa.; gadzirai@gmail.com
# On the Radiation Problem of High Mass Stars 00footnotetext: ∗Supported by
the Republic of South Africa’s National Research Foundation and the North West
University, and Germany’s DAAD Programme via the University of
K$\ddot{\rm{o}}$ln.
G. G. Nyambuya
(Received $4$ Jan. $2010$ ; Accepted $2$ March $2010$)
###### Abstract
A massive star is defined to be one with mass greater than $\sim
8-10\mathcal{M}_{\odot}$. Central to the on-going debate on how these objects
[massive stars] come into being is the so-called Radiation Problem. For nearly
forty years, it has been argued that the radiation field emanating from
massive stars is high enough to cause a global reversal of direct radial in-
fall of material onto the nascent star. We argue that only in the case of a
non-spinning isolated star does the gravitational field of the nascent star
overcome the radiation field. An isolated non-spinning star is a non-spinning
star without any circumstellar material around it, and the gravitational field
beyond its surface is described exactly by Newton’s inverse square law. The
supposed fact that massive stars have a gravitational field that is much
stronger than their radiation field is drawn from the analysis of an isolated
massive star. In this case the gravitational field is much stronger than the
radiation field. This conclusion has been erroneously extended to the case of
massive stars enshrouded in gas & dust. We find that, for the case of a non-
spinning gravitating body where we take into consideration the circumstellar
material, that at $\sim 8-10\mathcal{M}_{\odot}$, the radiation field will not
reverse the radial in-fall of matter, but rather a stalemate between the
radiation and gravitational field will be achieved, i.e. in-fall is halted but
not reversed. This picture is very different from the common picture that is
projected and accepted in the popular literature that at $\sim
8-10\mathcal{M}_{\odot}$, all the circumstellar material, from the surface of
the star right up to the edge of the molecular core, is expected to be swept
away by the radiation field. We argue that massive stars should be able to
start their normal stellar processes if the molecular core from which they
form has some rotation, because a rotating core exhibits an Azimuthally
Symmetric Gravitational Field which causes there to be an accretion disk and
along this disk. The radiation field cannot be much stronger than the
gravitational field, hence this equatorial accretion disk becomes the channel
via which the nascent massive star accretes all of its material.
PACS (2010): $97.10.$Bt, $97.10.$Gz, $97.10.$Fy
###### keywords:
(stars:) circumstellar matter – (stars:) formation – radiative transfer.
## 1 Introduction
According to current and prevailing wisdom, it is bona-fide scientific
knowledge that our current understanding of massive star formation is lacking.
This is due to the existing theoretical and observational dichotomy. In the
gestation period of a star’s life, its mass will grow via the in-falling
envelope (i.e., circumstellar material) and also through the formation of an
accretion disk lying along the plane of it’s equator. As far as our
theoretical understanding is concerned, this works well for stars less than
about $8-10\mathcal{M}_{\odot}$. In the literature, it is said that the
problem of massive stars ($\mathcal{M}_{star}>8-10\mathcal{M}_{\odot}$) arises
because as the central protostar’s mass grows, so does the radiation pressure
from it, and at about $8-10\,\mathcal{M}_{\odot}$, the star’s radiation
pressure becomes powerful enough to halt any further in-fall of matter onto
the protostar (Larson & Starfield [$1971$]; Kahn [$1974$]; Yorke &
Kr$\ddot{\textrm{u}}$gel [${1977}$]; Wolfire & Cassinelli [$1987$]; Palla &
Stahler [$1993$]; Yorke [$2002$]; Yorke & Sonnhalter [${2003}$]). So the
problem is: how does the star continue to accumulate more mass beyond the
$8-10\,\mathcal{M}_{\odot}$ limit? If the radiation field really did reverse
any further in-fall of matter and protostars exclusively accumulated mass via
direct radial in-fall of matter onto the nascent star and also via the
accretion disk, this would set a mass upper limit of
$8-10\,\mathcal{M}_{\odot}$ for any star in the Universe. Unfortunately (or
maybe fortunately) this is not what we observe. It therefore means that some
process(es) responsible for the formation of stars beyond the
$8-10\,\mathcal{M}_{\odot}$ limit must be at work. A solution to the problem
must be sought because observations dictate that it exists.
If this is the case, i.e. the radiation problem really did exist as stated
above, and our physics where complete viz gravitation and radiation transport,
then, the solution to the conundrum would be to seek a star formation model
that overcomes the radiation pressure problem while at the sametime allowing
for the star to form (accumulate all of its mass) before it exhausts its
nuclear fuel. Two such (competing) models have been set-forth: ($1$) the
Accelerated Accretion Model (AAM) (Yorke [$2002$], [${2004}$]) and, ($2$) the
Coalescence Model (CM) (Bonnell et al. [$1998$], [${2001}$], [${2004}$],
[$2006$], [$2007$]; Bonnell & Bate [${2002}$]).
The latter scenario, the CM, is born out of the observational fact that
massive stars are generally found in the centers of dense clusters (see e.g.
Hillenbrand [$1997$]; Clarke et al. [$2000$]). In these dense environments,
the probability of collision of proto-stellar objects is significant, leading
to the CM. This model easily by-passes the radiation pressure problem and,
despite the fact that not a single observation to date has confirmed it
(directly or indirectly), it [CM] appears111This relies on the assumption that
our understanding of gravitation and radiation transport is complete. to be
the most natural mechanism by which massive stars form given the said
observational fact about massive stars and their preferential environment.
The AAM is just a scaled up version of the accepted accretion paradigm
applicable to Low Mass Stars (LMSs). This accretion takes place via the
accretion disk and, for the reason mentioned above that the accretion
mechanism must be such that it allows for the star to form before it exhausts
its nuclear fuel, the accretion cannot take place at the same steady rate as
in the case of LMSs ($\mathcal{M}\leq\,3\mathcal{M}_{\odot}$) but must be
accelerated and significantly higher. While there exists many examples of
massive stars surrounded by accretion disks, one of the chief obstacles in
verifying this paradigm is that examples of HMSs tend to be relatively distant
($>1\,\textrm{kpc}$), deeply embedded, and confused with other emission
sources (see e.g. Mathews et al. [$2007$]). Additionally, HMSs evolve rapidly,
and by the time an unobstructed view of the young star emerges, the disk and
outflow structures may have been destroyed. Consequently, observations to date
have been unable to probe the $10-100\,\textrm{AU}$ spatial scales over which
outflows from the accretion disks are expected to be launched and collimated
(e.g. Mathews et al. [$2007$]).
The other alternative, which is less pursued, would be to seek a physical
mechanism that overcomes the radiation pressure problem as has been conducted
by the authors Krumholz et al. ([${2005}$], [${2009}$]). These authors
(Krumholz et al. [${2005}$], [${2009}$]) believe that the radiation problem
does not exist because radiation-driven bubbles that block accreting gas are
subject to Rayleigh-Taylor instability which occurs anytime a dense, heavy
fluid is being accelerated by lighter fluid, for example, when a cloud
receives a shock, or when a fluid of a certain density floats above a fluid of
lesser density, such as dense oil floating on water. The Rayleigh-Taylor
instabilities allow fingers of dense gas to break into the evacuated bubbles
and reach the stellar surface while in addition, outflows from massive stars
create optically thin cavities in the accreting envelope. These channel
radiation away from the bulk of the gas and reduce the radiation pressure it
experiences. In this case, the radiation pressure feedback is not the dominant
factor in setting the final size of massive stars and accretion will proceed,
albeit at much higher rates. Amongst others, the model by the authors Krumholz
et al. ([${2005}$], [${2009}$]) is ad hoc rather than natural, in that Nature
has to make a special arrangement or must configure herself in such a way that
massive stars have a way of starting their normal stellar processes. Does
there not exist a smooth and natural way to bring massive stars into
existence?
In this reading, we redefine the radiation problem (for the spherically
symmetric case) and we do this via a subtle and overlooked assumption made in
the analysis leading to the radiation problem: that the surroundings of the
protostar is a vacuum (see e.g. Yorke [$2002$]; Yorke & Sonnhalter [${2003}$];
Zinnecker & Yorke [${2007}$]); surely, this is clearly not true. The
researchers Yorke [$2002$]; Yorke & Sonnhalter [$2002$]; Zinnecker & Yorke
[${2007}$]; among others, hold the view that from a theoretical stand-point,
the radiation field is stronger than the gravitational field for massive stars
hence the in-fall process of material must be reversed; but this conclusion
has been reached, as will be shown in the next section; after comparing the
gravitational field strength at point $r$ of a star in empty space to its
radiation field strength at point $r$. In practice, stars are found embedded
inside a significant mass of gas and dust. The radiation problem is arguably
the most important problem of all in the study of the formation of stars,
thus, it is important to make sure that this problem is clearly defined and
understood.
Having taken into consideration the circumstellar material, we find that at
$\sim 8-10\mathcal{M}_{\odot}$, the radiation field will not reverse the
radial in-fall of matter but rather a stalemate between the radiation and
gravitational field will be achieved, where in-fall is halted but not
reversed. Certainly, this picture is not at all congruent (or somewhere near
there) to the common picture that is accepted in the popular literature where
at $\sim 8-10\mathcal{M}_{\odot}$, all the circumstellar material, from the
surface of the star right up to the edge of the molecular cloud core, is
expected to be swept away by the powerful radiation field. This finding is not
a complete but rather a partial solution to the radiation problem in that
beyond the $8-10\mathcal{M}_{\odot}$ limit, the nascent star will not accrete
any further. Under this model, its mass will stay at this value; it accretes
from the stagnant and frozen envelope once its mass drops below this
$8-10\mathcal{M}_{\odot}$ limit. A very important point to note is that this
is for a spherically symmetric gravitational setting where the gravitational
field only has the radial dependence and is exactly described by Newton’s
inverse square law.
In a different reading, Nyambuya ([${2010a}$]), an Azimuthally Symmetric
Theory of Gravitation (ASTG) was set-up and thereby a thesis was set-forth to
the effect that: ($1$) for a non-spinning star, its gravitational field is
spherically symmetric, so it depends on the radial distance from the central
body; ($2$) for a spinning gravitating body, the gravitational field of the
body in question is azimuthally symmetric, that is to say, it is dependent on
the radial distance from the central body and as-well the azimuthal angle. In
a follow-up reading of Nyambuya ([${2010b}$]), it has been shown that the ASTG
predicts ($1$) that bipolar outflows may very well be a purely gravitational
phenomenon and also that; ($2$) along the spin-equator of a spinning
gravitating body, gravity will channel matter onto the spinning nascent star
via the spin-equatorial disk without radiation having to reverse this inflow,
thus allowing stars beyond the critical mass $8-10\mathcal{M}_{\odot}$ to come
into existence.
If the ASTG proves itself, then the present reading together with Nyambuya
([${2010a}$], [${2010b}$]) comprise (in our view) a solution to the radiation
problem. Given that the solution to this problem has been sought via
sophisticated computer simulations and lengthy numerical solutions, and
additionally, given the simplicity and naïvity of the present approach which
seeks to further our understanding of this problem, perhaps this reading
presents not only my misunderstanding of the problem, but also of the approach
to the problem. But more on the optimistic side of things, I believe the
radiation problem as discussed herein has been understood and that the
approach is mathematically and physically legitimate, so much that we hold the
objective view that to they [i.e. other researchers] that seek a solution to
this problem, this reading is something worthwhile.
## 2 The Radiation Problem
Following Yorke [$2002$]; for direct radial accretion and accretion via the
disk to occur onto the nascent star, it is required that the Newtonian
gravitational force, $G\mathcal{M}_{star}(t)/r^{2}$, at a point distance $r$
from the star of mass $\mathcal{M}_{star}$ and luminosity
$\mathcal{L}_{star}(t)$ at any time $t$, must exceed the radiation force
$\kappa_{eff}\mathcal{L}_{star}(t)/4\pi cr^{2}$ i.e.:
$\frac{G\mathcal{M}_{star}(t)}{r^{2}}>\frac{\kappa_{eff}\mathcal{L}_{star}(t)}{4\pi
cr^{2}},$ (1)
where $c=2.99792458\times 10^{8}\,\textrm{ms}^{-1}$ is the speed of light in a
vacuum, $G=6.667\times
10^{-11}\,\textrm{kg}^{-1}\textrm{m}^{3}\textrm{s}^{-2}$ is Newton’s universal
constant of gravitation, $\kappa_{eff}$ is the effective opacity which is the
measure of the gas’ state of being opaque or a measure of the gas’
imperviousness to light rays and is measured in
$\textrm{m}^{2}\textrm{kg}^{-1}$. This analysis by Yorke ([$2002$]), which is
also reproduced in Zinnecker & Yorke ([${2007}$]), is a standard and well
accepted analysis that assumes spherical symmetry and, at the same time, it
does not take into account the nascent star’s circumstellar material. On the
other hand, star formation is not a truly spherically symmetric phenomenon
(see e.g. reviews by Zinnecker & Yorke [${2007}$]; McKee & Ostrikker [$2007$])
but this simple calculation suffices in as far as defining the curtain-region
of $8-10\mathcal{M}_{\odot}$ when radiation pressure is expected to become a
significant player in the star’s formation. What will be done in this reading
is simply to perform the same calculation albeit with the circumstellar
material taken into account. In the penultimate of this section, we shall make
our case based on the above statements.
Now, this calculation by Yorke ([$2002$]) and Zinnecker & Yorke ([${2007}$]),
proceeds as follows: the inequality (1), sets a maximum condition for
accretion of material, namely $\kappa_{eff}<4\pi
cG\mathcal{M}_{star}(t)/\mathcal{L}_{star}(t)$, and evaluating this we obtain:
$\kappa_{eff}<1.30\times
10^{4}\left(\frac{\mathcal{M}_{star}(t)}{\mathcal{M}_{\odot}}\right)\left(\frac{\mathcal{L}_{star}(t)}{\mathcal{L}_{\odot}}\right)^{-1},$
(2)
where $\mathcal{M}_{star}(t)$ and $\mathcal{L}_{star}(t)$ are in solar units.
Given that,
$\mathcal{L}_{star}(t)=\mathcal{L}_{\odot}\left(\mathcal{M}_{star}(t)/\mathcal{M}_{\odot}\right)^{3}$,
this implies that:
$\kappa_{eff}<1.30\times
10^{4}\left(\frac{\mathcal{M}_{star}(t)}{\mathcal{M}_{\odot}}\right)^{-2}\Rightarrow\left(\frac{\mathcal{M}_{star}}{\mathcal{M}_{\odot}}\right)>\left(\frac{1.30\times
10^{4}}{\kappa_{eff}}\right)^{1/2}.$ (3)
Now, given that the dusty Interstellar Medium’s (ISM) averaged opacity is
measured to be about $20.0$ $\textrm{m}^{2}\textrm{kg}^{-1}$ (Yorke [$2002$])
and using this (as an estimate to setting the minimum critical mass, see Yorke
[$2002$]; Zinnecker & Yorke [${2007}$]), we find that this sets a minimum
upper mass limit for stars of about $10\mathcal{M}_{\odot}$ for gravitation to
dominate the scene before radiation does. It is clear here that the opacity of
the molecular cloud material is what sets the critical mass, thus a cloud of
lower opacity will have a higher critical mass. It is expected that the
opacity inside the cloud will be lower than in the ISM. In adopting the value
$\kappa_{eff}=20.0\,\textrm{m}^{2}\textrm{kg}^{-1}$ (see Yorke [$2002$];
Zinnecker & Yorke [${2007}$]), this was done only to set a minimum lower bound
for massive stars. Dust and gas opacities are significantly frequency-
dependent and one has to take this into account for a more rigid constraint of
a minimum mass for when the radiation field is expected to overcome the
gravitational field.
As can be found in Yorke ([$2002$]), the AAM finds some of its ground around
the alteration of the opacity. For example, if the opacity inside the gas
cloud is significantly lower than the ISM value, then accretion can proceed
via the AAM. To reduce the opacity inside the gas & dust cloud, the AAM posits
as one of the its options that optical and Ultra-Violet (UV) radiation inside
the accreting material is shifted from the optical/UV into the far Infrared
(IR) and also that the opacity may be lower than the ISM value because the
opacity will be reduced by the accretion of optically thick material in the
blobs of the accretion disk. Thus reducing the opacity, or finding a physical
mechanism that reduces the opacity to values lower than the ISM, is a viable
solution to the radiation problem. The above mechanism to reduce the opacity
is rather ad hoc and dependent on the environment.
Now that we have presented the radiation problem as it is commonly understood,
we are ready to make our case by inspecting (1). Clearly and without any
doubt, the left hand side of this inequality is the gravitational field
intensity for a gravitating body in empty space while the right hand side is
the radiation field of this same star in empty space. From this, clearly, we
are actually comparing the radiation and gravitational field intensity of a
star in empty space, whereas for the real setting in Nature, stars are found
heavily enshrouded by gas and dust. Clearly, the conclusions that one finds
from (1) such as that, at about $8-10\mathcal{M}_{\odot}$, the radiation field
of the nascent star is powerful enough to not only halt but reverse the in-
fall of material onto the nascent star; this cannot be extended to the
scenario where a star is submerged in gas and dust. It is erroneous to do so.
Clearly, at this very simplistic, naive and fundamental level, there is a need
to redefine the radiation problem by including in the left hand side of (1),
the circumstellar material. Wolfire & Cassinelli ([$1987$]) among others, have
performed this calculation where they have taken into account the
circumstellar material and reached similar conclusions (as e.g. those of Yorke
[$2002$]). We reach a different conclusion to that of Wolfire & Cassinelli
([$1987$]) because, unlike these researchers, we use the observational fact
that molecular clouds and molecular cores are found exhibiting a well behaved
density profile $\rho\propto r^{-\alpha_{\rho}}$, and from this, we calculate
a general mass distribution ($\mathcal{M}\propto r^{-\alpha}$). We use this to
compare the gravitational and radiation field strengths at point $r$ and from
there draw our interesting conclusions.
## 3 Radiation and the Circumstellar Material
Neglecting thermal, magnetic effects, turbulence and any other forces (as will
be shown latter in this section, these forces do not change the essence of our
argument, hence we do not need to worry about them here) and considering only
the gravitational and radiation field from the nascent star, we assume here
that a star is formed from a gravitationally bound system of material enclosed
in a volume space of radius $\mathcal{R}_{core}(t)$ and we shall call this
system of material the core and further assume that this core shall have a
constant total mass $\mathcal{M}_{core}$ at all times. Now, as long as the
material enclosed in the sphere of radius $r<\mathcal{R}_{core}(t)$ is such
that:
$\frac{G\mathcal{M}(r,t)}{r^{2}}>\frac{\kappa_{eff}\mathcal{L}_{star}(t)}{4\pi
cr^{2}},$ (4)
then, radiation pressure will not exceed the gravitational force in the region
$r<\mathcal{R}_{core}(t)$, thus direct radial in-fall is expected to continue
in that region. If $\mathcal{M}_{csl}(r,t)$ is the mass of the circumstellar
material at time $t$ enclosed in the region stretching from the surface of the
star to the radius $r$, then,
$\mathcal{M}(r,t)=\mathcal{M}_{csl}(r,t)+\mathcal{M}_{star}(t)$. Hence, the
difference between (4) and (1) is that in (4) we have included the
circumstellar material. This is not the whole story.
Now, (4) can be written differently as:
$\mathcal{M}(r,t)>\frac{\kappa_{eff}\mathcal{L}_{star}(t)}{4\pi Gc},$ (5)
which basically says as long as the amount of matter enclosed in the region of
sphere radius $r$ satisfies the above condition, the radiation force will not
exceed the gravitational force in that region of radius $r$. In fact, (5) is
the Eddington limit applied to the region of radius $r$. This is identical to
equation ($10$) in Wolfire & Cassinelli ([$1987$]). In their work, Wolfire &
Cassinelli ([$1987$]) solve numerically the radiative transfer problem to
determine the effective opacity at the outer edge of the massive star forming
core and, from this, they determine the limits of grain-sizes that are needed
for the formation of massive stars. Wolfire & Cassinelli ([$1987$])’s approach
is a typical approach used to probe the conditions necessary for massive stars
to form.
Our approach is very different from that of Wolfire & Cassinelli ([$1987$])
and most typical approaches used to study the radiation problem where
sophisticated computer simulations and numerical solutions are used. Ours is a
simple and naïve approach needing no computer simulations nor numerical codes.
We shall insert
$\mathcal{M}(r,t)=\mathcal{M}_{csl}(r,t)+\mathcal{M}_{star}(t)$ into (4) and
after rearranging, one obtains:
$\mathcal{M}_{csl}(r,t)>\left[\frac{\kappa_{eff}\mathcal{L}_{star}(t)}{4\pi
cG\mathcal{M}_{star}(t)}-1\right]\mathcal{M}_{star}(t)=\left[\left(\frac{\mathcal{M}_{star}(t)}{10\mathcal{M}_{\odot}}\right)^{2}-1\right]\mathcal{M}_{star}(t){.}$
(6)
Our main thrust is to seek values of $r$ in the above inequality that satisfy
it. We shall do this by finding a form for $\mathcal{M}_{csl}(r,t)$.
Before doing this, let us apply (5) to the entire core, that is
$r=\mathcal{R}_{core}$. This must give us the condition when the star’s
radiation field is strong enough to sweep away all the circumstellar material
from the surface of the star right up to the outer edge of the core. In doing
so, one finds that the star’s luminosity should be such that:
$\mathcal{M}_{core}>\frac{\kappa_{eff}\mathcal{L}_{star}(t)}{4\pi Gc}.$ (7)
In making this calculation, we have made the tacit and fundamental assumption
that the star’s mass will continue to increase until the star reaches a
critical luminosity determined by the mass of the core. Let us denote this
critical luminosity by $\mathcal{L}^{*}_{core}$. From the above, it follows
that:
$\mathcal{L}^{*}_{core}=\frac{4\pi cG\mathcal{M}_{core}}{\kappa_{eff}}.$ (8)
With this definition, then for the radiation field to globally overcome the
gravitational field, the nascent star’s luminosity must exceed the critical
luminosity of the core, i.e.:
$\mathcal{L}_{star}(t)>\mathcal{L}^{*}_{core}.$ (9)
Now, knowing the mass-luminosity relationship of stars is given by
$\mathcal{L}_{star}(t)=\mathcal{L}_{\odot}\left(\mathcal{M}(t)/\mathcal{M}_{\odot}\right)^{3}$,
then the critical condition $\mathcal{L}_{star}(t)=\mathcal{L}^{*}_{core}$
will occur when:
$\left(\frac{\mathcal{M}_{star}}{\mathcal{M}_{\odot}}\right)=\left(\frac{\kappa_{eff}\mathcal{L}_{\odot}}{4\pi
G\mathcal{M}_{\odot}c}\right)^{-1/3}\left(\frac{\mathcal{M}_{core}}{\mathcal{M}_{\odot}}\right)^{1/3}.$
(10)
Given this and taking $\kappa_{eff}=20.0\,\textrm{m}^{2}\textrm{kg}^{-1}$ and
then plugging these and the other relevant values, such as $G,c$, etc, into
the above, we are lead to:
$\left(\frac{\mathcal{M}_{max}}{\mathcal{M}_{\odot}}\right)=\left(\frac{\mathcal{M}_{core}}{10\mathcal{M}_{\odot}}\right)^{1/3}.$
(11)
where we have set $\mathcal{M}_{star}=\mathcal{M}_{max}$. As we already said,
using $\kappa_{eff}=20.0\,\textrm{m}^{2}\textrm{kg}^{-1}$ gives us the minimum
lower bound. What this means is that the mass of the core from which a star is
formed may very well be crucial in deciding the final mass of the star because
the mass of the core determines the time when global in-fall reversal will
occur.
From this simplistic and rather naïve calculation, we can estimate the
efficiency of the core:
$\xi_{core}=\left(\frac{\mathcal{M}_{star}}{\mathcal{M}_{core}}\right)=0.10\left(\frac{\mathcal{M}_{core}}{10\mathcal{M}_{\odot}}\right)^{-2/3},$
(12)
thus a $100\mathcal{M}_{\odot}$ core will (according to the above) form a star
at an efficiency rate of about $2\%$ and it will produce a star of mass
$2\mathcal{M}_{\odot}$. A $10\mathcal{M}_{\odot}$ star will be produced by a
core of mass $10^{4}\mathcal{M}_{\odot}$ at an efficiency rate of about
$0.1\%$. A $10^{4}\mathcal{M}_{\odot}$ core is basically a fully-fledged
molecular cloud. The production of this $10\mathcal{M}_{\odot}$ star is based
on the assumption that the rest of the material
($10^{4}\mathcal{M}_{\odot}-10\mathcal{M}_{\odot}$ = $9.99\times
10^{3}\mathcal{M}_{\odot}$) will not form stars. In reality, some of the
material in this $10^{4}\mathcal{M}_{\odot}$ core will form many other stars.
Furthermore, a $100\mathcal{M}_{\odot}$ star will form in a GMC of mass about
$10^{7}\mathcal{M}_{\odot}$. The above deductions, that high mass stars will
need to form in clouds of mass $\geq 10^{4}\mathcal{M}_{\odot}$, is in
resonance with the observational fact that massive stars are not found in
isolation (e.g. Hillenbrand [$1997$]; Clarke et al. [$2000$]) since the other
material will form stars.
Relationship (11) is interesting because of its similarity to Larson’s $1982$
empirical discovery. With a handful of data, Larson ([$1982$]) was the first
to note that the maximum stellar mass of a given population of stars is
related to the total mass of the parent cloud from which the stellar
population has been born. That is to say, if $\mathcal{M}_{cl}$ is the mass of
a molecular cloud and $\mathcal{M}_{max}$ is the maximum stellar mass of the
population, then:
$\mathcal{M}_{max}=\left(\frac{\mathcal{M}_{cl}}{\mathcal{M}_{0}}\right)^{\alpha_{L}}$
(13)
where $\mathcal{M}_{0}=13.2\mathcal{M}_{\odot}$ and $\alpha_{L}=0.430$. This
law was obtained from a sample of molecular clouds whose masses are in the
range $1.30\leq\log_{10}\left(\mathcal{M}/\mathcal{M}_{\odot}\right)\leq
5.50$. Larson’s Law is thought to be a result of statistical sampling but we
are not persuaded to think that this is the case; such a coincidence is, in
our opinion and understanding, to good to be true. We believe Larson’s Law is
Nature’s subtle message to researchers; it is telling us something about the
underlaying dynamics of star formation. This said, could the relationship (11)
be related to Larson’s result? The indices of Larson’s relation and
relationship (11) have a deviation of about $33\%$ and the constant
$\mathcal{M}_{0}$ has a similar deviation of about $33\%$. Could Larson’s
fitting procedure be “tuned” to conform to relationship (11) and if so, does
that mean Larson’s relationship finds an explanation from this behavior?
Perhaps the deviation of our relation from that of Larson may well be that our
result is derived from an ideal situation where we have considered not the
other forces, such as the magnetic, thermal forces etc, but also we have
considered star formation as a spherically symmetric process, which it is not,
and this may also be a source of correction to this result in order to bring
it into agreement with Larson’s result. Let us represent all these other
forces by $\vec{\textbf{F}}_{other}$ (e.g. magnetic, turbulence, viscocity
etc). Clearly these forces will not aid gravity in its endeavor to squeeze all
the material to a single point but rather aid the radiation pressure in
opposing this. Given this, we must write Inequality (4) as:
$\frac{G\mathcal{M}(r,t)}{r^{2}}>\frac{\kappa_{eff}\mathcal{L}_{star}(t)}{4\pi
cr^{2}}+\frac{|\vec{\textbf{F}}_{other}|}{m},$ (14)
where $m$ is the average mass of the molecular species of the material
constituting the cloud. The above can be written in the form:
$\mathcal{L}_{star}(t)<\frac{4\pi
cG\left(\mathcal{M}(r,t)-r^{2}|\vec{\textbf{F}}_{other}|/m\right)}{\kappa_{eff}},$
(15)
and writing $\mathcal{M}^{\prime}(r,t)=r^{2}|\vec{\textbf{F}}_{other}|/m$, we
will have:
$\mathcal{L}_{star}(t)<\frac{4\pi
cG\left[\mathcal{M}(r,t)-\mathcal{M}^{\prime}(r,t)\right]}{\kappa_{eff}}{.}$
(16)
From this, it is clear that the other forces will act in a manner as to reduce
the critical luminosity of the core. Thus our result (11), when compared to
natural reality where these other forces are present, is expected to show that
deviation from the real observations must occur. As stated in the opening of
this section, the inclusion of the magnetic and thermal forces, etc, will not
change the essence of our argument, hence the above argument justifies why we
did not have to worry about these other forces because the essence of our
result still stands. The situation is only critical when these other forces
become significant in comparison to the gravitational force.
In the succeeding section, we compute the mass distribution function and then
show that one arrives at the same result as (5). Additionally and more
importantly, we are able to compute the boundaries where the radiation field
will be strong enough to overcome the gravitational field. Among other
interesting outcomes, we shall see that the radiation field will create a
cavity inside the star forming core and that this cavity grows with time in
proportion to the radiation field of the nascent star.
## 4 Mass Distribution Function
First, we compute the enclosed mass $\mathcal{M}(r,t)$. We know that stellar
systems such as molecular clouds and cores are found to exhibit radial density
profiles given by:
$\rho(r,t)=\rho_{0}(t)\left(\frac{r_{0}(t)}{r}\right)^{\alpha_{\rho}}$ (17)
where $\rho_{0}(t)$ and $r_{0}(t)$ are time dependent normalization constants
and $\alpha_{\rho}$ is the density index. In order to make sense of this
density profile (17), we have to calculate these normalization constants. In
its bare form, the power law (17) as it stands implies an infinite density at
$r=0$. In general, power laws have this property. Obviously, one has to deal
with this. The usual or typical way is to impose a minimum value for $r$, say
$r=r_{min}=r_{0}(t)$ and, assign a density there. Here, this minimum radius
has been made time dependent for the sole reason that if the cloud is
undergoing free fall as in the case in star formation regions, this quantity
will respond dynamically to this, so it will be time dependent.
Now, for a radially dependent density profile, the mass distribution is
calculated from the integral:
$\mathcal{M}(r,t)=\int^{r}_{r_{min}}4\pi r^{2}\rho(r,t)dr.$ (18)
Inserting the density function (17) into the above integral and then
evaluating the resultant integral, we are led to:
$\mathcal{M}(r,t)=\left(\frac{4\pi\rho_{0}(t)r^{\alpha_{\rho}}_{min}(t)}{3-\alpha_{\rho}}\right)\left(r^{3-\alpha_{\rho}}-r_{min}^{3-\alpha_{\rho}}(t)\right),$
(19)
and this formula does not apply to the case $\alpha_{\rho}=3$. This is valid
for $0\leq\alpha_{\rho}<3$. The case $\alpha_{\rho}=3$ is described by a
special MDF which is
$\mathcal{M}(r,t)=\left[4\pi\rho_{0}(t)r^{3}_{min}(t)\right]\ln\left(r/r_{min}(t)\right)$.
We shall not consider this case as it will not change the essence of our
argument.
Now, what we shall do here is to constrain $\alpha_{\rho}$ and show that:
$0\leq\alpha_{\rho}<3$. This exercise is being conducted to define the domain
in which our result has physical significance. First we shall establish that
$\alpha_{\rho}<3$ and this we shall do by using the method of proof by
contradiction. Let ($r_{2}>r_{1}$). For this setting, we expect that
[$\mathcal{M}(r_{2})>\mathcal{M}(r_{1})$] which is obvious because as one
zooms out of the molecular cloud, one would expect to have more matter in a
bigger sphere of radius $r_{2}$ than that enclosed in a smaller sphere of
radius $r_{1}$. Therefore, our condition is:
[$r_{2}>r_{1}\Longrightarrow\mathcal{M}(r_{2})-\mathcal{M}(r_{1})\geq 0$].
Using equation (19), we have:
$\mathcal{M}(r_{2})-\mathcal{M}(r_{1})=\left(\frac{4\pi\rho_{0}r^{\alpha_{\rho}}_{min}}{3-\alpha_{\rho}}\right)\left(r^{3-\alpha_{\rho}}_{2}-r_{1}^{3-\alpha_{\rho}}\right)>0,$
(20)
and for ($\alpha_{\rho}>3$) we have ($3-\alpha_{\rho}<0$) so when we divide by
the term $(4\pi\rho_{0}r^{\alpha}_{min})/(3-\alpha_{\rho})$ on both sides of
the inequality, we must change the sign of the inequality from $>$ to $<$
because $(4\pi\rho_{0}r^{\alpha_{\rho}}_{min})/(3-\alpha_{\rho})$ is a
negative number. In so doing, we will have:
$r^{3-\alpha_{\rho}}_{2}-r_{1}^{3-\alpha_{\rho}}<0$, and this implies
$r^{3-\alpha_{\rho}}_{2}<r_{1}^{3-\alpha_{\rho}}$ and from this directly
follows the relationship:
$r_{1}<r_{2}{.}$ (21)
This is a clear contradiction because it violates our initial condition
[$r_{2}>r_{1}\Longrightarrow\mathcal{M}(r_{2})>\mathcal{M}(r_{1})$] as this is
saying [$r_{2}<r_{1}\Longrightarrow\mathcal{M}(r_{2})>\mathcal{M}(r_{1})$]
which is certainly wrong. From a purely mathematical stand-point, we are
therefore forced to conclude that $\alpha_{\rho}<3$ if the condition
[$r_{2}>r_{1}\Longrightarrow\mathcal{M}(r_{2})>\mathcal{M}(r_{1})$] is to hold
– QED.
Now we shall establish that $\alpha_{\rho}\geq 0$ and we shall do this using
physical arguments. If $3-\alpha_{\rho}>3$, as one zooms out of the cloud from
the center, the cloud’s average material density increases. This scenario is
unphysical because gravity is an attractive inverse distance law and thus will
always pack more and more material in the center than in the outer regions.
Hence, the only material configuration that can emerge from this setting is
one in which the average density of material decreases as one zooms out of the
cloud. This implies $3-\alpha_{\rho}\leq 3$ which leads to $\alpha_{\rho}\geq
0$. Thus combining the two results, we are going to have:
$0\leq\alpha_{\rho}<3$. Now we have defined the physical boundaries of the
density profile.
Now we have to normalize the MDF by imposing some boundary conditions. The
usual or traditional boundary condition is to set $\mathcal{M}(r_{min},t)=0$
and this in fact means there will be a cavity of radius $r_{min}(t)$ in the
cloud. What we shall do next is different from this traditional normalization.
We shall set $\mathcal{M}(r_{min},t)=\mathcal{M}_{star}$ where
$\mathcal{M}_{star}$ is the mass of the central star, hence
$r_{min}(t)=\mathcal{R}_{star}(t)$. Thus what we have done is to place the
nascent star in the cavity which means we must write our MDF as:
$\mathcal{M}(r,t)=\left(\frac{4\pi\rho_{0}(t)\mathcal{R}_{star}^{\alpha_{\rho}}(t)}{3-\alpha_{\rho}}\right)\left(r^{3-\alpha_{\rho}}-\mathcal{R}_{star}^{3-\alpha_{\rho}}(t)\right)+\mathcal{M}_{star}(t),$
(22)
and this applies for $\mathcal{R}_{star}(t)\leq r\leq\mathcal{R}_{core}(t)$.
Now, if the mass enclosed inside the core remains constant throughout, then we
must have at $r=\mathcal{R}_{core}(t)$ the boundary condition
$\mathcal{M}(\mathcal{R}_{core},t)=\mathcal{M}_{core}$. We know that the sum
total of all the circumstellar material at any given time is given by:
$\mathcal{M}_{csl}(t)=\mathcal{M}_{core}-\mathcal{M}_{star}(t)$. Combining all
the information, we will have:
$\left(\frac{4\pi\rho_{0}(t)r^{\alpha_{\rho}}_{0}(t)}{3-\alpha_{\rho}}\right)=\frac{\mathcal{M}_{csl}(t)}{\mathcal{R}^{3-\alpha_{\rho}}_{core}(t)-\mathcal{R}_{star}^{3-\alpha_{\rho}}(t)},$
(23)
and this means the MDF can now be written as:
$\mathcal{M}(r,t)=\overbrace{\mathcal{M}_{csl}(t)\left(\frac{r^{3-\alpha_{\rho}}-\mathcal{R}_{star}^{3-\alpha_{\rho}}(t)}{\mathcal{R}^{3-\alpha_{\rho}}_{core}(t)-\mathcal{R}_{star}^{3-\alpha_{\rho}}(t)}\right)}^{\tiny\textbf{Circumstellar}\,\,\textbf{Material}\,\,\textbf{in}\,\,\textbf{Region}\,\,\textbf{Radius}\,\,\textbf{r}}+\overbrace{\mathcal{M}_{star}(t)}^{\tiny\textbf{Mass}\,\,\textbf{of}\,\,\textbf{the}\,\,\textbf{nascent}\,\,\textbf{star}}\,\,\,\,\,\,\,\,\,\textrm{for}\,\,\,\,\,\,r\geq\mathcal{R}_{star}(t).$
(24)
We shall take this as the final form of our mass distribution function. If the
reader accepts this, then what follows is a straight forward exercise and
leads to what we believe is a significant step forward in the resolution of
the radiation problem. The reader may want to query that we have overstretched
our boundary limits by making the assumption that the MDF be continuous from
the surface of the star right up to the edge of the core. In that event, we
need to make this point clear and reach an accord.
First, let us consider a serene molecular core long before a star begins to
form at the center. We know that the density is not a fundamental physical
quantity but a physical quantity derived from two fundamental physical
quantities which are mass and volume, i.e., density=mass/volume. We must note
that this is defined for (volume$>0$). We shall assume that this core exhibits
the density profile $\rho\propto r^{-\alpha_{\rho}}$. This fact that
$\rho\propto r^{-\alpha_{\rho}}$, when combined with the fact that density is
not a fundamental physical quantity but a quantity derived from two
fundamental quantities, suggests that at any given time the mass must be
distributed in proportion to the radius, i.e., $\mathcal{M}(r,t)\propto
r^{\alpha}$. The radial dependency of the density is an indicator that that
mass has a radial dependency. The relationship $\mathcal{M}(r,t)\propto
r^{\alpha}$ means we must have $\mathcal{M}(r,t)=ar^{\alpha}+b$ where ($a,b$)
are constants. We expect that $\mathcal{M}(0,t)=0$. If this is to hold (as it
must), then ($b=0$) and ($\alpha\geq 0$). We also expect the condition
$\mathcal{M}(\mathcal{R}_{core},t)=\mathcal{M}_{core}$ to hold. If this is to
hold (as it must), then we will have
$a=\mathcal{M}_{core}/\mathcal{R}_{core}^{\alpha}(t)$ hence
$\mathcal{M}(r,t)=\mathcal{M}_{core}(r/\mathcal{R}_{core}(t))^{\alpha}$. From
the definition of density this means:
$\rho(r,t)=\left(\frac{3\mathcal{M}_{core}}{4\pi\mathcal{R}_{core}^{\alpha}(t)}\right)r^{\alpha-3}\,\,\,\,\,\,\,\,\textrm{for}\,\,\,\,\,\,\,\,\,\,r>0.$
(25)
Now, if the density profile is to fall off as $r$ increases as is the case in
Nature, then ($\alpha-3\leq 0$) which implies ($\alpha\leq 3$). Combining this
with ($\alpha\geq 0$) we will have ($0\leq\alpha\leq 3$). Comparing this with
the profile ($\rho\propto r^{-\alpha_{\rho}}$), we have:
($-\alpha_{\rho}=\alpha-3$) and substituting this into ($0\leq\alpha\leq 3$),
one obtains ($0\leq 3-\alpha_{\rho}\leq 3$). From ($3-\alpha_{\rho}\leq 3$),
we have ($\alpha_{\rho}\geq 0$), and from ($0\leq 3-\alpha_{\rho}$), we have
($\alpha_{\rho}\leq 3$), hence ($0\leq\alpha_{\rho}\leq 3$).
Now, in this serene molecular cloud, a small lamp begins to form; let this
lamp have a radius $\mathcal{R}_{lamp}(t)$ and mass $\mathcal{M}_{lamp}$. I
shall pose a question: do we expect this lamp to cause any fundamental changes
to the mass distribution $\mathcal{M}(r,t)=ar^{\alpha}+b$? I think not. If
this is the case, then our mass distribution must now be defined up to the
radius of the lamp, $\mathcal{M}(\mathcal{R}_{lamp},t)=\mathcal{M}_{lamp}$ and
this condition leads to: $b=\mathcal{M}_{lamp}-a\mathcal{R}_{lamp}^{\alpha}$,
thus:
$\mathcal{M}(r,t)=a(r^{3-\alpha_{\rho}}-\mathcal{R}_{lamp}^{3-\alpha_{\rho}})+\mathcal{M}_{lamp}\,\,\,\,\,\,\,\,\,\textrm{for}\,\,\,\,\,\,r\geq\mathcal{R}_{lamp}(t).$
(26)
where we have substituted $\alpha=3-\alpha_{\rho}$. Now inserting the
condition that $\mathcal{M}(\mathcal{R}_{core},t)=\mathcal{M}_{core}$, we will
have:
$a=\left(\frac{\mathcal{M}(r,t)-\mathcal{M}_{lamp}}{\mathcal{R}^{3-\alpha_{\rho}}_{core}(t)-\mathcal{R}_{lamp}^{3-\alpha_{\rho}}}\right)\,\,\,\,\,\,\,\,\,\textrm{for}\,\,\,\,\,\,r\geq\mathcal{R}_{lamp}(t).$
(27)
and putting all this together we will have:
$\mathcal{M}(r,t)=\mathcal{M}_{csl}(t)\left(\frac{r^{3-\alpha_{\rho}}-\mathcal{R}_{lamp}^{3-\alpha_{\rho}}}{\mathcal{R}^{3-\alpha_{\rho}}_{core}(t)-\mathcal{R}_{lamp}^{3-\alpha_{\rho}}}\right)+\mathcal{M}_{lamp}\,\,\,\,\,\,\,\,\,\textrm{for}\,\,\,\,\,\,r\geq\mathcal{R}_{lamp}(t).$
(28)
where $\mathcal{M}_{csl}(t)=\mathcal{M}_{core}-\mathcal{M}_{lamp}(t)$.
Comparison of the above with (24) shows that the lamp in the above formula is
the star in (24).
We are certain that the reader will have no problem with (28) because the lamp
does not disrupt the mass distribution since it has no radiation. Hence, we
would expect a continuous distribution of mass right up to the surface of the
lamp as material will be flowing into the lamp. However, this same lamp is a
protostar and at somepoint it must switch on to become a star. At this moment,
assuming the correctness of the thesis that at $8-10\mathcal{M}_{\odot}$, the
radiation field begins to push material away from the nascent star, we could
from logic expect that the mass distribution must be continuous up till that
time when disruption starts at $8-10\mathcal{M}_{\odot}$. During the time when
the lamp’s (or protostar’s) mass is in the range
$0\leq\mathcal{M}_{lamp}(t)<8-10\mathcal{M}_{\odot}$, the MDF (28) must hold.
From this, we have just justified the formula (24) for the mass range:
$0\leq\mathcal{M}_{star}(t)<8-10\mathcal{M}_{\odot}$. When the radiation field
begins to be significant, we shall have to check and revise this formula.
Now, from the MDF (24), the gravitational field intensity, at any given time
$t$ and at any given point $r$ inside the core from the surface of the star,
will be given by:
$\vec{\textbf{g}}(r,t)=\overbrace{-\left(\frac{G\mathcal{M}_{csl}(t)}{r^{2}}\right)\left(\frac{r^{3-\alpha_{\rho}}-\mathcal{R}_{star}^{3-\alpha_{\rho}}(t)}{\mathcal{R}^{3-\alpha_{\rho}}_{core}(t)-\mathcal{R}_{star}^{3-\alpha_{\rho}}(t)}\right)\hat{\textbf{r}}}^{\tiny\textbf{Circumstellar}\,\,\textbf{Gravitation}}-\overbrace{\left(\frac{G\mathcal{M}_{star}(t)}{r^{2}}\right)\hat{\textbf{r}}}^{\tiny\textbf{
Star's}\,\,\textbf{Gravitation}}.$ (29)
Clearly, we have been able to separate the gravitation due to the star from
that due to the circumstellar material.
Now, from the above, the inequality (4) becomes:
$\left(\frac{G\mathcal{M}_{csl}(t)}{r^{2}}\right)\left(\frac{r^{3-\alpha_{\rho}}-\mathcal{R}_{star}^{3-\alpha_{\rho}}(t)}{\mathcal{R}^{3-\alpha_{\rho}}_{core}(t)-\mathcal{R}_{star}^{3-\alpha_{\rho}}(t)}\right)+\left(\frac{G\mathcal{M}_{star}(t)}{r^{2}}\right)>\frac{\kappa_{eff}\mathcal{L}_{star}(t)}{4\pi
r^{2}c},$ (30)
where the first term on the left hand-side of (30) is clearly the
gravitational field intensity of the circumstellar material and the second
term is the gravitational field of the nascent star.
## 5 Radiation Cavity
The inequality (5) gives us a condition that must be met before the radiation
field is powerful enough that it can push away (all) the circumstellar
material inside the shell of radius $r$. Beyond this radius, the radiation
field is not at all powerful enough to overcome the gravitational field.
Unfortunately, one cannot deduce this radius $r$ from (5). The inequality
(30), as does (5) and (30) tells us the conditions to be met before the
radiation field is powerful enough to halt in-fall. In addition to this, (30)
yields more information than (5) because in (30) we have quantified the MDF
for the circumstellar material and this allows us to compute the region $r$
where the radiation field is much stronger than the gravitational field. From
(30), we deduce that the radiation field will create a cavity in the star
forming core; in this cavity, the radiation field is much stronger than the
gravitational field, thus there will be a radiation cavity with no material
but only radiation, hence the term “radiation cavity”. To see that (30)
describes a cavity, we simply have to write (30) with $r$ as the subject of
the formula; after doing so, one arrives at:
$r>\left(\frac{\left(\kappa_{eff}\mathcal{L}_{star}(t)-{4}\pi
cG\mathcal{M}_{star}(t)\right)\left(\mathcal{R}^{3-\alpha_{\rho}}_{core}(t)-\mathcal{R}_{star}^{3-\alpha_{\rho}}(t)\right)}{{4}\pi
cG\mathcal{M}_{csl}(t)}+\mathcal{R}_{star}^{3-\alpha_{\rho}}(t)\right)^{\frac{1}{3-\alpha_{\rho}}}=\mathcal{R}_{cav}(t)$
(31)
Figure (1): For a non-spinning core at $\sim 8-10\mathcal{M}_{\odot}$, the
nascent stars’s accretion is halted (and importantly, in-fall is not reversed
but only halted) because when the radiation field tries to create a cavity in
which process the star is separated from its accretion source which is the
circumstellar material. This means the star’s mass accretion is halted because
its mass can no longer grow since there exists no other channel(s) via which
its mass feeds. Should the star’s mass fall below $\sim
8-10\mathcal{M}_{\odot}$, the circumstellar material will fall onto the
nascent star until its mass is restored to its previous value of $\sim
8-10\mathcal{M}_{\odot}$. In order for the radiation field to start pushing
the circumstellar material, its mass must exceed $\sim
8-10\mathcal{M}_{\odot}$. Since there is no way to do this, in-fall is only
halted and not reversed. Hence, the star’s mass for a non-spinning star stays
constant at $\sim 8-10\mathcal{M}_{\odot}$. As urged in Nyambuya
([${2010b}$]), this scenario is different for a spinning star because the ASGF
(which comes about due the spin of the nascent star) allows matter to continue
accreting via the equatorial disk inside the cavity as illustrated above. The
accretion disk will exist inside the radiation cavity and, according to the
azimuthally symmetric theory of gravitation (Nyambuya [${2010b}$]), this disk
should channel mass onto the nascent star right-up to the surface of the star
without radiation hindrances.
where $\mathcal{R}_{cav}(t)$ is the radius of the cavity. Now that there is a
cavity, let us pause so that we can revise the MDF. Clearly, in the case where
there are outflows, this must be given by:
$\mathcal{M}(r,t)=\mathcal{M}_{csl}(t)\left(\frac{r^{3-\alpha_{\rho}}-\mathcal{R}_{cav}^{3-\alpha_{\rho}}}{\mathcal{R}^{3-\alpha_{\rho}}_{core}(t)-\mathcal{R}_{cav}^{3-\alpha_{\rho}}}\right)+\mathcal{M}_{*}(t)\,\,\,\,\,\,\,\,\,\textrm{for}\,\,\,\,\,\,r\geq\mathcal{R}_{cav}(t).$
(32)
where $\mathcal{M}_{csl}(t)=\mathcal{M}_{core}-\mathcal{M}_{*}(t)$ and
$\mathcal{M}_{*}(t)=\mathcal{M}_{star}(t)+\mathcal{M}_{disk}(t)+\mathcal{M}_{outf}(t)$:
$\mathcal{M}_{disk}(t)$ is the disk mass inside the cavity at time $t$ and
$\mathcal{M}_{outf}$ the bipolar outflow contained in the cavity at time $t$.
Now, what this inequality (31) is “saying” is that, at any given moment when
the star has surpassed the critical mass ($8-10\mathcal{M}_{\odot}$), there
will exist a region $r<\mathcal{R}_{cav}(t)$ where the radiation field will
reverse the radially in-falling material and in the region
$r>\mathcal{R}_{cav}(t)$, for material therein, the radiation field has not
reached a state where it exceeds the gravitational field. Hence, in-fall
reversal in that region has not been achieved. This region [i.e.
$r<\mathcal{R}_{cav}(t)$] grows with time thus the radiation field slowly and
gradually pushes the material further and further away from the nascent star
until $\mathcal{R}_{cav}(t)=\mathcal{R}_{cl}$ where radial in-fall is
completely halted. This will occur when the star has reached the critical core
luminosity $\mathcal{L}^{*}_{core}$. The condition when the critical core
luminosity has been shown earlier leads to (12) which is a Larson-like
relation, i.e. (13), ipso facto, this strongly suggests that Larson’s Law may
not be a result of statistical sampling but a statement (about) and a fossil
record of the battle of forces between gravitation and the radiation field.
By saying that the nascent massive star will create a cavity, we have made a
tacit and fundamental assumption that its mass will continue to grow soon
after the cavity begins to form and that its mass will thereafter continue to
grow while in the cavity. However, how can this be since the cavity separates
the nascent star from the circumstellar matter? The nascent star now does not
have a channel to feed its mass, so there can be no growth in its mass unless
there exists a channel via which its mass feeds. At this juncture, we direct
the reader to the readings Nyambuya ([${2010b}$], [${2010a}$]).
In Nyambuya ([${2010a}$]), as already said in the introductory section, we
set-up the ASTG such that the thesis was advanced to the effect ($1$) that,
for a non-spinning star, its gravitational field is spherically symmetric (to
be specific, it only depends on the radial distance from the central body);
($2$) that, for a spinning gravitating body, the gravitational field of the
body in question is azimuthally symmetric, i.e., it depends on the radial
distance ($r$) from the central body and the azimuthal angle ($\theta$). In a
follow-up reading, Nyambuya ($2010b$); we showed that the ASTG predicts ($1$)
that bipolar outflows may very well be a purely gravitational phenomenon
(i.e., a repulsive gravitational phenomenon) and also that; ($2$) along the
spin-equator (defined Nyambuya [${2010b}$]) of a spinning gravitating body,
gravity will channel matter onto the spinning nascent star via the accretion
disk (lying along the spin-equator) thus allowing stars beyond the critical
mass $8-10\mathcal{M}_{\odot}$ to form and begin their stellar processes. It
should be said that accretion disks can also be formed by a number of
different mechanisms other than an Azimuthally Symmetric Gravitational Field
(ASGF).
The accretion of matter beyond the $8-10\mathcal{M}_{\odot}$ limit must only
be possible for a spinning star because it possesses the ASGF that is needed
to continue the channeling of matter onto the star via the accretion disk –
see the illustration in figure (1). For a non-spinning core, the nascent
stars’s accretion cannot proceed beyond $8-10\mathcal{M}_{\odot}$. It is
halted because the moment the radiation field tries to create a cavity, when
the (non-spinning) star’s mass is $8-10\mathcal{M}_{\odot}$, the (non-
spinning) star that very moment becomes separated from the surrounding
circumstellar material. This means the (non-spinning) star’s mass accretion is
halted because its mass can no longer grow since there exists no other
channel(s) via which its mass feeds. Should the (non-spinning) star’s mass
fall below $8-10\mathcal{M}_{\odot}$, the circumstellar material will fall
onto the nascent (non-spinning) star until its mass is restored to its
previous value of $8-10\mathcal{M}_{\odot}$. This means the star’s mass for a
non-spinning star stays constant at $8-10\mathcal{M}_{\odot}$. As explained in
the above paragraphs, this scenario is different for a spinning star because
the ASGF (which comes about due to the spin of the nascent star) allows matter
to continue accreting via the equatorial disk. The accretion disk will exist
inside the radiation cavity and this disk should according to the ASTG
(Nyambuya [${2010b}$]), channel mass right up to the surface of the star
without radiation hindrances. The scenario just presented is completely
different from that projected in much of the wider literature where, at
$8-10\,\mathcal{M}_{\odot}$, suddenly the radiation is so powerful that it
reverses any further in-fall. It is bona-fide knowledge that star formation is
not a spherically symmetric process and from the above, it follows that stars
beyond the $8-10\mathcal{M}_{\odot}$ limit must form with no hindrance, form
the radiation field and the only limit to their existence is if the
gravitationally bound core has enough mass to form them.
## 6 Discussion & Conclusions
This contribution coupled with Nyambuya ([${2010b}$]) seem to strongly point
to the possibility that the radiation problem of massive stars may not exist
as previously thought. In the present reading, we find that beginning at the
time when $\mathcal{M}_{star}(t)\simeq 8-10\mathcal{M}_{\odot}$, the radiation
field will create a cavity inside the star forming core and the circumstellar
material inside the region $\mathcal{R}_{cav}(t)<r\leq\mathcal{R}_{core}(t)$
is going to be pushed gradually (in particular, not blown away) as the
radiation field from the star grows until a point is reached when the cavity
is the size of the core itself. At this point, complete in-fall reversal is
attained. If the radiation field of the star is to grow, its mass must grow,
thus, the cavity must not prevent accretion of mass onto the nascent star and
this is possible for a spinning massive star. Once the cavity is created, the
mass of the nascent will, for a spinning massive star; feed via the accretion
disk and this disk is not affected by the radiation field. By saying the disk
is not affected by the radiation field, we mean the material on the disk is
not going to be pushed away by the radiation field as it pushes the other
material away because the azimuthally symmetric gravitational field of the
star is powerful enough along this plane to overcome the radiation field. This
has been shown or argued in Nyambuya ([${2010b}$]) that this must be the case.
The ASGF is only possible for a spinning star; since all known stars are
spinning, every star should, according to the ASTG, have the potential to grow
to higher masses. This means massive stars should start their stellar
processes because of their spin which brings about the much needed ASGF. A
non-spinning star will have no ASGF, hence there will be no disk around it to
channel material once the radiation field begins taking its toll. In this case
of a non-spinning star, once the star has reached the critical mass $\sim
8-10\mathcal{M}_{\odot}$, its mass cannot grow any further because at the
moment it tries to grow, the star and the circumstellar material become
separated due to the radiation field which, in this case, is stronger than the
gravitational field. In this event, any further growth in mass of the star is
stymied. This, in fact, means that as long as there is circumstellar material,
the mass of a non-spinning star will stay constant at $\sim
8-10\mathcal{M}_{\odot}$ because, the moment it falls slightly below $\sim
8-10\mathcal{M}_{\odot}$, gravity becomes more powerful, thus accreting only
enough mass to restore it to its previous value of $\sim
8-10\mathcal{M}_{\odot}$. In this case, we have an “eternal” stalemate between
the gravitational and radiation field.
An important and subtle difference between the present work and that of other
researchers (Larson & Starfield [$1971$]; Kahn [$1974$]; Yorke &
Kr$\ddot{\textrm{u}}$gel [${1977}$]; Wolfire & Cassinelli [$1987$]; Palla &
Stahler [$1993$]; Yorke [$2002$]; Yorke & Sonnhalter [${2003}$]) is that we
have seized on the observational fact that molecular clouds and cores are
found exhibiting well defined density profiles. From this we computed the MDF
which enabled us to exactly find the physical boundaries where the
gravitational field is expected to be much stronger than the radiation field
once the star exceed the critical mass. Additionally and more importantly is
that from Nyambuya ([${2010b}$]) we have been able to argue that even after
the cavity has been created mass will be channeled on to the star via the
accretion disk. Without the ideas presented in Nyambuya ([${2010b}$]), we
would have been stuck because we where going to find ourselves without a means
to justify how the mass accretion continues once the cavity has been created.
Importantly, we have pointed out a real problem in Yorke ([$2002$]), Yorke &
Sonnhalter ([${2003}$]) and Zinnecker & Yorke ([${2007}$]), namely that these
researchers have neglected the treatment of the circumstellar material in
their theoretical arguments leading to their definition of the radiation
problem because they used Newton’s inverse square law which clearly applies to
a non-rotating mass in empty space, so the inequality (4) applies only for a
star in empty space. In empty space, it is correct to say that the radiation
field for a star of mass $10\,\mathcal{M}_{\odot}$ and beyond, will exceed the
gravitational field everywhere in space beyond the nascent star’s surface, but
the same is not true for a star submerged in a pool of gas, which as is the
case for the stars that we observe.
Another important outcome is that it appears Larson’s Laws may well be a
signature and fossil record of the battle of forces between the radiation and
gravitational fields. At present, it is thought of as being a result of
statistical sampling. Thus the present brings us to start rethinking this
view. We are not persuaded to think this is a result of statistical sampling.
This view finds support from Weidner et al. ([$2009$])’s most recent and
exciting work. In this work, these researchers present a thorough literature
study of the most-massive stars in several young star clusters in order to
assess whether or not star clusters are populated from the stellar initial
mass function (IMF) by random sampling over the mass range
($0.01\mathcal{M}_{\odot}\leq\mathcal{M}_{star}\leq 150\mathcal{M}_{\odot}$)
without being constrained by the cluster mass. Their data reveal a partition
of the sample into lowest mass objects ($\mathcal{M}_{cl}\leq
100\mathcal{M}_{\odot}$), moderate mass clusters
($100\mathcal{M}_{\odot}\leq\mathcal{M}_{cl}\leq 1000\mathcal{M}_{\odot}$) and
rich clusters above ($\mathcal{M}_{cl}\geq 1000\mathcal{M}_{\odot}$) where
$\mathcal{M}_{cl}$ is the mass of the molecular cloud. Their statistical tests
of this data set reveal that the hypothesis of random sampling is highly
unlikely, thus strongly suggesting that there exists some well defined
physical cause.
In closing, allow us to say that we do not claim to have solved the radiation
problem but merely believe that what we have presented herein, together with
the readings Nyambuya ([${2010b}$], [${2010a}$]), is work that may very well
be a significant step forward in the endeavor to resolve this massive star
formation riddle.
Acknowledgments: I am grateful to my brother, Baba va Panashe, and his wife,
Amai va Panashe, for their kind hospitality they offered while working on this
reading and to Mr. Isak D. Davids & Ms. M. Christina Eddington for proof
reading the grammar, spelling & language editing. Last and certainly not
least, I am very grateful to my Professor, D. Johan van der Walt, and
Professor Pienaar Kobus, for the strength and courage that they have given me.
## References
* [$1998$] Bonnell I. A., Bate M., Zinnecker H., $1998$, MNRAS, $298$, $93$
* [${2001}$] Bonnell I. A., Clarke C. J., Bate M. R., Pringle J. E., $2001$, MNRAS, $324$, $573$
* [${2002}$] Bonnell I. A., Bate M. R., $2002$, MNRAS, $336$, $659$
* [${2004}$] Bonnell I. A., Vine S. G., Bate, M. R., $2004$, MNRAS, $349$, $735$
* [$2006$] Bonnell I. A., Clarke C. J., Bate M. R., $2006$, MNRAS, $368$, $1296$
* [$2007$] Bonnell, I. A., Larson, R. B., & Zinnecker, H. 2007, Protostars and Planets V, ed. V. B. Reipurth, D. Jewitt, & K. Keil (Tucson, AZ: Univ. Arizona Press), 149 (arXiv:0603447)
* [$2000$] Clarke C. J., Bonnell I. A., Hillenbrand L. A., $2000$, Protostars and Planets IV, $151$
* [$1997$] Hillenbrand L. A., $1997$, AJ, $113$, $1733$
* [${2009}$] Krumholz M. R., Klein R. I., McKee C. F. et al., $2005$, Science, $323$, ${754}$
* [${2005}$] Krumholz M. R., Klein R. I., McKee C. F., Offner S. S. R., Cunningham A. J., ${2005}$, Protostars and Planets V, $1286$, $8271$
* [$1974$] Kahn F. D., $1974$, A&A, $37$, $149-162$
* [${1972}$] Larson R. B., $1972$, MNRAS, 156,$437$
* [$1982$] Larson R. B., $1982$, MNRAS, $200$, $159$
* [$1971$] Larson R. B. & Starrfield S., $1971$, A&A, $13$, $190$
* [${2002}$] Maeder A., Behrend R., $2002$, ASP Conf. Ser., $267$, $179$
* [$2007$] Matthews L. D., Goddi C., Greenhill L. J., Chandler C. J., Reid M. J. & Humphreys E. M. L., $2007$, Astrophysical Masers and their Environments (IAU Symp. $242$), ed. J. M. Chapman & W. A. Baan (Dordrecht: Kluwer), $130$
* [$2007$] McKee C. F., Ostriker E. C., $2007$, ARA&A, $45$, $565$
* [${2010a}$] Nyambuya G. G., ${2010a}$, MNRAS, 403, Issue 3, 1381.
* [${2010b}$] Nyambuya G. G., ${2010b}$, RAA (Research in Astronomy and Astrophysics), Vol. $10$ No. $11$, $1151-1176$ ($viXra:0911.0025$).
* [$1993$] Palla F., Stahler S. W., $1993$, AJ, $418$, $414$
* [$1976$] Shu F. H., $1977$, Bulletin of the Amer. Astron. Soc., $8$, $547$
* [$1977$] Shu F. H., $1977$, AJ, $214$, $488-497$
* [$2002$] Yorke H. W., $2002$, ASP Conf. Series, $267$, $165$
* [${2004}$] Yorke, H. W. $2004$, Star Formation at High Angular Resolution (IAU Symp. $221$), ed. M. Burton, R. Jayawardhana, & T. Bourke (Dordrecht: Kluwer), $141$
* [${1977}$] Yorke H. W. & Kr${\ddot{\rm{u}}}$gel, $1977$, A&A, $54$, $183$
* [${2003}$] Yorke H. W., Sonnhalter, $2002$, ApJ, $569$, $846-862$
* [$2009$] Weidner C., Kroupa P., Bonnell I. A. D., $2009$, MNRAS, 401, 275.
* [$1987$] Wolfire M. G., Cassinelli J. P., $1987$, ApJ, $319$, $850$
* [${2007}$] Zinnecker H. & Yorke H. W., $2007$, ARA&A, $45$, $481$
| arxiv-papers | 2008-07-18T19:24:42 | 2024-09-04T02:48:56.876091 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "G. G. Nyambuya",
"submitter": "Golden Gadzirayi Nyambuya Mr.",
"url": "https://arxiv.org/abs/0807.3035"
} |
0807.3054 | # Infinitesimal deformations of symmetric simple modular Lie algebras and Lie
superalgebras
Sofiane Bouarroudj1, Pavel Grozman2, Dimitry Leites3 1Department of
Mathematics, United Arab Emirates University, Al Ain, PO. Box: 17551;
Bouarroudj.sofiane@uaeu.ac.ae
2Equa Simulation AB, Stockholm, Sweden; pavel@rixtele.com
3A.Salam International School of Mathematical Science Lahore, Pakistan,
on leave from Department of Mathematics, University of Stockholm, Roslagsv.
101, Kräftriket hus 6, SE-106 91 Stockholm, Sweden; mleites@math.su.se
###### Abstract.
Over algebraically closed fields of positive characteristic, infinitesimal
deformations of simple finite dimensional symmetric (the ones that with every
root have its opposite of the same multiplicity) Lie algebras and Lie
superalgebras are described for small ranks. The results are obtained by means
of the Mathematica based code SuperLie.
The infinitesimal deformation given by any odd cocycle is integrable. The
moduli of the deformations form, in general, a supervariety. Not each even
cocycle is integrable; but for those that are integrable, the global deforms
(the results of deformations) are linear with respect to the parameter.
In characteristic 2, the simple 3-dimensional Lie algebra admits a parametric
family of non-isomorphic simple deforms.
Some of Shen’s ”variations of G(2) theme” are interpreted as two global
deforms corresponding to the several of the 20 infinitesimal deforms first
found by Chebochko; we give their explicit form.
###### Key words and phrases:
modular Lie superalgebra
###### 1991 Mathematics Subject Classification:
17B50, 70F25
We are thankful to A. Lebedev for help; DL is thankful to MPIMiS, Leipzig, for
financial support and most creative environment. We are thankful to N.
Chebochko and M. Kuznetsov for helpful discussions of their unpublished
results pertaining to this paper.
## 1\. Introduction
Hereafter, ${\mathbb{K}}$ is an algebraically closed field of characteristic
$p>0$ and ${\mathfrak{g}}$ is a finite dimensional Lie (super)algebra;
${\mathbb{Z}}_{+}$ is the set of non-negative integers.
### 1.1. The quest for simple modular Lie algebras. The Kostrikin–Shafarevich
conjecture
Shafarevich, together with his student Kostrikin, first considered the
restricted simple modular Lie algebras. This self-restriction was, perhaps,
occasioned by the fact that only restricted Lie algebras correspond to
algebraic groups, and being a geometer, albeit an algebraic one, Shafarevich
did not see much reason to consider non-restricted Lie algebras. Recently, on
a different occasion, Deligne gave us an advice [LL] which we interpret as a
suggestion to look, if $p>0$, at the groups (geometry) rather than at Lie
algebras. Since only restricted Lie algebras correspond to algebraic groups,
we interpret this advice as a certain natural restriction of the
classification problem of simple Lie (super)algebras which makes the problem
more tangible but requires to select deforms (the results of deformations)
with $p$-structure as precious gems.
Be as it may, in late 1960s, Kostrikin and Shafarevich formulated a conjecture
describing all simple finite dimensional (not only restricted) modular Lie
algebras over algebraically closed fields of characteristic $p>7$. The
statement of the conjecture turned out to hold for $p=7$ as well.
After 30 years of work of several teams of researchers, Block, Wilson, Premet
and Strade not only proved the KSh-conjecture, but even completed the
classification of all simple finite dimensional Lie algebra over algebraically
closed fields of characteristic $p>3$, see [S].
It turns out that, for $p>3$, the classification can be formulated as follows
(for details and definitions, see [BGL, Lt]):
(1) Take any finite dimensional Lie algebra of the form ${\mathfrak{g}}(A)$
with indecomposable Cartan matrix $A$, or a trivial central extension
${\mathfrak{c}}({\mathfrak{g}}(A))$ thereof. In one of the
${\mathbb{Z}}$-gradings of ${\mathfrak{g}}:={\mathfrak{g}}(A)$ or
${\mathfrak{c}}({\mathfrak{g}}(A))$, take the Cartan-Tanaka-Shchepochkina
(CTS) prolong $({\mathfrak{g}}_{-},{\mathfrak{g}}_{0})_{\ast,\underline{N}}$
of the pair $({\mathfrak{g}}_{-},{\mathfrak{g}}_{0})$ whose dimension is
bounded by a shearing parameter $\underline{N}$. All simple finite dimensional
modular Lie algebras are among such prolongs or the quotients of their derived
algebras modulo the center, and deformations thereof.
Although for $p=2,3$, and for Lie superalgebras, certain other types of pairs
$({\mathfrak{g}}_{-},{\mathfrak{g}}_{0})$ also yield simple examples, the
importance of the algebras of the form ${\mathfrak{g}}(A)$ is clear; it is
also of independent interest.
Weisfeiler and Kac [WK] gave a classification of finite dimensional Lie
algebras ${\mathfrak{g}}(A)$ with indecomposable Cartan matrix $A$ for any
$p>0$, but although the idea of their proof is OK, the paper has several gaps
and vague notions (the Brown algebra ${\mathfrak{br}}(3)$ was missed [Br3,
KWK]; the definition of the Lie algebra with Cartan matrix given in [K] (and
applicable to Lie superalgebras and modular Lie (super)algebras) was not
properly developed at the time [WK] was written; the algebras
${\mathfrak{g}}(A)^{(1)}/{\mathfrak{c}}$ — which have no Cartan matrix — are
sometimes identified as having one, and so on). All these notions, and several
more, are clarified in [BGL, LCh].
There were also known several scattered examples of serial and (conjecturally,
in the absence of a classification) exceptional simple modular Lie algebras
for $p=3$ and 2, see [Lt]. For all these cases, there remained the problem of
description of deformations. In the latest paper on the topic, with difficult
results nicely explained, Natasha Chebochko gave an overview of the situation
for Lie algebras of the form ${\mathfrak{g}}(A)$ and
${\mathfrak{g}}(A)^{(1)}/{\mathfrak{c}}$. She writes:
(2) “According to [Dz], for $p=3$, the Lie algebra $C_{2}$ is the only algebra
among the series $A_{n}$, $B_{n}$, $C_{n}$, $D_{n}$ that admits non-trivial
deformations. In [Ru] it was established that over a field of characteristic
$p>3$ all the classical Lie superalgebras are rigid. In [KuCh] and [KKCh] a
new scheme was proposed for studying rigidity, and it was proved that the
classical Lie algebras of all types over a field of characteristic $p>2$ are
rigid, except for the Lie algebra of type $C_{2}$ for $p=3$. (“Classical” here
means any algebra with indecomposable Cartan matrix (or its quotient modulo
center), except Brown algebras: for ${\mathfrak{br}}(2)$ deformations were
known, for ${\mathfrak{br}}(3)$ the question was open. This makes our answer
concerning the Brown algebras ${\mathfrak{br}}(3)$ even more interesting.
BGL.) For $p=2$, some deformations of the Lie algebra of type $G_{2}$ were
constructed in [She1]. … The Lie algebras of type $A_{l}$ for $l+1\equiv
0\pmod{2}$, $D_{l}$ and $E_{7}$, have non-trivial centers. We shall say that
the corresponding quotient algebras are of type $\overline{A_{l}}$,
$\overline{D_{l}}$ and $\overline{E_{7}}$, respectively. …
(3) Theorem. Let $L$ be a Lie algebra over a field of characteristic $2$.
$(1)$ If $L$ is one of the types $A_{l}$ for $l>1$, $D_{l}$ for $l\equiv
1\pmod{2}$, $E_{6}$, $E_{7}$, $E_{8}$, $\overline{A_{l}}$ for $l\neq 3,5$, or
$\overline{D_{l}}$ for $l\equiv 0\pmod{2}$ and $l\neq 6$, then $H^{2}(L;L)=0$.
$(2)$ If $L$ is of type $\overline{A_{l}}$ for $l=3,5$, then $\dim
H^{2}(L;L)=20$.
$(3)$ If $L$ is of type $D_{4}$, then $\dim H^{2}(L;L)=24$.
$(4)$ If $L$ is of type $D_{l}$ for $l\equiv 0\pmod{2}$ and $l>4$ or
$\overline{D_{l}}$ for $l\equiv 1\pmod{2}$, then $\dim H^{2}(L;L)=2l$.
$(5)$ If $L$ is of type $\overline{D_{6}}$, then $\dim H^{2}(L;L)=64$.
$(6)$ If $L$ is of type $\overline{E_{7}}$, then $\dim H^{2}(L;L)=56$.”
#### 1.1.1. Elucidating certain moments unclear to us in the conventional
presentations
1) The term “classical” in the above quotation (and almost all other papers
and books on the topic) is applied to simple Lie algebras with Cartan matrix
of the same types that exist for $p=0$; their quotients modulo center are also
“classical”, but, for $p=2$, the simple Lie algebras of Brown, Weisfeiler and
Kac, and ${\mathfrak{o}}(2n+1)$ are left out. Clearly, the nomenclature should
be improved; Shen [She2] indicates other reasons for being unhappy with it.
Neither the $p=2$ analogs of ${\mathfrak{o}}_{I}(2n)$ nor its simple
subquotient have Cartan matrix. In the above quotation, they are not qualified
as “classical”, whereas we are sure they should be. So we have to consider
(4) ${\mathfrak{o}}_{I}(2n),\;\;{\mathfrak{o}}(2n+1)\;\text{(a.k.a. $B_{l}$),
the Brown algebras, and the Weisfeiler-Kac algebras.}$
2) The simple Lie algebras constitute a natural “core”, but some of their
relatives are “better” (are restricted, possess Cartan matrix, and are rigid,
whereas their simple cores may lack one or all of these features). In
particular, the abundance of deformations might point at an inner “flaw” of
the object whose relative’s behavior is impeccable …
Together with Lebedev, in [BGL] and in chapters in [LCh] due to Lebedev, we
clarified the above moments (we will reproduce the needed definitions in what
follows).
#### 1.1.2. Explicit cocycles v. ecology
Having written the first 15 pages of this paper we were appalled by the amount
of the space saturated by explicit cocycles. Who needs them?! Let us give just
the dimensions and save paper! However, it was only thanks to the explicit
form of the cocycles that we were able to interpret some of the mysterious
Shen’s “variations”. And only having the explicit form of the cocycle can one
check if the local deformation is integrable or be able to compute the global
one.
### 1.2. Main result
In [BGL], we have classified finite dimensional modular Lie superalgebras of
the form ${\mathfrak{g}}(A)$ with indecomposable symmetrizable Cartan matrix
$A$ over algebraically closed fields, in particular, the simple Lie
superalgebras of the form ${\mathfrak{g}}(A)$ and those for which
${\mathfrak{g}}(A)^{(1)}/{\mathfrak{c}}$, where ${\mathfrak{c}}$ is the
center, is simple. For brevity, we will sometimes address the latter algebras
also as having Cartan matrix, though, strictly speaking, they do not possess
any Cartan matrix, see Warning in [BGL]. In [Ch], Chebochko described
infinitesimal deformations of the “classical” Lie algebras with Cartan matrix,
except for (4) for which it remained an open problem.
The Lie superalgebras with Cartan matrix, and the queer algebras
${\mathfrak{psq}}(n)$ for $p\neq 2$ are “symmetric”:
(5) With each root $\alpha$ they have a root $-\alpha$ of the same
multiplicity.
(If $p=2$, there are more examples of queer type Lie superalgebras, serial and
exceptional, and some of them are not symmetric, see [LCh].)
In this note, a sequel to [BGL], we describe deformations of “symmetric” Lie
(super)algebras:
(1) the “left out” cases (4), and of Lie superalgebras of small rank, both
with Cartan matrix (and relatives thereof) classified in [BGL],
(2) of the so-called queer type,
(3) of certain Lie algebras without Cartan matrix: ${\mathfrak{o}}_{I}(2n)$,
leaving its super versions without Cartan matrix
(${\mathfrak{o}}{\mathfrak{o}}_{II}(2n+1|2m+1)$,
${\mathfrak{o}}{\mathfrak{o}}_{II}(2n|2m)$,
${\mathfrak{o}}{\mathfrak{o}}_{I\Pi}(2n|2m)$,
${\mathfrak{o}}{\mathfrak{o}}_{II}(2n+1|2m)$) for a time being; we intend to
do this elsewhere.
Conjecturally, the symmetric algebras of higher ranks over fields of
characteristic distinct from 2 are rigid, except for the simple relatives of
Lie (super)algebras of Hamiltonian vector fields; for the computers available
to us at the moment, the problem is out of reach even with this code.
#### 1.2.1. On Shen’s “variations”
Trying to append the results of Chebochko [Ch] by deforms of the Lie algebras
she did not consider but which we consider no less “classical” than the ones
she considered, we obtained, as a byproduct, an elucidation of Shen’s
“variations” [She1, LLg]. Shen described seven “variations” $V_{i}G(2)$ of
${\mathfrak{g}}(2)$ and three more “variations” of ${\mathfrak{sl}}(3)$; Shen
claimed that all his examples are simple and all but two
($V_{1}G(2):={\mathfrak{g}}(2)$ and $V_{7}G(2):={\mathfrak{wk}}(3;a)$) are
new. It was later found that the “variations” of ${\mathfrak{sl}}(3)$ are
isomorphic to ${\mathfrak{sl}}(3)$, the variations of dimension 15 are not
simple (this has to be verified). At the moment, the only verified claim on
simple and really new Lie algebras Shen has discovered concerns the
2-parameter deformations of ${\mathfrak{g}}(2)\simeq{\mathfrak{psl}}(4)$.
Several “variations” are depicted with typos: as stated, these algebras do not
satisfy Jacobi identity or have ideals. Unfortunately, we did not guess how to
amend the multiplication tables whereas our letters to Shen or his students
remain unanswered.
Chebochko [Ch] has found all infinitesimal deformations of
${\mathfrak{g}}(2)\simeq{\mathfrak{psl}}(4)$ and we have found that each
individual cocycle is integrable. In [Ch1], Chebochko found out that the space
of cocycles constitutes two (apart from the origin) orbits relative
$\mathop{\mathrm{{Aut\/}}}\nolimits({\mathfrak{psl}}(4))$, and described the
two respective Lie algebras.
We only consider the Lie algebras of small rank because our Mathematica-based
code SuperLie (see [Gr]) with which we got our results is incapable to process
higher ranks on computers available to us. However, since the higher the rank
the more rigid simple Lie (super)algebras are, we make the following
###### Conjecture.
In this paper we have found all (except for the analogs of the cases of
Chebochko’s theorem (3) that we have to add) the infinitesimal deformations of
finite dimensional Lie algebras and Lie superalgebras with indecomposable
Cartan matrices for $p=5$ and $3$ (and their simple subquotients) as well as
of queer Lie superalgebras (and their simple subquotients).
### 1.3. Motivations
The classification of simple finite dimensional Lie algebras ${\mathfrak{g}}$
over fields ${\mathbb{K}}$ of characteristic $p>3$ — a proof of the
generalized Kostrikin-Shafarevich conjecture — is now completed, see [S]. So
it is time now to consider the cases $p=3$, 2, and the super case, especially
in view of the hidden supersymmetry of the non-super $p=2$ case, see [ILL]).
Let us list the results known earlier.
Several examples of simple Lie algebras over fields ${\mathbb{K}}$ for $p=3$
not embraced by the approach of the Kostrikin-Shafarevich conjecture are given
in [S]; in [GL4], some of these examples, earlier considered “mysterious”, are
described as (generalized) Cartan prolongs provided they are
${\mathbb{Z}}$-graded; other examples from [S] are deformations of these
${\mathbb{Z}}$-graded ones. In [WK], the simple modular Lie algebras
possessing Cartan matrix are classified, some of them form parametric
families, but no complete study of deformations was performed.
It was clear since long ago that the smaller characteristic, the less rigid
the simple Lie algebras are; Rudakov gave an example of a 3-parameter family
of simple Lie algebras (for the first published description, see [Kos]; for an
exposition of the initial (never published) Rudakov’s approach — in terms of
the Cartan prolongation, — see [GL4]).
After Rudakov’s example became known, Kostrikin and Dzhumadildaev ([DK, Dz1,
Dz2, Dz3]) studied various (e.g., filtered and infinitesimal) deformations of
simple vectorial Lie algebras (sometimes dubbed algebras of Cartan type); for
a detailed summary of the part of their results with understandable proofs,
and some new results (all pertaining to the infinitesimal deformations), see
[Vi].
Rudakov’s paper [Ru] clearly showed that speaking about deformations it is
unnatural to consider the modular Lie algebras naively, as vector spaces: Lie
algebras should be viewed as algebras in the category of varieties. This
approach should, actually, be applied even over fields of characteristic 0,
but the simplicity of the situation with finite dimensional Lie algebras
obscures this (rather obvious) fact. Lie superalgebras can not be treated
otherwise in various problems of interest in applications, see [LSh].
Recently the deformations of the simple modular Lie algebras with Cartan
matrix, except for the Brown algebras we consider, were classified ([Ch, KK,
KuCh, KKCh]). The findings of Chebochko did not, however, elucidate several of
mysterious Shen’s “variations”, see [She1, LLg].
### 1.4. Disclaimer
1) Same as practically everybody ([Ch, KK, KuCh, KKCh, Vi]) we disregard, for
the moment, the danger of ignoring singular global deformations $=$ an
interesting but dangerous possibility pointed at by D. B. Fuchs with co-
authors [FF, FL]. This danger, however, may only concern even deformations
whereas most of the deformations we list here are odd.
2)In [Ch, KK, KuCh, KKCh], a complete description of deformations of the
“classical” simple Lie algebras with Cartan matrix is performed over fields
${\mathbb{K}}$ for $p=3$ and 2. Although in [LL] there are given reasons111The
overwhelming abundance of cocycles found in [Ch, Vi], is hardly a reason as we
thought before we have read [Ch, Ch1]: Although there are many nontrivial
cocycles, non-isomorphic deforms of ${\mathfrak{g}}$ correspond only to the
distinct orbits in the space of cocycles under the action of the Chevalley
group $\mathop{\mathrm{{Aut\/}}}\nolimits({\mathfrak{g}})$ and there are very
few such orbits. to doubt the possibility to liberally apply the conventional,
so far, definition of Lie algebra (co)homology ([Fu]) to the case where $p>0$,
these doubts seem to be inapplicable for $H^{2}({\mathfrak{g}};M)$ if
${\mathfrak{g}}$ possesses Cartan matrix, $\dim M<\infty$, and $p\neq 2$. In
particular, it seems safe to compute $H^{2}({\mathfrak{g}};{\mathfrak{g}})$
and interpret the result in terms of infinitesimal deformations.
Let us explain how definition of $U({\mathfrak{g}})$ affects the definition of
cohomology and the notion of (co)induced representations. The scientific
definition of the Lie (super)algebra cohomology is
(6)
$H^{i}({\mathfrak{g}};M):=\mathop{\mathrm{Ext}}\nolimits^{i}_{U({\mathfrak{g}})}({\mathbb{K}};M).$
So it is clear, actually, how to approach the problem, at least for the
modular Lie algebras obtained by means of the Kostrikin-Shafarevich approach
(and its super analog): Speaking about non-super cases, take any book (e.g.,
[St]) in which a convenient ${\mathbb{Z}}$-form
$U_{\mathbb{Z}}({\mathfrak{g}})$ of $U({\mathfrak{g}})$ is described for any
simple complex ${\mathfrak{g}}$, and introduce $\underline{N}$ (similar to the
$\underline{N}$ in the definition of the algebra of divided powers ${\cal
O}(n;\underline{N})$) by setting something like
(7) $\begin{array}[]{ll}U({\mathfrak{g}};\underline{N}):=&\text{subalgebra of
$U_{\mathbb{Z}}({\mathfrak{g}})$ constructed}\\\ &\text{``similarly to the
algebra of divided powers" ${\cal O}(n;\underline{N})$};\\\
H^{i}_{\underline{N}}({\mathfrak{g}};M):=&\mathop{\mathrm{Ext}}\nolimits^{i}_{U({\mathfrak{g}};\underline{N})}({\mathbb{K}};M).\end{array}$
How to perform this “similar construction” of “something like” is the whole
point.
Absolutely correct — in terms of the conventional definition (6) —
computations of Dzhumadildaev elucidated in [Vi] imply that
${\mathfrak{vect}}(n;\underline{N}):={\mathfrak{der}}({\cal
O}(n;\underline{N}))$, where only “special derivatives” are considered, is not
rigid; more precisely, ${\mathfrak{vect}}(n;\underline{N})$ has infinitesimal
deformations. We find this result “ideologically wrong” and believe that the
cause is buried in the definitions used. Recall our arguments in favor of
rigidity of ${\mathfrak{vect}}$, see [LL]:
Let ${\mathfrak{h}}$ be a subalgebra of ${\mathfrak{g}}$. For any
${\mathfrak{h}}$-module $V$, we define a series of coinduced
${\mathfrak{g}}$-modules:
(8)
$\begin{array}[]{l}\mathop{\mathrm{Coind}}\nolimits_{\mathfrak{h}}^{\mathfrak{g}}(V;\underline{N}):=\mathop{\mathrm{Hom}}\nolimits_{U({\mathfrak{h}};\underline{N})}(U({\mathfrak{g}};\underline{N}),V);\end{array}$
Then, in terms of the conjectural definition (7), we should have the following
analog of the well-known isomorphism:
(9)
$\begin{array}[]{l}H^{i}_{\underline{N}}({\mathfrak{g}};\mathop{\mathrm{Coind}}\nolimits_{\mathfrak{h}}^{\mathfrak{g}}(V;\underline{N}))\simeq
H^{i}({\mathfrak{h}};V).\end{array}$
This isomorphism would imply that, for
${\mathfrak{g}}={\mathfrak{vect}}(n;\underline{N})$, we should have
$H^{2}_{\underline{N}}({\mathfrak{g}};{\mathfrak{g}})\simeq
H^{2}({\mathfrak{gl}}(V);V)=0,\text{~{}~{}where $\dim V=n$,}$
at least, if $n$ is not divisible by $p$.
The situation is opposite, in a sense, to that with the Kac-Moody groups that
“did not exist” until a correct definition of cohomology was used; or with
Dirac’s $\delta$-function which is not a function in the conventional sense.
Dzhumadildaev [Dz2] (also Farnsteiner and Strade [FS]) showed that for $p>0$
the conventional analog of the statement (9), a.k.a. Shapiro’s lemma, should
be formulated differently because
$H^{i}({\mathfrak{g}};\mathop{\mathrm{Coind}}\nolimits_{\mathfrak{h}}^{\mathfrak{g}}(V))$
strictly contains $H^{i}({\mathfrak{h}};V)$. We hope that one can get rid of
these extra cocycles in an appropriate theory.
However, even if we are wrong here and any reasonable theory contains these
infinitesimal deformations, and nobody, as far as we know, investigated their
number after factorization by the action of the corresponding automorphism
group.
3) The Lie (super)algebra (co)homology can also be defined “naively”, as
generalizations (and dualizations) of the de Rham complex. In this approach
the enveloping algebra does not appear explicitly and the divided powers we
tried to introduce in item 2) of Disclaimer seem to disappear.
The divided powers of (co)chains naturally appear in the study of Lie
superalgebras for any $p$, even for $p=0$, but their meaning is unclear at the
moment. We suggest to denote the corresponding spaces of homology by
$DPH_{(n)}^{\underline{N}}({\mathfrak{g}};M)$ and cohomology by
$DPH^{(n)}_{\underline{N}}({\mathfrak{g}};M)$.
For $p=2$ and Lie superalgebras, there is an interpretation: Such divided
(co)chains (for $\underline{N}_{i}>1$ for every odd basis element
$X_{i}\in{\mathfrak{g}}$) are indispensable in the study of relations and
deformations of the Lie superalgebras because the bracket is now determined by
the squaring.
We denote the product of divided powers of cochains by wedge, the usual powers
are denoted $(dx)^{\wedge n}$ whereas the divided powers are denoted
$(dx)^{(\wedge n)}$. Actually, we should not write
$dx\in\Pi({\mathfrak{g}}^{*})$ but rather $x^{*}$ either for brevity or having
in mind the element of ${\mathfrak{g}}^{*}$ unless the wedge product is
needed.
4) On integrability. We compute infinitesimal deformations, i.e., we compute
$H^{2}({\mathfrak{g}};{\mathfrak{g}})$. However, if the cocycle is odd, it
certainly extends to a global deformation, albeit with an odd parameter.
Kostrikin and Dzhumadildaev ([DK]) claimed that every local (meaning
infinitesimal) deformation of the Lie algebra $W_{1}(m)$ (we denote this
algebra by ${\mathfrak{vect}}(1;\underline{m})$) is integrable.222A question
arises: in which sense is the classification of [S] complete?! It turns out:
“as stated”: for $p>3$, ALL simple finite dimensional algebras are classified
because although there are not only filtered deformations of Lie algebras of
vectorial type, the deforms are isomorphic to something known. Ain’t it a
miracle: if $p=2$, this definitely is not so. We were unable to follow their
proof (reduction to a paper of R. Amayo [Proc. London Math. Soc. (3) 33
(1976), no. 1, 28–64; MR0409573 (53 #13327a) ,b]) but, to our incredulity,
have encountered a similar phenomenon with most of the cocycles unearthed in
this paper: Most of the infinitesimal deformations of some of the Lie
(super)algebras considered here are integrable; moreover, the global
deformation corresponding to a given (homogeneous with respect to weight)
cocycle $c$ is often linear in parameter, i.e., it is of the form
(10) $[x,y]_{\text{new}}=[x,y]+\alpha c(x,y).$
The deformation of ${\mathfrak{osp}}(4|2)$ and Shen’s deformations of
${\mathfrak{psl}}(4)$ are linear in parameter, and so the corresponding
cocycles give the global deformations. These facts encouraged us to conjecture
that ALL cocycles we have found are integrable (this is not so) and directly
investigate which of the other cocycles determine the new bracket (10).
We also have to consider deformations of the deforms (we intend to do this
elsewhere).
#### 1.4.1. What ${\mathfrak{g}}(A)$ is
The Lie (super)algebras of the form ${\mathfrak{g}}(A)$ (sometimes called,
together with (super)algebras of certain other types, “contragredient”) are
determined as follows. Let $A=(A_{ij})$ be an arbitrary $n\times n$ matrix of
rank $l$ with entries in ${\mathbb{K}}$. Fix a vector space ${\mathfrak{h}}$
of dimension $2n-l$ and its dual ${\mathfrak{h}}^{*}$, select $n$ linearly
independent vectors $h_{i}\in{\mathfrak{h}}$ and $n$ linearly independent
vectors $\alpha_{j}\in{\mathfrak{h}}^{*}$ so that $\alpha_{i}(h_{j})=A_{ij}$.
Let $I=\\{i_{1},\dots,i_{n}\\}\subset({\mathbb{Z}}/2{\mathbb{Z}})^{n}$;
consider the free Lie superalgebra $\widetilde{{\mathfrak{g}}}(A,I)$ generated
by $e_{1}^{\pm},\dots,e_{n}^{\pm}$, where $p(e_{j}^{\pm})=i_{j}$, and
${\mathfrak{h}}$, subject to relations (here either all superscripts $\pm$ are
$+$ or all are $-$)
(11)
$[e_{i}^{+},e_{j}^{-}]=\delta_{ij}h_{i};\quad[h,e_{j}^{\pm}]=\pm\alpha_{j}(h)e_{j}^{\pm}\text{~{}~{}for
any $h\in{\mathfrak{h}}$ and any
$i,j$};\quad\;[{\mathfrak{h}},{\mathfrak{h}}]=0.$
Let $\tilde{\mathfrak{g}}_{\pm}$ be subalgebras generated by $e_{i}^{\pm}$.
###### Statement.
There exists a maximal ideal ${\mathfrak{r}}$ among the ideals of
$\tilde{\mathfrak{g}}(A,I)$ whose intersection with ${\mathfrak{h}}$ is zero,
and ${\mathfrak{r}}$ is the direct sum of the ideals
${\mathfrak{r}}\bigcap\tilde{\mathfrak{g}}_{+}$ and
${\mathfrak{r}}\bigcap\tilde{\mathfrak{g}}_{-}$.
The statement is well known over ${\mathbb{C}}$ for the finite dimensional and
certain infinite dimensional simple Lie algebras [K]; for an explicit
description of the ideal ${\mathfrak{r}}$ for simple Lie superalgebras over
${\mathbb{C}}$ of the same type, see [GL1]; for the modular case, see
Lebedev’s chapters in [LCh].
The change
$A\longmapsto\tilde{A}:=\mathop{\mathrm{diag}}\nolimits(\lambda_{1},\dots,\lambda_{n})A,\text{~{}~{}where
$\lambda_{1},\dots,\lambda_{n}\in{\mathbb{K}}\setminus\\{0\\}$},$
preserves the Lie (super)algebra $\tilde{\mathfrak{g}}(A,I)$. The Cartan
matrix $A$ is said to be normalized if $A_{jj}=0$ or 1, or 2 (only if
$i_{j}={\bar{0}}$, where ${\bar{0}}$ and ${\bar{1}}$ are residue classes
modulo 2); in order to distinguish the cases $i_{j}={\bar{0}}$ from
$i_{j}={\bar{1}}$, we write $A_{jj}={\bar{0}}$ or ${\bar{1}}$, instead of 0 or
1, if $i_{j}={\bar{0}}$. We will only consider normalized Cartan matrices; for
them, we can skip $I$.
The subalgebra ${\mathfrak{h}}$ diagonally acts on $\tilde{\mathfrak{g}}(A)$
and the nonzero eigenspaces with nonzero eigenvalues are called roots. In the
modular case, it is more appropriate to use the following definition. The
algebra $\tilde{\mathfrak{g}}(A)$, where $A$ is of size $n\times n$, is
naturally ${\mathbb{Z}}^{n}$-graded, where $\deg e_{i}^{\pm}=(0,\dots,0,\pm
1,0,\dots,0)$, where $\pm 1$ occupies the $i$th slot. The nonzero eigenspaces
with nonzero eigenvalues in this grading are said to be roots. (Over
${\mathbb{C}}$, this grading is equivalent to the usual grading by weights.)
For any subset $B=\\{\sigma_{1},\dots,\sigma_{n}\\}\subset R$, we set:
$R_{B}^{\pm}=\\{\alpha\in R\mid\alpha=\pm\sum
n_{i}\sigma_{i},\;\;n_{i}\in{\mathbb{Z}}_{+}=\\{0,1,2,\dots\\}\\}.$
The subset $B\subset R_{B}^{+}$ is said to be a system of simple roots of $R$
(or ${\mathfrak{g}}$) if $\sigma_{1},\dots,\sigma_{n}$ are linearly
independent and $R=R_{B}^{-}\coprod R_{B}^{+}$.
Set ${\mathfrak{g}}(A):=\tilde{\mathfrak{g}}(A)/{\mathfrak{r}}$. Observe that
the conditions sufficient for simplicity of Lie algebras ${\mathfrak{g}}(A)$
over ${\mathbb{C}}$ fail to ensure simplicity of modular Lie algebras
${\mathfrak{g}}(A)$ and Lie superalgebras ${\mathfrak{g}}(A)$.
###### Warning.
For $m-n\equiv 0\pmod{p}$ (for $m=n$ over ${\mathbb{C}}$), neither
${\mathfrak{sl}}(m|n)$ nor its simple quotient333If a Lie superalgebra
${\mathfrak{g}}\subset{\mathfrak{gl}}(V)$ contains the space of scalar
matrices ${\mathfrak{s}}={\mathbb{K}}1_{m|n}$, we denote
${\mathfrak{p}}{\mathfrak{g}}={\mathfrak{g}}/{\mathfrak{s}}$ the projective
version of ${\mathfrak{g}}$. ${\mathfrak{psl}}(m|n)$ possess Cartan matrix; it
is ${\mathfrak{gl}}(m|n)$ which possesses it.
## 2\. The queer series
In the text-book [Ls] it is demonstrated that there are two superizations of
${\mathfrak{gl}}(n)$: a naive one, ${\mathfrak{gl}}(n|m)$, and the “queer”
one, ${\mathfrak{q}}(n)$ which preserves an “odd” complex structure. Recall
that
${\mathfrak{q}}(n):=\left\\{X\in{\mathfrak{gl}}(n|n)\mid[X,J_{2n}]=0,\;\text{where
$J_{2n}=\begin{pmatrix}0&1_{n}\\\ -1_{n}&0\end{pmatrix}$}\right\\};$
explicitly,
${\mathfrak{q}}(n):=\left\\{\begin{pmatrix}A&B\\\ B&A\end{pmatrix}\right\\}.$
We set
${\mathfrak{sq}}(n):=\left\\{\begin{pmatrix}A&B\\\ B&A\end{pmatrix}\text{ such
that $\mathrm{tr}(B)=0$}\right\\}.$
Let $\mathfrak{psq}(n):=\mathfrak{sq}(n)/\mathfrak{center}$ be the
projectivisation of $\mathfrak{sq}(n)$. We denote the images of the
$A_{ij}$-elements in $\mathfrak{psq}(n)$ by $a_{ij}$ and the images of the
$B_{ij}$-elements by $b_{ij}$.
It is not difficult to show that the derived superalgebra
$\mathfrak{psq}^{(1)}(n)=[\mathfrak{psq}(n),\mathfrak{psq}(n)]$ is simple and
is only different from $\mathfrak{psq}(n)$ for $n=pk$, when $\mathfrak{sq}(n)$
contains $\begin{pmatrix}0&1_{n}\\\ -1_{n}&0\end{pmatrix}$.
## 3\. Results
Observe that, for our problem, it does not matter which of several Cartan
matrices that a given algebra ${\mathfrak{g}}(A)$ possesses we take: The
simplest incarnation will do.
Since our algebras are symmetric with respect to the change of the sign of the
roots, it suffices to consider cocycles of only non-negative (or non-positive)
degrees. We do not list cocycles of positive (negative) degrees.
We denote the positive Chevalley generators by $x_{i}$, the corresponding
negative ones by the $y_{i}$.
We underline the degrees of the odd cocycles.
###### Lemma.
For ${\mathfrak{g}}={\mathfrak{br}}(2;\alpha)$, where $\alpha\neq 0,-1$, the
Cartan matrix is
(12) $\begin{pmatrix}2&-1\\\ \alpha&2\end{pmatrix}\quad\text{and the basis
}\;x_{1},\;x_{2},\;x_{3}=[x_{1},x_{2}],\quad
x_{4}=-\mathop{\mathrm{ad}}\nolimits_{x_{2}}^{2}(x_{1}).$
Then $H^{2}({\mathfrak{g}};{\mathfrak{g}})$ is spanned by the cocycles
$(\ref{eqbr2a})$.
(13) $\tiny\begin{array}[]{ll}\deg=-6:&2\alpha(1+\alpha)y_{1}\otimes
dx_{3}{\wedge}dx_{4}+(1+\alpha)y_{3}\otimes dx_{1}{\wedge}dx_{4}+y_{4}\otimes
dx_{1}{\wedge}dx_{3}\\\ &\\\ \deg=-3:&\alpha(1+\alpha)x_{1}\otimes
dx_{2}{\wedge}dx_{4}+2(1+\alpha)y_{2}\otimes dx_{4}{\wedge}dy_{1}+y_{4}\otimes
dx_{2}{\wedge}dy_{1}\\\ &\\\ \deg=0:&h_{1}\otimes dx_{2}\wedge
dy_{2}+h_{1}\otimes dx_{3}\wedge dy_{3}+x_{1}\otimes dx_{3}\wedge
dy_{2}-y_{1}\otimes dx_{2}\wedge dy_{3}\end{array}$
###### Lemma.
For
${\mathfrak{g}}={\mathfrak{br}}(2)=\mathop{\lim}\limits_{-\frac{2}{a}\longrightarrow
0}~{}{\mathfrak{br}}(2,a)$, the Cartan matrix is
(14) $\begin{pmatrix}2&-1\\\ -1&{\bar{0}}\end{pmatrix}\quad\text{and the basis
}\;x_{1},\;x_{2},\;x_{3}=[x_{1},x_{2}],\quad
x_{4}=-\mathop{\mathrm{ad}}\nolimits_{x_{2}}^{2}(x_{1}).$
Then $H^{2}({\mathfrak{g}};{\mathfrak{g}})$ is spanned by the cocycles
$(\ref{eqbr2})$.
(15) $\tiny\begin{array}[]{ll}\deg=-6:&y_{1}\otimes
dx_{3}{\wedge}dx_{4}+y_{3}\otimes dx_{1}{\wedge}dx_{4}+2\,y_{4}\otimes
dx_{1}{\wedge}dx_{3};\\\ \deg=-3:&{x_{1}\otimes
dx_{2}{\wedge}dx_{4}+y_{2}\otimes dx_{4}{\wedge}dy_{1}+y_{4}\otimes
dx_{2}{\wedge}dy_{1}};\\\ \deg=0:&h_{1}\otimes
dx_{2}{\wedge}dy_{2}+h_{1}\otimes dx_{3}{\wedge}dy_{3}+x_{1}\otimes
dx_{3}{\wedge}dy_{2}-y_{1}\otimes dx_{2}{\wedge}dy_{3}\end{array}$
###### Comment.
Since we know that ${\mathfrak{br}}(2;a)$ admits a 3-parameter family of
deformations ([GL4]), three of these five cocycles (recall also the ones of
degrees 3 and 6) must be integrable. By symmetry the algebras obtained from
the cocycles of of degrees $\pm 3$ are isomorphic and so are the algebras
obtained from the cocycles of of degrees $\pm 6$. Anyway, the $\geq 3$
cocycles in the above case are expected, so all cocycles are integrable, as in
[DK]. Contrariwise, the next result is not expected; and we find it a bit too
much (are all these cocycle integrable?!), as in [Ch, Vi]. The application of
the group $\mathop{\mathrm{{Aut\/}}}\nolimits({\mathfrak{g}})$ by Kuznetsov
and Chebochko [KuCh, Ch1] makes the abundance of cocycles not so scary since
only the $\mathop{\mathrm{{Aut\/}}}\nolimits({\mathfrak{g}})$-orbits matter.
###### Lemma.
For ${\mathfrak{g}}={\mathfrak{br}}(3)$, we consider the Cartan matrix
(16) $\begin{pmatrix}2&-1&0\\\ -1&2&-1\\\
0&-1&{\bar{0}}\end{pmatrix}\quad\text{and the basis
}\;\tiny\begin{array}[]{l}x_{1},\;x_{2},\;x_{3};\\\
{}x_{4}=[x_{1},x_{2}],\quad x_{5}=[x_{2},x_{3}];\\\
{}x_{6}=[x_{3},[x_{1},x_{2}]],\quad x_{7}=[x_{3},[x_{2},x_{3}]];\\\
{}x_{8}=[x_{3},[x_{3},[x_{1},x_{2}]]];\\\
{}x_{9}=[[x_{2},x_{3}],[x_{3},[x_{1},x_{2}]]];\\\
{}x_{10}=[[x_{3},[x_{1},x_{2}]],[x_{3},[x_{2},x_{3}]]];\\\
{}x_{11}=[[x_{3},[x_{2},x_{3}]],[x_{3},[x_{3},[x_{1},x_{2}]]]];\\\
{}x_{12}=[[x_{3},[x_{2},x_{3}]],[[x_{2},x_{3}],[x_{3},[x_{1},x_{2}]]]];\\\
{}x_{13}=[[x_{3},[x_{3},[x_{1},x_{2}]]],[[x_{2},x_{3}],[x_{3},[x_{1},x_{2}]]]].\end{array}$
Then $H^{2}({\mathfrak{g}};{\mathfrak{g}})$ is spanned by the cocycles
$(\ref{eqbr3})$.
(17) $\tiny\begin{array}[]{ll}\deg=-18:&2\,y_{3}\otimes
dx_{12}{\wedge}dx_{13}+2\,y_{5}\otimes dx_{11}{\wedge}dx_{13}+2\,y_{6}\otimes
dx_{11}{\wedge}dx_{12}+y_{7}\otimes dx_{10}{\wedge}dx_{13}+y_{8}\otimes
dx_{10}{\wedge}dx_{12}+\\\ &y_{9}\otimes
dx_{10}{\wedge}dx_{11}+2\,y_{10}\otimes dx_{7}{\wedge}dx_{13}+y_{10}\otimes
dx_{8}{\wedge}dx_{12}+2\,y_{10}\otimes dx_{9}{\wedge}dx_{11}+2\,y_{11}\otimes
dx_{5}{\wedge}dx_{13}+\\\ &y_{11}\otimes
dx_{6}{\wedge}dx_{12}+2\,y_{11}\otimes dx_{9}{\wedge}dx_{10}+2\,y_{12}\otimes
dx_{3}{\wedge}dx_{13}+y_{12}\otimes dx_{6}{\wedge}dx_{11}+2\,y_{12}\otimes
dx_{8}{\wedge}dx_{10}\\\ &+2\,y_{13}\otimes
dx_{3}{\wedge}dx_{12}+y_{13}\otimes dx_{5}{\wedge}dx_{11}+2\,y_{13}\otimes
dx_{7}{\wedge}dx_{10};\\\ &\\\ \deg=-9:&2\,x_{3}\otimes
dx_{1}{\wedge}dx_{13}+x_{5}\otimes dx_{4}{\wedge}dx_{13}+x_{7}\otimes
dx_{6}{\wedge}dx_{13}+y_{1}\otimes dx_{4}{\wedge}dx_{10}+2\,y_{1}\otimes
dx_{6}{\wedge}dx_{9}+\\\ &2\,y_{1}\otimes dx_{13}{\wedge}dy_{3}+y_{4}\otimes
dx_{1}{\wedge}dx_{10}+y_{4}\otimes dx_{6}{\wedge}dx_{8}+y_{4}\otimes
dx_{13}{\wedge}dy_{5}+y_{6}\otimes dx_{1}{\wedge}dx_{9}+\\\ &y_{6}\otimes
dx_{4}{\wedge}dx_{8}+2\,y_{6}\otimes dx_{13}{\wedge}dy_{7}+2\,y_{8}\otimes
dx_{4}{\wedge}dx_{6}+y_{9}\otimes dx_{1}{\wedge}dx_{6}+2\,y_{10}\otimes
dx_{1}{\wedge}dx_{4}\\\ &+y_{13}\otimes dx_{1}{\wedge}dy_{3}+y_{13}\otimes
dx_{4}{\wedge}dy_{5}+y_{13}\otimes dx_{6}{\wedge}dy_{7};\\\ &\\\
\deg=-6:&2\,x_{1}\otimes dx_{2}{\wedge}dx_{10}+2\,x_{1}\otimes
dx_{5}{\wedge}dx_{9}+2\,x_{1}\otimes dx_{12}{\wedge}dy_{3}+x_{3}\otimes
dx_{12}{\wedge}dy_{1}+2\,x_{6}\otimes dx_{2}{\wedge}dx_{12}+\\\ &x_{8}\otimes
dx_{5}{\wedge}dx_{12}+2\,y_{2}\otimes dx_{5}{\wedge}dx_{7}+2\,y_{2}\otimes
dx_{10}{\wedge}dy_{1}+2\,y_{2}\otimes dx_{12}{\wedge}dy_{6}+2\,y_{5}\otimes
dx_{2}{\wedge}dx_{7}+\\\ &y_{5}\otimes dx_{9}{\wedge}dy_{1}+2\,y_{5}\otimes
dx_{12}{\wedge}dy_{8}+y_{7}\otimes dx_{2}{\wedge}dx_{5}+2\,y_{9}\otimes
dx_{5}{\wedge}dy_{1}+2\,y_{10}\otimes dx_{2}{\wedge}dy_{1}\\\
&+2\,y_{12}\otimes dx_{2}{\wedge}dy_{6}+y_{12}\otimes
dx_{5}{\wedge}dy_{8}+2\,y_{12}\otimes dy_{1}{\wedge}dy_{3};\\\ &\\\
\deg=-3:&2\,x_{2}\otimes dx_{3}{\wedge}dx_{7}+x_{2}\otimes
dx_{10}{\wedge}dy_{4}+2\,x_{2}\otimes dx_{11}{\wedge}dy_{6}+x_{4}\otimes
dx_{3}{\wedge}dx_{8}+x_{4}\otimes dx_{10}{\wedge}dy_{2}+\\\ &x_{4}\otimes
dx_{11}{\wedge}dy_{5}+2\,x_{5}\otimes dx_{11}{\wedge}dy_{4}+x_{6}\otimes
dx_{11}{\wedge}dy_{2}+x_{9}\otimes dx_{3}{\wedge}dx_{11}+2\,y_{3}\otimes
dx_{7}{\wedge}dy_{2}+\\\ &2\,y_{3}\otimes dx_{8}{\wedge}dy_{4}+2\,y_{3}\otimes
dx_{11}{\wedge}dy_{9}+2\,y_{7}\otimes dx_{3}{\wedge}dy_{2}+y_{8}\otimes
dx_{3}{\wedge}dy_{4}+y_{10}\otimes dy_{2}{\wedge}dy_{4}\\\ &+y_{11}\otimes
dx_{3}{\wedge}dy_{9}+y_{11}\otimes dy_{2}{\wedge}dy_{6}+y_{11}\otimes
dy_{4}{\wedge}dy_{5};\end{array}$
###### Lemma.
For ${\mathfrak{g}}={\mathfrak{brj}}(2;3)$, we consider the Cartan matrix
(18) $\begin{pmatrix}0&-1\\\ -2&1\end{pmatrix}\quad\begin{matrix}\text{and the
basis }\\\ \text{even $|$ odd
}\end{matrix}\;\tiny\begin{array}[]{l}\;|\;x_{1},\;x_{2},\\\
x_{3}=\left[x_{1},\,x_{2}\right]\;|\\\ x_{4}=\left[x_{2},\,x_{2}\right],\;|\\\
\;|\;x_{5}=\left[x_{2},\,\left[x_{1},\,x_{2}\right]\right]\\\
x_{6}=\left[\left[x_{1},\,x_{2}\right],\,\left[x_{2},\,x_{2}\right]\right],\;|\\\
\;|\;x_{7}=\left[\left[x_{2},\,x_{2}\right],\,\left[x_{2},\,\left[x_{1},\,x_{2}\right]\right]\right]\\\
x_{8}=\left[\left[x_{1},\,x_{2}\right],\,\left[\left[x_{1},\,x_{2}\right],\,\left[x_{2},\,x_{2}\right]\right]\right]\;|\end{array}$
Then $H^{2}({\mathfrak{g}};{\mathfrak{g}})$ is spanned by the cocycles
(19) $\tiny\begin{array}[]{ll}\deg=-12:&2\,y_{2}\otimes
dx_{7}{\wedge}dx_{8}+y_{3}\otimes dx_{7}{\wedge}dx_{7}+y_{4}\otimes
dx_{6}{\wedge}dx_{8}+y_{5}\otimes dx_{6}{\wedge}dx_{7}+2\,y_{6}\otimes
dx_{4}{\wedge}dx_{8}+\\\ &2\,y_{6}\otimes dx_{5}{\wedge}dx_{7}+y_{7}\otimes
dx_{2}{\wedge}dx_{8}+2\,y_{7}\otimes dx_{3}{\wedge}dx_{7}+2\,y_{7}\otimes
dx_{5}{\wedge}dx_{6}+2\,y_{8}\otimes dx_{2}{\wedge}dx_{7}+\\\ &2\,y_{8}\otimes
dx_{4}{\wedge}dx_{6}\\\ &\\\ \deg=-6:&x_{2}\otimes
dx_{1}{\wedge}dx_{8}+2\,x_{4}\otimes dx_{3}{\wedge}dx_{8}+2\,y_{1}\otimes
dx_{1}{\wedge}dx_{6}+2\,y_{1}\otimes dx_{3}{\wedge}dx_{5}+y_{1}\otimes
dx_{8}{\wedge}dy_{2}+\\\ &2\,y_{3}\otimes dx_{1}{\wedge}dx_{5}+2\,y_{3}\otimes
dx_{8}{\wedge}dy_{4}+y_{5}\otimes dx_{1}{\wedge}dx_{3}+y_{6}\otimes
dx_{1}{\wedge}dx_{1}+2\,y_{8}\otimes dx_{1}{\wedge}dy_{2}+\\\ &2\,y_{8}\otimes
dx_{3}{\wedge}dy_{4}\end{array}$
###### Lemma.
For ${\mathfrak{g}}={\mathfrak{brj}}(2;5)$, we consider the Cartan matrix
(20) $\begin{pmatrix}0&-1\\\ -2&1\end{pmatrix}\quad\begin{matrix}\text{and the
basis }\\\ \text{even $|$ odd
}\end{matrix}\;\tiny\begin{array}[]{l}\;|\;x_{1},\;x_{2},\\\
x_{3}=\left[x_{1},\,x_{2}\right],\;x_{4}=\left[x_{2},\,x_{2}\right],\;|\\\
\;|\;x_{5}=\left[x_{2},\,\left[x_{1},\,x_{2}\right]\right]\\\
x_{6}=\left[\left[x_{1},\,x_{2}\right],\,\left[x_{2},\,x_{2}\right]\right],\;|\\\
\;|\;x_{7}=\left[\left[x_{1},\,x_{2}\right],\,\left[x_{2},\,\left[x_{1},\,x_{2}\right]\right]\right],\;x_{8}=\left[\left[x_{2},\,x_{2}\right],\,\left[x_{2},\,\left[x_{1},\,x_{2}\right]\right]\right],\\\
x_{9}=\left[\left[x_{1},\,x_{2}\right],\,\left[\left[x_{1},\,x_{2}\right],\,\left[x_{2},\,x_{2}\right]\right]\right]\;|\\\
\;|\;x_{10}=\left[\left[x_{2},\,\left[x_{1},\,x_{2}\right]\right],\,\left[\left[x_{1},\,x_{2}\right],\,\left[x_{2},\,x_{2}\right]\right]\right]\end{array}$
Then $H^{2}({\mathfrak{g}};{\mathfrak{g}})$ is spanned by the cocycles
(21) $\tiny\begin{array}[]{ll}\underline{\deg=-15}&2\,y_{1}\otimes
dx_{10}\wedge dx_{10}+3\,y_{3}\otimes dx_{9}\wedge dx_{10}+y_{5}\otimes
dx_{7}\wedge dx_{10}+3\,y_{6}\otimes dx_{7}\wedge dx_{9}+3\,y_{7}\otimes
dx_{5}\wedge dx_{10}+\\\ &4\,y_{7}\otimes dx_{6}\wedge dx_{9}+y_{7}\otimes
dx_{7}\wedge dx_{8}+2\,y_{8}\otimes dx_{7}\wedge dx_{7}+4\,y_{9}\otimes
dx_{3}\wedge dx_{10}+y_{9}\otimes dx_{6}\wedge dx_{7}+\\\ &4\,y_{10}\otimes
dx_{1}\wedge dx_{10}+4\,y_{10}\otimes dx_{3}\wedge dx_{9}+2\,y_{10}\otimes
dx_{5}\wedge dx_{7}\\\ &\\\ \underline{\deg=-5}:&3\,x_{1}\otimes
dx_{2}{\wedge}dx_{8}+4\,x_{1}\otimes dx_{4}{\wedge}dx_{6}+3\,x_{1}\otimes
dx_{10}{\wedge}dy_{1}+4\,x_{3}\otimes dx_{4}{\wedge}dx_{8}+3\,x_{7}\otimes
dx_{8}){\wedge}dx_{8}+\\\ &y_{2}\otimes dx_{8}{\wedge}dy_{1}+4\,y_{4}\otimes
dx_{6}{\wedge}dy_{1}+2\,y_{4}\otimes dx_{8}{\wedge}dy_{3}+2\,y_{6}\otimes
dx_{4}{\wedge}dy_{1}+2\,y_{8}\otimes dx_{2}{\wedge}dy_{1}+\\\ &3\,y_{8}\otimes
dx_{4}{\wedge}dy_{3}+y_{8}\otimes dx_{8}{\wedge}dy_{7}+y_{10}\otimes
dy_{1}{\wedge}dy_{1}\end{array}$
###### Lemma.
For ${\mathfrak{g}}={\mathfrak{g}}(1,6)$, we consider the Cartan matrix
(22) $\begin{pmatrix}2&-1&0\\\ -1&1&-1\\\
0&-1&0\end{pmatrix}\quad\begin{matrix}\text{and the basis }\\\ \text{even $|$
odd }\end{matrix}\;\tiny\begin{array}[]{l}x_{1},\;|\;x_{2},\;x_{3}\\\
x_{5}=\left[x_{2},\,x_{2}\right],\;|\;x_{4}=\left[x_{1},\,x_{2}\right],\\\
x_{6}=\left[x_{2},\,x_{3}\right],\;x_{7}=\left[x_{2},\,\left[x_{1},\,x_{2}\right]\right]\;|\\\
x_{8}=\left[x_{3},\,\left[x_{1},\,x_{2}\right]\right],\;|\;x_{9}=\left[x_{3},\,\left[x_{2},\,x_{2}\right]\right]\\\
x_{10}=\left[\left[x_{1},\,x_{2}\right],\,\left[x_{1},\,x_{2}\right]\right],\;|\;x_{11}=\left[\left[x_{1},\,x_{2}\right],\,\left[x_{2},\,x_{3}\right]\right]\\\
\;|\;x_{12}=\left[\left[x_{1},\,x_{2}\right],\,\left[x_{3},\,\left[x_{1},\,x_{2}\right]\right]\right]\\\
x_{13}=\left[\left[x_{2},\,x_{3}\right],\,\left[x_{2},\,\left[x_{1},\,x_{2}\right]\right]\right]\;|\\\
x_{14}=\left[\left[x_{2},\,\left[x_{1},\,x_{2}\right]\right],\,\left[x_{3},\,\left[x_{1},\,x_{2}\right]\right]\right]\;|\\\
\;|\;x_{15}=\left[\left[x_{3},\,\left[x_{2},\,x_{2}\right]\right],\,\left[\left[x_{1},\,x_{2}\right],\,\left[x_{1},\,x_{2}\right]\right]\right]\\\
x_{16}=\left[\left[x_{3},\,\left[x_{2},\,x_{2}\right]\right],\,\left[\left[x_{1},\,x_{2}\right],\,\left[x_{3},\,\left[x_{1},\,x_{2}\right]\right]\right]\right]\;|\end{array}$
Then $H^{2}({\mathfrak{g}};{\mathfrak{g}})$ is spanned by the cocycles
$(\ref{eq1,3})$.
(23) $\tiny\begin{array}[]{ll}\underline{\deg=-12}:&2\,x_{2}\otimes
dx_{13}{\wedge}dx_{16}+2\,x_{4}\otimes dx_{14}{\wedge}dx_{16}+y_{2}\otimes
dx_{8}{\wedge}dx_{16}+2\,y_{3}\otimes dx_{13}{\wedge}dx_{14}+y_{4}\otimes
dx_{6}{\wedge}dx_{16}+\\\ &y_{6}\otimes dx_{4}{\wedge}dx_{16}+y_{6}\otimes
dx_{8}{\wedge}dx_{15}+2\,y_{6}\otimes dx_{11}{\wedge}dx_{14}+2\,y_{6}\otimes
dx_{12}{\wedge}dx_{13}+y_{8}\otimes dx_{2}{\wedge}dx_{16}+\\\ &y_{8}\otimes
dx_{6}{\wedge}dx_{15}+y_{8}\otimes dx_{9}{\wedge}dx_{14}+y_{8}\otimes
dx_{11}{\wedge}dx_{13}+y_{9}\otimes dx_{8}{\wedge}dx_{14}+y_{11}\otimes
dx_{6}{\wedge}dx_{14}\\\ &+y_{11}\otimes dx_{8}{\wedge}dx_{13}+y_{12}\otimes
dx_{6}{\wedge}dx_{13}+2\,y_{13}\otimes dx_{3}{\wedge}dx_{14}+2\,y_{13}\otimes
dx_{6}{\wedge}dx_{12}+2\,y_{13}\otimes dx_{8}{\wedge}dx_{11}+\\\
&y_{13}\otimes dx_{16}{\wedge}dy_{2}+2\,y_{14}\otimes
dx_{3}{\wedge}dx_{13}+y_{14}\otimes dx_{6}{\wedge}dx_{11}+y_{14}\otimes
dx_{8}{\wedge}dx_{9}+y_{14}\otimes dx_{16}{\wedge}dy_{4}+\\\ &2\,y_{15}\otimes
dx_{6}{\wedge}dx_{8}+y_{16}\otimes dx_{2}{\wedge}dx_{8}+2\,y_{16}\otimes
dx_{4}{\wedge}dx_{6}+y_{16}\otimes dx_{13}{\wedge}dy_{2}+2\,y_{16}\otimes
dx_{14}{\wedge}dy_{4};\\\ \underline{\deg=-6}:&x_{2}\otimes\left(dx_{7}\wedge
dx_{10}\right)+x_{3}\otimes\left(dx_{1}\wedge
dx_{14}\right)+x_{3}\otimes\left(dx_{4}\wedge
dx_{12}\right)+x_{3}\otimes\left(dx_{8}\wedge dx_{10}\right)\text{+
}2\,x_{6}\otimes\left(dx_{1}\wedge dx_{15}\right)+\\\
&2\,x_{6}\otimes\left(dx_{7}\wedge
dx_{12}\right)+x_{8}\otimes\left(dx_{10}\wedge
dx_{12}\right)+x_{9}\otimes\left(dx_{4}\wedge dx_{15}\right)\text{+
}2\,x_{9}\otimes\left(dx_{7}\wedge dx_{14}\right)\text{+
}2\,x_{9}\otimes\left(dx_{10}\wedge dx_{13}\right)\\\
&+x_{13}\otimes\left(dx_{10}\wedge dx_{15}\right)\text{+
}2\,y_{1}\otimes\left(dx_{2}\wedge dx_{10}\right)\text{+
}2\,y_{1}\otimes\left(dx_{4}\wedge
dx_{7}\right)+y_{1}\otimes\left(dx_{14}\wedge
dy_{3}\right)+y_{1}\otimes\left(dx_{15}\wedge dy_{6}\right)\\\
&+y_{2}\otimes\left(dx_{1}\wedge dx_{10}\right)\text{+
}2\,y_{4}\otimes\left(dx_{1}\wedge
dx_{7}\right)+y_{4}\otimes\left(dx_{12}\wedge
dy_{3}\right)+y_{4}\otimes\left(dx_{15}\wedge
dy_{9}\right)+y_{7}\otimes\left(dx_{1}\wedge dx_{4}\right)\\\
&+y_{7}\otimes\left(dx_{10}\wedge
dy_{2}\right)+y_{7}\otimes\left(dx_{12}\wedge dy_{6}\right)\text{+
}2\,y_{7}\otimes\left(dx_{14}\wedge dy_{9}\right)\text{+
}2\,y_{8}\otimes\left(dx_{10}\wedge
dy_{3}\right)+y_{10}\otimes\left(dx_{1}\wedge dx_{2}\right)\\\ &\text{+
}2\,y_{10}\otimes\left(dx_{7}\wedge dy_{2}\right)\text{+
}2\,y_{10}\otimes\left(dx_{8}\wedge
dy_{3}\right)+y_{10}\otimes\left(dx_{12}\wedge dy_{8}\right)\text{+
}2\,y_{10}\otimes\left(dx_{13}\wedge
dy_{9}\right)+y_{10}\otimes\left(dx_{15}\wedge dy_{13}\right)\\\ &\text{+
}2\,y_{12}\otimes\left(dx_{4}\wedge dy_{3}\right)\text{+
}2\,y_{12}\otimes\left(dx_{7}\wedge dy_{6}\right)\text{+
}2\,y_{12}\otimes\left(dx_{10}\wedge dy_{8}\right)\text{+
}2\,y_{13}\otimes\left(dx_{10}\wedge dy_{9}\right)\text{+
}2\,y_{14}\otimes\left(dx_{1}\wedge dy_{3}\right)\\\
&+y_{14}\otimes\left(dx_{7}\wedge dy_{9}\right)\text{+
}2\,y_{15}\otimes\left(dx_{1}\wedge dy_{6}\right)\text{+
}2\,y_{15}\otimes\left(dx_{4}\wedge dy_{9}\right)\text{+
}2\,y_{15}\otimes\left(dx_{10}\wedge dy_{13}\right);\\\
\underline{\deg=-3}:&2\,x_{1}\otimes\left(dx_{2}\wedge
dx_{7}\right)+2\,x_{1}\otimes\left(dx_{4}\wedge
dx_{5}\right)+2\,x_{1}\otimes\left(dx_{13}\wedge
dy_{3}\right)+2\,x_{1}\otimes\left(dx_{15}\wedge
dy_{8}\right)+x_{3}\otimes\left(dx_{2}\wedge dx_{9}\right)\\\
&+2\,x_{3}\otimes\left(dx_{5}\wedge
dx_{6}\right)+x_{3}\otimes\left(dx_{13}\wedge
dy_{1}\right)+2\,x_{4}\otimes\left(dx_{5}\wedge
dx_{7}\right)+x_{6}\otimes\left(dx_{5}\wedge dx_{9}\right)\\\
&+2\,x_{8}\otimes\left(dx_{7}\wedge
dx_{9}\right)+2\,x_{8}\otimes\left(dx_{15}\wedge
dy_{1}\right)+x_{12}\otimes\left(dx_{2}\wedge
dx_{15}\right)+x_{12}\otimes\left(dx_{5}\wedge
dx_{14}\right)+x_{12}\otimes\left(dx_{7}\wedge dx_{13}\right)\\\ &\text{+
}2\,x_{14}\otimes\left(dx_{5}\wedge dx_{15}\right)\text{+
}2\,y_{2}\otimes\left(dx_{7}\wedge
dy_{1}\right)+2\,y_{2}\otimes\left(dx_{9}\wedge
dy_{3}\right)+2\,y_{2}\otimes\left(dx_{15}\wedge
dy_{12}\right)+y_{4}\otimes\left(dx_{5}\wedge dy_{1}\right)\\\
&+y_{5}\otimes\left(dx_{4}\wedge
dy_{1}\right)+2\,y_{5}\otimes\left(dx_{6}\wedge
dy_{3}\right)+y_{5}\otimes\left(dx_{7}\wedge
dy_{4}\right)+2\,y_{5}\otimes\left(dx_{9}\wedge
dy_{6}\right)+y_{5}\otimes\left(dx_{14}\wedge dy_{12}\right)\\\
&+y_{5}\otimes\left(dx_{15}\wedge
dy_{14}\right)+y_{6}\otimes\left(dx_{5}\wedge
dy_{3}\right)+y_{7}\otimes\left(dx_{2}\wedge
dy_{1}\right)+2\,y_{7}\otimes\left(dx_{5}\wedge
dy_{4}\right)+2\,y_{7}\otimes\left(dx_{9}\wedge dy_{8}\right)\\\
&+y_{7}\otimes\left(dx_{13}\wedge dy_{12}\right)\text{+
}2\,y_{9}\otimes\left(dx_{2}\wedge
dy_{3}\right)+y_{9}\otimes\left(dx_{5}\wedge
dy_{6}\right)+y_{9}\otimes\left(dx_{7}\wedge
dy_{8}\right)+y_{13}\otimes\left(dx_{7}\wedge dy_{12}\right)\\\ &\text{+
}2\,y_{13}\otimes\left(dy_{1}\wedge
dy_{3}\right)+2\,y_{14}\otimes\left(dx_{5}\wedge dy_{12}\right)\text{+
}2\,y_{15}\otimes\left(dx_{2}\wedge dy_{12}\right)\text{+
}2\,y_{15}\otimes\left(dx_{5}\wedge
dy_{14}\right)+y_{15}\otimes\left(dy_{1}\wedge dy_{8}\right).\end{array}$
###### Lemma.
For ${\mathfrak{g}}={\mathfrak{g}}(2,3)$, we consider the Cartan matrix
(24) $\begin{pmatrix}2&-1&-1\\\ -1&2&-1\\\
-1&-1&0\end{pmatrix}\quad\begin{matrix}\text{and the basis }\\\ \text{even $|$
odd }\end{matrix}\;\tiny\begin{array}[]{l}x_{1},\;x_{2},\;|\;x_{3}\\\
x_{4}=\left[x_{1},\,x_{2}\right],\;|\;x_{5}=\left[x_{1},\,x_{3}\right]\\\
\;|\;x_{6}=\left[x_{2},\,x_{3}\right],\;x_{7}=\left[x_{3},\,\left[x_{1},\,x_{2}\right]\right],\\\
\;|\;x_{8}=\left[\left[x_{1},\,x_{2}\right],\,\left[x_{1},\,x_{3}\right]\right],\;x_{9}=\left[\left[x_{1},\,x_{2}\right],\,\left[x_{2},\,x_{3}\right]\right]\\\
\;|\;x_{10}=\left[\left[x_{1},\,x_{2}\right],\,\left[x_{3},\,\left[x_{1},\,x_{2}\right]\right]\right]\\\
x_{11}=\left[\left[x_{2},\,x_{3}\right],\,\left[\left[x_{1},\,x_{2}\right],\,\left[x_{1},\,x_{3}\right]\right]\right]\;|\end{array}$
Then $H^{2}({\mathfrak{g}};{\mathfrak{g}})$ is spanned by the cocycles
$(\ref{eqg2,3})$.
(25) $\tiny\begin{array}[]{ll}\underline{\deg=-9}:&2\,x_{2}\otimes
dx_{9}{\wedge}dx_{11}+2\,x_{4}\otimes dx_{10}{\wedge}dx_{11}+y_{2}\otimes
dx_{5}{\wedge}dx_{11}+y_{3}\otimes dx_{4}{\wedge}dx_{11}+2\,y_{3}\otimes
dx_{7}{\wedge}dx_{10}+\\\ &y_{3}\otimes dx_{8}{\wedge}dx_{9}+y_{4}\otimes
dx_{3}{\wedge}dx_{11}+y_{5}\otimes dx_{2}{\wedge}dx_{11}+2\,y_{5}\otimes
dx_{6}{\wedge}dx_{10}+y_{5}\otimes dx_{7}{\wedge}dx_{9}+\\\ &2\,y_{6}\otimes
dx_{5}){\wedge}dx_{10}+y_{7}\otimes dx_{3}{\wedge}dx_{10}+y_{7}\otimes
dx_{5}{\wedge}dx_{9}+2\,y_{8}\otimes dx_{3}{\wedge}dx_{9}+2\,y_{9}\otimes
dx_{3}{\wedge}dx_{8}+\\\ &y_{9}\otimes dx_{5}{\wedge}dx_{7}+y_{9}\otimes
dx_{11}{\wedge}dy_{2}+2\,y_{10}\otimes dx_{3}{\wedge}dx_{7}+y_{10}\otimes
dx_{5}{\wedge}dx_{6}+y_{10}\otimes dx_{11}{\wedge}dy_{4}+\\\ &2\,y_{11}\otimes
dx_{2}{\wedge}dx_{5}+y_{11}\otimes dx_{3}{\wedge}dx_{4}+y_{11}\otimes
dx_{9}{\wedge}dy_{2}+2\,y_{11}\otimes dx_{10}{\wedge}dy_{4}\end{array}$
###### Lemma.
For ${\mathfrak{g}}={\mathfrak{g}}(3,3)$ and
${\mathfrak{g}}={\mathfrak{g}}(4,3)$, we have
$H^{2}({\mathfrak{g}};{\mathfrak{g}})=0$.
###### Lemma.
The “classical” (serial) simple Lie superalgebras of rank $=2$ and $3$ and
with Cartan matrix are rigid if $p=3$ and $5$, except for
${\mathfrak{osp}}(4|2)$. The corresponding cocycles are (26) for $p\geq 5$ and
$p=0$:
(26) $\tiny\begin{array}[]{l}2h_{1}\otimes dx_{2}\wedge dy_{2}+3h_{1}\otimes
dx_{3}\wedge dy_{3}+2h_{1}\otimes dx_{4}\wedge dy_{4}+h_{1}\otimes
dx_{6}\wedge dy_{6}+3h_{1}\otimes dx_{7}\wedge dy_{7}+\\\ h_{2}\otimes
dx_{3}\wedge dy_{3}+4h_{2}\otimes dx_{5}\wedge dy_{5}+h_{2}\otimes
dx_{6}\wedge dy_{6}+h_{2}\otimes dx_{7}\wedge dy_{7}+4x_{1}\otimes
dh_{3}\wedge dx_{1}+\\\ x_{1}\otimes dx_{4}\wedge dy_{2}+x_{1}\otimes
dx_{6}\wedge dy_{5}+2x_{2}\otimes dh_{2}\wedge dx_{2}+x_{2}\otimes
dh_{3}\wedge dx_{2}+3x_{3}\otimes dh_{2}\wedge dx_{3}+\\\ 2x_{4}\otimes
dh_{2}\wedge dx_{4}+x_{5}\otimes dh_{3}\wedge dx_{5}+3x_{5}\otimes
dx_{2}\wedge dx_{3}+2x_{6}\otimes dx_{3}\wedge dx_{4}+2x_{7}\otimes
dh_{2}\wedge dx_{7}+\\\ x_{7}\otimes dh_{3}\wedge dx_{7}+3x_{7}\otimes
dx_{2}\wedge dx_{6}+3x_{7}\otimes dx_{4}\wedge dx_{5}+y_{1}\otimes
dh_{3}\wedge dy_{1}+4y_{1}\otimes dx_{2}\wedge dy_{4}+\\\ 4y_{1}\otimes
dx_{5}\wedge dy_{6}+3y_{2}\otimes dh_{2}\wedge dy_{2}+4y_{2}\otimes
dh_{3}\wedge dy_{2}+3y_{2}\otimes dx_{3}\wedge dy_{5}+2y_{2}\otimes
dx_{6}\wedge dy_{7}+\\\ 2y_{3}\otimes dh_{2}\wedge dy_{3}+2y_{3}\otimes
dx_{2}\wedge dy_{5}+2y_{3}\otimes dx_{4}\wedge dy_{6}+3y_{4}\otimes
dh_{2}\wedge dy_{4}+2y_{4}\otimes dx_{3}\wedge dy_{6}+\\\ 3y_{4}\otimes
dx_{5}\wedge dy_{7}+4y_{5}\otimes dh_{3}\wedge dy_{5}+3y_{5}\otimes
dx_{4}\wedge dy_{7}+2y_{6}\otimes dx_{2}\wedge dy_{7}+3y_{7}\otimes
dh_{2}\wedge dy_{7}+\\\ 4y_{7}\otimes dh_{3}\wedge dy_{7}\end{array}$
and (27) for $p=3$:
(27) $\tiny\begin{array}[]{l}h_{1}\otimes dx_{2}\wedge dy_{2}+h_{1}\otimes
dx_{3}\wedge dy_{3}+h_{1}\otimes dx_{4}\wedge dy_{4}+h_{1}\otimes dx_{5}\wedge
dy_{5}+\\\ h_{1}\otimes dx_{7}\wedge dy_{7}+2h_{2}\otimes dx_{3}\wedge
dy_{3}+h_{2}\otimes dx_{7}\wedge dy_{7}+x_{1}\otimes dh_{3}\wedge dx_{1}+\\\
x_{1}\otimes dx_{4}\wedge dy_{2}+x_{1}\otimes dx_{6}\wedge dy_{5}+x_{2}\otimes
dh_{2}\wedge dx_{2}+2x_{2}\otimes dh_{3}\wedge dx_{2}+\\\ 2x_{2}\otimes
dx_{5}\wedge dy_{3}+x_{2}\otimes dx_{7}\wedge dy_{6}+2x_{3}\otimes
dh_{2}\wedge dx_{3}+x_{4}\otimes dh_{2}\wedge dx_{4}+\\\ x_{4}\otimes
dx_{6}\wedge dy_{3}+2x_{4}\otimes dx_{7}\wedge dy_{5}+2x_{5}\otimes
dh_{3}\wedge dx_{5}+2x_{5}\otimes dx_{2}\wedge dx_{3}+\\\ x_{6}\otimes
dx_{3}\wedge dx_{4}+x_{7}\otimes dh_{2}\wedge dx_{7}+2x_{7}\otimes
dh_{3}\wedge dx_{7}+2x_{7}\otimes dx_{2}\wedge dx_{6}+\\\ 2x_{7}\otimes
dx_{4}\wedge dx_{5}+2y_{1}\otimes dh_{3}\wedge dy_{1}+2y_{1}\otimes
dx_{2}\wedge dy_{4}+2y_{1}\otimes dx_{5}\wedge dy_{6}+\\\ 2y_{2}\otimes
dh_{2}\wedge dy_{2}+y_{2}\otimes dh_{3}\wedge dy_{2}+y_{2}\otimes dx_{3}\wedge
dy_{5}+y_{3}\otimes dh_{2}\wedge dy_{3}+\\\ y_{3}\otimes dx_{2}\wedge
dy_{5}+y_{3}\otimes dx_{4}\wedge dy_{6}+2y_{4}\otimes dh_{2}\wedge
dy_{4}+2y_{4}\otimes dx_{3}\wedge dy_{6}+\\\ y_{5}\otimes dh_{3}\wedge
dy_{5}+2y_{5}\otimes dx_{4}\wedge dy_{7}+y_{6}\otimes dx_{2}\wedge
dy_{7}+2y_{7}\otimes dh_{2}\wedge dy_{7}+y_{7}\otimes dh_{3}\wedge
dy_{7}\end{array}$
###### Lemma.
For ${\mathfrak{g}}={\mathfrak{psq}}(3)$, we have
$H^{2}({\mathfrak{g}};{\mathfrak{g}})=0$ for $p=5$.
For $p=3$, when ${\mathfrak{g}}={\mathfrak{psq}}(3)$ is not simple, we take
the simple algebra ${\mathfrak{g}}^{\prime}={\mathfrak{psq}}^{(1)}(3)$ and
$H^{2}({\mathfrak{g}}^{\prime};{\mathfrak{g}}^{\prime})$ is spanned by the
$2$-cocycle
(28)
$\tiny\begin{array}[]{ll}\underline{\deg=0}:&2\,a_{2,2}\otimes\text{da}_{2,2}\wedge\text{db}_{2}+2\,a_{2,2}\otimes\text{da}_{3,3}\wedge\text{db}_{2}+2\,a_{2,2}\otimes\text{da}_{1,2}\wedge\text{db}_{2,1}+a_{2,2}\otimes\text{da}_{1,3}\wedge\text{db}_{3,1}+a_{2,2}\otimes\text{da}_{3,2}\wedge\text{db}_{2,3}+\\\
&a_{2,2}\otimes\text{da}_{3,1}\wedge\text{db}_{1,3}+2\,a_{3,3}\otimes\text{da}_{2,2}\wedge\text{db}_{1}+2\,a_{3,3}\otimes\text{da}_{2,2}\wedge\text{db}_{2}+2\,a_{3,3}\otimes\text{da}_{3,3}\wedge\text{db}_{1}+2\,a_{3,3}\otimes\text{da}_{3,3}\wedge\text{db}_{2}+\\\
&a_{3,3}\otimes\text{da}_{2,1}\wedge\text{db}_{1,2}+2\,a_{3,3}\otimes\text{da}_{3,2}\wedge\text{db}_{2,3}+a_{3,3}\otimes\text{da}_{3,1}\wedge\text{db}_{1,3}+2\,a_{1,2}\otimes\text{da}_{1,2}\wedge\text{db}_{2})+2\,a_{1,2}\otimes\text{da}_{3,2}\wedge\text{db}_{1,3}\\\
&+a_{2,3}\otimes\text{da}_{2,3}\wedge\text{db}_{2}+a_{1,3}\otimes\text{da}_{2,2}\wedge\text{db}_{1,3}+a_{1,3}\otimes\text{da}_{3,3}\wedge\text{db}_{1,3}+a_{1,3}\otimes\text{da}_{2,3}\wedge\text{db}_{1,2}+a_{1,3}\otimes\text{da}_{1,3}\wedge\text{db}_{2}+\\\
&2\,a_{2,1}\otimes\text{da}_{2,1}\wedge\text{db}_{1}+a_{3,2}\otimes\text{da}_{3,2}\wedge\text{db}_{1}+a_{3,1}\otimes\text{da}_{2,2}\wedge\text{db}_{3,1}+a_{3,1}\otimes\text{da}_{3,3}\wedge\text{db}_{3,1}+2\,a_{3,1}\otimes\text{da}_{3,1}\wedge\text{db}_{1}+\\\
&a_{3,1}\otimes\text{da}_{3,1}\wedge\text{db}_{2}+2\,b_{1}\otimes\text{db}_{1}\wedge\text{db}_{1}+b_{1}\otimes\text{db}_{1,3}\wedge\text{db}_{3,1}+b_{2}\otimes\text{db}_{2}\wedge\text{db}_{2}+b_{1,2}\otimes\text{da}_{2,2}\wedge\text{da}_{1,2}\\\
&+b_{1,2}\otimes\text{da}_{3,3}\wedge\text{da}_{1,2}+b_{1,2}\otimes\text{da}_{1,3}\wedge\text{da}_{3,2}+b_{1,2}\otimes\text{db}_{1,3}\wedge\text{db}_{3,2}+2\,b_{2,3}\otimes\text{da}_{2,2}\wedge\text{da}_{2,3}+2\,b_{2,3}\otimes\text{da}_{3,3}\wedge\text{da}_{2,3}+\\\
&2\,b_{2,3}\otimes\text{da}_{1,3}\wedge\text{da}_{2,1}+b_{2,3}\otimes\text{db}_{1}\wedge\text{db}_{2,3}+2\,b_{2,3}\otimes\text{db}_{2}\wedge\text{db}_{2,3}+2\,b_{1,3}\otimes\text{da}_{1,2}\wedge\text{da}_{2,3}+2\,b_{1,3}\otimes\text{db}_{1}\wedge\text{db}_{1,3}+\\\
&2\,b_{2,1}\otimes\text{da}_{2,2}\wedge\text{da}_{2,1}+2\,b_{2,1}\otimes\text{da}_{3,3}\wedge\text{da}_{2,1}+2\,b_{2,1}\otimes\text{da}_{2,3}\wedge\text{da}_{3,1}+b_{2,1}\otimes\text{db}_{1}\wedge\text{db}_{2,1}+2\,b_{2,1}\otimes\text{db}_{2}\wedge\text{db}_{2,1}\\\
&+b_{2,1}\otimes\text{db}_{2,3}\wedge\text{db}_{3,1}+b_{3,2}\otimes\text{da}_{2,2}\wedge\text{da}_{3,2}+b_{3,2}\otimes\text{da}_{3,3}\wedge\text{da}_{3,2}+b_{3,2}\otimes\text{da}_{1,2}\wedge\text{da}_{3,1}+b_{3,2}\otimes\text{db}_{1,2}\wedge\text{db}_{3,1}+\\\
&2\,b_{3,1}\otimes\text{da}_{2,1}\wedge\text{da}_{3,2}\end{array}$
###### Lemma.
For ${\mathfrak{g}}={\mathfrak{psq}}(4)$, we have
$H^{2}({\mathfrak{g}};{\mathfrak{g}})=0$ for $p=3$ and for $p=5$.
We did not insert here the (interesting) result of deformations of
${\mathfrak{psq}}(2)$ because it is not simple.
## 4\. Results for $p=2$
If the shapes of the answers for the usual cohomology and divided power one do
not differ, we only give the answer for the usual one.
###### Lemma.
For ${\mathfrak{g}}={\mathfrak{o}}^{(1)}(3)$, the space
$H^{2}({\mathfrak{g}};{\mathfrak{g}})$ is spanned by the cocycles
$(\ref{eqo3})$:
(29) $\footnotesize\begin{array}[]{ll}\deg=-2:&y\otimes dh\wedge
dx\end{array}$
###### Lemma.
For ${\mathfrak{g}}={\mathfrak{o}}{\mathfrak{o}}_{I\Pi}^{(1)}(1|2)$, the space
$H^{2}({\mathfrak{g}};{\mathfrak{g}})$ is spanned by the cocycles
$(\ref{eqoo12})$:
(30) $\footnotesize\begin{array}[]{ll}\deg=-2:&h_{1}\otimes
d\left(x_{1}\right){}^{\wedge 2}+x_{1}\otimes\left(dx_{1}\wedge
dx_{2}\right)+y_{2}\otimes\left(dx_{1}\wedge
dy_{1}\right)+y_{2}\otimes\left(dx_{2}\wedge dy_{2}\right)\end{array}$
###### Lemma.
For ${\mathfrak{g}}={\mathfrak{sl}}(3)$, we have
$H^{2}({\mathfrak{g}};{\mathfrak{g}})=0$ for any $p$.
###### Lemma.
For ${\mathfrak{g}}={\mathfrak{o}}^{(1)}_{I}(5)$, we take the Cartan matrix
(31) $\begin{pmatrix}\overline{1}&1\\\ 1&{\bar{0}}\\\
\end{pmatrix}\qquad\begin{array}[]{l}\text{and the basis }\\\
x_{1},x_{2},x_{3}=[x_{1}\text{,$\,$}x_{2}],x_{4}=[x_{1}\text{,$\,$}[x_{1}\text{,$\,$}x_{2}]].\end{array}$
Then $H^{2}({\mathfrak{g}};{\mathfrak{g}})$ is spanned by the cocycles
$(\ref{eqo5})$.
(32) $\tiny\begin{array}[]{ll}\deg=-4:&h_{1}\otimes dx_{2}\wedge
dx_{4}+x_{1}\otimes dx_{3}\wedge dx_{4}+y_{1}\otimes dx_{2}\wedge dx_{3}+\\\
&y_{2}\otimes dh_{2}\wedge dx_{4}+y_{3}\otimes dh_{2}\wedge
dx_{3}+y_{4}\otimes dh_{2}\wedge dx_{2}\\\ &\\\ \deg=-2:&h_{1}\otimes
dx_{4}\wedge dy_{2}+x_{2}\otimes dh_{1}\wedge dx_{4}+x_{2}\otimes dh_{2}\wedge
dx_{4}+x_{3}\otimes dx_{1}\wedge dx_{4}+y_{1}\otimes dh_{1}\wedge dx_{1}+\\\
&y_{1}\otimes dh_{2}\wedge dx_{1}+y_{1}\otimes dx_{3}\wedge
dy_{2}+y_{4}\otimes dh_{1}\wedge dy_{2}+y_{4}\otimes dh_{2}\wedge
dy_{2}+y_{4}\otimes dx_{1}\wedge dy_{3}\end{array}$
###### Lemma.
For ${\mathfrak{g}}={\mathfrak{wk}}(3;\alpha)$, where $\alpha\neq 0,-1$, we
take the Cartan matrix
(33) $\begin{pmatrix}{\bar{0}}&\alpha&0\\\ \alpha&{\bar{0}}&1\\\
0&1&{\bar{0}}\end{pmatrix}\qquad\tiny\begin{array}[]{l}\text{and the basis
}\\\ x_{1},\;\;x_{2},\;\;x_{3},\\\ x_{4}=[x_{1}\text{,$\,$}x_{2}],\;\
x_{5}=[x_{2}\text{,$\,$}x_{3}],\\\
x_{6}=[x_{3}\text{,$\,$}[x_{1}\text{,$\,$}x_{2}]],\\\
x_{7}=[[x_{1}\text{,$\,$}x_{2}]\text{,$\,$}[x_{2}\text{,$\,$}x_{3}]].\end{array}$
Then $H^{2}({\mathfrak{g}};{\mathfrak{g}})$ is spanned by the cocycles
$(\ref{eqwk3a})$.
(34) $\tiny\begin{array}[]{ll}\deg=-6:&\alpha(1+\alpha)y_{1}\otimes
dx_{3}\wedge dx_{7}+\alpha\,y_{1}\otimes dx_{5}\wedge
dx_{6}+\alpha(1+\alpha)y_{3}\otimes dx_{1}\wedge
dx_{7}+\alpha^{2}\,y_{3}\otimes dx_{4}\wedge dx_{6}+\\\ &\alpha\,y_{4}\otimes
dx_{3}\wedge dx_{6}+\alpha\,y_{5}\otimes dx_{1}\wedge dx_{6}+y_{6}\otimes
dx_{1}\wedge dx_{5}+\alpha\,y_{6}\otimes dx_{3}\wedge dx_{4}+y_{7}\otimes
dx_{1}\wedge dx_{3}\\\ &\\\ \deg=-4:&\alpha(1+\alpha)x_{1}\otimes dx_{3}\wedge
dx_{7}+\alpha\,x_{1}\otimes dx_{5}\wedge dx_{6}+\alpha\,y_{2}\otimes
dx_{3}\wedge dx_{5}+\alpha\,y_{3}\otimes dx_{2}\wedge
dx_{5}+\alpha(1+\alpha)y_{3}\otimes dx_{7}\wedge dy_{1}+\\\
&\alpha\,y_{5}\otimes dx_{2}\wedge dx_{3}+\alpha\,y_{5}\otimes dx_{6}\wedge
dy_{1}+y_{6}\otimes dx_{5}\wedge dy_{1}+y_{7}\otimes dx_{3}\wedge dy_{1}\\\
&\\\ \deg=-4:&\alpha(1+\alpha)x_{3}\otimes dx_{1}\wedge
dx_{7}+\alpha^{2}\,x_{3}\otimes dx_{4}\wedge dx_{6}+\alpha\,y_{1}\otimes
dx_{2}\wedge dx_{4}+\alpha(1+\alpha)y_{1}\otimes dx_{7}\wedge
dy_{3}+\alpha\,y_{2}\otimes dx_{1}\wedge dx_{4}+\\\ &y_{4}\otimes dx_{1}\wedge
dx_{2}+\alpha\,y_{4}\otimes dx_{6}\wedge dy_{3}+\alpha\,y_{6}\otimes
dx_{4}\wedge dy_{3}+y_{7}\otimes dx_{1}\wedge dy_{3}\\\ &\\\
\deg=-2:&\alpha\,x_{1}\otimes dx_{2}\wedge dx_{4}+\alpha(1+\alpha)x_{1}\otimes
dx_{7}\wedge dy_{3}+\alpha\,x_{3}\otimes dx_{2}\wedge
dx_{5}+\alpha(1+\alpha)x_{3}\otimes dx_{7}\wedge dy_{1}+\alpha\,y_{2}\otimes
dx_{4}\wedge dy_{1}+\\\ &\alpha\,y_{2}\otimes dx_{5}\wedge dy_{3}+y_{4}\otimes
dx_{2}\wedge dy_{1}+\alpha\,y_{5}\otimes dx_{2}\wedge dy_{3}+y_{7}\otimes
dy_{1}\wedge dy_{3}\\\ &\\\ \deg=0:&h_{1}\otimes\left(dx_{4}\wedge
dy_{4}\right)+h_{1}\otimes\left(dx_{6}\wedge
dy_{6}\right)+h_{1}\otimes\left(dx_{7}\wedge
dy_{7}\right)+\alpha\,h_{3}\otimes\left(dx_{7}\wedge
dy_{7}\right)+h_{4}\otimes\left(dx_{2}\wedge dy_{2}\right)\\\
&+\alpha\,h_{4}\otimes\left(dx_{4}\wedge
dy_{4}\right)+h_{4}\otimes\left(dx_{5}\wedge
dy_{5}\right)+\alpha\,h_{4}\otimes\left(dx_{6}\wedge
dy_{6}\right)+x_{1}\otimes\left(dx_{4}\wedge
dy_{2}\right)+x_{1}\otimes\left(dx_{6}\wedge dy_{5}\right)\\\
&+\alpha\,x_{2}\otimes\left(dx_{7}\wedge
dy_{6}\right)+x_{4}\otimes\left(dx_{7}\wedge
dy_{5}\right)+\alpha\,x_{5}\otimes\left(dx_{7}\wedge
dy_{4}\right)+x_{6}\otimes\left(dx_{7}\wedge
dy_{2}\right)+y_{1}\otimes\left(dx_{2}\wedge dy_{4}\right)\\\
&+y_{1}\otimes\left(dx_{5}\wedge
dy_{6}\right)+\alpha\,y_{2}\otimes\left(dx_{6}\wedge
dy_{7}\right)+y_{4}\otimes\left(dx_{5}\wedge
dy_{7}\right)+\alpha\,y_{5}\otimes\left(dx_{4}\wedge
dy_{7}\right)+y_{6}\otimes\left(dx_{2}\wedge dy_{7}\right)\end{array}$
###### Lemma.
For ${\mathfrak{g}}={\mathfrak{bgl}}(3;\alpha)$, where $\alpha\neq 0,-1$, (the
super analog of ${\mathfrak{wk}}(3;\alpha)$ and an analog of non-existing
${\mathfrak{osp}}(4|2;\alpha)$ for $p=2$), we take the Cartan matrix
(35) $\begin{pmatrix}0&\alpha&0\\\ \alpha&{\bar{0}}&1\\\
0&1&{\bar{0}}\end{pmatrix}\qquad\tiny\begin{array}[]{l}\text{and the basis:
even $\mid$ odd}\\\ x_{2},\;\;x_{3}\mid x_{1},\\\
x_{5}=[x_{2}\text{,$\,$}x_{3}]\mid x_{4}=[x_{1}\text{,$\,$}x_{2}],\\\ \mid
x_{6}=[x_{3}\text{,$\,$}[x_{1}\text{,$\,$}x_{2}]],\\\ \mid
x_{7}=[[x_{1}\text{,$\,$}x_{2}]\text{,$\,$}[x_{2}\text{,$\,$}x_{3}]]\end{array}$
Then $H^{2}({\mathfrak{g}};{\mathfrak{g}})$ is spanned by the cocycles
$(\ref{eqosp42a})$.
(36) $\tiny\begin{array}[]{ll}\deg=-8:&h_{1}\otimes dx_{7}^{\wedge
2}+\alpha\,h_{3}\otimes dx_{7}^{\wedge 2}\\\ &\\\ \deg=-6:&h_{1}\otimes
dx_{6}^{\wedge 2}+\alpha\,h_{3}\otimes dx_{6}^{\wedge 2}\\\ &\\\
\deg=-4:&\alpha(1+\alpha)x_{1}\otimes dx_{3}\wedge dx_{7}+\alpha\,x_{1}\otimes
dx_{5}\wedge dx_{6}+\alpha\,y_{2}\otimes dx_{3}\wedge
dx_{5}+\alpha\,y_{3}\otimes dx_{2}\wedge dx_{5}+\\\
&\alpha(1+\alpha)y_{3}\otimes dx_{7}\wedge dy_{1}+\alpha\,y_{5}\otimes
dx_{2}\wedge dx_{3}+\alpha\,y_{5}\otimes dx_{6}\wedge dy_{1}+y_{6}\otimes
dx_{5}\wedge dy_{1}+y_{7}\otimes dx_{3}\wedge dy_{1}\\\ &\\\
\deg=-4:&h_{1}\otimes dx_{4}^{\wedge 2}+\alpha\,h_{3}\otimes dx_{4}^{\wedge
2}\\\ &\\\ \deg=-4:&\alpha^{2}\,h_{2}\otimes dx_{4}{}^{\wedge
2}+\alpha^{2}\,h_{3}\otimes dx_{4}^{\wedge
2}+(\alpha^{3}+\alpha^{2})\,h_{4}\otimes dx_{4}^{\wedge
2}+(\alpha^{2}+\alpha)\,x_{3}\otimes\left(dx_{1}\wedge
dx_{7}\right)+\alpha^{2}\,x_{3}\otimes\left(dx_{4}\wedge dx_{6}\right)\\\
&+\alpha\,y_{1}\otimes\left(dx_{2}\wedge
dx_{4}\right)+(\alpha^{2}+\alpha)\,y_{1}\otimes\left(dx_{7}\wedge
dy_{3}\right)+\alpha\,y_{2}\otimes\left(dx_{1}\wedge
dx_{4}\right)+y_{4}\otimes\left(dx_{1}\wedge dx_{2}\right)\\\
&+\alpha\,y_{4}\otimes\left(dx_{6}\wedge
dy_{3}\right)+\alpha\,y_{6}\otimes\left(dx_{4}\wedge
dy_{3}\right)+y_{7}\otimes\left(dx_{1}\wedge dy_{3}\right)\\\ &\\\
\deg=-2:&h_{1}\otimes dx_{1}^{\wedge 2}+\alpha\,h_{3}\otimes dx_{1}^{\wedge
2}\\\ &\\\ \deg=-2:&\alpha\,x_{1}\otimes dx_{2}\wedge
dx_{4}+\alpha(1+\alpha)x_{1}\otimes dx_{7}\wedge dy_{3}+\alpha\,x_{3}\otimes
dx_{2}\wedge dx_{5}+\alpha(1+\alpha)x_{3}\otimes dx_{7}\wedge dy_{1}+\\\
&\alpha\,y_{2}\otimes dx_{4}\wedge dy_{1}+\alpha\,y_{2}\otimes dx_{5}\wedge
dy_{3}+y_{4}\otimes dx_{2}\wedge dy_{1}+\alpha\,y_{5}\otimes dx_{2}\wedge
dy_{3}+y_{7}\otimes dy_{1}\wedge dy_{3}\\\ &\\\
\deg=0:&h_{1}\otimes\left(dx_{4}\wedge
dy_{4}\right)+h_{1}\otimes\left(dx_{6}\wedge
dy_{6}\right)+h_{1}\otimes\left(dx_{7}\wedge
dy_{7}\right)+\alpha\,h_{3}\otimes\left(dx_{7}\wedge
dy_{7}\right)+h_{4}\otimes\left(dx_{2}\wedge dy_{2}\right)\\\
&+\alpha\,h_{4}\otimes\left(dx_{4}\wedge
dy_{4}\right)+h_{4}\otimes\left(dx_{5}\wedge
dy_{5}\right)+\alpha\,h_{4}\otimes\left(dx_{6}\wedge
dy_{6}\right)+x_{1}\otimes\left(dx_{4}\wedge
dy_{2}\right)+x_{1}\otimes\left(dx_{6}\wedge dy_{5}\right)\\\
&+\alpha\,x_{2}\otimes\left(dx_{7}\wedge
dy_{6}\right)+x_{4}\otimes\left(dx_{7}\wedge
dy_{5}\right)+\alpha\,x_{5}\otimes\left(dx_{7}\wedge
dy_{4}\right)+x_{6}\otimes\left(dx_{7}\wedge
dy_{2}\right)+y_{1}\otimes\left(dx_{2}\wedge dy_{4}\right)\\\
&+y_{1}\otimes\left(dx_{5}\wedge
dy_{6}\right)+\alpha\,y_{2}\otimes\left(dx_{6}\wedge
dy_{7}\right)+y_{4}\otimes\left(dx_{5}\wedge
dy_{7}\right)+\alpha\,y_{5}\otimes\left(dx_{4}\wedge
dy_{7}\right)+y_{6}\otimes\left(dx_{2}\wedge dy_{7}\right)\end{array}$
###### Lemma.
For ${\mathfrak{g}}={\mathfrak{wk}}(4;\alpha)$, where $\alpha\neq 0,-1$, we
take the Cartan matrix and the basis
(37) $\begin{pmatrix}{\bar{0}}&\alpha&1&0\\\ \alpha&{\bar{0}}&0&0\\\
1&0&{\bar{0}}&1\\\
0&0&1&{\bar{0}}\end{pmatrix}\qquad\tiny\begin{array}[]{l}x_{1},\;\;x_{2},\;\;x_{3},\;\;x_{4},\\\
{}[x_{1}\text{,$\,$}x_{2}],\;\;[x_{1}\text{,$\,$}x_{3}],\;\;[x_{3}\text{,$\,$}x_{4}],\\\
{}[x_{3}\text{,$\,$}[x_{1}\text{,$\,$}x_{2}]],\;\;{}[x_{4}\text{,$\,$}[x_{1}\text{,$\,$}x_{3}]],\\\
{}[[x_{1}\text{,$\,$}x_{2}]\text{,$\,$}[x_{1}\text{,$\,$}x_{3}]],\;\;{}[[x_{1}\text{,$\,$}x_{2}]\text{,$\,$}[x_{3}\text{,$\,$}x_{4}]],\\\
{}[[x_{1}\text{,$\,$}x_{2}]\text{,$\,$}[x_{4}\text{,$\,$}[x_{1}\text{,$\,$}x_{3}]]],\\\
{}[[x_{3}\text{,$\,$}[x_{1}\text{,$\,$}x_{2}]]\text{,$\,$}[x_{4}\text{,$\,$}[x_{1}\text{,$\,$}x_{3}]]],\\\
{}[[x_{4}\text{,$\,$}[x_{1}\text{,$\,$}x_{3}]]\text{,$\,$}[[x_{1}\text{,$\,$}x_{2}]\text{,$\,$}[x_{1}\text{,$\,$}x_{3}]]],\\\
{}[[[x_{1}\text{,$\,$}x_{2}]\text{,$\,$}[x_{1}\text{,$\,$}x_{3}]]\text{,$\,$}[[x_{1}\text{,$\,$}x_{2}]\text{,$\,$}[x_{3}\text{,$\,$}x_{4}]]]\end{array}$
Then $H^{2}({\mathfrak{g}};{\mathfrak{g}})$ is spanned by the cocycles which
for polygraphical reasons are divided into three groups: $(\ref{eqwk4a})$,
$(\ref{eqwk4ac1})$, $(\ref{eqwk4ac2})$, $(\ref{eqpsl4c3})$.
(38) $\tiny\begin{array}[]{ll}\deg=-12&\alpha^{3}(1+\alpha)y_{3}\otimes
dx_{9}\wedge dx_{15}+\alpha^{3}(1+\alpha)y_{3}\otimes dx_{11}\wedge
dx_{14}+\alpha^{2}(1+\alpha)y_{3}\otimes dx_{12}\wedge dx_{13}+\\\
&\alpha^{3}(1+\alpha)y_{6}\otimes dx_{7}\wedge
dx_{15}+\alpha^{2}(1+\alpha)y_{6}\otimes dx_{11}\wedge
dx_{13}+\alpha^{3}(1+\alpha)y_{7}\otimes dx_{6}\wedge dx_{15}+\\\
&\alpha^{3}(1+\alpha)y_{7}\otimes dx_{8}\wedge
dx_{14}+\alpha^{2}(1+\alpha)y_{7}\otimes dx_{10}\wedge
dx_{13}+\alpha^{2}(1+\alpha)y_{8}\otimes dx_{7}\wedge dx_{14}+\\\
&\alpha(1+\alpha)y_{8}\otimes dx_{9}\wedge
dx_{13}+\alpha^{3}(1+\alpha)y_{9}\otimes dx_{3}\wedge
dx_{15}+\alpha^{2}(1+\alpha)y_{9}\otimes dx_{8}\wedge dx_{13}+\\\
&\alpha(1+\alpha)y_{10}\otimes dx_{7}\wedge
dx_{13}+\alpha^{2}(1+\alpha)y_{11}\otimes dx_{3}\wedge
dx_{14}+\alpha(1+\alpha)y_{11}\otimes dx_{6}\wedge dx_{13}+\\\
&\alpha(1+\alpha)y_{12}\otimes dx_{3}\wedge
dx_{13}+\alpha(1+\alpha)y_{13}\otimes dx_{3}\wedge
dx_{12}+\alpha\,y_{13}\otimes dx_{6}\wedge dx_{11}+\\\
&\alpha(1+\alpha)y_{13}\otimes dx_{7}\wedge dx_{10}+\alpha\,y_{13}\otimes
dx_{8}\wedge dx_{9}+\alpha\,y_{14}\otimes dx_{3}\wedge dx_{11}+\\\
&\alpha\,y_{14}\otimes dx_{7}\wedge dx_{8}+y_{15}\otimes dx_{3}\wedge
dx_{9}+y_{15}\otimes dx_{6}\wedge dx_{7}\\\ &\\\
\deg=-10&\alpha^{3}(1+\alpha)x_{3}\otimes dx_{9}\wedge
dx_{15}+\alpha^{3}(1+\alpha)x_{3}\otimes dx_{11}\wedge
dx_{14}+\alpha^{2}(1+\alpha)x_{3}\otimes dx_{12}\wedge dx_{13}+\\\
&\alpha^{3}(1+\alpha)y_{1}\otimes dx_{4}\wedge
dx_{15}+\alpha^{2}(1+\alpha)y_{1}\otimes dx_{11}\wedge
dx_{12}+\alpha^{3}(1+\alpha)y_{4}\otimes dx_{1}\wedge dx_{15}+\\\
&\alpha^{3}(1+\alpha)y_{4}\otimes dx_{5}\wedge
dx_{14}+\alpha^{2}(1+\alpha)y_{4}\otimes dx_{10}\wedge
dx_{12}+\alpha^{2}(1+\alpha)y_{5}\otimes dx_{4}\wedge dx_{14}+\\\
&\alpha(1+\alpha)y_{5}\otimes dx_{9}\wedge
dx_{12}+\alpha^{2}(1+\alpha)y_{9}\otimes dx_{5}\wedge
dx_{12}+\alpha^{3}(1+\alpha)y_{9}\otimes dx_{15}\wedge dy_{3}+\\\
&\alpha(1+\alpha)y_{10}\otimes dx_{4}\wedge
dx_{12}+\alpha(1+\alpha)y_{11}\otimes dx_{1}\wedge dx_{12}+\\\
&\alpha^{2}(1+\alpha)y_{11}\otimes dx_{14}\wedge dy_{3}+\alpha\,y_{12}\otimes
dx_{1}\wedge dx_{11}+\alpha(1+\alpha)y_{12}\otimes dx_{4}\wedge dx_{10}+\\\
&\alpha\,y_{12}\otimes dx_{5}\wedge dx_{9}+\alpha(1+\alpha)y_{12}\otimes
dx_{13}\wedge dy_{3}+\alpha(1+\alpha)y_{13}\otimes dx_{12}\wedge dy_{3}+\\\
&\alpha\,y_{14}\otimes dx_{4}\wedge dx_{5}+\alpha\,y_{14}\otimes dx_{11}\wedge
dy_{3}+y_{15}\otimes dx_{1}\wedge dx_{4}+y_{15}\otimes dx_{9}\wedge dy_{3}\\\
&\\\ \deg=-8&\alpha^{3}(1+\alpha)x_{4}\otimes dx_{1}\wedge
dx_{15}+\alpha^{3}(1+\alpha)x_{4}\otimes dx_{5}\wedge
dx_{14}+\alpha^{2}(1+\alpha)x_{4}\otimes dx_{10}\wedge dx_{12}+\\\
&\alpha^{3}(1+\alpha)x_{7}\otimes dx_{6}\wedge
dx_{15}+\alpha^{3}(1+\alpha)x_{7}\otimes dx_{8}\wedge
dx_{14}+\alpha^{3}(1+\alpha)x_{7}\otimes dx_{10}\wedge dx_{13}+\\\
&\alpha^{2}(1+\alpha)y_{1}\otimes dx_{8}\wedge
dx_{10}+\alpha^{3}(1+\alpha)y_{1}\otimes dx_{15}\wedge
dy_{4}+\alpha(1+\alpha)y_{5}\otimes dx_{6}\wedge dx_{10}+\\\
&\alpha^{2}(1+\alpha)y_{5}\otimes dx_{14}\wedge
dy_{4}+\alpha^{2}(1+\alpha)y_{6}\otimes dx_{5}\wedge
dx_{10}+\alpha^{3}(1+\alpha)y_{6}\otimes dx_{15}\wedge dy_{7}+\\\
&\alpha(1+\alpha)y_{8}\otimes dx_{1}\wedge
dx_{10}+\alpha^{2}(1+\alpha)y_{8}\otimes dx_{14}\wedge
dy_{7}+\alpha\,y_{10}\otimes dx_{1}\wedge dx_{8}+\\\ &\alpha\,y_{10}\otimes
dx_{5}\wedge dx_{6}+\alpha(1+\alpha)y_{10}\otimes dx_{12}\wedge
dy_{4}+\alpha(1+\alpha)\,y_{10}\otimes dx_{13}\wedge dy_{7}+\\\
&\alpha(1+\alpha)y_{12}\otimes dx_{10}\wedge
dy_{4}+\alpha(1+\alpha)y_{13}\otimes dx_{10}\wedge
dy_{7}+\alpha\,y_{14}\otimes dx_{5}\wedge dy_{4}+\\\ &\alpha\,y_{14}\otimes
dx_{8}\wedge dy_{7}+y_{15}\otimes dx_{1}\wedge dy_{4}+y_{15}\otimes
dx_{6}\wedge dy_{7}\\\ &\\\ \end{array}$ (39)
$\tiny\begin{array}[]{ll}\deg=-8&\alpha^{3}(1+\alpha)x_{1}\otimes dx_{4}\wedge
dx_{15}+\alpha^{2}(1+\alpha)x_{1}\otimes dx_{11}\wedge
dx_{12}+\alpha^{3}(1+\alpha)x_{6}\otimes dx_{7}\wedge dx_{15}+\\\
&\alpha^{2}(1+\alpha)x_{6}\otimes dx_{11}\wedge
dx_{13}+\alpha^{2}(1+\alpha)y_{2}\otimes dx_{4}\wedge
dx_{13}+\alpha^{2}(1+\alpha)y_{2}\otimes dx_{7}\wedge dx_{12}+\\\
&\alpha^{2}(1+\alpha)y_{4}\otimes dx_{2}\wedge
dx_{13}+\alpha^{2}(1+\alpha)y_{4}\otimes dx_{8}\wedge
dx_{11}+\alpha^{3}(1+\alpha)y_{4}\otimes dx_{15}\wedge dy_{1}+\\\
&\alpha(1+\alpha)y_{5}\otimes dx_{7}\wedge
dx_{11}+\alpha^{2}(1+\alpha)y_{7}\otimes dx_{2}\wedge
dx_{12}+\alpha^{2}(1+\alpha)y_{7}\otimes dx_{5}\wedge dx_{11}+\\\
&\alpha^{3}(1+\alpha)y_{7}\otimes dx_{15}\wedge
dy_{6}+\alpha(1+\alpha)y_{8}\otimes dx_{4}\wedge
dx_{11}+\alpha(1+\alpha)y_{11}\otimes dx_{4}\wedge dx_{8}+\\\
&\alpha(1+\alpha)y_{11}\otimes dx_{5}\wedge
dx_{7}+\alpha(1+\alpha)y_{11}\otimes dx_{12}\wedge
dy_{1}+\alpha(1+\alpha)\,y_{11}\otimes dx_{13}\wedge dy_{6}+\\\
&\alpha\,y_{12}\otimes dx_{2}\wedge dx_{7}+\alpha\,y_{12}\otimes dx_{11}\wedge
dy_{1}+\alpha\,y_{13}\otimes dx_{2}\wedge dx_{4}+\\\ &\alpha\,y_{13}\otimes
dx_{11}\wedge dy_{6}+y_{15}\otimes dx_{4}\wedge dy_{1}+y_{15}\otimes
dx_{7}\wedge dy_{6}\\\ &\\\ \deg=-6&\alpha(1+\alpha)x_{2}\otimes dx_{4}\wedge
dx_{13}+\alpha(1+\alpha)x_{2}\otimes dx_{7}\wedge
dx_{12}+\alpha(1+\alpha)x_{5}\otimes dx_{4}\wedge dx_{14}+\\\
&(\alpha+1)\,x_{5}\otimes dx_{9}\wedge dx_{12}+\alpha(1+\alpha)x_{8}\otimes
dx_{7}\wedge dx_{14}+(\alpha+1)\,x_{8}\otimes dx_{9}\wedge dx_{13}+\\\
&(\alpha+1)\,y_{1}\otimes dx_{7}\wedge dx_{9}+(\alpha+1)\,y_{4}\otimes
dx_{6}\wedge dx_{9}+\alpha(1+\alpha)y_{4}\otimes dx_{13}\wedge dy_{2}+\\\
&\alpha^{2}(1+\alpha)y_{4}\otimes dx_{14}\wedge
dy_{5}+(\alpha+1)\,y_{6}\otimes dx_{4}\wedge dx_{9}+(\alpha+1)\,y_{7}\otimes
dx_{1}\wedge dx_{9}+\\\ &\alpha(1+\alpha)y_{7}\otimes dx_{12}\wedge
dy_{2}+\alpha^{2}(1+\alpha)y_{7}\otimes dx_{14}\wedge
dy_{8}+(\alpha+1)\,y_{9}\otimes dx_{1}\wedge dx_{7}+\\\
&(\alpha+1)\,y_{9}\otimes dx_{4}\wedge dx_{6}+\alpha(1+\alpha)y_{9}\otimes
dx_{12}\wedge dy_{5}+\alpha(1+\alpha)y_{9}\otimes dx_{13}\wedge dy_{8}+\\\
&y_{12}\otimes dx_{7}\wedge dy_{2}+y_{12}\otimes dx_{9}\wedge
dy_{5}+y_{13}\otimes dx_{4}\wedge dy_{2}+y_{13}\otimes dx_{9}\wedge dy_{8}+\\\
&y_{14}\otimes dx_{4}\wedge dy_{5}+y_{14}\otimes dx_{7}\wedge dy_{8}\\\ &\\\
\deg=-6&\alpha^{2}(1+\alpha)x_{1}\otimes dx_{8}\wedge
dx_{10}+\alpha^{3}(1+\alpha)x_{1}\otimes dx_{15}\wedge
dy_{4}+\alpha^{2}(1+\alpha)x_{4}\otimes dx_{2}\wedge dx_{13}+\\\
&\alpha^{2}(1+\alpha)x_{4}\otimes dx_{8}\wedge
dx_{11}+\alpha^{3}(1+\alpha)x_{4}\otimes dx_{15}\wedge
dy_{1}+\alpha^{3}(1+\alpha)x_{9}\otimes dx_{3}\wedge dx_{15}+\\\
&\alpha^{2}(1+\alpha)x_{9}\otimes dx_{8}\wedge
dx_{13}+\alpha^{2}(1+\alpha)y_{2}\otimes dx_{3}\wedge
dx_{10}+\alpha^{2}(1+\alpha)y_{2}\otimes dx_{13}\wedge dy_{4}+\\\
&\alpha^{2}(1+\alpha)y_{3}\otimes dx_{2}\wedge
dx_{10}+\alpha^{2}(1+\alpha)y_{3}\otimes dx_{5}\wedge
dx_{8}+\alpha^{3}(1+\alpha)y_{3}\otimes dx_{15}\wedge dy_{9}+\\\
&\alpha(1+\alpha)y_{5}\otimes dx_{3}\wedge dx_{8}+\alpha(1+\alpha)y_{8}\otimes
dx_{3}\wedge dx_{5}+\alpha(1+\alpha)y_{8}\otimes dx_{10}\wedge dy_{1}+\\\
&\alpha(1+\alpha)y_{8}\otimes dx_{11}\wedge
dy_{4}+\alpha(1+\alpha)y_{8}\otimes dx_{13}\wedge dy_{9}+\alpha\,y_{10}\otimes
dx_{2}\wedge dx_{3}+\\\ &\alpha\,y_{10}\otimes dx_{8}\wedge
dy_{1}+\alpha(1+\alpha)y_{11}\otimes dx_{8}\wedge dy_{4}+\alpha\,y_{13}\otimes
dx_{2}\wedge dy_{4}+\\\ &\alpha\,y_{13}\otimes dx_{8}\wedge
dy_{9}+y_{15}\otimes dx_{3}\wedge dy_{9}+y_{15}\otimes dy_{1}\wedge dy_{4}\\\
\end{array}$
(40) $\tiny\begin{array}[]{ll}\deg=0&\alpha\,h_{1}\otimes dx_{14}\wedge
dy_{14}+\alpha^{2}\,h_{1}\otimes dx_{15}\wedge dy_{15}+h_{2}\otimes
dx_{5}\wedge dy_{5}+h_{2}\otimes dx_{8}\wedge dy_{8}+\\\ &h_{2}\otimes
dx_{10}\wedge dy_{10}+h_{2}\otimes dx_{11}\wedge dy_{11}+h_{2}\otimes
dx_{12}\wedge dy_{12}+h_{2}\otimes dx_{13}\wedge dy_{13}+\\\
&\alpha^{2}\,h_{2}\otimes dx_{14}\wedge dy_{14}+h_{3}\otimes dx_{1}\wedge
dy_{1}+\alpha\,h_{3}\otimes dx_{5}\wedge dy_{5}+h_{3}\otimes dx_{6}\wedge
dy_{6}+\\\ &\alpha\,h_{3}\otimes dx_{8}\wedge dy_{8}+h_{3}\otimes dx_{9}\wedge
dy_{9}+\alpha\,h_{3}\otimes dx_{10}\wedge dy_{10}+\alpha\,h_{3}\otimes
dx_{11}\wedge dy_{11}+\\\ &\alpha\,h_{3}\otimes dx_{12}\wedge
dy_{12}+\alpha^{2}(1+\alpha)h_{3}\otimes dx_{14}\wedge
dy_{14}+\alpha^{3}(1+\alpha)h_{3}\otimes dx_{15}\wedge dy_{15}+\\\
&\alpha\,h_{4}\otimes dx_{12}\wedge dy_{12}+\alpha\,h_{4}\otimes dx_{13}\wedge
dy_{13}+\alpha\,h_{4}\otimes dx_{14}\wedge dy_{14}+\alpha^{2}\,h_{4}\otimes
dx_{15}\wedge dy_{15}+\\\ &x_{1}\otimes dh_{1}\wedge
dx_{1}+\alpha\,x_{1}\otimes dx_{10}\wedge dy_{8}+\alpha\,x_{1}\otimes
dx_{12}\wedge dy_{11}+\alpha\,x_{1}\otimes dx_{14}\wedge dy_{13}+\\\
&x_{2}\otimes dh_{1}\wedge dx_{2}+x_{2}\otimes dx_{5}\wedge
dy_{1}+x_{2}\otimes dx_{8}\wedge dy_{6}+x_{2}\otimes dx_{11}\wedge
dy_{9}+\alpha^{3}\,x_{2}\otimes dx_{15}\wedge dy_{14}+\\\
&\alpha\,x_{3}\otimes dx_{13}\wedge dy_{12}+x_{4}\otimes dh_{1}\wedge
dx_{4}+\alpha\,x_{4}\otimes dx_{12}\wedge dy_{10}+x_{5}\otimes dx_{10}\wedge
dy_{6}+\\\ &x_{5}\otimes dx_{12}\wedge dy_{9}+\alpha\,x_{5}\otimes
dx_{15}\wedge dy_{13}+x_{6}\otimes dh_{1}\wedge dx_{6}+\alpha\,x_{6}\otimes
dx_{10}\wedge dy_{5}+\\\ &\alpha\,x_{6}\otimes dx_{13}\wedge
dy_{11}+\alpha\,x_{6}\otimes dx_{14}\wedge dy_{12}+x_{7}\otimes dh_{1}\wedge
dx_{7}+\alpha\,x_{7}\otimes dx_{13}\wedge dy_{10}+\\\ &x_{8}\otimes
dx_{10}\wedge dy_{1}+x_{8}\otimes dx_{13}\wedge dy_{9}+\alpha\,x_{8}\otimes
dx_{15}\wedge dy_{12}+\\\ &\alpha\,x_{9}\otimes dx_{12}\wedge
dy_{5}+\alpha\,x_{9}\otimes dx_{13}\wedge dy_{8}+\alpha\,x_{9}\otimes
dx_{14}\wedge dy_{10}+\\\ &x_{10}\otimes dh_{1}\wedge dx_{10}+x_{10}\otimes
dx_{14}\wedge dy_{9}+\alpha\,x_{10}\otimes dx_{15}\wedge dy_{11}+\\\
&x_{11}\otimes dh_{1}\wedge dx_{11}+x_{11}\otimes dx_{12}\wedge
dy_{1}+x_{11}\otimes dx_{13}\wedge dy_{6}+\alpha\,x_{11}\otimes dx_{15}\wedge
dy_{10}+\\\ &x_{12}\otimes dx_{14}\wedge dy_{6}+\alpha\,x_{12}\otimes
dx_{15}\wedge dy_{8}+x_{13}\otimes dx_{14}\wedge dy_{1}+\alpha\,x_{13}\otimes
dx_{15}\wedge dy_{5}+\\\ &x_{14}\otimes dh_{1}\wedge dx_{14}+y_{1}\otimes
dh_{1}\wedge dy_{1}+\alpha\,y_{1}\otimes dx_{8}\wedge
dy_{10}+\alpha\,y_{1}\otimes dx_{11}\wedge dy_{12}+\\\ &\alpha\,y_{1}\otimes
dx_{13}\wedge dy_{14}+y_{2}\otimes dh_{1}\wedge dy_{2}+y_{2}\otimes
dx_{1}\wedge dy_{5}+y_{2}\otimes dx_{6}\wedge dy_{8}+\\\ &y_{2}\otimes
dx_{9}\wedge dy_{11}+\alpha^{3}\,y_{2}\otimes dx_{14}\wedge
dy_{15}+\alpha\,y_{3}\otimes dx_{12}\wedge dy_{13}+y_{4}\otimes dh_{1}\wedge
dy_{4}\\\ &\alpha\,y_{4}\otimes dx_{10}\wedge dy_{12}+y_{5}\otimes
dx_{6}\wedge dy_{10}+y_{5}\otimes dx_{9}\wedge dy_{12}+\alpha\,y_{5}\otimes
dx_{13}\wedge dy_{15}+\\\ &y_{6}\otimes dh_{1}\wedge
dy_{6}+\alpha\,y_{6}\otimes dx_{5}\wedge dy_{10}+\alpha\,y_{6}\otimes
dx_{11}\wedge dy_{13}+\alpha\,y_{6}\otimes dx_{12}\wedge dy_{14}+\\\
&y_{7}\otimes dh_{1}\wedge dy_{7}+\alpha\,y_{7}\otimes dx_{10}\wedge
dy_{13}+y_{8}\otimes dx_{1}\wedge dy_{10}+y_{8}\otimes dx_{9}\wedge
dy_{13}+\\\ &\alpha\,y_{8}\otimes dx_{12}\wedge dy_{15}+\alpha\,y_{9}\otimes
dx_{5}\wedge dy_{12}+\alpha\,y_{9}\otimes dx_{8}\wedge
dy_{13}+\alpha\,y_{9}\otimes dx_{10}\wedge dy_{14}+\\\ &y_{10}\otimes
dh_{1}\wedge dy_{10}+y_{10}\otimes dx_{9}\wedge dy_{14}+\alpha\,y_{10}\otimes
dx_{11}\wedge dy_{15}+y_{11}\otimes dh_{1}\wedge dy_{11}+\\\ &y_{11}\otimes
dx_{1}\wedge dy_{12}+y_{11}\otimes dx_{6}\wedge dy_{13}+\alpha\,y_{11}\otimes
dx_{10}\wedge dy_{15}+\\\ &y_{12}\otimes dx_{6}\wedge
dy_{14}+\alpha\,y_{12}\otimes dx_{8}\wedge dy_{15}+y_{13}\otimes dx_{1}\wedge
dy_{14}+\\\ &\alpha\,y_{13}\otimes dx_{5}\wedge dy_{15}+y_{14}\otimes
dh_{1}\wedge dy_{14}\end{array}$ (41)
$\tiny\begin{array}[]{ll}\deg=-4&\alpha(1+\alpha)x_{2}\otimes dx_{3}\wedge
dx_{10}+\alpha(1+\alpha)x_{2}\otimes dx_{13}\wedge
dy_{4}+(\alpha+1)\,x_{4}\otimes dx_{6}\wedge dx_{9}+\\\
&\alpha(1+\alpha)x_{4}\otimes dx_{13}\wedge
dy_{2}+\alpha^{2}(1+\alpha)x_{4}\otimes dx_{14}\wedge
dy_{5}+(\alpha+1)\,x_{5}\otimes dx_{6}\wedge dx_{10}+\\\
&\alpha(1+\alpha)x_{5}\otimes dx_{14}\wedge
dy_{4}+\alpha(1+\alpha)x_{11}\otimes dx_{3}\wedge
dx_{14}+(\alpha+1)\,x_{11}\otimes dx_{6}\wedge dx_{13}+\\\
&(\alpha+1)\,y_{1}\otimes dx_{3}\wedge dx_{6}+(\alpha+1)\,y_{3}\otimes
dx_{1}\wedge dx_{6}+\alpha(1+\alpha)y_{3}\otimes dx_{10}\wedge dy_{2}+\\\
&\alpha^{2}(1+\alpha)y_{3}\otimes dx_{14}\wedge
dy_{11}+(\alpha+1)\,y_{6}\otimes dx_{1}\wedge dx_{3}+(\alpha+1)\,y_{6}\otimes
dx_{9}\wedge dy_{4}+\\\ &\alpha(1+\alpha)y_{6}\otimes dx_{10}\wedge
dy_{5}+\alpha(1+\alpha)y_{6}\otimes dx_{13}\wedge
dy_{11}+(\alpha+1)\,y_{9}\otimes dx_{6}\wedge dy_{4}+\\\ &y_{10}\otimes
dx_{3}\wedge dy_{2}+y_{10}\otimes dx_{6}\wedge dy_{5}+y_{13}\otimes
dx_{6}\wedge dy_{11}+\\\ &y_{13}\otimes dy_{2}\wedge dy_{4}+y_{14}\otimes
dx_{3}\wedge dy_{11}+y_{14}\otimes dy_{4}\wedge dy_{5}\\\ &\\\
\deg=-4&\alpha^{2}(1+\alpha)x_{3}\otimes dx_{2}\wedge
dx_{10}+\alpha^{2}(1+\alpha)x_{3}\otimes dx_{5}\wedge
dx_{8}+\alpha^{3}(1+\alpha)x_{3}\otimes dx_{15}\wedge dy_{9}+\\\
&\alpha^{2}(1+\alpha)x_{6}\otimes dx_{5}\wedge
dx_{10}+\alpha^{3}(1+\alpha)x_{6}\otimes dx_{15}\wedge
dy_{7}+\alpha^{2}(1+\alpha)x_{7}\otimes dx_{2}\wedge dx_{12}+\\\
&\alpha^{2}(1+\alpha)x_{7}\otimes dx_{5}\wedge
dx_{11}+\alpha^{3}(1+\alpha)x_{7}\otimes dx_{15}\wedge
dy_{6}+\alpha^{2}(1+\alpha)x_{9}\otimes dx_{5}\wedge dx_{12}+\\\
&\alpha^{3}(1+\alpha)x_{9}\otimes dx_{15}\wedge
dy_{3}+\alpha^{2}(1+\alpha)y_{2}\otimes dx_{10}\wedge
dy_{3}+\alpha^{2}(1+\alpha)y_{2}\otimes dx_{12}\wedge dy_{7}+\\\
&\alpha(1+\alpha)y_{5}\otimes dx_{8}\wedge dy_{3}+\alpha(1+\alpha)y_{5}\otimes
dx_{10}\wedge dy_{6}+\alpha(1+\alpha)y_{5}\otimes dx_{11}\wedge dy_{7}+\\\
&\alpha(1+\alpha)y_{5}\otimes dx_{12}\wedge
dy_{9}+\alpha(1+\alpha)y_{8}\otimes dx_{5}\wedge dy_{3}+\alpha\,y_{10}\otimes
dx_{2}\wedge dy_{3}+\\\ &\alpha\,y_{10}\otimes dx_{5}\wedge
dy_{6}+\alpha^{2}+\alpha\,y_{11}\otimes dx_{5}\wedge
dy_{7}+\alpha\,y_{12}\otimes dx_{2}\wedge dy_{7}+\\\ &\alpha\,y_{12}\otimes
dx_{5}\wedge dy_{9}+y_{15}\otimes dy_{3}\wedge dy_{9}+y_{15}\otimes
dy_{6}\wedge dy_{7}\\\ &\\\ \deg=-2&\alpha(1+\alpha)x_{2}\otimes dx_{10}\wedge
dy_{3}+\alpha(1+\alpha)x_{2}\otimes dx_{12}\wedge
dy_{7}+(\alpha+1)\,x_{3}\otimes dx_{1}\wedge dx_{6}+\\\
&\alpha^{2}+\alpha\,x_{3}\otimes dx_{10}\wedge
dy_{2}+\alpha^{3}+\alpha^{2}\,x_{3}\otimes dx_{14}\wedge
dy_{11}+(\alpha+1)\,x_{7}\otimes dx_{1}\wedge dx_{9}+\\\
&\alpha(1+\alpha)x_{7}\otimes dx_{12}\wedge
dy_{2}+\alpha^{2}(1+\alpha)x_{7}\otimes dx_{14}\wedge
dy_{8}+(\alpha+1)\,x_{8}\otimes dx_{1}\wedge dx_{10}+\\\
&\alpha(1+\alpha)x_{8}\otimes dx_{14}\wedge dy_{7}+(\alpha+1)\,x_{11}\otimes
dx_{1}\wedge dx_{12}+\alpha(1+\alpha)x_{11}\otimes dx_{14}\wedge dy_{3}+\\\
&(\alpha+1)\,y_{1}\otimes dx_{6}\wedge dy_{3}+(\alpha+1)\,y_{1}\otimes
dx_{9}\wedge dy_{7}+\alpha^{2}+\alpha\,y_{1}\otimes dx_{10}\wedge dy_{8}+\\\
&\alpha(1+\alpha)y_{1}\otimes dx_{12}\wedge dy_{11}+(\alpha+1)\,y_{6}\otimes
dx_{1}\wedge dy_{3}+(\alpha+1)\,y_{9}\otimes dx_{1}\wedge dy_{7}+\\\
&y_{10}\otimes dx_{1}\wedge dy_{8}+y_{10}\otimes dy_{2}\wedge
dy_{3}+y_{12}\otimes dx_{1}\wedge dy_{11}+y_{12}\otimes dy_{2}\wedge
dy_{7}+\\\ &y_{14}\otimes dy_{3}\wedge dy_{11}+y_{14}\otimes dy_{7}\wedge
dy_{8}\end{array}$
###### Lemma.
For ${\mathfrak{g}}={\mathfrak{bgl}}(4;\alpha)$, where $\alpha\neq 0,-1$, we
take the Cartan matrix and the basis
(42) $\begin{pmatrix}0&\alpha&1&0\\\ \alpha&{\bar{0}}&0&0\\\
1&0&{\bar{0}}&1\\\
0&0&1&{\bar{0}}\end{pmatrix}\qquad\tiny\begin{array}[]{l}x_{1},\;\;x_{2},\;\;x_{3},\;\;x_{4},\\\
{}[x_{1}\text{,$\,$}x_{2}],\;\;[x_{1}\text{,$\,$}x_{3}],\;\;[x_{3}\text{,$\,$}x_{4}],\\\
{}[x_{3}\text{,$\,$}[x_{1}\text{,$\,$}x_{2}]],\;\;{}[x_{4}\text{,$\,$}[x_{1}\text{,$\,$}x_{3}]],\\\
{}[[x_{1}\text{,$\,$}x_{2}]\text{,$\,$}[x_{1}\text{,$\,$}x_{3}]],\;\;{}[[x_{1}\text{,$\,$}x_{2}]\text{,$\,$}[x_{3}\text{,$\,$}x_{4}]],\\\
{}[[x_{1}\text{,$\,$}x_{2}]\text{,$\,$}[x_{4}\text{,$\,$}[x_{1}\text{,$\,$}x_{3}]]],\\\
{}[[x_{3}\text{,$\,$}[x_{1}\text{,$\,$}x_{2}]]\text{,$\,$}[x_{4}\text{,$\,$}[x_{1}\text{,$\,$}x_{3}]]],\\\
{}[[x_{4}\text{,$\,$}[x_{1}\text{,$\,$}x_{3}]]\text{,$\,$}[[x_{1}\text{,$\,$}x_{2}]\text{,$\,$}[x_{1}\text{,$\,$}x_{3}]]],\\\
{}[[[x_{1}\text{,$\,$}x_{2}]\text{,$\,$}[x_{1}\text{,$\,$}x_{3}]]\text{,$\,$}[[x_{1}\text{,$\,$}x_{2}]\text{,$\,$}[x_{3}\text{,$\,$}x_{4}]]]\end{array}$
Then $H^{2}({\mathfrak{g}};{\mathfrak{g}})$ is spanned by the cocycles:
(43)
$\tiny\begin{array}[]{ll}\deg=-12&\left(\alpha^{4}+\alpha^{3}\right)\,y_{3}\otimes\left(dx_{9}\wedge
dx_{15}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,y_{3}\otimes\left(dx_{11}\wedge
dx_{14}\right)\text{+
}\left(\alpha^{4}+\alpha^{2}\right)\,y_{3}\otimes\left(dx_{12}\wedge
dx_{13}\right)\\\ &\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,y_{6}\otimes\left(dx_{7}\wedge
dx_{15}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{6}\otimes\left(dx_{11}\wedge
dx_{13}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,y_{7}\otimes\left(dx_{6}\wedge
dx_{15}\right)\\\ &\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,y_{7}\otimes\left(dx_{8}\wedge
dx_{14}\right)\text{+
}\left(\alpha^{4}+\alpha^{2}\right)\,y_{7}\otimes\left(dx_{10}\wedge
dx_{13}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{8}\otimes\left(dx_{7}\wedge
dx_{14}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{8}\otimes\left(dx_{9}\wedge
dx_{13}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,y_{9}\otimes\left(dx_{3}\wedge
dx_{15}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{9}\otimes\left(dx_{8}\wedge
dx_{13}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{10}\otimes\left(dx_{7}\wedge
dx_{13}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{11}\otimes\left(dx_{3}\wedge
dx_{14}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{11}\otimes\left(dx_{6}\wedge
dx_{13}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{12}\otimes\left(dx_{3}\wedge
dx_{13}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{13}\otimes\left(dx_{3}\wedge
dx_{12}\right)\text{+ }\alpha\,y_{13}\otimes\left(dx_{6}\wedge
dx_{11}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{13}\otimes\left(dx_{7}\wedge
dx_{10}\right)\text{+ }\alpha\,y_{13}\otimes\left(dx_{8}\wedge
dx_{9}\right)\text{+ }\alpha\,y_{14}\otimes\left(dx_{3}\wedge
dx_{11}\right)\text{+ }\alpha\,y_{14}\otimes\left(dx_{7}\wedge
dx_{8}\right)\\\ &\text{+ }y_{15}\otimes\left(dx_{3}\wedge
dx_{9}\right)\text{+ }y_{15}\otimes\left(dx_{6}\wedge dx_{7}\right)\\\ &\\\
\deg=-10&\left(\alpha^{4}+\alpha^{3}\right)\,x_{3}\otimes\left(dx_{9}\wedge
dx_{15}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,x_{3}\otimes\left(dx_{11}\wedge
dx_{14}\right)\text{+
}\left(\alpha^{4}+\alpha^{2}\right)\,x_{3}\otimes\left(dx_{12}\wedge
dx_{13}\right)\\\ &\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,y_{1}\otimes\left(dx_{4}\wedge
dx_{15}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{1}\otimes\left(dx_{11}\wedge
dx_{12}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,y_{4}\otimes\left(dx_{1}\wedge
dx_{15}\right)\\\ &\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,y_{4}\otimes\left(dx_{5}\wedge
dx_{14}\right)\text{+
}\left(\alpha^{4}+\alpha^{2}\right)\,y_{4}\otimes\left(dx_{10}\wedge
dx_{12}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{5}\otimes\left(dx_{4}\wedge
dx_{14}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{5}\otimes\left(dx_{9}\wedge
dx_{12}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{9}\otimes\left(dx_{5}\wedge
dx_{12}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,y_{9}\otimes\left(dx_{15}\wedge
dy_{3}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{10}\otimes\left(dx_{4}\wedge
dx_{12}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{11}\otimes\left(dx_{1}\wedge
dx_{12}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{11}\otimes\left(dx_{14}\wedge
dy_{3}\right)\\\ &\text{+ }\alpha\,y_{12}\otimes\left(dx_{1}\wedge
dx_{11}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{12}\otimes\left(dx_{4}\wedge
dx_{10}\right)\text{+ }\alpha\,y_{12}\otimes\left(dx_{5}\wedge
dx_{9}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{12}\otimes\left(dx_{13}\wedge
dy_{3}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{13}\otimes\left(dx_{12}\wedge
dy_{3}\right)\text{+ }\alpha\,y_{14}\otimes\left(d_{4}\wedge
dx_{5}\right)\text{+ }\alpha\,y_{14}\otimes\left(dx_{11}\wedge
dy_{3}\right)\\\ &\text{+ }y_{15}\otimes\left(dx_{1}\wedge
dx_{4}\right)\text{+ }y_{15}\otimes\left(d_{9}\wedge dy_{3}\right)\\\ &\\\
\deg=-8&\left(\alpha^{4}+\alpha^{3}\right)\,x_{4}\otimes\left(dx_{1}\wedge
dx_{15}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,x_{4}\otimes\left(dx_{5}\wedge
dx_{14}\right)\text{+
}\left(\alpha^{4}+\alpha^{2}\right)\,x_{4}\otimes\left(dx_{10}\wedge
dx_{12}\right)\\\ &\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,x_{7}\otimes\left(dx_{6}\wedge
dx_{15}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,x_{7}\otimes\left(dx_{8}\wedge
dx_{14}\right)\text{+
}\left(\alpha^{4}+\alpha^{2}\right)\,x_{7}\otimes\left(dx_{10}\wedge
dx_{13}\right)\\\ &\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{1}\otimes\left(dx_{8}\wedge
dx_{10}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,y_{1}\otimes\left(dx_{15}\wedge
dy_{4}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{5}\otimes\left(dx_{6}\wedge
dx_{10}\right)\\\ &\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{5}\otimes\left(dx_{14}\wedge
dy_{4}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{6}\otimes\left(dx_{5}\wedge
dx_{10}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,y_{6}\otimes\left(dx_{15}\wedge
dy_{7}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{8}\otimes\left(dx_{1}\wedge
dx_{10}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{8}\otimes\left(dx_{14}\wedge
dy_{7}\right)\text{+ }\alpha\,y_{10}\otimes\left(dx_{1}\wedge dx_{8}\right)\\\
&\text{+ }\alpha\,y_{10}\otimes\left(dx_{5}\wedge dx_{6}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{10}\otimes\left(dx_{12}\wedge
dy_{4}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{10}\otimes\left(dx_{13}\wedge
dy_{7}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{12}\otimes\left(dx_{10}\wedge
dy_{4}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{13}\otimes\left(dx_{10}\wedge
dy_{7}\right)\text{+ }\alpha\,y_{14}\otimes\left(dx_{5}\wedge dy_{4}\right)\\\
&\text{+ }\alpha\,y_{14}\otimes\left(dx_{8}\wedge dy_{7}\right)\text{+
}y_{15}\otimes\left(dx_{1}\wedge dy_{4}\right)\text{+
}y_{15}\otimes\left(dx_{6}\wedge dy_{7}\right)\end{array}$
(44)
$\tiny\begin{array}[]{ll}\deg=-6&(\alpha+1)\,h_{3}\otimes\left(dx_{9}\right){}^{\wedge
2}\text{+ }\left(\alpha^{2}+\alpha\right)\,x_{2}\otimes\left(dx_{4}\wedge
dx_{13}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,x_{2}\otimes\left(dx_{7}\wedge
dx_{12}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,x_{5}\otimes\left(dx_{4}\wedge
dx_{14}\right)\text{+ }(\alpha+1)\,x_{5}\otimes\left(dx_{9}\wedge
dx_{12}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,x_{8}\otimes\left(dx_{7}\wedge
dx_{14}\right)\\\ &\text{+ }(\alpha+1)\,x_{8}\otimes\left(dx_{9}\wedge
dx_{13}\right)\text{+ }(\alpha+1)\,y_{1}\otimes\left(dx_{7}\wedge
dx_{9}\right)\text{+ }(\alpha+1)\,y_{4}\otimes\left(dx_{6}\wedge
dx_{9}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{4}\otimes\left(dx_{13}\wedge
dy_{2}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{4}\otimes\left(dx_{14}\wedge
dy_{5}\right)\text{+ }(\alpha+1)\,y_{6}\otimes\left(dx_{4}\wedge
dx_{9}\right)\\\ &\text{+ }(\alpha+1)\,y_{7}\otimes\left(dx_{1}\wedge
dx_{9}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{7}\otimes\left(dx_{12}\wedge
dy_{2}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{7}\otimes\left(dx_{14}\wedge
dy_{8}\right)\\\ &\text{+ }(\alpha+1)\,y_{9}\otimes\left(dx_{1}\wedge
dx_{7}\right)\text{+ }(\alpha+1)\,y_{9}\otimes\left(dx_{4}\wedge
dx_{6}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{9}\otimes\left(dx_{12}\wedge
dy_{5}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{9}\otimes\left(dx_{13}\wedge
dy_{8}\right)\text{+ }y_{12}\otimes\left(dx_{7}\wedge dy_{2}\right)\text{+
}y_{12}\otimes\left(dx_{9}\wedge dy_{5}\right)\text{+
}y_{13}\otimes\left(dx_{4}\wedge dy_{2}\right)\\\ &\text{+
}y_{13}\otimes\left(dx_{9}\wedge dy_{8}\right)\text{+
}y_{14}\otimes\left(dx_{4}\wedge dy_{5}\right)\text{+
}y_{14}\otimes\left(dx_{7}\wedge dy_{8}\right)\\\ &\\\
\deg=-6&\left(\alpha^{3}+\alpha^{2}\right)\,x_{1}\otimes\left(dx_{8}\wedge
dx_{10}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,x_{1}\otimes\left(dx_{15}\wedge
dy_{4}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,x_{4}\otimes\left(dx_{2}\wedge
dx_{13}\right)\\\ &\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,x_{4}\otimes\left(dx_{8}\wedge
dx_{11}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,x_{4}\otimes\left(dx_{15}\wedge
dy_{1}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,x_{9}\otimes\left(dx_{3}\wedge
dx_{15}\right)\\\ &\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,x_{9}\otimes\left(dx_{8}\wedge
dx_{13}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{2}\otimes\left(dx_{3}\wedge
dx_{10}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{2}\otimes\left(dx_{13}\wedge
dy_{4}\right)\\\ &\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{3}\otimes\left(dx_{2}\wedge
dx_{10}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{3}\otimes\left(dx_{5}\wedge
dx_{8}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,y_{3}\otimes\left(dx_{15}\wedge
dy_{9}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{5}\otimes\left(dx_{3}\wedge
dx_{8}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{8}\otimes\left(dx_{3}\wedge
dx_{5}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{8}\otimes\left(dx_{10}\wedge
dy_{1}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{8}\otimes\left(d_{11}\wedge
dy_{4}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{8}\otimes\left(dx_{13}\wedge
dy_{9}\right)\text{+ }\alpha\,y_{10}\otimes\left(dx_{2}\wedge
dx_{3}\right)\text{+ }\alpha\,y_{10}\otimes\left(dx_{8}\wedge dy_{1}\right)\\\
&\text{+ }\left(\alpha^{2}+\alpha\right)\,y_{11}\otimes\left(dx_{8}\wedge
dy_{4}\right)\text{+ }\alpha\,y_{13}\otimes\left(dx_{2}\wedge
dy_{4}\right)\text{+ }\alpha\,y_{13}\otimes\left(dx_{8}\wedge
d_{9}\right)\text{+ }y_{15}\otimes\left(dx_{3}\wedge dy_{9}\right)\text{+
}y_{15}\otimes\left(dy_{1}\wedge dy_{4}\right)\\\ &\\\
\deg=-4&(\alpha+1)\,h_{3}\otimes\left(dx_{6}\right){}^{\wedge 2}\text{+
}(\alpha+1)\,h_{4}\otimes\left(dx_{6}\right){}^{\wedge 2}\text{+
}\left(\alpha^{2}+\alpha\right)\,x_{2}\otimes\left(dx_{3}\wedge
dx_{10}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,x_{2}\otimes\left(dx_{13}\wedge
dy_{4}\right)\\\ &\text{+ }(\alpha+1)\,x_{4}\otimes\left(dx_{6}\wedge
dx_{9}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,x_{4}\otimes\left(dx_{13}\wedge
dy_{2}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,x_{4}\otimes\left(dx_{14}\wedge
dy_{5}\right)\\\ &\text{+ }(\alpha+1)\,x_{5}\otimes\left(dx_{6}\wedge
dx_{10}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,x_{5}\otimes\left(dx_{14}\wedge
dy_{4}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,x_{11}\otimes\left(dx_{3}\wedge
dx_{14}\right)\\\ &\text{+ }(\alpha+1)\,x_{11}\otimes\left(dx_{6}\wedge
dx_{13}\right)+(\alpha+1)\,y_{1}\otimes\left(dx_{3}\wedge dx_{6}\right)\text{+
}(\alpha+1)\,y_{3}\otimes\left(dx_{1}\wedge dx_{6}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{3}\otimes\left(dx_{10}\wedge
dy_{2}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{3}\otimes\left(d_{14}\wedge
dy_{11}\right)\text{+ }(\alpha+1)\,y_{6}\otimes\left(dx_{1}\wedge
d_{3}\right)\text{+ }(\alpha+1)\,y_{6}\otimes\left(d_{9}\wedge
dy_{4}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{6}\otimes\left(dx_{10}\wedge
dy_{5}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{6}\otimes\left(dx_{13}\wedge
dy_{11}\right)\text{+ }(\alpha+1)\,y_{9}\otimes\left(dx_{6}\wedge
dy_{4}\right)\text{+ }y_{10}\otimes\left(dx_{3}\wedge dy_{2}\right)\\\
&\text{+ }y_{10}\otimes\left(dx_{6}\wedge dy_{5}\right)\text{+
}y_{13}\otimes\left(dx_{6}\wedge dy_{11}\right)\text{+
}y_{13}\otimes\left(dy_{2}\wedge dy_{4}\right)\text{+
}y_{14}\otimes\left(dx_{3}\wedge dy_{11}\right)\text{+
}y_{14}\otimes\left(dy_{4}\wedge dy_{5}\right)\\\ &\\\
\deg=-4&\left(\alpha^{3}+\alpha^{2}\right)\,h_{4}\otimes\left(dx_{5}\right){}^{\wedge
2}\text{+ }\left(\alpha^{3}+\alpha^{2}\right)\,x_{3}\otimes\left(dx_{2}\wedge
dx_{10}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,x_{3}\otimes\left(dx_{5}\wedge
dx_{8}\right)\\\ &\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,x_{3}\otimes\left(dx_{15}\wedge
dy_{9}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,x_{6}\otimes\left(dx_{5}\wedge
dx_{10}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,x_{6}\otimes\left(dx_{15}\wedge
dy_{7}\right)\\\ &\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,x_{7}\otimes\left(dx_{2}\wedge
dx_{12}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,x_{7}\otimes\left(dx_{5}\wedge
dx_{11}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,x_{7}\otimes\left(dx_{15}\wedge
dy_{6}\right)\\\ &\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,x_{9}\otimes\left(dx_{5}\wedge
dx_{12}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,x_{9}\otimes\left(dx_{15}\wedge
dy_{3}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{2}\otimes\left(dx_{10}\wedge
dy_{3}\right)\\\ &\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,y_{2}\otimes\left(dx_{12}\wedge
dy_{7}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{5}\otimes\left(dx_{8}\wedge
dy_{3}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{5}\otimes\left(dx_{10}\wedge
dy_{6}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{5}\otimes\left(dx_{11}\wedge
dy_{7}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{5}\otimes\left(d(x_{12}\wedge
dy_{9}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{8}\otimes\left(dx_{5}\wedge
dy_{3}\right)\\\ &\text{+ }\alpha\,y_{10}\otimes\left(dx_{2}\wedge
dy_{3}\right)\text{+ }\alpha\,y_{10}\otimes\left(dx_{5}\wedge
dy_{6}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{11}\otimes\left(dx_{5}\wedge
dy_{7}\right)\text{+ }\alpha\,y_{12}\otimes\left(dx_{2}\wedge dy_{7}\right)\\\
&\text{+ }\alpha\,y_{12}\otimes\left(dx_{5}\wedge dy_{9}\right)\text{+
}y_{15}\otimes\left(dy_{3}\wedge dy_{9}\right)\text{+
}y_{15}\otimes\left(dy_{6}\wedge dy_{7}\right)\\\ &\\\
\deg=-2&(\alpha+1)\,h_{4}\otimes\left(dx_{1}\right){}^{\wedge 2}\text{+
}\left(\alpha^{2}+\alpha\right)\,x_{2}\otimes\left(d(x_{10}\wedge
dy_{3}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,x_{2}\otimes\left(dx_{12}\wedge
dy_{7}\right)\text{+ }(\alpha+1)\,x_{3}\otimes\left(dx_{1}\wedge
dx_{6}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,x_{3}\otimes\left(dx_{10}\wedge
dy_{2}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,x_{3}\otimes\left(dx_{14}\wedge
dy_{11}\right)\text{+ }(\alpha+1)\,x_{7}\otimes\left(dx_{1}\wedge
dx_{9}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,x_{7}\otimes\left(dx_{12}\wedge
dy_{2}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,x_{7}\otimes\left(d_{14}\wedge
dy_{8}\right)\text{+ }(\alpha+1)\,x_{8}\otimes\left(dx_{1}\wedge
dx_{10}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,x_{8}\otimes\left(dx_{14}\wedge
dy_{7}\right)\text{+ }(\alpha+1)\,x_{11}\otimes\left(dx_{1}\wedge
dx_{12}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,x_{11}\otimes\left(dx_{14}\wedge
dy_{3}\right)\\\ &\text{+ }(\alpha+1)\,y_{1}\otimes\left(dx_{6}\wedge
dy_{3}\right)\text{+ }(\alpha+1)\,y_{1}\otimes\left(dx_{9}\wedge
dy_{7}\right)\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{1}\otimes\left(dx_{10}\wedge
dy_{8}\right)\\\ &\text{+
}\left(\alpha^{2}+\alpha\right)\,y_{1}\otimes\left(dx_{12}\wedge
dy_{11}\right)\text{+ }(\alpha+1)\,y_{6}\otimes\left(dx_{1}\wedge
dy_{3}\right)\text{+ }(\alpha+1)\,y_{9}\otimes\left(dx_{1}\wedge
dy_{7}\right)\text{+ }y_{10}\otimes\left(dx_{1}\wedge dy_{8}\right)\\\
&\text{+ }y_{10}\otimes\left(dy_{2}\wedge dy_{3}\right)\text{+
}y_{12}\otimes\left(dx_{1}\wedge dy_{11}\right)\text{+
}y_{12}\otimes\left(dy_{2}\wedge dy_{7}\right)\text{+
}y_{14}\otimes\left(dy_{3}\wedge dy_{11}\right)\text{+
}y_{14}\otimes\left(dy_{7}\wedge dy_{8}\right)\\\ &\\\
\deg=0&\alpha\,h_{1}\otimes\left(d_{14}\wedge dy_{14}\right)\text{+
}\alpha^{2}\,h_{1}\otimes\left(dx_{15}\wedge dy_{15}\right)\text{+
}h_{2}\otimes\left(dx_{5}\wedge dy_{5}\right)\text{+
}h_{2}\otimes\left(dx_{8}\wedge dy_{8}\right)\text{+
}h_{2}\otimes\left(dx_{10}\wedge dy_{10}\right)\\\ &\text{+
}h_{2}\otimes\left(dx_{11}\wedge dy_{11}\right)\text{+
}h_{2}\otimes\left(dx_{12}\wedge dy_{12}\right)\text{+
}h_{2}\otimes\left(dx_{13}\wedge dy_{13}\right)\text{+
}\alpha^{2}\,h_{2}\otimes\left(dx_{14}\wedge dy_{14}\right)\text{+
}h_{3}\otimes\left(dx_{1}\wedge dy_{1}\right)\\\ &\text{+
}\alpha\,h_{3}\otimes\left(dx_{5}\wedge dy_{5}\right)\text{+
}h_{3}\otimes\left(dx_{6}\wedge dy_{6}\right)\text{+
}\alpha\,h_{3}\otimes\left(dx_{8}\wedge dy_{8}\right)\text{+
}h_{3}\otimes\left(dx_{9}\wedge dy_{9}\right)\text{+
}\alpha\,h_{3}\otimes\left(dx_{10}\wedge dy_{10}\right)\\\ &\text{+
}\alpha\,h_{3}\otimes\left(dx_{11}\wedge dy_{11}\right)\text{+
}\alpha\,h_{3}\otimes\left(dx_{12}\wedge dy_{12}\right)\text{+
}\left(\alpha^{3}+\alpha^{2}\right)\,h_{3}\otimes\left(dx_{14}\wedge
dy_{14}\right)\text{+
}\left(\alpha^{4}+\alpha^{3}\right)\,h_{3}\otimes\left(dx_{15}\wedge
dy_{15}\right)\\\ &\text{+ }\alpha\,h_{4}\otimes\left(dx_{12}\wedge
dy_{12}\right)\text{+ }\alpha\,h_{4}\otimes\left(dx_{13}\wedge
dy_{13}\right)\text{+ }\alpha\,h_{4}\otimes\left(dx_{14}\wedge
dy_{14}\right)\text{+ }\alpha^{2}\,h_{4}\otimes\left(dx_{15}\wedge
dy_{15}\right)\text{+ }x_{1}\otimes\left(dh_{1}\wedge dx_{1}\right)\\\
&\text{+ }\alpha\,x_{1}\otimes\left(dx_{10}\wedge dy_{8}\right)\text{+
}\alpha\,x_{1}\otimes\left(dx_{12}\wedge dy_{11}\right)\text{+
}\alpha\,x_{1}\otimes\left(dx_{14}\wedge dy_{13}\right)\text{+
}x_{2}\otimes\left(dh_{1}\wedge dx_{2}\right)\text{+
}x_{2}\otimes\left(dx_{5}\wedge dy_{1}\right)\\\ &\text{+
}x_{2}\otimes\left(dx_{8}\wedge dy_{6}\right)\text{+
}x_{2}\otimes\left(dx_{11}\wedge dy_{9}\right)\text{+
}\alpha^{3}\,x_{2}\otimes\left(dx_{15}\wedge dy_{14}\right)\text{+
}\alpha\,x_{3}\otimes\left(dx_{13}\wedge dy_{12}\right)\text{+
}x_{4}\otimes\left(dh_{1}\wedge dx_{4}\right)\\\ &\text{+
}\alpha\,x_{4}\otimes\left(dx_{12}\wedge dy_{10}\right)\text{+
}x_{5}\otimes\left(dx_{10}\wedge dy_{6}\right)\text{+
}x_{5}\otimes\left(dx_{12}\wedge dy_{9}\right)\text{+
}\alpha\,x_{5}\otimes\left(dx_{15}\wedge dy_{13}\right)\text{+
}x_{6}\otimes\left(dh_{1}\wedge dx_{6}\right)\\\ &\text{+
}\alpha\,x_{6}\otimes\left(dx_{10}\wedge dy_{5}\right)\text{+
}\alpha\,x_{6}\otimes\left(dx_{13}\wedge dy_{11}\right)\text{+
}\alpha\,x_{6}\otimes\left(dx_{14}\wedge dy_{12}\right)\text{+
}x_{7}\otimes\left(dh_{1}\wedge dx_{7}\right)\text{+
}\alpha\,x_{7}\otimes\left(dx_{13}\wedge dy_{10}\right)\\\ &\text{+
}x_{8}\otimes\left(dx_{10}\wedge dy_{1}\right)\text{+
}x_{8}\otimes\left(dx_{13}\wedge dy_{9}\right)\text{+
}\alpha\,x_{8}\otimes\left(dx_{15}\wedge dy_{12}\right)\text{+
}\alpha\,x_{9}\otimes\left(dx_{12}\wedge dy_{5}\right)\text{+
}\alpha\,x_{9}\otimes\left(dx_{13}\wedge dy_{8}\right)\\\ &\text{+
}\alpha\,x_{9}\otimes\left(dx_{14}\wedge dy_{10}\right)\text{+
}x_{10}\otimes\left(dh_{1}\wedge dx_{10}\right)\text{+
}x_{10}\otimes\left(dx_{14}\wedge dy_{9}\right)\text{+
}\alpha\,x_{10}\otimes\left(dx_{15}\wedge dy_{11}\right)\text{+
}x_{11}\otimes\left(dh_{1}\wedge dx_{11}\right)\\\ &\text{+
}x_{11}\otimes\left(dx_{12}\wedge dy_{1}\right)\text{+
}x_{11}\otimes\left(dx_{13}\wedge dy_{6}\right)\text{+
}\alpha\,x_{11}\otimes\left(dx_{15}\wedge dy_{10}\right)\text{+
}x_{12}\otimes\left(dx_{14}\wedge dy_{6}\right)\text{+
}\alpha\,x_{12}\otimes\left(dx_{15}\wedge dy_{8}\right)\\\ &\text{+
}x_{13}\otimes\left(dx_{14}\wedge dy_{1}\right)\text{+
}\alpha\,x_{13}\otimes\left(dx_{15}\wedge dy_{5}\right)\text{+
}x_{14}\otimes\left(dh_{1}\wedge dx_{14}\right)\text{+
}y_{1}\otimes\left(dh_{1}\wedge dy_{1}\right)\text{+
}\alpha\,y_{1}\otimes\left(dx_{8}\wedge dy_{10}\right)\\\ &\text{+
}\alpha\,y_{1}\otimes\left(dx_{11}\wedge dy_{12}\right)\text{+
}\alpha\,y_{1}\otimes\left(dx_{13}\wedge dy_{14}\right)\text{+
}y_{2}\otimes\left(dh_{1}\wedge dy_{2}\right)\text{+
}y_{2}\otimes\left(dx_{1}\wedge dy_{5}\right)\text{+
}y_{2}\otimes\left(dx_{6}\wedge dy_{8}\right)\\\ &\text{+
}y_{2}\otimes\left(dx_{9}\wedge dy_{11}\right)\text{+
}\alpha^{3}\,y_{2}\otimes\left(dx_{14}\wedge dy_{15}\right)\text{+
}\alpha\,y_{3}\otimes\left(dx_{12}\wedge dy_{13}\right)\text{+
}y_{4}\otimes\left(dh_{1}\wedge dy_{4}\right)\text{+
}\alpha\,y_{4}\otimes\left(dx_{10}\wedge dy_{12}\right)\\\ &\text{+
}y_{5}\otimes\left(dx_{6}\wedge dy_{10}\right)\text{+
}y_{5}\otimes\left(dx_{9}\wedge dy_{12}\right)\text{+
}\alpha\,y_{5}\otimes\left(dx_{13}\wedge dy_{15}\right)\text{+
}y_{6}\otimes\left(dh_{1}\wedge dy_{6}\right)\text{+
}\alpha\,y_{6}\otimes\left(dx_{5}\wedge dy_{10}\right)\\\ &\text{+
}\alpha\,y_{6}\otimes\left(dx_{11}\wedge dy_{13}\right)\text{+
}\alpha\,y_{6}\otimes\left(dx_{12}\wedge dy_{14}\right)\text{+
}y_{7}\otimes\left(dh_{1}\wedge dy_{7}\right)\text{+
}\alpha\,y_{7}\otimes\left(dx_{10}\wedge dy_{13}\right)\text{+
}y_{8}\otimes\left(dx_{1}\wedge dy_{10}\right)\\\ &\text{+
}y_{8}\otimes\left(dx_{9}\wedge dy_{13}\right)\text{+
}\alpha\,y_{8}\otimes\left(dx_{12}\wedge dy_{15}\right)\text{+
}\alpha\,y_{9}\otimes\left(dx_{5}\wedge dy_{12}\right)\text{+
}\alpha\,y_{9}\otimes\left(dx_{8}\wedge dy_{13}\right)\text{+
}\alpha\,y_{9}\otimes\left(dx_{10}\wedge dy_{14}\right)\\\ &\text{+
}y_{10}\otimes\left(dh_{1}\wedge dy_{10}\right)\text{+
}y_{10}\otimes\left(dx_{9}\wedge dy_{14}\right)\text{+
}\alpha\,y_{10}\otimes\left(dx_{11}\wedge dy_{15}\right)\text{+
}y_{11}\otimes\left(dh_{1}\wedge dy_{11}\right)\text{+
}y_{11}\otimes\left(dx_{1}\wedge dy_{12}\right)\\\ &\text{+
}y_{11}\otimes\left(dx_{6}\wedge dy_{13}\right)\text{+
}\alpha\,y_{11}\otimes\left(dx_{10}\wedge dy_{15}\right)\text{+
}y_{12}\otimes\left(dx_{6}\wedge dy_{14}\right)\text{+
}\alpha\,y_{12}\otimes\left(dx_{8}\wedge dy_{15}\right)\text{+
}y_{13}\otimes\left(dx_{1}\wedge dy_{14}\right)\\\ &\text{+
}\alpha\,y_{13}\otimes\left(dx_{5}\wedge dy_{15}\right)\text{+
}y_{14}\otimes\left(dh_{1}\wedge dy_{14}\right)\end{array}$
### 4.9. ${\mathfrak{gl}}(4)$ and its simple relative
The Lie algebra444Observe that ${\mathfrak{psl}}(4):={\mathfrak{g}}(2)$ in the
basis given in [FH], p. 346. Shen himself observed this, and therefore the
property of the root systems to have all roots of “equal length” (whatever
that might mean if $p=2$) is not an invariant if $p=2$.
${\mathfrak{g}}={\mathfrak{psl}}(4)$ has no Cartan matrix; its relative that
has a Cartan matrix which is $(\ref{gl4cm})$ is ${\mathfrak{gl}}(4)$.
###### Lemma.
In order to compare with Shen’s “variations of ${\mathfrak{g}}(2)$ theme” we
give, in parentheses, together with the generators of ${\mathfrak{sl}}(4)$,
their weights in terms of the root systems of ${\mathfrak{gl}}(4)$ ($\beta$’s)
and ${\mathfrak{g}}(2)$ ($\alpha$’s):
(45) $\begin{pmatrix}{\bar{0}}&1&0\\\ 1&{\bar{0}}&1\\\
0&1&{\bar{0}}\end{pmatrix}\qquad\tiny\begin{array}[]{l}x_{1}=E_{12}\;(\beta_{1}=\alpha_{2}),\;\;x_{2}=E_{23}\;(\beta_{2}=\alpha_{1}),\;\;x_{3}=E_{34}\;(\beta_{3}=2\alpha_{1}+\alpha_{2}),\\\
{}x_{4}:=[x_{1}\text{,$\,$}x_{2}]=E_{13}\;(\beta_{1}+\beta_{2}=\alpha_{1}+\alpha_{2}),\;\;x_{5}:=[x_{2}\text{,$\,$}x_{3}]=E_{24}\;(\beta_{2}+\beta_{3}=3\alpha_{1}+\alpha_{2}),\\\
{}x_{6}:=[x_{3}\text{,$\,$}[x_{1}\text{,$\,$}x_{2}]]=E_{14}\;(\beta_{1}+\beta_{2}+\beta_{3}=3\alpha_{1}+2\alpha_{2})\end{array}$
Then $H^{2}({\mathfrak{g}};{\mathfrak{g}})$ is spanned by the cocycles
$(\ref{eqpsl4})$ and $(\ref{eqpsl4bis})$. The parameters $a$, $b$ in
parentheses near the degree correspond to Shen’s parameters of his
$V_{4}{\mathfrak{g}}(2;a,b)$.
(46) $\tiny\begin{array}[]{lll}\deg=-6&-4(\alpha_{1}+\alpha_{2})&y_{1}\otimes
dx_{4}\wedge dx_{6}+y_{4}\otimes dx_{1}\wedge dx_{6}+y_{6}\otimes dx_{1}\wedge
dx_{4}\\\\[8.53581pt] &&\\\ \deg=-6&-4(2\alpha_{1}+\alpha_{2})&y_{3}\otimes
dx_{5}\wedge dx_{6}+y_{5}\otimes dx_{3}\wedge dx_{6}+y_{6}\otimes dx_{3}\wedge
dx_{5}\\\ &&\\\ \deg=-4\;(b)&-2(2\alpha_{1}+\alpha_{2})&h_{2}\otimes
dx_{4}\wedge dx_{5}+y_{2}\otimes dh_{3}\wedge dx_{6}+y_{2}\otimes dx_{1}\wedge
x_{5}+y_{2}\otimes dx_{3}\wedge dx_{4}+y_{4}\otimes dh_{3}\wedge dx_{5}+\\\
&&+y_{5}\otimes dh_{3}\wedge dx_{4}+y_{6}\otimes dh_{3}\wedge
dx_{2}+y_{6}\otimes dx_{4}\wedge dy_{1}+y_{6}\otimes dx_{5}\wedge dy_{3}\\\
&&\\\ \deg=-4&-3(\alpha_{1}+\alpha_{2})&h_{3}\otimes dx_{2}\wedge
dx_{6}+x_{1}\otimes dx_{4}\wedge dx_{6}+x_{3}\otimes dx_{5}\wedge
dx_{6}+y_{2}\otimes dh_{2}\wedge dx_{6}\\\ &&+y_{2}\otimes dh_{3}\wedge
dx_{6}+y_{2}\otimes dx_{1}\wedge dx_{5}+y_{2}\otimes dx_{3}\wedge
dx_{4}+y_{4}\otimes dh_{2}\wedge dx_{5}+\\\ &&y_{4}\otimes dh_{3}\wedge
dx_{5}+y_{4}\otimes dx_{2}\wedge dx_{3}+y_{5}\otimes dh_{2}\wedge
dx_{4}+y_{5}\otimes dh_{3}\wedge dx_{4}\\\ &&+y_{5}\otimes dx_{1}\wedge
dx_{2}+y_{6}\otimes dh_{2}\wedge dx_{2}+y_{6}\otimes dh_{3}\wedge dx_{2}\\\
\end{array}$ (47) $\tiny\begin{array}[]{lll}\deg=-2&0&x_{3}\otimes
dx_{2}\wedge dx_{4}+y_{2}\otimes dx_{4}\wedge dy_{3}+y_{4}\otimes dx_{2}\wedge
dy_{3}\\\ &&\\\ \deg=-2&-4\alpha_{1}&x_{1}\otimes dx_{2}\wedge
dx_{5}+y_{2}\otimes dx_{5}\wedge dy_{1}+y_{5}\otimes dx_{2}\wedge dy_{1}\\\
&&\\\ \deg=-2&-2(\alpha_{1}+\alpha_{2})&h_{3}\otimes dx_{6}\wedge
dy_{2}+x_{2}\otimes dh_{2}\wedge dx_{6}+y_{1}\otimes dh_{2}\wedge
dx_{3}+y_{1}\otimes dx_{5}\wedge dy_{2}+y_{1}\otimes dx_{6}\wedge
dy_{4}\\\\[8.53581pt] &&+y_{3}\otimes dh_{2}\wedge dx_{1}+y_{3}\otimes
dx_{4}\wedge dy_{2}+y_{3}\otimes dx_{6}\wedge dy_{5}+y_{6}\otimes dh_{2}\wedge
dy_{2}\\\ &&\\\ \deg=-2&-2(\alpha_{1}+\alpha_{2})&h_{2}\otimes dx_{1}\wedge
dx_{3}+x_{2}\otimes dh_{3}\wedge dx_{6}+x_{4}\otimes dx_{1}\wedge
dx_{6}+x_{5}\otimes dx_{3}\wedge dx_{6}+y_{1}\otimes dh_{3}\wedge
dx_{3}\\\\[8.53581pt] &&+y_{1}\otimes dx_{5}\wedge dy_{2}+y_{1}\otimes
dx_{6}\wedge dy_{4}+y_{3}\otimes dh_{3}\wedge dx_{1}+y_{3}\otimes dx_{4}\wedge
dy_{2}+y_{3}\otimes dx_{6}\wedge dy_{5}+\\\ &&y_{4}\otimes dx_{3}\wedge
dy_{2}+y_{5}\otimes dx_{1}\wedge dy_{2}+y_{6}\otimes dh_{3}\wedge dy_{2}\\\
&&\\\ \deg=0&2\alpha_{1}&h_{3}\otimes dx_{4}\wedge dy_{5}+x_{3}\otimes
dh_{2}\wedge dx_{1}+x_{3}\otimes dx_{4}\wedge dy_{2}+x_{3}\otimes dx_{6}\wedge
dy_{5}+x_{5}\otimes dh_{2}\wedge dx_{4}\\\\[8.53581pt] &&+y_{1}\otimes
dh_{2}\wedge dy_{3}+y_{1}\otimes dx_{2}\wedge dy_{5}+y_{1}\otimes dx_{4}\wedge
dy_{6}+y_{4}\otimes dh_{2}\wedge dy_{5}\\\ &&\\\
\deg=0&-2\alpha_{1}&h_{3}\otimes dx_{5}\wedge dy_{4}+x_{1}\otimes dh_{2}\wedge
dx_{3}+x_{1}\otimes dx_{5}\wedge dy_{2}+x_{1}\otimes dx_{6}\wedge
dy_{4}+x_{4}\otimes dh_{2}\wedge dx_{5}\\\\[8.53581pt] &&+y_{3}\otimes
dh_{2}\wedge dy_{1}+y_{3}\otimes dx_{2}\wedge dy_{4}+y_{3}\otimes dx_{5}\wedge
dy_{6}+y_{5}\otimes dh_{2}\wedge dy_{4}\\\ &&\\\
\deg=0\;(a)&2\alpha_{1}&h_{2}\otimes dx_{1}\wedge dy_{3}+x_{2}\otimes
dx_{4}\wedge dy_{3}+x_{3}\otimes dh_{2}\wedge dx_{1}+x_{3}\otimes dh_{3}\wedge
dx_{1}+x_{5}\otimes dh_{2}\wedge dx_{4}\\\\[8.53581pt] &&+x_{5}\otimes
dh_{3}\wedge dx_{4}+x_{5}\otimes dx_{1}\wedge dx_{2}+y_{1}\otimes dh_{2}\wedge
dy_{3}+y_{1}\otimes dh_{3}\wedge dy_{3}+y_{1}\otimes dx_{2}\wedge dy_{5}+\\\
&&y_{1}\otimes dx_{4}\wedge dy_{6}+y_{4}\otimes dh_{2}\wedge
dy_{5}+y_{4}\otimes dh_{3}\wedge dy_{5}+y_{4}\otimes dx_{1}\wedge
dy_{6}+y_{6}\otimes dy_{3}\wedge dy_{5}\\\ &&\\\
\deg=0&-2\alpha_{1}&h_{2}\otimes dx_{3}\wedge dy_{1}+x_{1}\otimes dh_{2}\wedge
dx_{3}+x_{1}\otimes dh_{3}\wedge dx_{3}+x_{2}\otimes dx_{5}\wedge
dy_{1}+x_{4}\otimes dh_{2}\wedge dx_{5}\\\\[8.53581pt] &&+x_{4}\otimes
dh_{3}\wedge dx_{5}+x_{4}\otimes dx_{2}\wedge dx_{3}+y_{3}\otimes dh_{2}\wedge
dy_{1}+y_{3}\otimes dh_{3}\wedge dy_{1}+y_{3}\otimes dx_{2}\wedge dy_{4}+\\\
&&y_{3}\otimes dx_{5}\wedge dy_{6}+y_{5}\otimes dh_{2}\wedge
dy_{4}+y_{5}\otimes dh_{3}\wedge dy_{4}+y_{5}\otimes dx_{3}\wedge
dy_{6}+y_{6}\otimes dy_{1}\wedge dy_{4}\end{array}$
## 5\. On integrability
Consider the bracket
(48) $[\cdot,\cdot]_{c}=[\cdot,\cdot]+tc.$
###### Lemma.
For $p=3$, the bracket $(\ref{lindef})$ satisfies the Jacobi identity in the
following cases:
1) For ${\mathfrak{br}}(2)$, ${\mathfrak{brj}}(2,3)$, ${\mathfrak{br}}(2,5)$
and ${\mathfrak{br}}(3)$, for every of the above homogeneous cocycles $c$;
2) For ${\mathfrak{g}}(2,3)$, never;
3) For ${\mathfrak{g}}(1,6)$, only for the cocycle $c$ of degree $-12$.
###### Remark.
Although JI is not satisfied for the bracket (48) for some cocycles $c$, it
does not mean that these cocycles are not integrable: The Jacobi identity
might be satisfied by a bracket of the form
$[\cdot,\cdot]_{c}=[\cdot,\cdot]+tc_{1}+t^{2}c_{2}+\dots$. We have started to
investigate this option elsewhere, together with integrability study of
multiparameter deformations, except for the simplest cases where the global
deform is linear in all parameters simultaneously.
###### Lemma.
For $p=2$, the bracket $(\ref{lindef})$ satisfies the Jacobi identity in the
following cases:
1) For ${\mathfrak{o}}^{(1)}(3)$, ${\mathfrak{o}}^{(1)}(5)$,
${\mathfrak{psl}}(4)$, and ${\mathfrak{psl}}(6)$, for every of the above
homogeneous cocycles $c$.
2) For ${\mathfrak{wk}}(3;\alpha)$ and ${\mathfrak{wk}}(4;\alpha)$, and also
for ${\mathfrak{bgl}}(3;\alpha)$ and ${\mathfrak{bgl}}(4;\alpha)$, for every
of the above homogeneous cocycles $c$, except the ones of degree $0$.
5) For ${\mathfrak{o}}{\mathfrak{o}}_{I\Pi}^{(1)}(1|2)$, for every of the
above homogeneous cocycles $c$.
###### Lemma.
For $p=2$, the Lie algebra ${\mathfrak{o}}^{(1)}_{I}(3)$ admits the global
deformation given by
$[\cdot,\cdot]_{\alpha,\beta}=[\cdot,\cdot]+\alpha c_{1}+\beta c_{2}.$
Denote the deform by ${\mathfrak{o}}^{(1)}_{I}(3,\alpha,\beta)$. This Lie
algebra is simple if and only if $\alpha\beta\not=1$.
If $\alpha\beta=1$, then ${\mathfrak{o}}^{(1)}_{I}(3,\alpha,\beta)$ has an
ideal $I=\mathop{\mathrm{{Span}}}\nolimits\\{h,x+\alpha y\\}$.
For $\alpha\beta\not=1$, the Lie algebra
${\mathfrak{o}}^{(1)}_{I}(3,\alpha,\beta)$ is simple with two outer
derivations given by
$\alpha h\otimes dh+\alpha x\otimes dx+x\otimes dy,\quad\beta h\otimes
dh+\beta x\otimes dx+y\otimes dx.$
Proof of the statements of this Lemma are straightforward.
Reducing the operator $\mathop{\mathrm{ad}}\nolimits_{h}$ to the Jordan normal
form, we immediately see (in the eigenbasis) that the deform depends,
actually, on one parameter, not two; and all of them are non-isomorphic.
## References
* [BGL] Bouarroudj S., Grozman P., Leites D., Classification of simple finite dimensional modular Lie superalgebras with Cartan matrix. arXiv: 0710.5149
* [Br3] Brown, G., Properties of a 29-dimensional simple Lie algebra of characteristic three, Math. Ann. 261 (1982), 487–492
* [Ch] Chebochko, N. G. Deformations of classical Lie algebras with a homogeneous root system in characteristic two. I. (Russian) Mat. Sb. 196 (2005), no. 9, 125–156; translation in Sb. Math. 196 (2005), no. 9-10, 1371–1402
* [Ch1] Chebochko, N. G. Deformations of the Lie algebra of type $G_{2}$ in characteristic two.
* [DK] Dzhumadil’daev, A.S.; Kostrikin, A.I. Deformations of the Lie Algebra $W_{1}(m)$ (Russian). Trudy Mat. Inst. Steklov 1978, 148, 141–155.
* [Dz] Dzhumadil’daev, A.S. On deformations of classical simple Lie algebras (Russian) Uspehi Mat. Nauk 31 (1976), 211–212
* [Dz1] Dzhumadil’daev, A.S. Deformations of the Lie Algebras $W_{n}(m)$ (Russian). Mat. Sbornik 1989, 180, 168–186
* [Dz2] Dzhumadil’daev, A.S. Cohomology of truncated co-induced representations of Lie algebras of positive characteristic. (Russian) Mat. Sb. 180 (1989), no. 4, 456–468, 559; translation in Math. USSR-Sb. 66 (1990), no. 2, 461–473
* [Dz3] Dzhumadil’daev, A.S. On the cohomology of modular Lie algebras. Math. USSR-Sb 47 (1984), N1, 127-143 Mat. Sb. 119 (1982), no. 1, 132–149
* [FS] Farnsteiner, R.; Strade, H. Shapiro’s lemma and its consequences in the cohomology theory of modular Lie algebras. Math. Z. 206 (1991), no. 1, 153–168
* [FF] Fialowski, A.; Fuchs, D. Singular deformations of Lie algebras. Example: deformations of the Lie algebra $L_{1}$. Topics in singularity theory, Amer. Math. Soc. Transl. Ser. 2, 180, Amer. Math. Soc., Providence, RI, 1997, 77–92; arXiv:q-alg/9706027
* [FG] Frohardt D., Griess R.L. Jr., Automorphisms of modular Lie algebras, Nova J. Algebra and Geometry, v.1, 1992, no.4, 339–345
* [FL] Fuchs D., Lang L., Massey products and deformations; arXiv: q-alg/9602024
* [Fu] Fuks (Fuchs) D., Cohomology of infinite dimensional Lie algebras, Consultants Bureau, NY, 1986
* [FH] Fulton, W., Harris, J., Representation theory. A first course. Graduate Texts in Mathematics, 129. Readings in Mathematics. Springer-Verlag, New York, 1991. xvi+551 pp.
* [Gr] Grozman P., SuperLie, http://www.equaonline.com/math/SuperLie
* [GL1] Grozman P., Leites D., Defining relations for classical Lie superalgebras with Cartan matrix, Czech. J. Phys., Vol. 51, 2001, no. 1, 1–22; arXiv: hep-th/9702073
* [GL4] Grozman P., Leites D., Structures of $G(2)$ type and nonintegrable distributions in characteristic $p$. Lett. Math. Phys. 74 (2005), no. 3, 229–262; arXiv: math.RT/0509400
* [ILL] Iyer U., Lebedev A., Leites D., Prolongs of orthogonal Lie (super)algebras in characteristic $2$. WORK IN PROGRESS
* [K] Kac V. Infinite-dimensional Lie algebras. Third edition. Cambridge University Press, Cambridge, 1990. xxii+400 pp.
* [KWK] Kac, V. G. Corrections to: “Exponentials in Lie algebras of characteristic $p$” [Izv. Akad. Nauk SSSR 35 (1971), no. 4, 762–788; MR0306282 (46 #5408)] by B. Yu. Veisfeiler and Kac. (Russian) Izv. Ross. Akad. Nauk Ser. Mat. 58 (1994), no. 4, 224; translation in Russian Acad. Sci. Izv. Math. 45 (1995), no. 1, 229
* [Kapp] Kaplansky I., Graded Lie algebras, preprints, Univ. Chicago, Chicago, 1975 http://justpasha.org/math/links/subj/lie/kaplansky
* [KKCh] Kirillov, S. A.; Kuznetsov, M. I.; Chebochko, N. G. Deformations of a Lie algebra of type $G_{2}$ of characteristic three. (Russian) Izv. Vyssh. Uchebn. Zaved. Mat. 2000, , no. 3, 33–38; translation in Russian Math. (Iz. VUZ) 44 (2000), no. 3, 31–36
* [Kos] Kostrikin, A. I. A parametric family of simple Lie algebras. (Russian) Izv. Akad. Nauk SSSR Ser. Mat. 34 1970 744–756
* [KK] Kostrikin, A. I.; Kuznetsov, M. I. On deformations of classical Lie algebras of characteristic three. (Russian) Dokl. Akad. Nauk 343 (1995), no. 3, 299–301.
* [KuCh] Kuznetsov, M. I.; Chebochko, N. G. Deformations of classical Lie algebras. (Russian) Mat. Sb. 191 (2000), no. 8, 69–88; translation in Sb. Math. 191 (2000), no. 7-8, 1171–1190
* [LL] Lebedev A., Leites D., (with an appendix by Deligne P.) On realizations of the Steenrod algebras, J. of Prime Research in Mathematics, v. 2 no. 1 (2006), 1–13
* [Lt] Leites D., Towards classification of simple finite dimensional modular Lie superalgebras in characteristic $p$. J. Prime Res. Math., v. 3, 2007, 101–110 MPIMiS preprint 132/2006 (http://www.mis.mpg.de); arXiv: 0710.5638
* [Ls] Leites D. (ed.) Seminar on Supersymmetries, MCCME, Moscow, 2008, ca 500 pp. (in Russian)
* [LCh] Leites D. (ed.) Chevalley supergroups and simple modular Lie superalgebras(S. Bouarroudj, B. Clarke, P. Grozman, A. Lebedev, D. Leites, I. Shchepochkina), ca 350 pp. WORK IN PROGRESS
* [LSh] Leites D., Shchepochkina I., Classification of the simple Lie superalgebras of vector fields, preprint MPIM-2003-28 (http://www.mpim-bonn.mpg.de)
* [LLg] Liu, Dong; Lin, Lei. On the variations of $G_{2}$ Chin. Ann. Math. 24B (2003), N3, 387–294
* [Ru] Rudakov, A. N. Deformations of simple Lie algebras. (Russian) Izv. Akad. Nauk SSSR Ser. Mat. 35 (1971), 1113–1119
* [She1] Shen, Guang Yu, Variations of the classical Lie algebra $G_{2}$ in low characteristics. Nova J. Algebra Geom. 2 (1993), no. 3, 217–243.
* [She2] Shen, Guang Yu, Lie algebras of CL type, J. Algebra, 249 (2002), 95–109
* [S] Strade, H. Simple Lie algebras over fields of positive characteristic. I. Structure theory. de Gruyter Expositions in Mathematics, 38. Walter de Gruyter & Co., Berlin, 2004\. viii+540 pp.
* [St] Steinberg, R. Lectures on Chevalley groups. Notes prepared by John Faulkner and Robert Wilson. Yale University, New Haven, Conn., 1968\. iii+277 pp.
* [Vi] Viviani F., Deformations of Simple Restricted Lie Algebras I, II. arXiv: math.RA/0612861, math.RA/0702499; Deformations of the restricted Melikian Lie algebra; arXiv: math.RA/0702594; Restricted simple Lie algebras and their infinitesimal deformations; arXiv: math.RA/0702755
* [WK] Weisfeiler B., Kac V. Exponentials in Lie algebras of characteristic $p$. Izv. Akad. Nauk SSSR 35 (1971), no. 4, 762–788
| arxiv-papers | 2008-07-18T21:41:24 | 2024-09-04T02:48:56.883900 | {
"license": "Creative Commons - Attribution Share-Alike - https://creativecommons.org/licenses/by-sa/4.0/",
"authors": "Sofiane Bouarroudj, Pavel Grozman, Dimitry Leites",
"submitter": "Dimitry Leites",
"url": "https://arxiv.org/abs/0807.3054"
} |
0807.3067 | # Spin field effect transistors with ultracold atoms
J. Y. Vaishnav Joint Quantum Institute, National Institute of Standards and
Technology, Gaithersburg MD 20899 USA Julius Ruseckas Institute of
Theoretical Physics and Astronomy of Vilnius University, A. Goštauto 12,
Vilnius 01108, Lithuania Charles W. Clark Joint Quantum Institute, National
Institute of Standards and Technology, Gaithersburg MD 20899 USA Gediminas
Juzelūnas Institute of Theoretical Physics and Astronomy of Vilnius
University, A. Goštauto 12, Vilnius 01108, Lithuania
(August 27, 2024)
###### Abstract
We propose a method of constructing cold atom analogs of the spintronic device
known as the Datta-Das transistor (DDT), which despite its seminal conceptual
role in spintronics, has never been successfully realized with electrons. We
propose two alternative schemes for an atomic DDT, both of which are based on
the experimental setup for tripod stimulated Raman adiabatic passage. Both
setups involve atomic beams incident on a series of laser fields mimicking the
relativistic spin orbit coupling for electrons that is the operating mechanism
of the DDT.
###### pacs:
37.10.Vz, 37.10.Jk,85.75.Hh
The emerging technology of semiconductor spintronics exploits the electron’s
spin degree of freedom, as well as its charge state. The first scheme for a
semiconductor spintronic device was a spin field-effect transistor known as
the Datta-Das transistor (DDT) (Fig. 1a) (dattadas, ). The eighteen years
since the theoretical proposal have seen numerous experimental efforts to
construct the DDT. Various experimental obstacles, such as difficulties in
spin injection, stray electric fields and insufficient quality of spin-orbit
coupling, have prevented successful implementation of the DDT (zutic2004sfa,
).
Cold atom systems, in contrast with their electronic counterparts, are highly
controllable and tunable. This suggests the possibility of designing precise
atomic analogs of electronic systems which, due either to fundamental physical
limits or technological difficulties, are experimentally inaccessible in their
original manifestations. The idea grows out of recent interest in
“atomtronics,” or building cold atom analogs of ordinary electronic materials,
devices and circuits (Atom-Diode, ; seaman:023615, ; experimentaldiode, ). In
particular, an atom diode has been proposed (Atom-Diode, ) and realized
(experimentaldiode, ).
In this Letter, we identify a method for constructing a cold atom analog of a
Datta-Das transistor. The setup is based on a four level “tripod” scheme of
atom-light coupling (unanyan1998rca, ; unanyan99pra, ; theuer1999nlc, ;
stirap, ; ruseckas-2005-95, ; lietfiz-2007, ) involving three atomic ground
states and one excited state (see Fig. 1b). Such tripod schemes are an
extension of the usual three-level $\Lambda$-type setup for stimulated Raman
adiabatic passage (STIRAP) (Atom-Diode, ; spielman-2008, ), and are
experimentally accessible in metastable Ne, ${}^{87}\mbox{Rb}$ and a number of
other gases (theuer1999nlc, ; stirap, ). The proposed device provides a robust
method for atomic state manipulation that is immune to the inhomogeneities
intrinsic to programmed Rabi pulses.
Figure 1: (a) Schematic of a DDT. “S” and “D” are ferromagnetic source and
drain electrodes. In between is a semiconducting gate region, where the spin
precesses by an amount which depends periodically on the tunable gate voltage
$V_{g}.$ This precession results in a controllable current modulation at $D.$
(b) A tripod scheme of atomic energy levels, coupled by laser fields with Rabi
frequencies $\Omega_{i}$. (c,d) Two alternative setups for an atomic version
of the DDT. Here, the source is a state-polarized atomic beam (blue), the gate
is the intersection region of a configuration of laser beams (red), and the
drain is an atomic state analyzer (green).
The source terminal of an electronic DDT (Fig. 1a) is a ferromagnetic
electrode that emits spin-polarized electrons. The DDT drain terminal, a
ferromagnetic analyzer, acts as a spin filter. Between source and drain is a
semiconducting gate region, in which the gate-induced electric field produces
a Rashba spin-orbit coupling (Rashba60, ) for electrons. While passing through
the gate region, the electron’s spin precesses; the electron emerges at the
drain having undergone a spin rotation which is tunable via the gate voltage.
Since the drain passes only a certain spin direction, the drain current is an
oscillating function of the gate voltage.
Our atomic analog of the DDT (Figs. 1c,d) uses a beam of atoms in place of
electrons. The two dark states in the tripod setup play the role of the
electron’s spin states, and the “source” is a dilute atomic beam. The “gate”
region consists of crossed laser beams engineered to mimic Rashba or Rashba-
like spin orbit couplings (galitski, ; Jacob07, ; jayzb, ; prar08, ;
juz08-neg-refl, ); the analog of the gate voltage can be tuned by varying the
relative strengths of the lasers. The drain is a state-selective atomic
filter, such as a Stern-Gerlach device or radio-frequency or Raman outcoupler
(Edwards, ). While the goal of this paper is to explore the possibility of
constructing the atomic analog of spintronic devices, the two dark states of
the tripod atom can be considered qubit states (qi2, ; qi0, ; unanyan04PRA, ;
qi3, ); in this context the atomic DDT represents a single-qubit phase gate
for a dilute atomic beam. In contrast to typical single qubit gates, this
setup does not involve time-dependent pulses, and the amount of the qubit
rotation within the gate region is independent of the atom’s velocity, due to
the geometric nature of the process.
## Tripod scheme
The proposed DDT implementations exploit the tripod scheme (Fig. 1b,c)
(unanyan1998rca, ; unanyan99pra, ; theuer1999nlc, ; stirap, ;
ruseckas-2005-95, ; lietfiz-2007, ), in which a four level atom feels two
counterpropagating stationary laser beams and a third orthogonal beam
(Jacob07, ; prar08, ; juz08-neg-refl, ). The lasers induce transitions between
the ground states $|j\rangle$ ($j=1,2,3$) and an excited state $|0\rangle$
with spatially dependent Rabi frequencies
$\Omega_{1}=|\Omega_{1}|e^{-i\kappa_{0}x}$ ,
$\Omega_{2}=|\Omega_{2}|e^{i\kappa_{0}x}$ and
$\Omega_{3}=|\Omega_{3}|e^{i\kappa_{0}z}$ , $\kappa_{0}$ being a wave-number.
The electronic Hamiltonian of a tripod atom is, in the interaction
representation and rotating wave approximation,
$\hat{H}_{e}=-\hbar\Omega|B\rangle\langle 0|+\mathrm{H.c.}$, where
$|B\rangle=\left(|1\rangle\Omega_{1}^{*}+|2\rangle\Omega_{2}^{*}+|3\rangle\Omega_{3}^{*}\right)/\Omega$
and $\Omega^{2}=|\Omega_{1}|^{2}+|\Omega_{2}|^{2}+|\Omega_{3}|^{2}$.
$\hat{H}_{e}$ has two degenerate dark states $|D_{j}\rangle$ containing no
excited state contribution: $\hat{H}_{e}|D_{j}\rangle=0$, $j=1,2.$ An
additional pair of bright eigenstates
$|\pm\rangle=\left(|B\rangle\pm|0\rangle\right)/\sqrt{2}$ is separated from
the dark states by $\pm\hbar\Omega$. For the light fields of interest, the
dark states can be chosen as:
$\displaystyle|D_{1}\rangle$ $\displaystyle=$
$\displaystyle\left(\sin\varphi|1\rangle^{\prime}-\cos\varphi|2\rangle^{\prime}\right)\,,$
(1) $\displaystyle|D_{2}\rangle$ $\displaystyle=$
$\displaystyle\varepsilon\left(\cos\varphi|1\rangle^{\prime}+\sin\varphi|2\rangle^{\prime}\right)-\sqrt{1-\varepsilon^{2}}|3\rangle,$
(2)
with $|1\rangle^{\prime}=|1\rangle e^{i\kappa_{0}(z+x)}$ and
$|2\rangle^{\prime}=|2\rangle e^{i\kappa_{0}(z-x)}$, where
$\varepsilon=|\Omega_{3}|/\Omega\,,\quad\varphi=\arctan(|\Omega_{1}|/|\Omega_{2}|)$
(3)
characterize the relative intensities of the laser beams. The dark states
$|D_{j}\rangle\equiv|D_{j}(\mathbf{r})\rangle$ are position-dependent due to
the spatial variation of the Rabi frequencies $\Omega_{j}(\mathbf{r})$.
Let us adiabatically eliminate the bright states, so that the atom evolves
within the dark-state manifold. The full atomic state vector can then be
expanded as
$|\Psi(\mathbf{r},t)\rangle=\sum_{n=1}^{2}\chi_{n}(\mathbf{r},t)|D_{n}(\mathbf{r})\rangle$,
where $\chi_{n}(\mathbf{r},t)$ describes the motion of an atom in the dark
state $|D_{n}(\mathbf{r})\rangle$. The atomic center of mass motion is thus
represented by a two-component wavefunction $\chi=(\chi_{1},\chi_{2})^{T}$
obeying (ruseckas-2005-95, )
$i\hbar\frac{\partial}{\partial
t}\chi=\left[\frac{1}{2M}(-i\hbar\nabla-\mathbf{A})^{2}+U\right]\chi\,,$ (4)
where $\mathbf{A}$ is the effective vector potential (ruseckas-2005-95, ;
berry84, ; wilczek84, ; mead91, ) representing a $2\times 2$ matrix whose
elements are vectors, $\mathbf{A}_{n,m}=i\hbar\langle D_{n}(\mathbf{r})|\nabla
D_{m}(\mathbf{r})\rangle$. The particular light field configuration we have
chosen yields
$\displaystyle\mathbf{A}_{11}$ $\displaystyle=$
$\displaystyle-\hbar\kappa_{0}(\mathbf{e}_{z}-\cos(2\varphi)\mathbf{e}_{x}),$
(5) $\displaystyle\mathbf{A}_{12}$ $\displaystyle=$
$\displaystyle-\hbar\varepsilon(\kappa_{0}\sin(2\varphi)\mathbf{e}_{x}+i\nabla\varphi),$
(6) $\displaystyle\mathbf{A}_{22}$ $\displaystyle=$
$\displaystyle-\hbar\kappa_{0}\varepsilon^{2}(\mathbf{e}_{z}+\cos(2\varphi)\mathbf{e}_{x}),$
(7)
with $\mathbf{e}_{x}$ and $\mathbf{e}_{z}$ the unit Cartesian vectors. The
$2\times 2$ matrix $U$ with elements $U_{nm}=(\hbar^{2}/2M)\langle
D_{n}(\mathbf{r})|\nabla B(\mathbf{r})\rangle\langle B(\mathbf{r})|\nabla
D_{m}(\mathbf{r})\rangle$ is an effective scalar potential; both $\mathbf{A}$
and $U$ arise due to the spatial dependence of the atomic dark states.
Suppose the incident atom has a velocity $\mathbf{v}$ much greater than the
recoil velocity $v_{\mathrm{rec}}=\hbar\kappa_{0}/M\approx 0.5\mbox{cm/s}$ for
87Rb. In this limit, the laser beams do not significantly change the atom’s
velocity, permitting a simplified semiclassical approach with no reflected
waves. We apply a gauge transformation
$\chi(\mathbf{r},t)=e^{iM\mathbf{v}\cdot\mathbf{r}/\hbar-
iM\mathbf{v}^{2}t/2\hbar}\tilde{\chi}(\mathbf{r},t)$, implying transition to a
reference frame moving with velocity $\mathbf{v}$, where the two-component
envelope function $\tilde{\chi}$ varies slowly with $\mathbf{r}$ over the
atom’s wavelength $\lambda=h/(Mv)$. Keeping only terms containing $\mathbf{v}$
(or its time derivatives), we arrive at the following approximate equation for
$\tilde{\chi}$:
$i\hbar\left(\partial/\partial
t+\mathbf{v}\cdot\nabla\right)\tilde{\chi}(\mathbf{r},t)=-\mathbf{v}\cdot\mathbf{A}(\mathbf{r})\tilde{\chi}(\mathbf{r},t).$
(8)
As the omitted scalar potential $U$ and the $A^{2}$ term are of the order of
the recoil energy $\hbar\omega_{\mathrm{rec}}=\hbar^{2}\kappa_{0}^{2}/2M\ll
Mv^{2}/2$, the fast moving atoms will not feel these potentials. For incident
velocities $v$ of the order of $v_{\mathrm{rec}}$ or smaller, the atomic
motion will undergo a _Zitterbewegung_ (jayzb, ; patrikzb, ) which is beyond
the scope of the present study. While the atoms must move much faster than the
recoil velocity, they should also be slow enough to avoid coupling to the
bright states. We provide a quantitative analysis of these limitations near
the end of the Letter.
In both of the DDT schemes to be presented, the operator
$\mathbf{v}\cdot\mathbf{A}$ commutes with itself at different times. Going to
a moving frame of reference $\mathbf{r}^{\prime}=\mathbf{r}-\mathbf{v}t$, we
can thus relate the wavefunction $\tilde{\chi}$ at time $t=t_{f}$ to the
wavefunction at a previous time $t=t_{i}$ through
$\tilde{\chi}(\mathbf{r}^{\prime},t_{f})=\exp(i\Theta)\tilde{\chi}(\mathbf{r}^{\prime},t_{i})\,.$
(9)
The $2\times 2$ Hermitian matrix
$\Theta=-\hbar^{-1}\int_{t_{i}}^{t_{f}}\mathbf{A}(\mathbf{r}^{\prime}+\mathbf{v}t)\cdot\mathbf{v}dt$
describes the evolution of the internal state of the atom as it traverses the
path from $\mathbf{r}_{i}=\mathbf{r}^{\prime}+\mathbf{v}t_{i}$ to
$\mathbf{r}_{f}=\mathbf{r}^{\prime}+\mathbf{v}t_{f}$,
$\Theta=-\frac{1}{\hbar}\int_{\mathbf{r}_{i}}^{\mathbf{r}_{f}}\mathbf{A}(\mathbf{r})\cdot
d\mathbf{r}\,.$ (10)
Our subsequent analysis of the atomic dynamics will center on Eqs. (9)-(10)
and (5)-(7).
## Atomic analogs of the DDT
We first consider the setup depicted in Figs. 1c and 2a. The atoms are
incident along the $y$ axis, along which laser beams $1$ and $2$ are
relatively shifted (unanyan1998rca, ; theuer1999nlc, ; unanyan99pra, ;
lietfiz-2007, ), so that
$A_{y}=\hbar\sigma_{y}\varepsilon(y)\partial\varphi(y)/\partial y\,.$ (11)
Equations (11) and (10) yield
$\Theta=\alpha\sigma_{y}\,,\qquad\alpha=-\int_{y_{i}}^{y_{f}}\varepsilon(y)\frac{\partial}{\partial
y}\varphi(y)dy\,,$ (12)
where $\alpha$ is the mixing angle, $\sigma_{y}$ (or $\sigma_{x}$) being the
usual Pauli matrix. By taking the initial and final times sufficiently large,
we have $y_{i}\rightarrow-\infty$ and $y_{f}\rightarrow+\infty$.
Figure 2: Schematics of the first (a) and second (b) setups for an atomic
transistor: The atom, along its trajectories (shown in Figs. 1c,d) sees the
above profile of laser fields.
As Figs. 1c and 2a show, the first laser beam dominates as the atom enters the
gate region, while the second dominates as it exits the region. In between,
the atom also feels the third beam. This configuration results in a gate-
induced rotation of the atom’s internal state by a mixing angle $\alpha$.
Specifically, suppose the atom enters the gate region in the internal state
$|3\rangle=-|D_{2}(\mathbf{r}^{\prime},t_{i})\rangle$, with center of mass
wave-function $\Phi(\mathbf{r}^{\prime})$. The atom then exits the gate region
in the rotated state
$\tilde{\chi}(\mathbf{r}^{\prime},t_{f})=-\Phi(\mathbf{r}^{\prime})\left(\begin{array}[]{c}\sin\alpha\\\
\cos\alpha\end{array}\right).$ (13)
Thus, the probability for the atom to emerge in the second dark state is
$\cos^{2}\alpha$. Note that the second dark state coincides with the third
internal ground state upon exit:
$|D_{2}(\mathbf{r}^{\prime},t_{f})\rangle=-|3\rangle$. This gate-controlled
state rotation is an atomic analog of the action of the DDT. Define
$\eta=|\Omega_{3}|/|\Omega_{1}|$ as the relative amplitude of the third laser
at the central point. The specific relation between $\alpha$ and $\eta$
depends on the particular choice of light field configuration and is readily
derived from Eqs. (3) and (12). For arbitrary light field configurations,
$\alpha$ is a complicated space-dependent function. However for the particular
laser configuration we examine here, $\alpha$ simplifies to a function solely
depending on $\eta$, and $\eta$ controls $\alpha$. Fig. 3 shows the dependence
of $\alpha$ on $\eta$ for Gaussian laser beams. As in the electronic DDT, the
transmission coefficient $\cos\alpha$ is independent of the velocity of the
incident atoms, so that the transistor properties are robust to a spread in
atomic velocities. We estimate the regime of validity of this independence
near the end of the Letter.
Figure 3: The mixing angle $\alpha$ vs. the relative amplitude of the third
field $\eta$ for the first (solid line) and the second (dashed line) setups.
The amplitudes of the beams are Gaussian:
$|\Omega_{1}|=a\exp(-(u+\delta)^{2}/w_{1}^{2})$,
$|\Omega_{2}|=a\exp(-(u-\delta)^{2}/w_{2}^{2})$, and
$|\Omega_{3}|=a\eta\exp(-u^{2}/w_{3}^{2}-\delta^{2}/w_{1}^{2})$, with $u=y$
(first setup) or $u=x$ (second setup). In the first setup,
$w_{1}=w_{2}=w_{3}=\delta=2\lambda$, with $\lambda=600\,\mathrm{nm}$ being the
laser wave length. For the second setup, all the beams are centered at the
same point ($\delta=0$) and have the widths $w_{1}=w_{2}=10w_{3}=20\lambda$.
Since $\varepsilon(y)\leq 1$, the mixing angle given by Eq. (12) ranges from
$0$ to $\pi/2$, and the sensitivity $|\Delta\alpha|/|\Delta\eta|$ of the DDT
is on the order of unity. Small changes in the relative Rabi frequency $\eta$
will thus lead to small changes in the mixing angle:
$|\Delta\alpha|\sim|\Delta\eta|$. We next analyze an alternative setup which
enables us to create a more sensitive DDT.
Now suppose that the first two light beams counterpropagate along the $x$ axis
with equal intensities (Fig. 1d), i.e., $\varphi=\pi/4$ in Eqs. (5)-(7) for
$\mathbf{A}$. After the trivial gauge transformation
$\exp[i\hbar\kappa_{0}(1+\varepsilon^{2})z\mathbf{I}]$, the light-induced
vector potential resembles the Rashba spin orbit coupling which is the spin
rotation mechanism of the electronic DDT:
$\displaystyle A_{z}$ $\displaystyle=$
$\displaystyle-\frac{\hbar\kappa_{0}}{2}(1-\varepsilon^{2})\sigma_{z}$ (14)
$\displaystyle A_{x}$ $\displaystyle=$
$\displaystyle-\hbar\kappa_{0}\varepsilon\sigma_{x},\qquad A_{y}=0.$ (15)
The atomic beam crosses the lasers at an angle in the $x-y$ plane, with
initial velocity components $v_{x}\neq 0$ and $v_{y}$. Although the atomic
motion in the $y$ direction does not affect the internal state rotation
$(A_{y}=0)$ , sending the beam in at an angle removes the experimental
difficulty of having the atoms incident from the same direction as the laser
beams. Along its trajectory, the atom feels the laser beam profile illustrated
in Fig. 2b. The evolution matrix of Eq. (10) is then
$\Theta=\alpha\sigma_{x}\,,\qquad\alpha=\kappa_{0}\int_{x_{i}}^{x_{f}}\varepsilon(x)dx\,.$
(16)
Initial and final times are taken sufficiently large that the spatial
integration runs from $x_{i}=-\infty$ to $x_{f}=+\infty$.
As in the previous scheme, the intensity of the third laser vanishes
($\varepsilon\rightarrow+0$) outside the gate region (see Fig. 2b). Only the
third laser’s intensity has significant spatial dependence inside the gate
region, the intensities of the first two lasers being nearly constant there.
In both setups, the controlled state rotation arises from the spatial
dependence of the beams in the gate region. In the first setup the variation
is in the lasers’ relative intensities. Contrastingly, in the second setup,
the intensities of the first two lasers are constant in the gate region, so
the controlled state rotation is driven by only the relative _phases_ of the
counterpropagating laser beams.
As in the previous setup, the atom enters the gate region in the internal
state $|3\rangle=-|D_{2}(\mathbf{r}^{\prime},t_{i})\rangle$ and with center of
mass wave-function $\Phi(\mathbf{r}^{\prime})$. The atom exits in the rotated
state
$\tilde{\chi}(\mathbf{r}^{\prime},t_{f})=-\Phi(\mathbf{r}^{\prime})\left(\begin{array}[]{c}i\sin\alpha\\\
\cos\alpha\end{array}\right),$ (17)
where the mixing angle $\alpha$ is controlled by the variation of the relative
intensity of the third laser beam.
To estimate the mixing angle, suppose that $\Omega_{3}$, and hence
$\varepsilon$, do not change significantly in the gate region. Equation (16)
then gives $\alpha=\kappa_{0}\bar{\varepsilon}L$, where $L$ (see Fig. 2b) is
the length of the area in which the third laser has the strongest intensity.
Note that the mixing angle is now proportional to $L$, as well as to the
average strength $\kappa_{0}\bar{\varepsilon}$ of the spin-orbit coupling.
This behavior is in direct analogy to the electronic DDT (dattadas, ). As in
Eq. (2) of (dattadas, ), the output power of the atoms in the internal state
$|3\rangle$ is $P=\cos^{2}\alpha=\cos^{2}(\kappa_{0}\bar{\varepsilon}L)$.
Using this atomic setup, $\alpha=\kappa_{0}\bar{\varepsilon}L$ can be much
larger than $\pi/2$, provided $L\gg(\kappa_{0}\bar{\varepsilon})^{-1}$, as
shown in Fig. 3. Small changes in the relative amplitude of the third laser
$\eta=|\Omega_{3}|/|\Omega_{1}|$ can therefore yield substantial changes in
the mixing angle: $|\Delta\alpha|\sim|\Delta\eta|\kappa_{0}L$. The sensitivity
of such a DDT, $|\Delta\alpha|/|\Delta\eta|\sim\kappa_{0}L$, can far exceed
unity if $L$ is much greater than the optical wave-length
$\lambda=2\pi/\kappa_{0}$.
Let us estimate the range of atomic beam velocities for which our
approximations are valid. The atom crosses the gate region in a time
$\tau=L/v$. Due to nonadiabatic coupling to the bright states, the dark state
atoms have the finite lifetime $\tau_{D}=\Omega^{2}/\gamma\Delta\omega^{2}$
(Juz06PRA, ), where $\gamma$ is the excited state decay rate and
$\Delta\omega=v\partial\varphi/\partial y\sim v\pi/L$ (first setup), or
$\Delta\omega=v\kappa_{0}$ (second setup). The frequency shift $\Delta\omega$
represents the two-photon detuning due to the finite time of the atom-light
interaction (first setup) or the two-photon Doppler shift (second setup). To
avoid decay, we require the beam to be in the adiabatic limit, i.e.
$\tau/\tau_{D}\ll 1$. Taking $\Omega=2\pi\times 10^{7}\,\mathrm{Hz}$ (Hau99,
), $\gamma=10^{7}\,\mathrm{s}^{-1}$, $\kappa_{0}=2\pi/\lambda$,
$\lambda=600\,\mathrm{nm}$ and $L=4\lambda$, we require atomic velocities
$v\ll 100\,\mathrm{m/s}$ for the first setup and $v\ll 1\,\mathrm{m/s}$ for
the second setup. The increased sensitivity in the second scheme thus comes at
the expense of increased non-adiabatic losses.
Ultracold atoms are highly tunable and controllable, and can thus serve as
quantum simulators for a variety of other systems, including systems which
have yet to be experimentally accessed in their original manifestations. In
this Letter, we have identified an atomic analog of one such system, the spin
field-effect transistor. Our atomic transistors, like their electronic
counterpart, provide controllable state manipulation that is relatively
insensitive to the thermal spread of beam velocities. The devices we have
proposed are based on the familiar tripod STIRAP configuration, and appear to
be feasible within current experimental procedures.
## References
* (1) S. Datta and B. Das, Appl. Phys. Lett. 56, 665 (1990).
* (2) I. Zutic, J. Fabian, and S. Das Sarma, Rev. Mod. Phys. 76, 323 (2004).
* (3) A. Ruschhaupt, J. G. Muga, and M. G. Raizen, J. Phys. B: At. Mol. Opt. Phys. 39, L133 (2006).
* (4) B. T. Seaman, M. Krämer, D. Z. Anderson, and M. J. Holland, Phys. Rev. A 75, 023615 (2007).
* (5) J. J. Thorn, E. A. Schoene, T. Li, and D. A. Steck, Phys. Rev. Lett. 100, 240407 (2008).
* (6) R. G. Unanyan, M. Fleischhauer, B. W. Shore, and K. Bergmann, Opt. Commun. 155, 144 (1998).
* (7) R. G. Unanyan, B. W. Shore, and K. Bergmann, Phys. Rev. A 59, 2910 (1999).
* (8) H. Theuer et al., Opt. Express 4, 77 (1999).
* (9) F. Vewinger et al., Phys. Rev. Lett. 91, 213001 (2003).
* (10) J. Ruseckas, G. Juzeliūnas, P. Öhberg, and M. Fleischhauer, Phys. Rev. Lett. 95, 010404 (2005).
* (11) G. Juzeliūnas, J. Ruseckas, P. Öhberg, and M. Fleischhauer, Lithuanian J. Phys 47, 351 (2007).
* (12) Y.-J. Lin et al., arXiv:0809.2976 (2008).
* (13) E. I. Rashba, Sov. Phys. Sol. St. 2, 1224 (1960).
* (14) T. D. Stanescu, C. Zhang, and V. Galitski, Phys. Rev. Lett. 99, 110403 (2007).
* (15) A. Jacob, P. Öhberg, G. Juzeliūnas, and L. Santos, Appl. Phys. B 89, 439 (2007).
* (16) J. Y. Vaishnav and C. W. Clark, Phys. Rev. Lett. 100, 153002 (2008).
* (17) G. Juzeliūnas et al., Phys. Rev. A 77, 011802(R) (2008).
* (18) G. Juzeliūnas et al., Phys. Rev. Lett. 100, 200405 (2008).
* (19) M. Edwards et al., J. Phys. B: At. Mol. Opt. Phys. 32, 2935 (1999).
* (20) L. M. Duan, J. I. Cirac, and P. Zoller, Science 292, 1695 (2001).
* (21) Z. Kis and F. Renzoni, Phys. Rev. A 65, 032318 (2002).
* (22) R. G. Unanyan and M. Fleischhauer, Phys. Rev. A 69, 050302(R) (2004).
* (23) S. Rebić et al., Phys. Rev. A 70, 032317 (2004).
* (24) M. V. Berry, Proc. R. Soc. A 392, 45 (1984).
* (25) F. Wilczek and A. Zee, Phys. Rev. Lett. 52, 2111 (1984).
* (26) C. A. Mead, Rev. Mod. Phys. 64, 51 (1992).
* (27) M. Merkl, F. E. Zimmer, G. Juzeliūnas, and P. Öhberg, Europhys. Lett. 83, 54002 (2008).
* (28) G. Juzeliūnas, J. Ruseckas, P. Öhberg, and M. Fleischhauer, Phys. Rev. A 73, 025602 (2006).
* (29) L. Hau, S. E. Harris, Z. Dutton, and C. Behrooz, Nature 397, 594 (1999).
| arxiv-papers | 2008-07-20T16:00:27 | 2024-09-04T02:48:56.896832 | {
"license": "Public Domain",
"authors": "J. Y. Vaishnav, Julius Ruseckas, Charles W. Clark, Gediminas\n Juzeliunas",
"submitter": "J. Y. Vaishnav",
"url": "https://arxiv.org/abs/0807.3067"
} |
0807.3138 | # Ground-state properties of gapped graphene using the random phase
approximation
Alireza Qaiumzadeh Institute for Advanced Studies in Basic Sciences (IASBS),
Zanjan, 45195-1159, Iran School of physics, Institute for research in
fundamental sciences, IPM, 19395-5531 Tehran, Iran Reza Asgari School of
physics, Institute for research in fundamental sciences, IPM, 19395-5531
Tehran, Iran
###### Abstract
We study the effect of band gap on the ground-state properties of Dirac
electrons in a doped graphene within the random phase approximation at zero
temperature. Band gap dependence of the exchange, correlation and ground-state
energies and the compressibility are calculated. In addition, we show that the
conductance in the gapped graphene is smaller than gapless one. We also
calculate the band gap dependence of charge compressibility and it decreases
with increasing the band gap values.
###### pacs:
73.63.-b, 72.10.-d, 71.10.-w, 73.50.Fq
## I Introduction
Graphene is a flat monolayer of carbon atoms tightly packed into a two-
dimensional (2D) honeycomb lattice and it is a basic building block for all
nanostructured carbon. This stable structure has attracted considerable
attention because of experimental progress novoselov and because of exotic
chiral feature in its electronic properties and promising applications geim .
Very recent experiments on both a suspended graphene and a graphene on
substrate have found remarkably high mobility $2\times 10^{5}$ cm${}^{2}/$Vs
for carrier transport mobility at room temperature which is two order of
magnitude higher than the mobility of silicon wafer used in microprocessors.
dean
An interesting feature of graphene which makes it very applicable in
semiconductor technology is the opening a gap in the band energy structure of
graphene. There are several scenarios to open a gap in the band energy
structure of graphene. One is the finite size effect by using graphene
nanoribbons, where gaps give rise to constriction of the electrons in the
ribbon and it depends on the detailed structure of ribbon edges. son More
precisely, the band gaps with armchair shaped edges originate from quantum
confinement and the value of the gap depends on the width of the ribbon. For
zigzag shaped edges, on the other hand, the band gaps arise from a staggered
sublattice potential due to magnetization at the edges. abanin Importantly,
edge effects plays a crucial role in transport properties. The gap engineering
can be also achieved through doping the graphene with chemical species ohta
due to the translational symmetry breaking. The electronic properties of a
graphene interacting with CrO3 molecules has been calculated by using ab
initio calculations. zanella This type of calculations, predicts opening a
gap about $0.12$ eV at the Dirac point. Another scenario is graphene by
placing it on top of an appropriate substrate which breaks the graphene
sublattice symmetry and, therefore generates an intrinsic Dirac mass for the
charge carriers. substarte Typical substrate is made of hexagonal $\rm SiC$
with a gap about $0.26$ eV. A recent band structure calculation for a graphene
on top of a hexagonal boron nitride crystal giovannetti has been shown a band
gap about $53$ meV. The gap can also be generated dynamically by applying a
magnetic field. gusynin Moreover, when both mono-and bilayer graphene
material are covered with water and ammonia molecules, a gap induce in the
spectrum of energy. ribeiro Interestingly, the mechanism that electrons
hopping on a honeycomb lattice with textured tight-binding hopping amplitudes,
the Kekulè texture, generates a Dirac gap. hou Eventually, It has been
suggested that a small gap can be opened on the Dirac points due to spin-orbit
coupling or Rashba effect yao which makes the system a spin Hall insulator
with quantized spin Hall conductances. murakami
Recently, the local compressibility of graphene has been measured martin
using a scannable single electron transistor. The measured compressibility is
claimed to be well described by the kinetic energy contribution and it is also
suggested that exchange and correlation effects have canceling contributions.
From the theoretical point of view, the compressibility was first calculated
by Peres et al. peres considering the exchange contribution to a
noninteracting doped or undoped graphene flake. A related quantity
$\partial\mu/\partial n$ (where $\mu$ is the chemical potential and $n$ is the
electron density) is recently considered by Hwang et al.hwang_dmu within the
same approximation and they stated that correlations and disorder effects
would introduce only small corrections. This statement is only true in quite
large density doped values. To go beyond the exchange contribution, the
correlation effects were taken into account by Barlas et al. yafis based on
an evaluation of graphene’s exchange and random phase approximation (RPA)
correlation energies. Moreover, Sheehy and Schmalian sheehy by exploiting the
proximity to relativistic electron quantum critical point, derived explicit
expressions for the temperature and density dependence of the compressibility
properties of graphene. Importantly, the effect of disorder and many-body
interactions on the compressibility has been recently studied by us. asgari
We successfully demonstrated the importance of including correlation effects
together with disorder effects in the thermodynamic quantities. It should be
noticed that all these theoretical efforts have been carried out for a gapless
graphene.
Our aim in this work is to study the ground-state properties in the presence
of Dirac gap and electron-electron interactions. For this purpose, we derive
the gap dependence of the dynamic polarization function for a doped graphene
to calculate the scattering rate, ground-state energies and the
compressibility of the system at the level of RPA including the opening gap at
Dirac point.
We consider different on-site energies for atoms in two sublattices in
graphene which is established experimentally to be important when an
appropriate substrate such a boron nitride or ${\rm SiC}$ is used. The
compressibility decreases by increasing the band gap values due to the
sublattice symmetry breaking.
The rest of this paper is organized as follows. In Sec. II, we introduce the
models for dynamic polarization function and ground-state energy calculations
. We then outline the calculation of d.c conductivity and compressibility.
Section III contains our numerical calculations of ground state properties. We
conclude in Sec. IV with a brief summary.
## II Theoretical Approach
We consider a Dirac-like electron in a continuum model interacting via a
Coulomb potential $e^{2}/\epsilon r$ and its Fourier transform $v_{q}=2\pi
e^{2}/(\epsilon q)$ where $\epsilon$ is the average background dielectric
constant (for instance, $\epsilon\simeq 5.5$ for graphene placed on ${\rm
SiC}$ with the other side being exposed to air) having an isotropic band gap
at Dirac points. If one assumes that the sublattice symmetry is broken and
$\alpha_{a}$, $\alpha_{b}$ are on-site energies of atoms $A$ and $B$,
respectively, then the contribution of the on-site energies in Hamiltonian of
graphene can be written gusynin as
$\displaystyle\hat{H}_{1}$ $\displaystyle=$ $\displaystyle\sum_{{\bf
k},\sigma}[\alpha_{a}a_{\sigma}^{\dagger}({\bf k})a_{\sigma}({\bf
k})+\alpha_{b}b_{\sigma}^{\dagger}({\bf k})b_{\sigma}({\bf k})]$ (1)
$\displaystyle=$ $\displaystyle\sum_{{\bf k},\sigma}\hat{\Psi}^{\dagger}_{{\bf
k},\sigma}[\alpha_{+}\tau^{0}-\alpha_{-}\tau^{3}]\otimes\sigma^{3}\hat{\Psi}_{{\bf
k},\sigma}$
where $\alpha_{+}=(\alpha_{a}+\alpha_{b})/2$ that corresponds to the same
carrier density on two sublattices, $\alpha_{-}=(\alpha_{a}-\alpha_{b})/2$ is
carrier imbalance on two sublattices that leads to break the inversion
symmetry and $\tau^{0}$ is $2\times 2$ unit matrix, $\tau^{3}$ is a Pauli
matrix that acts on $K_{+}$ and $K_{-}$ two-degenerate valleys at which $\pi$
and $\pi^{*}$ bands touch and $\sigma^{3}$ is Pauli matrix that act on
graphene’s pseudospin degrees of freedom. Consequently, the noninteracting
Hamiltonian for a gapped graphene is given by $\hat{H}_{0}=\sum_{{\bf
k},\sigma}\Psi^{\dagger}_{{\bf k},\sigma}\mathcal{\hat{H}}_{0}\Psi_{{\bf
k},\sigma}$ where
$\displaystyle\mathcal{\hat{H}}_{0}=\left(\begin{array}[]{cccc}\Delta&\hbar
v\hat{k}^{*}&0&0\\\ \hbar v\hat{k}&-\Delta&0&0\\\ 0&0&-\Delta&-\hbar
v\hat{k}^{*}\\\ 0&0&-\hbar v\hat{k}&\Delta\\\ \end{array}\right)$ (6)
where $\hat{k}=k_{x}+ik_{y}$ and $\sigma$ is the spin of charge carrier. Here,
$v=3ta/2$ is the Fermi velocity, $t$ is the tight-binding hopping integral,
$a$ is the spacing of the honeycomb lattice. For the hexagonal crystal
structure of graphene, $a=1.42$Å is the carbon-carbon distance, the tight-
bonding hopping energy is $t=2.8$ eV and the bare Fermi velocity is $v=10^{6}$
$m/s$. In the noninteracting Hamiltonian, $\hat{H}_{0}$ the reference energy
$(\alpha_{a}+\alpha_{b})\tau_{0}/2$ is subtracted and the energy gap is
defined as $2\Delta=(\alpha_{b}-\alpha_{a})$ where we expect $\Delta<t$. The
corresponding four components pseudospinor of the noninteracting Hamiltonian
is $\Psi^{\dagger}_{{\bf
k},\sigma}=\left(\begin{array}[]{cccc}\psi^{b}_{+,\sigma},\psi^{a}_{+,\sigma},\psi^{a}_{-,\sigma},\psi^{b}_{-,\sigma}\end{array}\right)$.
It is easy to diagonalize the noninteracting Hamiltonian based on
pseudospinors in the conduction and valance band of energies with eigenvalues
given by $\pm\sqrt{\hbar^{2}v^{2}k^{2}+\Delta^{2}}$. Importantly, the low
energy quasiparticle excitations in a gapless graphene are linearly dispersing
and it is valid for energy less than 1 eV. Accordingly, the validity of the
noninteracting Hamiltonian given by Eq. (6) to explore graphene properties is
to the case which $\sqrt{\hbar^{2}v^{2}k^{2}+\Delta^{2}}<1eV$. On the other
hand, we shall achieve to a conventional two-dimensional electron gas system
by setting $\hbar vk/\Delta\ll 1$.
Finally, the total Hamiltonian including the electron-electron repulsion
interaction is given by
$\hat{H}=\hat{H}_{0}+\frac{1}{2S}\sum_{{\bf q}\neq 0}v_{q}({\hat{n}}_{\bf
q}{\hat{n}}_{-{\bf q}}-{\hat{N}}),$ (7)
where $S$ is the sample area and ${\hat{N}}$ is the total number operator. The
presence of a neutralizing background of positive charge is explicit in Eq.
(7). As we mentioned in introduction section, this kind of Hamiltonian can be
used in graphene by placing it on top of an appropriate substrate that breaks
the graphene sublattice symmetry and generates an intrinsic Dirac gap.
A central quantity in the theoretical formulation of the many-body effects in
Dirac fermions is the noninteracting dynamical polarizability function yafis ;
hwang ; others $\chi^{(0)}({\bf q},i\Omega,\mu\geq\Delta)$ where $\mu$ is
chemical potential. Here, we would like to emphasize that we have calculated
the gap dependence of the noninteracting polarization function for doped
graphene however the vacuum polarization function in which $\mu=\Delta$ has
been calculated by Kotov et al kotov . They studied the distribution of
polarization charge induced by a Coulomb impurity for undoped graphene.
However, we would like to study the ground-state properties for doped graphene
sheets. To achieve this goal, we write the dynamical polarizability function
in terms of one-body noninteracting Green’s function
$\chi^{(0)}({\bf q},\Omega,\mu)=-i\int\frac{d^{2}{\bf
k}}{(2\pi)^{2}}\int\frac{d\omega}{2\pi}{\rm Tr}[i\gamma_{0}G^{(0)}({\bf
k}+{\bf q},\omega+\Omega,\mu)i\gamma_{0}G^{(0)}({\bf k},\omega,\mu)]~{},$ (8)
where one-body noninteracting Green’s function chin by using the
noninteracting Hamiltonian is given by
$G^{(0)}({\bf k},\omega,\mu)=i\frac{-\gamma_{0}\omega+\hbar v{\bf\gamma\cdot
k}+i\Delta}{-\omega^{2}+\hbar^{2}v^{2}k^{2}+\Delta^{2}-i\eta}-\pi\frac{-\gamma_{0}\omega+\hbar
v{\bf\gamma\cdot
k}+i\Delta}{\sqrt{\hbar^{2}v^{2}k^{2}+\Delta^{2}}}\delta(\hbar\omega-\sqrt{\hbar^{2}v^{2}k^{2}+\Delta^{2}})\theta(k-{\rm
k_{F}})~{},$ (9)
in which $\gamma$-matrices are related to Pauli matrices by
$\sigma^{3}=-i\gamma_{0}$ and $\sigma^{j}=(-1)^{j}\sigma^{3}\gamma_{j}$ for
$j=1,2$ and ${\rm k_{F}}$ is the Fermi momentum related to the density of
electron as given by ${\rm k_{F}}=(4\pi n/g)^{1/2}$. $g=g_{v}~{}g_{s}=4$ is
valley and spin degeneracy and $\theta$ is the Heaviside step function. The
chemical potential is given by $\mu=\sqrt{\hbar^{2}v^{2}{\rm
k_{F}}^{2}+\Delta^{2}}$ at zero temperature. After implementing $G^{(0)}({\bf
k},\omega,\mu)$ in Eq. (8) and calculating the traces and integrals, the
result is given by the follow expression
$\displaystyle\chi^{(0)}({\bf q},i\omega,\mu)$ $\displaystyle=$
$\displaystyle-\frac{g}{2\pi
v^{2}}\\{\mu-\Delta+\frac{\varepsilon_{q}^{2}}{2}\left[\frac{\Delta}{{\varepsilon_{q}^{2}+\hbar^{2}\omega^{2}}}+\frac{1}{2\sqrt{\varepsilon_{q}^{2}+\hbar^{2}\omega^{2}}}(1-\frac{4\Delta^{2}}{\varepsilon_{q}^{2}+\hbar^{2}\omega^{2}})\tan^{-1}(\frac{\sqrt{\varepsilon_{q}^{2}+\hbar^{2}\omega^{2}}}{2\Delta})\right]$
(10) $\displaystyle-$
$\displaystyle\frac{\varepsilon_{q}^{2}}{4\sqrt{\hbar^{2}\omega^{2}+\varepsilon_{q}^{2}}}\Re
e\left[(1-\frac{4\Delta^{2}}{\varepsilon_{q}^{2}+\hbar^{2}\omega^{2}})\\{\sin^{-1}(\frac{2\mu+i\hbar\omega}{\varepsilon_{q}\sqrt{1+\frac{4\Delta^{2}}{\varepsilon_{q}^{2}+\hbar^{2}\omega^{2}}}})-\sin^{-1}(\frac{2\Delta+i\hbar\omega}{\varepsilon_{q}\sqrt{1+\frac{4\Delta^{2}}{\varepsilon_{q}^{2}+\hbar^{2}\omega^{2}}}})\\}\right]$
$\displaystyle-$
$\displaystyle\frac{\varepsilon_{q}^{2}}{4\sqrt{\hbar^{2}\omega^{2}+\varepsilon_{q}^{2}}}\Re
e\left[(\frac{2\mu+i\hbar\omega}{\varepsilon_{q}})\sqrt{(1+\frac{4\Delta^{2}}{\varepsilon_{q}^{2}+\hbar^{2}\omega^{2}})-(\frac{2\mu+i\hbar\omega}{\varepsilon_{q}})^{2}}\right]$
$\displaystyle+$
$\displaystyle\frac{\varepsilon_{q}^{2}}{4\sqrt{\hbar\omega^{2}+\varepsilon_{q}^{2}}}\Re
e\left[(\frac{2\Delta+i\hbar\omega}{\varepsilon_{q}})\sqrt{(1+\frac{4\Delta^{2}}{\varepsilon_{q}^{2}+\hbar^{2}\omega^{2}})-(\frac{2\Delta+i\hbar\omega}{\varepsilon_{q}})^{2}}\right]\\}~{},$
where $\varepsilon_{q}=\hbar vq$. By setting $\Delta=0$, it is easy to
determine that Eq. (10) reduces to the noninteracting dynamic polarization
function of the gapless graphene sheet. yafis Furthermore, for the half-filed
gapped graphene sheet, the noninteracting dynamic polarization function,
vacuum polarization, is given by kotov
$\chi^{(0)}({\bf
q},i\omega,\mu=\Delta)=-g\frac{\varepsilon_{q}^{2}}{4v^{2}\pi}\left[\frac{\Delta}{{\varepsilon_{q}^{2}+\hbar^{2}\omega^{2}}}+\frac{1}{2\sqrt{\varepsilon_{q}^{2}+\hbar^{2}\omega^{2}}}(1-\frac{4\Delta^{2}}{\varepsilon_{q}^{2}+\hbar^{2}\omega^{2}})\tan^{-1}(\frac{\sqrt{\varepsilon_{q}^{2}+\hbar^{2}\omega^{2}}}{2\Delta})\right]~{}.$
(11)
Using the above results for the noninteracting polarization function on the
imaginary frequency axis, the density of state at Fermi energy is calculated
as
$D(\varepsilon_{F})=D^{0}(\varepsilon_{F})\left[(1+\Delta^{2}/\varepsilon_{\rm
F}^{2})^{1/2}\right]\theta(\varepsilon_{F})~{},$ (12)
where $D^{0}(\varepsilon_{F})=g\varepsilon_{\rm F}/2\pi\hbar v^{2}$ is the
density of states of gapless graphene. hwang Note that we define
$\varepsilon_{\rm F}=\hbar v{\rm k_{F}}$. The linear correction of expanded
gapped polarization function for $\Delta$ is zero, however the quadratic
correction is easy obtained
$\chi^{(0)}({q},i\omega,\mu)\simeq\chi^{(0)}({q},i\omega,\mu)\Big{|}_{\Delta=0}-\frac{g}{2\pi
v^{2}}\left[\frac{1}{2}+\frac{\varepsilon_{q}^{2}}{\sqrt{\varepsilon_{q}^{2}+\omega^{2}}}\Re
e\\{\frac{2\sin^{-1}(\frac{2\mu}{\varepsilon_{q}})-\pi}{\varepsilon_{q}^{2}+\omega^{2}}-\frac{\sqrt{\varepsilon_{q}^{2}-(2+i\omega)^{2}}}{4\varepsilon_{q}^{2}}\\}\right]\Delta^{2}+O(\Delta^{3})~{},$
(13)
Where the explicit expression of
$\chi^{(0)}({q},i\omega,\mu)\Big{|}_{\Delta=0}$ is given by our group. yafis
Now, we are in the stage to use the noninteracting polarization function given
by Eq. (10) to calculate some physical quantities.
### II.1 Transport scattering time in a gapped graphene
As a first application of the noninteracting polarization function, we would
like to calculate the gapped graphene transport scattering time by randomly
distributed impurity centers in the relaxation time approximation. adam The
validity of the Born approximation is discussed by Novikov born and here we
use this approximation to calculate qualitatively the graphene transport
scattering time. To this purpose, the transport scattering time is given by
Boltzmann theory,
$\displaystyle\frac{1}{\tau(\varepsilon_{F})}=\frac{2\pi}{\hbar}\sum_{{\bf
q},s,s^{\prime}}n_{i}\frac{<|v_{i}(q)|^{2}>}{\epsilon(q)^{2}}(1-\cos\theta_{{\bf
q},{\bf q+{\rm k_{F}}}})F^{s,s^{\prime}}({\bf q},{\bf q+{\rm
k_{F}}})\delta(s\sqrt{\varepsilon_{\rm
k_{F}}^{2}+\Delta^{2}}-s^{\prime}\sqrt{\varepsilon_{{\bf q+{\rm
k_{F}}}}^{2}+\Delta^{2}})~{},$ (14)
where $v_{i}(q)=\frac{2\pi e^{2}}{\epsilon q}\exp(-qd)$ is the Coulomb
scattering potential between an electron and an out of plane impurity,
$\epsilon(q)$ is the static RPA dielectric function appropriate for graphene,
$\epsilon(q)=1-v_{q}\chi^{(0)}(q,0,\mu)$, $n_{i}$ is the density of impurities
and $d$ is the setback distance from the graphene sheet and $s,s^{\prime}$
being $\pm$. Since we consider large charge carrier density and elastic
scattering, we can therefore neglect interband scattering process.
$F^{\beta}(\bf{q},\bf{q+k}_{F})$ is the overlap of states ($\beta=\pm$), which
can be easily calculated from the pseudospinors of Hamiltonian, Eq. (6). The
result is as follow
$\displaystyle F^{\pm}({\bf q},{\bf
q+k})=\frac{1}{2}\left[1\pm\frac{1}{\sqrt{\varepsilon_{k+q}^{2}+\Delta^{2}}}\\{\sqrt{\varepsilon_{k}^{2}+\Delta^{2}}+\frac{\varepsilon_{q}\varepsilon_{k}cos\phi}{\sqrt{\varepsilon_{k}^{2}+\Delta^{2}}}\\}\right]~{},$
(15)
where $\phi$ is an angle between ${\bf k}$ and ${\bf q}$. Graphene
conductivity can then be calculated by the Boltzmann transport theory with
$\sigma=(e^{2}/h)2\tau(\varepsilon_{\rm F})v{\rm k_{F}}$. The properties of
graphene’s Dirac fermions depends on the dimensionless coupling constant
$\alpha_{gr}=g{e^{2}/\upsilon\epsilon\hbar}$.
### II.2 RPA ground state energy in a gapped graphene
The ground-state energies is calculated by using the coupling constant
integration technique, which has the contributions $E_{tot}=E_{kin}+E_{\rm
x}+E_{\rm c}$. The kinetic energy per particle is easy calculated as
$2\varepsilon_{F}[(1+\Delta^{2}/\varepsilon_{F}^{2})^{3/2}-\Delta^{3}/\varepsilon_{F}^{3}]/3$.
The first-order, exchange contribution per particle is given by
$\varepsilon_{\rm x}=\frac{E_{\rm x}}{N}=\frac{1}{2}\int\frac{d^{2}{\bf
q}}{(2\pi)^{2}}~{}v_{q}\left[-\frac{1}{\pi
n}\int_{0}^{+\infty}d\Omega~{}\chi^{(0)}({\bf q},i\omega,\mu)-1\right]\,.$
(16)
To evaluate the correlation energy in the RPA, we follow a standard strategy
for uniform continuum models Giuliani_and_Vignale
$\varepsilon^{\rm RPA}_{\rm c}=\frac{E_{\rm c}}{N}=\frac{1}{2\pi
n}\int\frac{d^{2}{\bf
q}}{(2\pi)^{2}}\int_{0}^{+\infty}d\omega\left\\{v_{q}\chi^{(0)}({\bf
q},i\omega,\mu)\right.\left.+\ln{\left[1-v_{q}\chi^{(0)}({\bf
q},i\omega,\mu)\right]}\right\\}\,.$ (17)
Since $\chi^{(0)}({\bf q},i\omega,\mu)$ is linearly proportional to ${\bf q}$
at large ${\bf q}$ and decrease only like $\omega^{-1}$ at large $\omega$ in
both gapped and gapless graphene, accordingly the exchange and correlation
energy built by Eqs. (16) and (17) are divergent. yafis ; asgari In order to
improve convergence, it is convenient at this point to add and subtract vacuum
polarization, $\chi^{(0)}({\bf q},i\omega,\mu=\Delta)$, inside the frequency
integral and regularize the exchange and correlation energy. Therefore, these
ultraviolet divergences can be cured calculating
$\delta\varepsilon_{\rm x}=-\frac{1}{2\pi n}\int\frac{d^{2}{\bf
q}}{(2\pi)^{2}}~{}v_{q}\int_{0}^{+\infty}d\omega~{}\delta\chi^{(0)}({\bf
q},i\omega,\mu)$ (18)
and
$\delta\varepsilon^{\rm RPA}_{\rm c}=\frac{1}{2\pi n}\int\frac{d^{2}{\bf
q}}{(2\pi)^{2}}\int_{0}^{+\infty}d\omega\left\\{v_{q}\delta\chi^{(0)}({\bf
q},i\omega,\mu)\right.\left.+\ln{\left[\frac{1-v_{q}\chi^{(0)}({\bf
q},i\omega,\mu)}{1-v_{q}\chi^{(0)}({\bf
q},i\omega,\mu=\Delta)}\right]}\right\\}~{},$ (19)
where $\delta\chi^{(0)}$ is the difference between the doped ($\mu>\Delta$)
and undoped ($\mu=\Delta$) polarization functions. With this regularization,
the $q$ integrals have logarithmic ultraviolet divergences. yafis we can
introduce an ultraviolet cutoff for the wave vector integrals
$k_{c}=\Lambda{\rm k_{F}}$ which is the order of the inverse lattice spacing
and $\Lambda$ is dimensionless quantity. Once the ground-state is obtained the
compressibility $\kappa$ can easily be calculated from
$\kappa^{-1}=n^{2}\frac{\partial^{2}(n\delta\varepsilon_{\rm tot})}{\partial
n^{2}}~{},$ where the total ground-state energy per particle is given by
$\delta\varepsilon_{\rm tot}=\delta\varepsilon_{\rm
kin}+\delta\varepsilon_{\rm x}+\delta\varepsilon^{\rm RPA}_{\rm c}$. The
compressibility of noninteracting gapless graphene is
$\kappa_{0}^{0}=2/(n\varepsilon_{F})$ and the compressibility of
noninteracting gapped graphene is given by
$\kappa_{0}=2/n\varepsilon_{F}(1+\Delta^{2}/\varepsilon_{F}^{2})^{1/2}$.
## III Numerical results
In this section, we present our calculations for the ground-state properties
of gapped graphene in a continuum model at low energy, using the model
described in the previous section. Our results are considered a n-doped
graphene sheet with $\alpha_{\rm gr}=1$ and $2$ where it is a typical value
thought to apply to graphene sheets on the surface of a ${\rm SiC}$ or boron
nitride substrates. We expect that the physics of gapped graphene is different
from gapless graphene due to the sublattice symmetry breaking. In the follow,
we will investigate these differences, quantitatively.
Fig. 1 shows the noninteracting dynamic polarization function,
$\chi^{(0)}({\bf q},i\omega,\mu)$ of both gapped and gapless graphene in units
of the noninteracting density of state at the Fermi surface,
$D(\varepsilon_{F})$ as functions of $q/{\rm k_{F}}$ and
$\omega/\varepsilon_{F}$. In both cases, $\chi^{(0)}({\bf q},i\omega,\mu)$
linearly diverges with $q$ at small wavelength region and decays as $1/\omega$
at large frequency for finite $\Delta$ values due to interband fluctuations in
contrast to the ordinary 2D electron gas.
The function $-\chi^{(0)}({\bf q},0,\mu)$ contains a number of noteworthy
features is shown in Fig. 2(a). First, as we mentioned before, the
$q\rightarrow 0$ limit of the static polarization function is a measure of the
number of excited states. Second, the derivative of $\chi^{(0)}({\bf
q},0,\mu)$ at $q=2{\rm k_{F}}$ is singular at finite $\Delta$ values the same
as the normal 2D electron gas. Note that $\chi^{(0)}({\bf q},0,\mu)$ at
$\Delta=0$ is a smooth function. We stress here that the second order
correction of the noninteracting polarization function is mostly responsible
to this behavior. This singular behavior is responsible for several
interesting phenomena such as Friedel oscillations and the associated RKKY
interaction. Giuliani_and_Vignale Interestingly, $\chi^{(0)}({\bf q},0,\mu)$
reduces to the contentional 2D noninteracting dynamic polarization function at
very large $\Delta$ values. In this figure, we have shown $-\chi^{(0)}({\bf
q},0,\mu)$ at $\Delta=10\varepsilon_{F}$ which is exactly the same as the
conventional 2D noninteracting dynamic polarization function up to mid $q$
values. The behavior of $\chi^{(0)}({\rm k_{F}},i\omega,\mu)$ in unit of the
density of state of gapped graphene for various $\Delta$ is displayed in Fig.
2(b).
As an application of the noninteracting polarization function given by Eq.
(10), we calculate the electric conductivity using the Boltzmann equation. we
assume $d=1$Å and $\alpha_{\rm gr}=2$. Band gap and density dependence of d.c.
conductivity are shown in Fig. 3. Increasing disorder (increasing $n_{i}$ or
decreasing $d$ for charge-disorder potential ) decrease the $\sigma$ however
increasing the gapped value decreases the d.c. conductivity. Our calculations
show that $\sigma$ decreases by increasing $\Delta$ as a function of
$n/n_{i}$. Moreover, the density dependence of $\sigma$ is linear at small $d$
and $\triangle$ values and deviates from linearity at large $d$ values.hwang1
In the inset, we have shown the results for $d=10$Å which physically determine
that the value of $\sigma$ increases by increasing $d$. Interestingly, a large
value of $\sigma$ will be obtained for suspended graphene or by using the
${\rm SiO_{2}}$ substrate instead of using boron nitride or ${\rm SiC}$ which
result in the opening a gap due to symmetry breaking between sublattices.
We also calculated the exchange and correlation energies as a function of
$\Delta$ for various values of the cutoff $\Lambda$. The results are
summarized in Fig. 4. we have found that the band gap effects become more
appreciable at large cutoff values. The exchange energy is positive yafis
because our regularization procedure implicitly selects the chemical potential
of undoped graphene as the half gap energy; doping either occupies
quasiparticle states with energies larger than $\Delta$, or empties
quasiparticles with energies smaller than $-\Delta$. Figure 4(b) shows the
correlation energy $\delta\varepsilon_{c}$ as a function of $\Delta$. Note
that $\delta\varepsilon_{c}$ has the same density dependence as
$\delta\varepsilon_{x}$ apart from the weak dependence on $\Delta$. In
contrast to the exchange energy, Figure 4(a), the correlation energy is
negative yafis . It is important to note that there is a similar behavior
between the kinetic energy and the exchange-correlation energy as a function
of $\Delta$. The kinetic energy is a slowly varying function to $\Delta$ at
small gap values and increases by increasing $\Delta$ in middle and large
values. Consequently, the total energy increases as a function of $\Delta$.
Fig. 5 is shown the total ground-state energy. In the inset, the total energy
per particle is shown as a function of $\Delta$ for various values
$\alpha_{gr}$ at $\Lambda=50$.
Figure 6 shows the charge compressibility, $\kappa/\kappa^{0}_{0}$ scaled by
its noninteracting gapless compressibility as a function of $\Delta$ for
different $\Lambda$ values. The behavior of $\kappa$ suggests some novel
physics qualitatively different from the physics known in the conventional 2D
electron gas. yafis ; asgari Kinetic energy and the exchange-correlation
energy make negative contributions to the compressibility and therefore
reduces the compressibility by increasing the $\Delta$.
## IV Conclusion
We have studied the ground-state thermodynamic properties of a gapped graphene
sheet within the random phase approximation (RPA). Note that for a doped
graphene the Fermi liquid description is valid. Our aim in this paper is
investigating the ground-state properties of a gapped graphene sheet by going
from a system with a linear dispersion relation with vanishing the energy gap,
$\triangle=0$ to a system with a parabolic dispersion relation where
$\triangle\rightarrow\infty$. To achieve this goal, we have calculated the
band gap dependence of noninteracting dynamic polarization function for doped
graphene sheet. As a consequence, We have presented results for the
conductivity suppression over a wide range of energy gap. We have presented
results of ground-state energies by incorporating many-body electron-electron
interactions via RPA for gapped graphene sheet. The total ground-sate energy
increases by increasing the band gap values. This manner occurs based on our
model Hamiltonian. We have finally presented results for the charge
compressibility suppression over the energy gap. Importantly, the impact of
gap energy on the thermodynamic properties would be noticeable for
$\triangle\geq 0.2\varepsilon_{\rm F}$.
Our results demonstrate the importance of including correlation effects
together with the gap effects in the thermodynamic quantities of a gapped
graphene. It should be possible to extend our work to include disorder
effects. Another direction would be to consider the effects of temperature in
the thermodynamic quantities.
###### Acknowledgements.
We would like to thank M. M. Vazifeh and Kh. Hassani for helpful discussions.
A. Q. is supported by IPM grant.
## References
* (1) K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva, and A. A. Firsov, Science 306, 666 (2004); K. S. Novoselov, D. Jiang, F. Schedin, T. J. Booth, V. V. Khotkevich, S. V. Morozov and A. K. Geim, Prog. Natl. Acad. Soc. USA102, 10451 (2005); Y. Zhang, Joshua P. Small, Michael E. S. Amori and Philip Kim, Phys. Rev. Lett.94, 176803 (2005) .
* (2) A. K. Geim and K. S. Novoselov, Nature Mater. 6, 183 (2007) ; A. K. Geim and A. H. MacDonald, Phys. Today60, 35 (2007); A. H. Castro Neto, F. Guinea, N. M. Peres, K. S. Novoselov and A. K. Geim, cond-mat/0709.1163 .,
* (3) S. V. Morozov, K. S. Novoselov, M. I. Katsnelson, F. Schedin, D. C. Elias, J. A. Jaszczak, and A. K. Geim, Phys. Rev. Lett. 100, 016602 (2008); Xu Du, Ivan Skachko, Anthony Barker, Eva Y. Andrei, Nature Nanotechnology 3, 491 (2008); K. I. Bolotin, K. J. Sikes, Z. Jiang, M. Klima, G. Fudenberg, J. Hone, P. Kim, H. L. Stormer, Solid State Commun.146, 351 (2008) .
* (4) C. R. Dean, B. A. Piot, P. Hayden, S. Das sarma, G. Gervais, L. N. Pfeiffer and K. W. West, Phys. Rev. Lett. 100, 146803 (2008) .
* (5) Y. W. Son, M. L. Cohen and S. G. Louise, Phys. Rev. Lett. 97, 216803 (2006); M. Y. Han, B. Ozyilmaz, Y. Zhang and P. Kim, ibid 98, 206805 (2007) .
* (6) D. A. Abanin, P. A. Lee, L. S. Levitov, Phys. Rev. Lett. 96, 176801 (2006); Y.-W. Son, M. L. Cohen and S. G. Louie, Nature 444, 347 (2006) .
* (7) Taisuke Ohta, Aaron Bostwick, Thomas Seyller, Karsten Horn, and Eli Rotenberg, Science 313, 951(2006) .
* (8) I. Zanella, S. Guerini, S. B. Fagan, J. Mendes Filho and A. G. Souza Filho. Phy. Rev. B77, 073404 (2008) .
* (9) Fran ois Varchon , R. Feng , J. Hass, X. Li, Bich N. Nguyen, C cile Naud, Pierre Mallet, Jean Yves Veuillen , Claire Berger, E. H. Conrad, Laurence Magaud, Phys. Rev. Lett. 99, 126805 (2007); D. S. L. Abergel, A. Russell, Vladimir I. Fal’ko, Applied Physics Letters 91, 063125 (2007); S.Y. Zhou, G.-H. Gweon, A.V. Fedorov, P.N. First, W.A. de Heer, D.-H. Lee, F. Guinea, A.H. Castro Neto, A. Lanzara, Nature Materials6, 770 (2007); Jakub Kedzierski, Pei-Lan Hsu, Paul Healey, Peter Wyatt, Craig Keast, Mike Sprinkle, Claire Berger, Walt de Heer, arXiv: IEEE 55, 2078 (2008) .
* (10) Gianluca Giovannetti, Pet A. Khomyako, Geert Brocks, Paul J. Kelly and Jeoroen Van den Brink, Phys. Rev. B 76, 073103 (2007) .
* (11) V. P. Gusynin, S. G. Sharapov and J. P. Carbotte, Int. J. M. Phys. B21, 4611 (2007) and references therein .
* (12) R. M. Ribeiro, N. M. R. Peres, J. Coutinho and P. R. Briddon, Phys. Rev. B 78, 075442 (2008); Eduardo V. Castro, K. S. Novoselov, S. V. Morozov, N. M. R. Peres, J. M. B. Lopes dos Santos, Johan Nilsson, F. Guinea, A. K. Geim and A. H. Castro Neto, Phys. Rev. Lett. 99, 216802 (2007) .
* (13) Chang-Yu Hou, Claudio Chamon and Christopher Mudry, Phys. Rev. Lett.98, 186809 (2007) .
* (14) Yugui Yao, Xiao-Liaug Qi, Shou-Cheng Zhang and Zhang Fang, Phys. Rev. B75, 041401(R) (2007); C. L. Kane and E. J. Mele, Phys. Rev. Lett.95, 226801 (2005); Hongki Min, J. E. Hill, N. A. Sinitsyn, B. R. Sahu, Leonard Kleinman and A. H. MacDonald, Phys. Rev. B74, 165310 (2006) .
* (15) S. Murakami, N. Nagaos and S.-C. Zhang, Phys. Rev. Lett.93, 156804 (2004) .
* (16) J. Martin, N. Akerman, G. Ulbricht, T. Lohmann, J. H. Smet, K. von Klitzing, and A. Yacoby, Nature Physics4, 144 (2008) .
* (17) N.M.R. Peres, F. Guinea, and A.H. Castro Neto, Phys. Rev. B 72, 174406 (2005) .
* (18) E.H. Hwang, B.Y.-K. Hu, and S. Das Sarma, Phys. Rev. Lett. 99, 226801 (2007).
* (19) Y. Barlas, T. Pereg-Barnea, M. Polini, R. Asgari and A. H. MacDonald, Phys. Rev. Lett. 98, 236601 (2007) .
* (20) D.E. Sheehy and J. Schmalian, Phys. Rev. Lett. 99, 226803 (2007) .
* (21) R. Asgari M. M. Vazifeh, M. R. Ramezanali, E. Davoudi and Tanatar, Phys. Rev. B 77, 125432 (2008) .
* (22) E.H. Hwang, and S. Das Sarma, Phys. Rev. B75, 205418 (2007) .
* (23) K.W.-K. Shung, Phys. Rev. B, 34, 979 (1986); J. González, F. Guinea, and M.A.H. Vozmediano, Nucl. Phys. B424, 595 (1994); B. Wunsch, T. Stauber, F. Sols, and F. Guinea, New J. Phys. 8, 318 (2006); X.-F. Wang and T. Chakraborty, Phys. Rev. B 75, 041404 (2007) .
* (24) Valeri N. Kotov, Vitor M. Pereira, Bruno Uchoa, Physical Review B 78, 075433 (2008) .
* (25) S. A. Chin, Ann. Phys. 108, 301 (1977) .
* (26) S. Adam, E.H. Hwang, and S. Das Sarma, Physica E 40, 1022 (2008); S. Adam, E.H. Hwang, V.H. Galitski, and S. Das Sarma, Proc. Natl. Acad. Sci. 104, 18392 (2007) .
* (27) D. Novikov, Appl. Phys. Lett. 91, 102102 (2007); Phys. Rev. B76, 245435 (2007) .
* (28) G.F. Giuliani and G. Vignale, Quantum Theory of the Electron Liquid (Cambridge University Press, Cambridge, 2005) .
* (29) E.H. Hwang, S. Adam, and S. Das Sarma, Phys. Rev. Lett. 98, 186806 (2007) .
Figure 1: (Color online) Noninteracting dynamic polarization function for both
gapped and gapless graphene in units of density of state, $D(\varepsilon_{\rm
F})$ as functions of $q/{\rm k_{F}}$ and $\omega/\varepsilon_{F}$.
Figure 2: (Color online) (a): Static noninteracting polarization function as a
function of $q/{\rm k_{F}}$ for various $\Delta$. (b): $\chi^{(0)}(q=k_{\rm
F},i\omega,\mu)$ as a function of $\omega/\varepsilon_{F}$ for various
$\Delta$.
Figure 3: (Color online) (a): Conductivity as a function of $n/n_{i}$ for
several energy gaps at $d=1$ Å, in the inset at $d=10$ Å(b): Conductivity as a
function of $\Delta$ for various electron densities per impurity density.
Figure 4: (Color online) Exchange (a) and correlation (b) energies as a
function of $\Delta$ for various cutoff $\Lambda$ at $\alpha_{gr}=2$. Figure
5: (Color online) Total ground-state energy per particle as a function of gap
energy, $\Delta$ for various values of the cutoff $\Lambda$ at
$\alpha_{gr}=2$. In the inset the total ground-state energy per particle is
shown as a function of $\Delta$ for various values of $\alpha_{gr}$ at
$\Lambda=50$. Figure 6: (Color online) Compressibility $\kappa/\kappa^{0}_{0}$
sacled by that of a noninteracting gapless system as a function of $\Delta$
for various $\alpha_{\rm gr}$ and $\Lambda$.
| arxiv-papers | 2008-07-20T06:37:52 | 2024-09-04T02:48:56.901480 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Alireza Qaiumzadeh and Reza Asgari",
"submitter": "Reza Asgari",
"url": "https://arxiv.org/abs/0807.3138"
} |
0807.3311 | # CP Violation in the Neutrino Sector
Stephen J. Parke Theoretical Physics Department, Fermi National Accelerator
Laboratory
P.O.Box 500, Batavia, IL 60510, USA
parke@fnal.gov
###### Abstract
This talk will address how various experiments will address the following
issues: the $\nu_{e}$ flavor content of the 3rd neutrino mass eigenstates,
$\sin^{2}\theta_{13}$, the mass ordering of the neutrinos, ${\rm sign}(\delta
m^{2}_{31})$ and whether CP is violated in the neutrino sector,
$\sin\delta\neq 0$.
## I Introduction
Fig.1 summarizes are current knowledge of the flavor content as well as the
mass ordering of the neutrino mass eigenstates assuming that there are only
three flavors of neutrinos.
Figure 1: Flavor fraction of the three neutrino mass eigenstates showing the
dependence on the cosine of the CP violating phase, $\delta$. If CPT is
conserved, the flavor fraction must be the same for neutrinos and anti-
neutrinos. This figure was adapted from Ref. pmns-fig .
Our current knowledge of the $\delta m^{2}$’s can be summarized as follows
$\displaystyle|\delta m^{2}_{32}|$ $\displaystyle=$ $\displaystyle 2.4\pm
0.4\times 10^{-3}{\rm eV^{2}}$ $\displaystyle{\rm and}\quad\quad\delta
m^{2}_{21}$ $\displaystyle=$ $\displaystyle+7.6\pm 0.4\times 10^{-5}{\rm
eV^{2}}$ (1)
where the measurement of $|\delta m^{2}_{32}|$ comes from the MINOS experiment
and that of $|\delta m^{2}_{12}|$ from the KamLAND experiment, see Nu2008 .
The sign of $\delta m^{2}_{21}$ is determined from the SNO experiment.
The mixing angles and phase, using the particle data book convention, are
given by
$\displaystyle\sin^{2}\theta_{12}$ $\displaystyle=$ $\displaystyle 0.31\pm
0.02$ $\displaystyle\sin^{2}\theta_{23}$ $\displaystyle=$ $\displaystyle
0.50\pm 0.12$ $\displaystyle\sin^{2}\theta_{13}$ $\displaystyle<$
$\displaystyle 0.04$ (2) $\displaystyle 0\leq$ $\displaystyle\delta$
$\displaystyle<2\pi.$
The best constraints on $\sin^{2}\theta_{12},~{}\sin^{2}\theta_{23}~{}{\rm
and}~{}\sin^{2}\theta_{13}$ come from SNO, SuperK’s L/E analysis and Chooz
respectively, see Nu2008 . Global fits make only marginal improvements on our
knowledge of any of these parameters.
## II The Unknowns
The unknowns that can be addressed via neutrino oscillation experiments are
* •
The $\nu_{e}$ fraction in 3nd mass eigenstate: $\sin^{2}\theta_{13}$
* •
The neutrino mass hierarchy: $sign(\delta m^{2}_{32})$
* •
Is CP violated: $\sin\delta\neq 0$
* •
Is $|U_{\mu 3}|^{2}<,=,>|U_{\tau 3}|^{2}$: $\sin^{2}\theta_{23}<,=,>1/2$
* •
Unitarity of the MNS mixing matrix: # of light sterile $\nu$’s
* •
New Interactions and Surprises (the unknown unknowns)
The other important question is whether the light neutrinos are Majorana or
Dirac which can be addressed in neutrinoless double beta decay.
## III $\nu_{e}$ Disappearance
Figure 2: The $\nu_{e}$ survival probability as a function of L/E showing both
the atmospheric and solar oscillations..
Figure 3: The left and middle panels are the iso-probability contours for T2K
as a % for the neutrino (left) and anti-neutrino (middle) channels. The solid
(blue) line is for the normal hierarchy whereas the dashed (red) line is for
the inverted hierarchy. The right panel is the bi-probability plot showing the
correlation between the two probabilities. The matter effect is small but non-
negligible for T2K.
The most direct way to address the $\nu_{e}$ fraction of the 3rd neutrino mass
eigenstate is via reactor neutrino disappearance experiments at the first
atmospheric oscillation minimum, that is 1 to 2 km from the reactor core. The
$\nu_{e}$ survival probability in vacuum is given by (see Fig.2)
$\displaystyle
P(\bar{\nu}_{e}\rightarrow\bar{\nu}_{e})=1-\cos^{4}\theta_{13}\sin^{2}2\theta_{12}\sin^{2}\Delta_{21}$
$\displaystyle-\sin^{2}2\theta_{13}(\cos^{2}\theta_{12}\sin^{2}\Delta_{31}+\sin^{2}\theta_{12}\sin^{2}\Delta_{32})$
which can be rewritten as
$\displaystyle P(\bar{\nu}_{e}\rightarrow\bar{\nu}_{e})$
$\displaystyle\approx$ $\displaystyle
1-\sin^{2}2\theta_{13}\sin^{2}\left(\frac{\delta m^{2}_{ee}L}{4E}\right)-{\cal
O}(\Delta_{21})^{2}.$
Where $\Delta_{jk}$ is used as a shorthand for the the kinematic phase,
$\delta m^{2}_{jk}L/4E$ and
$\displaystyle\delta m^{2}_{ee}$ $\displaystyle=$
$\displaystyle\cos^{2}\theta_{12}|\delta
m^{2}_{31}|+\sin^{2}\theta_{12}|\delta m^{2}_{32}|$ (3)
is the atmospheric $\delta m^{2}$ for the $\nu_{e}$ survival probability. This
is the electron flavor weighted average of $|\delta m^{2}_{31}|$ and $|\delta
m^{2}_{32}|$.
Three experiments are being constructed to look for small values of
$\sin^{2}\theta_{13}$. These are Double Chooz (France), Daya Bay (China) and
Reno (South Korea) Nu2008 . Double Chooz will start data taking at the end of
2008 with only the far detector with the near detector coming on line in 2009.
The ultimate sensitivity of the Double Chooz experiment is
$\sin^{2}2\theta_{13}=0.03$ whereas Daya Bay which will start in 2009 has an
ultimate sensitivity for $\sin^{2}2\theta_{13}<0.01$. Reno’s sensitivity is
comparable to that of Double Chooz. Neutrino 2012 will be an interesting time
for results from these experiments.
The strength of these experiments is that they directly measure
$\sin^{2}\theta_{13}$. However, they have no sensitivity to the mass
hierarchy, the size of $\sin^{2}\theta_{23}$ or whether or not CP is violated
in the neutrino sector.
Figure 4: The left and middle panels are the iso-probability contours for
NO$\nu$A as a % for the neutrino (left) and anti-neutrino (middle) channels.
The solid (blue) line is for the normal hierarchy whereas the dashed (red)
line is for the inverted hierarchy. The right panel is the bi-probability plot
showing the correlation between the two probabilities. The matter effects and
hence the separation between the hierarchies is 3 times large for T2K than
NO$\nu$A primarily due to the fact NO$\nu$A has three times the baseline as
T2K. The difference in the matter effect between T2K and NO$\nu$A can be used
to untangle CP violation and the mass hierarchy om-sp .
## IV Appearance Channels: $\nu_{\mu}\rightarrow\nu_{e}$
To address the size of $\sin^{2}\theta_{13}$, the mass hierarchy, the size of
$\sin^{2}\theta_{23}$ and whether or not CP is violated in the neutrino
sector, the appearance process $\nu_{\mu}\rightarrow\nu_{e}$ and/or one of its
CP and T conjugate processes will need to be measured. That is, in one of
following transitions
| | CP | |
---|---|---|---|---
| $\nu_{\mu}\to\nu_{e}$ | $\Longleftrightarrow$ | $\bar{\nu}_{\mu}\to\bar{\nu}_{e}$ |
T | $\Updownarrow$ | CPT across diagonals | $\Updownarrow$ T |
| $\nu_{e}\to\nu_{\mu}$ | $\Longleftrightarrow$ | $\bar{\nu}_{e}\to\bar{\nu}_{\mu}$ |
| | CP | |
Processes across the diagonal are related by CPT. The first row will be
explored in very powerful conventional beams, Superbeams, whereas the second
row could be explored in Nu-Factories or Beta Beams.
The amplitude for $\nu_{\mu}\rightarrow\nu_{e}$ can be simple written a sum of
three amplitudes, one associated with each neutrino mass eigenstate,
$\displaystyle U_{\mu 1}^{*}e^{-im^{2}_{1}L/2E}U_{e1}+U_{\mu
2}^{*}e^{-im^{2}_{2}L/2E}U_{e2}+U_{\mu 3}^{*}e^{-im^{2}_{3}L/2E}U_{e3}.$
The first term can be eliminated using the unitarity of the MNS matrix and
thus the appearance probability can be written as followsCervera:2000kp
$\displaystyle P(\nu_{\mu}\rightarrow\nu_{e})$ $\displaystyle=$
$\displaystyle|~{}{2U^{*}_{\mu 3}U_{e3}\sin\Delta_{31}e^{-i\Delta_{32}}}$ (4)
$\displaystyle\quad\quad\quad\quad+{2U^{*}_{\mu
2}U_{e2}\sin\Delta_{21}}~{}|^{2}$ $\displaystyle\approx$
$\displaystyle\left|{\sqrt{P_{atm}}}e^{-i(\Delta_{32}+\delta)}+{\sqrt{P_{sol}}}\right|^{~{}2}.$
As the notation suggests the amplitude $\sqrt{P_{atm}}$ only depends on
$\delta m^{2}_{31}$ and $\sqrt{P_{sol}}$ only depends on $\delta m^{2}_{21}$.
For propagation in the matter, these amplitudes are simple given by
$\displaystyle\sqrt{P_{atm}}$ $\displaystyle=$
$\displaystyle\sin\theta_{23}~{}\sin
2\theta_{13}~{}\frac{\sin(\Delta_{31}-aL)}{(\Delta_{31}-aL)}~{}\Delta_{31}$
$\displaystyle\sqrt{P_{sol}}$ $\displaystyle=$
$\displaystyle~{}\cos\theta_{23}~{}\sin
2\theta_{12}~{}\frac{\sin(aL)}{(aL)}~{}\Delta_{21}.$ (5)
Figure 5: Left panel: The 90% sensitivity to $\sin^{2}2\theta_{13}$ for T2K
for 5 years of neutrino running. Right panel: The 90% sensitivity to
$\sin^{2}2\theta_{13}$ for NO$\nu$A assuming 3 years of running time for
neutrinos and anti-neutrinos. The blue (red) curves is for the normal
(inverted) hierarchy. The three lines from right to left are for 0.7, 1,2 and
2.3 MW of protons on target respectively. These curves correspond to
$P(\nu_{\mu}\rightarrow\nu_{e})$ and
$P(\bar{\nu}_{\mu}\rightarrow\bar{\nu}_{e})$ at the sub 1% level.
The matter potential is given by
$a=G_{F}N_{e}/\sqrt{2}\approx(4000~{}km)^{-1}$ and the sign of $\Delta_{31}$
(and $\Delta_{32}$) determines the hierarchy; normal $\Delta_{31}>0$ whereas
inverted $\Delta_{31}<0$. When $a$ is set to zero one recovers the vacuum
result.
Figure 6: The left panel shows the parameters in $\sin^{2}2\theta_{13}$ v
$\delta$ plane that NO$\nu$A determines the hierarchy assuming it is normal.
The three lines from right to left are for 0.7, 1,2 and 2.3 MW of protons on
target respectively. The right panel shows the enhancement in the sensitivity
when combined with T2K neutrino running only. For the inverted hierarchy the
curves are flipped about $\delta=\pi$. Figure 7: The left panel shows the
effects of matter on $P_{atm}$ and $P_{sol}$ holding the baseline fixed and
varying the energy like the Fermilab to DUSEL proposal. The right panel is the
corresponding figure holding the energy fixed and varying the baseline like
for the T2KK proposal. Clearly the effects of matter are very different for
these two ways of getting to the second oscillation maximum. This figure was
adapted from Ref. NPV .
For anti-neutrinos $a\rightarrow-a$ and $\delta\rightarrow-\delta$. Thus the
phase between $\sqrt{P_{atm}}$ and $\sqrt{P_{sol}}$ changes from
$(\Delta_{32}+\delta)$ to $(\Delta_{32}-\delta)$. This changes the
interference term from
$\displaystyle 2\sqrt{P_{atm}}\sqrt{P_{sol}}\cos(\Delta_{32}+\delta)$
$\displaystyle\Rightarrow$ $\displaystyle
2\sqrt{P_{atm}}\sqrt{P_{sol}}\cos(\Delta_{32}-\delta).$
Expanding $\cos(\Delta_{32}\pm\delta)$, one has a CP conserving part
$2\sqrt{P_{atm}}\sqrt{P_{sol}}\cos\Delta_{32}\cos\delta$ and the CP violating
part
$\mp 2\sqrt{P_{atm}}\sqrt{P_{sol}}\sin\Delta_{32}\sin\delta.$ (6)
Therefore CP violation is maximum when $\Delta_{32}=(2n+1)\frac{\pi}{2}$ and
grows as n grows. Notice also, that for this term to be non-zero the
kinematical phase $\Delta_{32}$ cannot be $n\pi$. This is the neutrino counter
part to the non-zero strong phase requirement for CP violation in the quark
sector.
The asymmetry between $P(\nu_{\mu}\rightarrow\nu_{e})$ and
$P(\bar{\nu}_{\mu}\rightarrow\bar{\nu}_{e})$ is a maximum when
$\sqrt{P_{atm}}=\sqrt{P_{sol}}$. At the first oscillation maximum,
$\Delta_{31}=\pi/2$, this occurs when $\sin^{2}2\theta_{13}=0.002$ in vacuum.
For values of $\sin^{2}2\theta_{13}<0.002$ the oscillation probabilities are
dominated by $P_{sol}$ and thus observing the effects of non-zero
$\sin^{2}2\theta_{13}$ become increasing more challenging.
Fig. 3 and Fig. 4 give the iso-probability contours for the T2K and NO$\nu$A
experiments, see NPV . The third panel in these figures shows the allowed
region in neutrino and anti-neutrino bi-probability plane for these
experiments. For NO$\nu$A these allowed regions are significantly separated at
large values of $\sin^{2}\theta_{13}$.
Fig. 5 shows the sensitivity for non-zero $\sin^{2}2\theta_{13}$ for the T2K
and NO$\nu$A experiments, see Nu2008 . Whereas Fig.6 shows the region in the
$\sin^{2}2\theta_{13}$ v $\delta$ plane for which the hirarchy is determined
for NO$\nu$A and the combination of NO$\nu$A with T2K.
Beyond T2K and NO$\nu$A there are a number of proposals to explore the
oscillation probability, $P(\nu_{\mu}\rightarrow\nu_{e})$, at the second
oscillation maximum. One of these, T2KK, consists of building a second very
large detector in Korea so that it is down stream from a similar new large
detector at Kamioka, see Nu2008 . Another proposal is to build a new neutrino
beamline at Fermilab to send a broad band neutrino beam to a new very large
detector at DUSEL in the Homestake mine, see Nu2008 . Here the detector could
either be a large version of a water Cerenkov like SuperK or be a very large
liquid Argon detector if this is feasible. The liquid Argon detector has
better $\pi^{0}$ rejection than the water Cerenkov detectors and also has
higher sensitivity to the proton decay channel $p\rightarrow K^{+}+\nu$. The
matter effects for these two proposals are quite different. T2KK experiment
gets to the second oscillation peak by using the same energy but three times
the baseline whereas Fermilab to DUSEL uses the same baseline but the energy
of the neutrinos at the second oscillation peak is one third that of the first
oscillation peak. Fig. 7 shows the difference in the effects of matter on the
$P_{atm}$ by varying the energy (fixed baseline) and varying the baseline
(fixed energy). Thus the difference in the matter effect on the full
oscillation probability makes these two proposals complementary and can be
used to untangle the effects of matter and CP violation on the oscillation
probabilities and thus determining the neutrino mass hierarchy and whether or
not CP is violated in the neutrino sector.
## References
* (1) O. Mena and S. J. Parke, Phys. Rev. D 69, 117301 (2004) [arXiv:hep-ph/0312131].
* (2) See talks and proceedings of Neutrino 2008.
* (3) A. Cervera, A. Donini, M. B. Gavela, J. J. Gomez Cadenas, P. Hernandez, O. Mena and S. Rigolin, Nucl. Phys. B 579, 17 (2000) [Erratum-ibid. B 593, 731 (2001)] [arXiv:hep-ph/0002108].
* (4) O. Mena and S. J. Parke, Phys. Rev. D 70, 093011 (2004) [arXiv:hep-ph/0408070].
* (5) H. Nunokawa, S. J. Parke and J. W. F. Valle, Prog. Part. Nucl. Phys. 60, 338 (2008) [arXiv:0710.0554 [hep-ph]].
| arxiv-papers | 2008-07-21T17:18:40 | 2024-09-04T02:48:56.908380 | {
"license": "Public Domain",
"authors": "Stephen J. Parke",
"submitter": "Stephen Parke",
"url": "https://arxiv.org/abs/0807.3311"
} |
0807.3315 | # THE REPRESENTATION OF BOL ALGEBRAS
Ndoune and Thomas B. Bouetou
###### Abstract.
We construct a representation of Bol algebras by using the notion of Bol
modules.
> Keywords: Bol algebras, representation of Bol Algebras, Bol modules.
> AMS subject classification 2000: 05B07, 17D05, 17D10, 17D99, 20N05, 22E60,
> 22E67,
## 1\. Introduction
It is well known that the algebraic systems which characterize locally a
totally geodesic subspace is a Lie triple system [1], [2], [3], [4]. A Bol
algebra is realized by equipping Lie triple System with an additional binary
skew operation which satisfying a speudo-differentiation property [5], [6].
More generally, the algebraic structures which characterize locally Bol loops
are Bol algebras [7]. Till date, the representation of these algebras has not
yet been given. And we know that, the representation of these algebras may
help solve some problems in geometry and Physics. As far as geometry is
concerned, this is used to construct the homology and the cohomology of Bol
algebras and loops. As far as physics is concerned, it is used for the
description of invariant properties of physical systems and the concomitant
conservation laws as a result. J.Mostovoy and J.M.P rez-Izquierdo showed that,
Malcev algebras and Lie triple systems are particular subclasses of Bol
algebras [8]. The representation of Malcev algebras can be found in [9], and
the Lie triple systems by T.L.Hodge and B.J.Parshall with a categorical
approach [10], and the one done by W.Bertrand and M.Didry uses symmetric
bundles of symmetric spaces[11]. A tangent algebra of a local analytic Moufang
loop is a Malcev algebra. Now, there already exists representations of other
classes of non-associative algebras; the case of alternative algebras was
constructed by R.D.Schafer [14], and the one of Leibniz algebras by P.
Kolesnikov [15]. The purpose of this paper is to present a construction of a
representation of Bol algebras and some important properties. Our approach
uses the notion of Bol modules defined in the paper, Ideals of Bol algebras
introduced in [5] and the one of Lie triple systems [16]. This paper will be
organized as follow: first and fore, we will talk about Bol algebras, Next, a
representation of these algebras will be presented. Dual representation and
the duality principle will be introduced in the last section.
## 2\. Bol algebras
###### Definition 2.0.1.
A Bol algebra is a vector space $\mathfrak{B}$ over a field $K$ of which is
closed with respect a trilinear operation $(x;y,z)$ and with additional
bilinear skew-symetric operation $x\cdot y$ satisfying:
1. (i)
$(x;x,y)=0$
2. (ii)
$(x;y,z)+(z;x,y)+(y;z,x)=0.$
3. (iii)
$((x;y,z);\alpha,\beta)=((x;\alpha,\beta);y,z)+(x;{y;\alpha,\beta},z)+(x;y,(\alpha,\beta,z))$
4. (iv)
$(x\cdot y;\alpha,\beta)=(x;\alpha,\beta)\cdot
y+x\cdot(y;\alpha,\beta)+(\alpha\cdot\beta;x,y)+(x\cdot
y)\cdot(\alpha\cdot\beta)$
for all $x$,$y$,$z$,$\alpha$,and $\beta$ in $\mathfrak{B}$.
In other words, a Bol algebra is a lie triple system $(\mathfrak{B},(-;-,-))$
with an additional bilinear skew-symmetric operation $x\cdot y$ such that, the
derivation$\mathfrak{D}_{\alpha,\beta}:x\longrightarrow(x;\alpha,\beta)$ on a
ternary operation is a pseudo-differentiation with components $\alpha,\beta$
on a binary operation, that is; for all $x$,$y$ and $z$ in $\mathfrak{B}$, we
have
$\mathfrak{D}_{\alpha,\beta}(x\cdot
y)=\big{(}\mathfrak{D}_{\alpha,\beta}(x)\big{)}\cdot
y+x\cdot\big{(}\mathfrak{D}_{\alpha,\beta}(y)\big{)}+(\alpha\cdot\beta;x,y)+(x\cdot
y)\cdot(\alpha\cdot\beta).$
$\mathfrak{D}_{\alpha,\beta}$ is differentiation on ternary operation
$(x;y,z)$ means that,
$\mathfrak{D}_{\alpha,\beta}(x;y,w)=(\mathfrak{D}_{\alpha,\beta}(x);y,w)+(x;\mathfrak{D}_{\alpha,\beta}(y),w)+(x;y,\mathfrak{D}_{\alpha,\beta}(w)).$
In fact, the Bol algebra defined above is called rigth Bol algebra [6].
Similarly, we define a left Bol algebra as a Lie triple system with an
additional bilinear skew-symetric operation $x\cdot y$ such that the
derivation
$\mathfrak{\triangle}_{\alpha,\beta}:x\longrightarrow(\alpha;\beta,x)$ on a
ternary operation is a pseudo-differentiation with components $\alpha$ and
$\beta$; That is for all $x,y,z$ in $\mathfrak{B},$
$\mathfrak{\triangle}_{\alpha,\beta}(x\cdot
y)=\big{(}\mathfrak{\triangle}_{\alpha,\beta}(x)\big{)}\cdot
y+x\cdot\big{(}\mathfrak{\triangle}_{\alpha,\beta}(y)\big{)}+(\alpha\cdot\beta;x,y)+(\alpha\cdot\beta)\cdot(x\cdot
y).$
From a right Bol algebra $\mathfrak{B}$ we can obtain a left Bol algebra
$\mathfrak{B}^{op}$ by considering $\mathfrak{B}$ with the operations
$[x;y,z]=-(z;x,y)$ and $[x,y]=-x\cdot y$.
Therefore, in all what follows we will simply talk about Bol algebras.
We will note that Bol algebras can be realized as the tangent algebras to a
Bol loops with the left Bol identity, and they allow embedding in Lie algebras
[7], [12].
The subset $S$ of a Bol algebra $\mathfrak{B}$ is a sub-Bol algebra if it is
closed under the ternary and the binary operations of $\mathfrak{B}$.
In all what follows, except otherwise stated, we will consider Bol algebras on
a field $K$ of characteristic zero.
###### Definition 2.0.2.
A homomorphism $\varphi:\mathfrak{B_{1}}\rightarrow\mathfrak{B_{2}}$ between
two Bol algebras is a linear map preserving the ternary and the binary
operations.
Let us give some definitions and constructions on Bol algebras.
###### Definition 2.0.3.
Given a Bol algebra $(\mathfrak{B},(-;-,-),\cdot)$ over a field $K$ of
characteristic zero, a pseudo-differentiation is a linear map
$D:\mathfrak{B}\longrightarrow\mathfrak{B}$ for which, there exists
$z\in\mathfrak{B}$ (a companion of $D$) such that $D(x\cdot y)=D(x)\cdot
y+x\cdot D(y)+(z;x,y)+(x\cdot y)\cdot z;$ The companion is not necessarily
unique.
The set of all companions of $D$ is denoted $Com(D).$
###### Example 2.0.1.
$\mathfrak{D}_{\alpha,\beta}:x\longrightarrow(x;\alpha,\beta)$ is a pseudo-
differentiation with companion $\alpha\cdot\beta$ for all $\alpha,\beta$ in
$\mathfrak{B}$
Denote by pder$\mathfrak{B}$ the algebra of all pseudo-differentiations of
$\mathfrak{B}$ and ipder$\mathfrak{B}$ the algebra of inner pseudo-
differentiations of $\mathfrak{B}$.
The enlarged algebra of pseudo-differentiations of $\mathfrak{B}$ is defined
as $Pder\mathfrak{B}=\\{(D,z),D\in pder\mathfrak{B},z\in Com(D)\\}$ and the
enlarged algebra of inner pseudo-differentiation is defined as
$Ipder\mathfrak{B}=\\{(D,z),D\in ipder\mathfrak{B},z\in Com(D)\\}$. Jos M. P
rez-Isiquiro shows in [6] that, those algebras defined below are the Lie
algebras with appropriate brackets.
###### Definition 2.0.4.
Let $\mathfrak{B}$ be a Bol algebra, and $\mathcal{I}$ a subspace of
$\mathfrak{B}$, $\mathfrak{B}$ is an ideal of $\mathfrak{B}$ if
$\mathcal{I}\cdot\mathfrak{B}\subset\mathcal{I}$ and
$(\mathcal{I};\mathfrak{B},\mathfrak{B})\subset\mathcal{I}$
###### Proposition 2.0.1.
Let $(\mathfrak{B},(-;-,-),\cdot)$ be a Bol algebra, $\mathcal{I}$ an ideal of
$\mathfrak{B}$, then $\mathfrak{B}/\mathcal{I}$ is a Bol algebra.
###### Proof.
We know that, $\mathfrak{B}/\mathcal{I}$ is an algebra. Define on
$\mathfrak{B}/\mathcal{I}$ the two operations:
$[x+\mathcal{I},y+\mathcal{I}]=x\cdot y+\mathcal{I}$ and
$[x+\mathcal{I};y+\mathcal{I},z+\mathcal{I}]=(x;y,z)+\mathcal{I}$
It is clear that, $(\mathfrak{B}/\mathcal{I},[-;-,-],[-,-])$ is a Bol algebra.
∎
###### Proposition 2.0.2.
Let $\mathfrak{B_{1}}$ and $\mathfrak{B_{2}}$ be Bol Algebras,
$\varphi:\mathfrak{B_{1}}\rightarrow\mathfrak{B_{2}}$ a morphism of Bol
algebras. The kernel of $\varphi$ is an ideal of $\mathfrak{B_{1}}$ and the
image of $\varphi$ is a sub-Bol algebra of $\mathfrak{B_{2}}.$
###### Proof.
$\varphi:\mathfrak{B_{1}}\rightarrow\mathfrak{B_{2}}$ is a morphism of Bol
algebras.
We show that, $[Ker\varphi,\mathfrak{B}]\subset Ker\varphi$ and
$[Ker\varphi;\mathfrak{B},\mathfrak{B}]\subset Ker\varphi$. Let $x\in
ker\varphi$, $y\in\mathfrak{B_{1}}$ and $z\in\mathfrak{B_{1}}$; we have
$\displaystyle\varphi([x,y])$ $\displaystyle=$
$\displaystyle[\varphi(x),\varphi(y)]$ $\displaystyle=$
$\displaystyle[0,\varphi(y)]$ $\displaystyle=$ $\displaystyle 0$
and
$\displaystyle\varphi([x;y,z])$ $\displaystyle=$
$\displaystyle[\varphi(x);\varphi(y),\varphi(z)]$ $\displaystyle=$
$\displaystyle[0;\varphi(y),\varphi(z)]$ $\displaystyle=$ $\displaystyle 0$
Next, we show that, $[Im\varphi,Im\varphi]\subset Im\varphi$ and
$[Im\varphi;Im\varphi,Im\varphi]\subset Im\varphi$.
Let $a,b,c$ in $Im\varphi$, there exists $x,y,z$ in $\mathfrak{B_{1}}$ such
that, $a=\varphi(x)$, $b=\varphi(y)$ and $c=\varphi(z)$. We have
$\displaystyle\varphi([a,b])$ $\displaystyle=$
$\displaystyle[\varphi(x),\varphi(y)]$ $\displaystyle=$
$\displaystyle[\varphi(x),\varphi(y)]$ $\displaystyle=$
$\displaystyle\varphi([x,y])$
and
$\displaystyle\varphi([a;b,c])$ $\displaystyle=$
$\displaystyle[\varphi(x);\varphi(y),\varphi(z)]$ $\displaystyle=$
$\displaystyle[\varphi(x);\varphi(y),\varphi(z)]$ $\displaystyle=$
$\displaystyle\varphi([x;y,z])$
Hence $ker\varphi$ is an ideal of $\mathfrak{B_{1}}$ and $Im\varphi$ is a sub-
Bol algebra of $\mathfrak{B_{2}}.$ ∎
Let’s remark here that the first isomorphism theorem remains unchanged for Bol
algebras,i.e., $\mathfrak{B_{1}}/ker\varphi\cong Im\varphi.$
###### Proposition 2.0.3.
Let $(\mathfrak{B},[-;-,-],[-,-])$ be a Bol algebra and $X$ a subset of
$\mathfrak{B}$. the intersection of all ideals containing $X$ is the smallest
ideal of $\mathfrak{B}$ containing $X$, called the ideal generate by $X$ and
denoted $\langle X\rangle$.
###### Proof.
$(\mathfrak{B},[-;-,-],[-,-])$ is Bol algebra, and $X$ a subset of
$\mathfrak{B}.$ Set $\mathcal{I}=\underset{i\in I}{\bigcap}\mathcal{I}_{i}$
the intersection of all ideals of $\mathfrak{B}$ containing $X$. Show that,
$\mathcal{I}$ is an ideal of $\mathfrak{B}.$
We have $[\mathcal{I},\mathfrak{B},\mathfrak{B}]\subset\mathcal{I}_{i}$ for
all $i\in I$, because $\mathcal{I}\subset\mathcal{I}$. Then, We have
$[\mathcal{I},\mathfrak{B},\mathfrak{B}]\subset\underset{i\in
I}{\bigcap}\mathcal{I}_{i}$; and
$[\mathcal{I},\mathfrak{B},\mathfrak{B}]\subset\mathcal{I}.$ Likewise, we have
$[\mathcal{I},\mathfrak{B}]\subset\mathcal{I}.$ In addition, we know that for
all $i\in I$, $X\subset\mathcal{I}_{i}$; therefore, $X\subset\mathcal{I}$.
Let, $\mathcal{L}$ be another ideal containing $X$, it is clear that
$\mathcal{I}\subset\mathcal{L}.$
∎
Bol algebras over a field $K$ and their morphisms form a category denoted by
$Bal(K)$
###### Proposition 2.0.4.
The category $Bal(K)$ is complete and cocomplete.
###### Proof.
We are going to show first that $Bal(K)$ has products and equalizers of
parallel pair of morphisms.
Let $(\mathfrak{B}_{i},[-;-,-]_{i},[-,-]_{i})_{i\in I}$ be a family of Bol
algebras, $I$ is a set. We define on $\underset{i\in
I}{\prod}\mathfrak{B}_{i}$ the ternary operation $[-;-,-]$ and the binary one
$[-,-]$ by:
$[(x_{i});(y_{i}),(z_{i})]=([x_{i};y_{i},z_{i}])_{i\in I}$ and
$[(x_{i}),(y_{i})]=([x_{i},y_{i}])_{i\in I},$ for all $(x_{i})$, $(y_{i})$ and
$(z_{i})$ in $\underset{i\in I}{\prod}\mathfrak{B}_{i}$. We recall that, the
product of algebras is an algebra; it remains to prove the properties of
binary and ternary operations.
We have:
$\displaystyle[(x_{i}),(x_{i})]$ $\displaystyle=$
$\displaystyle([x_{i},x_{i}])$ $\displaystyle=$ $\displaystyle 0$
The binary operation is linear and skew-symmetric. In other words,
$\displaystyle[(x_{i});(x_{i}),(y_{i})]$ $\displaystyle=$
$\displaystyle([x_{i};x_{i},z_{i}])$ $\displaystyle=$ $\displaystyle 0$
It is easy to see that, $[,,]$ verifies the identity of Jaccobi and
$\mathfrak{D}_{(\alpha_{i}),(\beta_{i})}:(x_{i})\longrightarrow[(x_{i});(\alpha_{i}),(\beta_{i})]$
is a pseudo-differentiation with components $(\alpha_{i})$ and $(\beta_{i})$
on a binary operation $[,]$. The projections $p_{j}:\underset{i\in
I}{\prod}\mathfrak{B}_{i}\longrightarrow\mathfrak{B}_{j}$ for all $i\in I$ are
morphisms of Bol algebras.
Now, we construct the equalizer of parallel pair of $Bal(K).$ Let us consider
$f$ and $g$ two morphisms of Bol algebra $\mathfrak{B}_{1}$ in the Bol algebra
$\mathfrak{B}_{2}$. Set
$E=\\{x\in\mathfrak{B}_{1}|f(x)=g(x)\\}$, it is clear that $E$ is closed by
the two operations $[-;-,-]$ and $[-,-];$ then $(E,[-;-,-],[-,-])$ is a sub-
Bol algebra. We define the inclusion map $e:E\longrightarrow\mathfrak{B}_{1}$.
$e$ is a morphism of Bol algebras which satisfies $f\circ e=g\circ e$. Let
$E^{\prime}$ be another Bol algebra and
$e^{\prime}:E^{\prime}\longrightarrow\mathfrak{B}_{1}$ a morphism of Bol
algebra which satisfies the equality $f\circ e^{\prime}=g\circ e^{\prime}$,
show that there exists a unique $\varphi:E^{\prime}\longrightarrow E$ morphism
of Bol algebras such that $e\circ\varphi=e^{\prime}$; see the diagram:
$\lx@xy@svg{\hbox{\raise 0.0pt\hbox{\kern
7.74915pt\hbox{\ignorespaces\ignorespaces\ignorespaces\hbox{\vtop{\kern
0.0pt\offinterlineskip\halign{\entry@#!@&&\entry@@#!@\cr&&&\\\&&&\\\\}}}\ignorespaces{\hbox{\kern-3.0pt\raise
0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise
0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern 31.74915pt\raise
0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise
0.0pt\hbox{$\textstyle{E\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern
53.55904pt\raise 4.50694pt\hbox{{}\hbox{\kern 0.0pt\raise
0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern
0.0pt\raise-1.50694pt\hbox{$\scriptstyle{e}$}}}\kern
3.0pt}}}}}}\ignorespaces{\hbox{\kern 69.70746pt\raise 0.0pt\hbox{\hbox{\kern
0.0pt\raise
0.0pt\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern
69.70746pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern
3.0pt\raise
0.0pt\hbox{$\textstyle{\mathfrak{B}_{1}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern
92.79587pt\raise 7.1875pt\hbox{{}\hbox{\kern 0.0pt\raise
0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern
0.0pt\raise-0.8264pt\hbox{$\scriptstyle{{g}}$}}}\kern
3.0pt}}}}}}\ignorespaces{\hbox{\kern 109.59082pt\raise 2.0pt\hbox{\hbox{\kern
0.0pt\raise
0.0pt\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern
92.50053pt\raise-8.1111pt\hbox{{}\hbox{\kern 0.0pt\raise
0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern
0.0pt\raise-1.75pt\hbox{$\scriptstyle{{f}}$}}}\kern
3.0pt}}}}}}\ignorespaces{\hbox{\kern 109.59082pt\raise-2.0pt\hbox{\hbox{\kern
0.0pt\raise
0.0pt\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern
109.59082pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern
3.0pt\raise
0.0pt\hbox{$\textstyle{\mathfrak{B}_{2}}$}}}}}}}{\hbox{\kern-7.74915pt\raise-38.5133pt\hbox{\hbox{\kern
0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise
0.0pt\hbox{$\textstyle{E^{\prime}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces{}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern
8.83536pt\raise-13.14555pt\hbox{{}\hbox{\kern 0.0pt\raise
0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern
0.0pt\raise-1.75pt\hbox{$\scriptstyle{\psi}$}}}\kern
3.0pt}}}}}}\ignorespaces{\hbox{\kern 35.71072pt\raise-3.0pt\hbox{\hbox{\kern
0.0pt\raise
0.0pt\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces\ignorespaces{}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern
36.2145pt\raise-24.31915pt\hbox{{}\hbox{\kern 0.0pt\raise
0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern
0.0pt\raise-2.0625pt\hbox{$\scriptstyle{{e^{\prime}}}$}}}\kern
3.0pt}}}}}}\ignorespaces{\hbox{\kern
69.70746pt\raise-3.9398pt\hbox{\hbox{\kern 0.0pt\raise
0.0pt\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\lx@xy@drawline@}}{\hbox{\kern
35.7283pt\raise-38.5133pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern
3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern
74.64914pt\raise-38.5133pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern
3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern
114.5325pt\raise-38.5133pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern
3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}\ignorespaces}}}}\ignorespaces.$
We set for all $x\in E^{\prime}$, $\varphi(x)=e^{\prime}(x)$. $\varphi$ is a
morphism of Bol algebra; and
$\displaystyle e\circ\varphi(x)$ $\displaystyle=$ $\displaystyle
e(e^{\prime}(x))$ $\displaystyle=$ $\displaystyle e^{\prime}(x).$
Let $\psi:E^{\prime}\longrightarrow E$ be another morphism of Bol algebra
which has the same properties as $\varphi.$ we have,
$\displaystyle\psi(x)$ $\displaystyle=$ $\displaystyle e(\varphi(x))$
$\displaystyle=$ $\displaystyle\varphi(x).$
Then $\psi=\varphi.$
We set now $\underset{i\in I}{\coprod}\mathfrak{B}_{i}=\\{(x_{i}),i\in
I_{0}\\}$ where $I_{0}$ is a finite subset of $I$. We define same as for
products, the ternary operation $[-;-,-]$ and the binary operation $[-,-].$ It
is easy to see that, $(\underset{i\in
I}{\coprod}\mathfrak{B}_{i},[-;-,-],[-,-])$ is a Bol algebra, named the
coproduct of the family $(\mathfrak{B}_{i},[-;-,-]_{i},[-,-]_{i})_{i\in I}.$
Now construct the coequalizer of morphisms $f$ and $g$ defined above. We know
that, $Imf+Img$ is sub-Bol space of Bol algebra $\mathfrak{B}_{1};$ But not
necessarly an ideal of Bol algebra. According to proposition (2.0.3), let us
consider $\mathcal{I}=\langle Imf+Img\rangle$ to be a smallest ideal generated
by $Imf+Img.$ $\mathfrak{B}/\mathcal{I}$ is a Bol algebra by
proposition(2.0.1). Let us show that, $(\mathfrak{B}/\mathcal{I},p)$ is a
coequalizer of $f$ and $g$, where $p$ is the canonical projection.
Let $x\in\mathfrak{B}_{1},$
$\displaystyle p\circ f(x)$ $\displaystyle=$ $\displaystyle f(x)+\mathcal{I}$
$\displaystyle=$ $\displaystyle\mathcal{I}$ $\displaystyle=$ $\displaystyle
p\circ g(x)$
We have $p\circ f=p\circ g.$ Let $(Q,h)$ be another pair, with $Q$ a Bol
algebra and $h:\mathfrak{B}_{2}\longrightarrow Q$ a morphism of Bol algebras
such that $h\circ f=h\circ g.$ We show that, there exists a unique morphism of
Bol algebras $\varphi:\mathfrak{B}/\mathcal{I}\longrightarrow\mathfrak{B}_{2}$
such that $\varphi\circ p=h.$
we define $\varphi:\mathfrak{B}/\mathcal{I}\longrightarrow Q$ such that,
$\varphi(x+\mathcal{I})=h(x)$; $\varphi$ is a morphism of Bol algebras.
Let $\psi:\mathfrak{B}/\mathcal{I}\longrightarrow Q$ be another morphism of
Bol algebras such that $\psi\circ p=h.$ We have for all
$x\in\mathfrak{B}_{2},$
$\displaystyle\psi(x+\mathcal{I})$ $\displaystyle=$ $\displaystyle\psi\circ
p(x)$ $\displaystyle=$ $\displaystyle\varphi\circ p(x)$ $\displaystyle=$
$\displaystyle\varphi(x+\mathcal{I})$
Then $\varphi$ is unique. $(\mathfrak{B}/\mathcal{I},p)$ is a coequalizer of
the pair $f,g.$ Hence $Bal(K)$ is complete and cocomplete see [13]
∎
## 3\. representation of Bol algebras
Before constructing the representation of Bol algebra, we must first remark
that it is not done in an analoguous manner to that of the Lie triple systems.
In [11] W. Bertrand and M. Didry defined a representation of Lie triple system
$T$, in End(V) by giving two bilinear maps:
$\begin{array}[]{ccc}\mathfrak{D}&:T\times T&\longrightarrow End(V)\\\
&(\alpha,\beta)&\longmapsto\mathfrak{D}_{\alpha,\beta}\end{array}$
$\begin{array}[]{ccc}\Delta&:T\times T&\longrightarrow End(V)\\\
&(\alpha,\beta)&\longmapsto\Delta_{\alpha,\beta}\end{array}$
Which satisfy certain properties; where $V$ is a vector space called
$T$-module. Talking of Bol algebras, one must remark that the presence of the
second binary operation brings about supplementary conditions which may not be
fulfilled by the above maps.
Let $(\mathfrak{B},(;,),(;))$ be a Bol algebra and $V$ be a vector space over
a field $K$, we consider the following bilinear maps as:
$\begin{array}[]{ccc}\cdot&V\times\mathfrak{B}&\longrightarrow V\\\
&(v,\alpha)&\longmapsto v\cdot\alpha\end{array}$
$\begin{array}[]{ccc}\cdot&\mathfrak{B}\times V&\longrightarrow V\\\
&(\alpha,v)&\longmapsto\alpha\cdot v\end{array}$
Likewise, we define the trilinear maps :
$\begin{array}[]{ccc}[-;-,-]:&V\times\mathfrak{B}\times\mathfrak{B}&\longrightarrow
V\\\ &(v,\alpha,\beta)&\longmapsto[v;\alpha,\beta]\end{array}$
$\begin{array}[]{ccc}[-;-,-]:&\mathfrak{B}\times
V\times\mathfrak{B}&\longrightarrow V\\\
&(\alpha,v,\beta)&\longmapsto[\alpha;v,\beta]\end{array}$
$\begin{array}[]{ccc}[-;-,-]:&\mathfrak{B}\times\mathfrak{B}\times
V&\longrightarrow V\\\
&(\alpha,\beta,v)&\longmapsto[\alpha;\beta,v]\end{array}$
###### Definition 3.0.1.
A vector space $V$ is a Bol module if for all $\alpha,\beta,\gamma,\tau$ in
$\mathfrak{B}$ and $v$ in $V$, the following properties are true:
1. (1)
$\alpha\cdot v=-v\cdot\alpha$
2. (2)
$[v;\alpha,\beta]+[v;\beta,\alpha]=0$
3. (3)
$[v;\alpha,\beta]+[\alpha;v,\beta]+[\alpha;\beta,v]=0$
4. (4)
$[(\alpha;\beta,\gamma);v,\tau]=[\alpha;\beta,[\gamma;v,\tau]]+[\alpha;[\beta,v,\tau],\gamma]+[[\alpha;v,\tau];\beta,\gamma]$
5. (5)
$[(\alpha;\beta,\gamma),v\cdot\tau]=v\cdot(\alpha;\beta,\tau)+[v;\alpha,\beta]\cdot\tau+[(\beta;\alpha);\tau,v]+(v\cdot\tau)\cdot(\beta;\alpha)$
The operations defined above are called the $\mathfrak{B}$-actions or actions
of $\mathfrak{B}$ on $V.$ Actually, the Bol module defined above is called
left Bol module.
###### Example 3.0.1.
Let $(\mathfrak{B},(-;-,-),(-;-))$ be a Bol algebra, $\mathfrak{B}$ seen as a
vector space is a Bol module.
We remark here that, the notion of Bol module is not the analogous notion
defined by Terrel L.Hodge and Brian J.Parshall for Lie triple systems [10].
According to them, a vector space $V$ is a $\mathfrak{B}$-module provided
that, $E_{V}=\mathfrak{B}\oplus V$ possesses a structure of Bol algebra such
that: (a) $\mathfrak{B}$ is a sub-Bol algebra of $E_{V}$, (b) $V$ is an
abelian ideal of $E_{V}$, (c)
$[V;V,\mathfrak{B}]=[V;\mathfrak{B},V]=[\mathfrak{B};V,V]=0$
###### Proposition 3.0.1.
Every $\mathfrak{B}$-module $V$ is a Bol module.
###### Proof.
Let $V$ be a $\mathfrak{B}$-module, $E_{V}=\mathfrak{B}\oplus V$ has a
structure of Bol algebra with brackets $[-;-,-]$ and $[-,-].$
We define the bilinear maps:
$\begin{array}[]{ccc}[-,-]_{1}:&V\times\mathfrak{B}&\longrightarrow V\\\
&(v,\alpha)&\longmapsto[v,\alpha]\end{array}$
$\begin{array}[]{ccc}[-,-]_{1}&\mathfrak{B}\times V&\longrightarrow V\\\
&(\alpha,v)&\longmapsto[\alpha,v]\end{array}$
Likewise, we define the trilinear maps :
$\begin{array}[]{ccc}[-;-,-]_{1}:&V\times\mathfrak{B}\times\mathfrak{B}&\longrightarrow
V\\\ &(v,\alpha,\beta)&\longmapsto[v;\alpha,\beta]\end{array}$
$\begin{array}[]{ccc}[-;-,-]_{1}:&\mathfrak{B}\times
V\times\mathfrak{B}&\longrightarrow V\\\
&(\alpha,v,\beta)&\longmapsto[\alpha;v,\beta]\end{array}$
$\begin{array}[]{ccc}[-;-,-]_{1}:&\mathfrak{B}\times\mathfrak{B}\times
V&\longrightarrow V\\\
&(\alpha,\beta,v)&\longmapsto[\alpha;\beta,v]\end{array}$
The properties $(1)-(5)$ of definition of a Bol module hold. ∎
###### Definition 3.0.2.
Every linear map between two Bol modules $V$ and $W$ which preserve the
$\mathfrak{B}$-actions is call morphism of Bol modules.
###### Example 3.0.2.
Let $V$ be a Bol module, the identity map of $V$ is a morphism of Bol module.
$V$ is a Bol module, $l_{\tau}:V\longrightarrow V$ defined by
$L_{\tau}(v)=v\cdot\tau$, is a morphism of Bol module.
Before studying the general representation of Bol algebras, we are going to
investigate a particular case of such representation that N.Jacobson called
regular representation in the case of Jordan Algebras [17].
$V$ is a Bol module, Let’s consider a linear map $l_{\tau}:V\longrightarrow V$
defined by $L_{\tau}(v)=v\cdot\tau.$ $L_{\tau}$ is an endomorphism of Bol
module $V$ into itself, which we call left transformation. we can not define a
right transformation here because , every right transformation
$R_{\tau}(v)=\tau\cdot v$ is also left transformation as:
$\displaystyle R_{\tau}(v)$ $\displaystyle=$ $\displaystyle\tau\cdot v$
$\displaystyle=$ $\displaystyle-v\cdot\tau$ $\displaystyle=$ $\displaystyle
L_{-\tau}(v)$
We define two operations in left transformations by:
for all $\alpha,\beta,\gamma\in\mathfrak{B}$ and $v\in V,$
$[L_{\alpha},L_{\beta}](v)=L_{(\alpha;\beta)}(v)$ and
$[L_{\alpha};L_{\beta},L_{\gamma}](v)=L_{(\alpha;\beta,\gamma)}(v).$
Now, we consider the map
$\begin{array}[]{ccc}\rho:&\mathfrak{B}&\longrightarrow End(V)\\\
&\tau&\longmapsto L_{\tau}\end{array}$
$\rho$ is a morphism of Bol algebras. $\rho$ is a representation of Bol
algebras, called regular representation.
Let us construct the general representation of Bol algebras.
Let $V$ be a Bol module, we define the following maps:
$\begin{array}[]{ccc}m(\alpha,\beta):&V&\longrightarrow V\\\
&v&\longmapsto[\alpha;\beta,v]\end{array}$
$\begin{array}[]{ccc}c(\alpha,\beta):&V&\longrightarrow V\\\
&v&\longmapsto[\alpha;v,\beta]\end{array}$
$\begin{array}[]{ccc}r(\alpha,\beta):&V&\longrightarrow V\\\
&v&\longmapsto[v;\alpha,\beta]\end{array}$
$\begin{array}[]{ccc}L_{\tau}:&V&\longrightarrow V\\\ &v&\longmapsto\tau\cdot
v\end{array}$
$\begin{array}[]{ccc}R_{\tau}:&V&\longrightarrow V\\\ &v&\longmapsto
v\cdot\tau\end{array}$
We have the following properties:
###### lemma 3.0.1.
1. The following properties are satisfied:
$(p_{1})$ $R_{\tau}=L_{-\tau}$
$(p_{2})$ $m(\alpha,\beta)+r(\alpha,\beta)=0$
$(p_{3})$ $m(\alpha,\beta)+c(\alpha,\beta)+r(\alpha,\beta)=0$
$(P_{4})$ $c(\mathfrak{D}_{\alpha,\beta}(\gamma),\tau)=m(\alpha,\beta)\circ
c(\gamma,\tau)+c(\alpha,\gamma)\circ c(\beta,\tau)+r(\beta,\gamma)\circ
c(\alpha,\tau)$
$(p_{5})$ $m(\alpha,\beta)\circ
L_{\tau}=L_{(\mathfrak{D}_{\alpha,\beta})(\tau)}+L_{\tau}\circ
r(\alpha,\beta)+m((\beta;\alpha),\tau)+L_{(\beta;\alpha)}\circ L_{\tau}$
###### Proof.
$(p_{1})$ was shown in regular representation.
For $(p_{2}),(p_{3}),(p_{4}),$ we write the relations (1)and (2) in the
definition of Bol module in the sense of $m(\alpha,\beta)$, $c(\alpha,\beta)$
and $r(\alpha,\beta).$
For $(p_{5}),$ we translate the relation (5) in the definition by the morphism
$m(\alpha,\beta)$,$r(\alpha,\beta),L_{\tau}$ and
$\mathfrak{D}_{\alpha,\beta}(\tau).$ ∎
###### Definition 3.0.3.
A general representation of a Bol algebra $\mathfrak{B}$ in $End(V)$, the
space of endomorphisms of a Bol module $V$, is given by two linear maps:
$\begin{array}[]{ccc}L_{\tau}:&\mathfrak{B}&\longrightarrow End(V)\\\
&\tau&\longmapsto L_{\tau}\end{array}$
$\begin{array}[]{ccc}R:&\mathfrak{B}&\longrightarrow End(V)\\\
&\tau&\longmapsto R_{\tau}\end{array}$
and tree bilinear maps:
$\begin{array}[]{ccc}m:&\mathfrak{B}\times\mathfrak{B}&\longrightarrow
End(V)\\\ &(\alpha,\beta)&\longmapsto m(\alpha,\beta)\end{array}$
$\begin{array}[]{ccc}c:&\mathfrak{B}\times\mathfrak{B}&\longrightarrow
End(V)\\\ &(\alpha,\beta)&\longmapsto c(\alpha,\beta)\end{array}$
$\begin{array}[]{ccc}r:&\mathfrak{B}\times\mathfrak{B}&\longrightarrow
End(V)\\\ &(\alpha,\beta)&\longmapsto r(\alpha,\beta)\end{array}$
such that $(p_{1})$-$(p_{5})$ hold. $V$ is called module of representation.
We can replace $End(V)$ in the definition above by a unitary associative
algebra $A.$
###### Proposition 3.0.2.
Let $V$ be a Bol module, for all $\alpha,\beta,\gamma$ in $\mathfrak{B}$, the
following equality holds:
$m(\alpha,\beta)\circ m(\gamma,\tau)+m(\alpha,\beta)\circ
r(\gamma,\tau)+r(\beta,\alpha)\circ m(\alpha,\tau)+r(\beta,\gamma)\circ
r(\alpha,\tau)=m(\mathfrak{D}_{\alpha,\beta}(\gamma),\tau)+r(\mathfrak{D}_{\alpha,\beta}(\gamma),\tau)+m(\alpha,\gamma)\circ
m(\beta,\tau)+m(\alpha,\gamma)\circ r(\beta,\tau)+r(\alpha,\gamma)\circ
m(\beta,\tau)+r(\alpha,\gamma)\circ r(\beta,\tau)$.
###### Proof.
We use the property $(p_{4})$ of lemma 3.0.1 and we replace $c(\alpha,\beta)$
by its value $c(\alpha,\beta)=-m(\alpha,\beta)-r(\alpha,\beta)$ in the
identity $(p_{2})$ of lemma 3.0.1 ∎
###### Proposition 3.0.3.
Let $\mathfrak{B}$ be a Bol algebra over a field $k$, and $V$ a Bol module. If
the extension $E_{V}=\mathfrak{B}\oplus V$ has a structure of Bol algebra,
then $\mathfrak{B}$ is a sub-Bol algebra of $E_{V}$ and $V$ is an ideal of
$E_{V}.$
###### Proof.
Let $P:E_{V}\longrightarrow\mathfrak{B}$ be the projection on $\mathfrak{B}$
with direction $V.$ $P$ is a morphism of Bol algebras; $V=kerP$ and
$\mathfrak{B}=imP,$ according to proposition 2.02, $V$ is an ideal of $E_{V}$
and $\mathfrak{B}$ is a sub-Bol algebra of $E_{V}.$ ∎
###### Proposition 3.0.4.
If $(V,L,R,m,c,r)$ and
$(W,L^{\prime},R^{\prime},m^{\prime},c^{\prime},r^{\prime})$ are two
representations of Bol algebras $\mathfrak{B}$, then $(V\oplus W,L\oplus
L^{\prime},R\oplus R^{\prime},m\oplus m^{\prime},c\oplus c^{\prime},r\oplus
r^{\prime})$ is again a representation of $\mathfrak{B},$ called a direct sum.
###### Proof.
we define the following linear maps:
$\begin{array}[]{ccc}(L\oplus L^{\prime})_{\tau}:&V\oplus W&\longrightarrow
V\oplus W\\\ &v\oplus w&\longmapsto L_{\tau}(v)\oplus
L^{\prime}_{\tau}(w)\end{array}$
$\begin{array}[]{ccc}(R\oplus R^{\prime})_{\tau}:&V\oplus W&\longrightarrow
V\oplus W\\\ &v\oplus w&\longmapsto R_{\tau}(v)\oplus
R^{\prime}_{\upsilon}(w)\end{array}$
$\begin{array}[]{ccc}(m\oplus m^{\prime})(\alpha,\beta):&V\oplus
W&\longrightarrow V\oplus W\\\ &v\oplus w&\longmapsto m_{\tau}(v)\oplus
m^{\prime}_{\tau}(w)\end{array}$
$\begin{array}[]{ccc}(c\oplus c^{\prime})(\alpha,\beta):&V\oplus
W&\longrightarrow V\oplus W\\\ &v\oplus w&\longmapsto c(\alpha,\beta)(v)\oplus
c^{\prime}(\alpha,\beta)(w)\end{array}$
$\begin{array}[]{ccc}(r\oplus r^{\prime})(\alpha,\beta):&V\oplus
W&\longrightarrow V\oplus W)\\\ &v\oplus w&\longmapsto
r(\alpha,\beta)(v)\oplus r^{\prime}(\alpha,\beta)(w)\end{array}$
It is clear that properties $(p_{1})-(p_{5})$ of lemma3.0.1 hold.
∎
## 4\. Dual representation and duality principle
### 4.1. The dual representation
Let $(V,L,R,m,c,r)$ be a representation of Bol algebra, we wish to define a
structure of Bol module in the dual space $V^{\ast}$.We set:
$\begin{array}[]{ccc}m^{\ast}(\alpha,\beta):&V^{\ast}&\longrightarrow
V^{\ast}\\\ &f&\longmapsto m^{\ast}(\alpha,\beta)(f)\end{array}$
with
$\begin{array}[]{ccc}m^{\ast}(\alpha,\beta)(f):&V&\longrightarrow K\\\
&v&\longmapsto f([\beta;\alpha,v])\end{array}$
$\begin{array}[]{ccc}c^{\ast}(\alpha,\beta):&V^{\ast}&\longrightarrow
V^{\ast}\\\ &f&\longmapsto c^{\ast}(\alpha,\beta)(f)\end{array}$
with
$\begin{array}[]{ccc}c^{\ast}(\alpha,\beta)(f):&V&\longrightarrow K\\\
&v&\longmapsto f([\beta;v,\alpha])\end{array}$
$\begin{array}[]{ccc}r^{\ast}(\alpha,\beta):&V^{\ast}&\longrightarrow
V^{\ast}\\\ &f&\longmapsto r^{\ast}(\alpha,\beta)(f)\end{array}$
with
$\begin{array}[]{ccc}r^{\ast}(\alpha,\beta)(f):&V&\longrightarrow K\\\
&v&\longmapsto f([v;\beta,\alpha])\end{array}$
$\begin{array}[]{ccc}L^{\ast}_{\tau}:&V^{\ast}&\longrightarrow V^{\ast}\\\
&f&\longmapsto L^{\ast}(\tau)(f)\end{array}$
with
$\begin{array}[]{ccc}L^{\ast}(\tau)(f):&V&\longrightarrow K\\\ &v&\longmapsto
f(L_{\tau}(v)\end{array}$
$\begin{array}[]{ccc}R^{\ast}_{\tau}:&V^{\ast}&\longrightarrow V^{\ast}\\\
&f&\longmapsto R^{\ast}(\tau)(f)\end{array}$
with
$\begin{array}[]{ccc}R^{\ast}(\tau)(f):&V&\longrightarrow K\\\ &v&\longmapsto
f(L_{\tau}(v)\end{array}$
The properties $(1)-(5)$ of the definition of Bol module are satisfied. Hence,
$V^{\ast}$ is a Bol module called the dual module of $V.$
$(V^{\ast},L^{\ast},R^{\ast},m^{\ast},c^{\ast},r^{\ast})$ is also a
representation of Bol module $\mathfrak{B}$ that we call dual representation.
Note that, if $V$ is a right Bol module, $V^{\ast}$ is a left Bol Module.
If $A:V\longrightarrow V$ is an operator in $V,$ its dual operator $A^{\ast}$
is defined as follows:
$\begin{array}[]{ccc}A^{\ast}:&V^{\ast}&\longrightarrow V^{\ast}\\\
&f&\longmapsto f\circ A\end{array}$
### 4.2. The duality principle
We know that, for finite dimensional module $V$ over a field, $V$ is the dual
of its dual module $V^{\ast}.$
More generally, we can define for any general representation of Bol algebra
$\mathfrak{B}$ in an algebra $A$ its opposite representation in the algebra
$A^{op}$ by setting:
$r^{op}(\alpha,\beta)=r(\beta,\alpha)$
$c^{op}(\alpha,\beta)=c(\beta,\alpha)$
$m^{op}(\alpha,\beta)=m(\beta,\alpha)$; $L^{op}_{\tau}=R_{\tau}$ and
$R^{op}_{\tau}=L_{\tau}$.
As above it is seen that this is again a representation. As an application of
this, we get a duality principle similar to the one of Jordan pairs formulated
by O.Loos[18] and another for Lie triple systems formulate by W.Bertrand and
M.Didry [11].
###### Theoreme 4.2.1.
If $I$ is an identity in $L_{\tau},R_{\varsigma},r(\alpha,\beta),c(X,Y)$ and
$m(S,T)$ valid for all left Bol algebras over $\mathbb{R}$, Then its dual
identity $I^{\ast}$, obtained by replacing $L_{\tau}$ by $R_{\tau}$;
$R_{\varsigma}$ by $L_{\varsigma}$;$r(\alpha,\beta)$ by $r(\beta,\alpha)$;
$c(X,Y)$ by $c(Y,X)$ and $m(S,T)$ by $m(T,S)$ and reversing the order of all
factors, is also valid for all right Bol algebras over $\mathbb{R}.$
###### Proof.
We remark that, if $A$ is the representation algebra of Bol algebra
$\mathfrak{B},$ $A^{op}$ is the representation algebra of $\mathfrak{B}^{op}.$
We denote by $Bal_{g}$ the category of left Bol algebras over $\mathbb{R}$ and
by $Bal_{d}$ the category of right Bol algebras over $\mathbb{R}.$ We define
the functor:
$\begin{array}[]{ccc}\mathcal{F}:&Bal_{d}&\longrightarrow Bal_{g}\\\
&\mathfrak{B}&\longmapsto\mathfrak{B}^{op}\end{array}$
Let $(\mathfrak{B},(-;-,-),(-;-))$ be a right Bol algebra; from right Bol
algebra $\mathfrak{B}$ we can obtain left Bol algebra $\mathfrak{B}^{op}$ by
considering $\mathfrak{B}$ with the operations
$[x;y,z]=-(x;y,z)$ and $[x,y]=-x\cdot y$.
The functor $\mathcal{F}$ realizes an isomorphism of category; and the
category of right Bol modules is equivalent to a category of left Bol modules.
∎
One can embed, $\mathfrak{B}$ into a Lie algebra $\mathfrak{G}$ as;
$\mathfrak{G}=\mathfrak{B}\oplus\mathfrak{h}$ where, $\mathfrak{h}$ is a Lie
subalgebra of $\mathfrak{G}.$ and we have the short sequence
$\mathfrak{h}\rightarrow\mathfrak{G}\rightarrow\mathfrak{G}/\mathfrak{h}\simeq\mathfrak{B}.$
The operation in $\mathfrak{B}$ from the Lie algebra $\mathfrak{G}$ is
obtained by means of projection on $\mathfrak{B}$ parallel to $\mathfrak{h}$
that is
$\begin{CD}\mathfrak{G}\times\mathfrak{G}@>{}>{}>\mathfrak{G}\\\
(\xi,\eta)@>{}>{}>[\xi,\eta]\\\ @V{}V{}V\\\
\Pi_{[\xi,\eta]}=\xi\cdot\eta\end{CD}.$
We have
$0\rightarrow\mathfrak{h}\leftrightarrow\mathfrak{G}\rightarrow\mathfrak{G}/\mathfrak{h}\simeq\mathfrak{B}\rightarrow
0$ which is the short exact sequence of $\mathfrak{h}$-module. Therefor we
have the following proposition
###### Proposition 4.2.1.
If $\mathfrak{B}$ is a Bol algebra with the operation $(-\cdot-)$, $(-;-,-).$
Then the set $\mathfrak{G}=\mathfrak{B}\oplus\wedge^{2}\mathfrak{B}$ is the
enveloping Lie algebra for the Bol algebra $\mathfrak{B}.$
Proof Set $[x,y]_{\mathfrak{G}}=x\cdot y+x\wedge y$ and from this, the Jacobi
identity follows.
Another way of proving this theorem is to consider the set $Pder\mathfrak{B},$
of the pseudo derivation of $\mathfrak{B}$ and $IPder\mathfrak{B},$ the set of
inner pseudo-differentiation of $\mathfrak{B}.$ As it is shown in [12], the
first is the maximal standard enveloping Lie algebra while second is the
standard minimal enveloping Lie algebra. Hence the proof.
###### Proposition 4.2.2.
[7] Any pair of Lie algebras $(\mathfrak{G},\mathfrak{h})$ such that
$\mathfrak{G}=\mathfrak{B}\oplus\mathfrak{h}$,
$[\mathfrak{B},[\mathfrak{B},\mathfrak{B}]]\subset\mathfrak{B},$
$[\mathfrak{B},\mathfrak{B}]\cap\mathfrak{B}=\\{0\\}$ defines on
$\mathfrak{B}$ a Bol algebra with the operations
$\xi\cdot\eta=\sqcap[\xi,\eta],$ the projection on $\mathfrak{B}$ parallel to
$\mathfrak{h},$ and $(\zeta;\xi,\eta)=[\zeta,[\xi,\eta]],$
$\forall\zeta,\xi,\eta\in\mathfrak{B}.$
## References
* [1] E.Cartan, La g om trie des groupes de transformations, J.Math. Pures et Appliqu es,6(1927) 1-119.
* [2] K.Nomizu, Invariant affine connections on homogeneous spaces, Amer.J.Math., 76(1954)33-65.
* [3] K.Yamaguti,On algebras of totally geodesic spaces (Lie triple systems), J.Sci.Hiroshima Univ.Ser.A,21 (1957-1958) 107-113.
* [4] O.Loos, Symetric SpacesII: Compacts spaces and classification (University of Minnesota and Univerit t m nchen) 1969, 183p.
* [5] Thomas.B. Bouetou, On the structure of Bol algebras Arxiv:Maths.
* [6] J.M. Perez-Izqquierdo, An envelope for Bol algebras J. Algebra 284 (2005)-N.2-p. 480-493.
* [7] L.V. Sabinin, Analytic quasigroups and Geometry, Monograph, Friendship of Nations University press, Moscow, 1991, 112p.
* [8] J. Mostovoy and J.M.P rez-Izquierdo, Ideals in non-associative universal enveloping algebras of Lie triples systems, Arxiv:math.RA/0506179v1.
* [9] E.K.Loginov and A.N.Grishkov, On a construction of self dual gauge-fields in seven dimensions, Journal of nonlinear mathematical phisics V.14, N.4 (2007), 570-577.
* [10] T.L.Hodge and B.J.Parshall, On the representation theory of Lie triple systems, Trans.Amer.Math.Soc.354(2002)N.11, 4359-4391.
* [11] Wolfgang Bertrand and Mamon Didry, 8 oct 2007, Symetric bundles and representation of Lie triple systems, arxiv.math.
* [12] L.V. Sabinin, Smooth Quasigroups and Loops. Mathematics and its Applications, 492. Kluwer Academic Publishers, Dordrecht, (1999) 249p.
* [13] Mac Lane, Categories for the working Mathematicien, Springer-Verlag S.(1971), 262p.
* [14] R.D.Schafer, Representations of alternative algebras, Trans.Amer.Math.Soc.72, 1-17(1952).
* [15] P.Kolesnikov, 17 Aug 2007 _Conformal representations of Leibniz algebras_ Arxiv:0708.2315v1 [math.RA].
* [16] W.G.Lister (1952), _A structure theory of Lie triple systems_ , Trans.Amer.Math.Soc. 72:217-242.
* [17] N.Jacobson (1950), General representation theory of Jordan algebra, Trans.Amer.Math.Soc. 70(1951), 509-530.
* [18] O.Loos, 8 junly 2002 , Jordan pairs, Springer LNM460, Berlin (1975), 218p.
DEPARTMENT OF MATHEMATICS, University of Yaounde I, P.O.Box: 812 Yaounde-
Cameroon.
E-mail: ndouneko@yahoo.fr
DEPARTMENT OF MATHEMATICS AND COMPUTER SCIENCES, POLYTECHNIC, Yaounde,
P.O.Box: 8390 Yaounde-Cameroon.
E-mail: tbouetou@yahoo.fr
| arxiv-papers | 2008-07-21T18:08:02 | 2024-09-04T02:48:56.912904 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/",
"authors": "Ndoune Ndoune and Thomas Bouetou Bouetou",
"submitter": "Bouetou Bouetou Thomas",
"url": "https://arxiv.org/abs/0807.3315"
} |
0807.3374 | # The Dynamics of Internet Traffic: Self-Similarity, Self-Organization, and
Complex Phenomena
Reginald D. Smith PO Box 10051
Rochester, NY 14610
rsmith@bouchet-franklin.org
(August 30, 2010)
###### Abstract
The Internet is the most complex system ever created in human history.
Therefore, its dynamics and traffic unsurprisingly take on a rich variety of
complex dynamics, self-organization, and other phenomena that have been
researched for years. This paper is a review of the complex dynamics of
Internet traffic. Departing from normal treatises, we will take a view from
both the network engineering and physics perspectives showing the strengths
and weaknesses as well as insights of both. In addition, many less covered
phenomena such as traffic oscillations, large-scale effects of worm traffic,
and comparisons of the Internet and biological models will be covered.
###### keywords:
complex networks; Internet; internet traffic; self-similarity; fractals; self-
organization; phase transition; critical phenomena; congestion;Internet
Protocols
## 1 Introduction
In the last ten years, research of networks, especially the Internet, has
exploded amongst physicists. This follows at least twenty years of research on
the same and similar questions among computer scientists and network
engineering researchers. Interest in the physics community began with several
seminal papers on small-world and scale-free networks [57, 194, 15, 153, 141].
Since then, this research has progressed at a rapid pace borrowing well-
developed tools from statistical mechanics and thermodynamics, spectral graph
theory, and percolation theory among others to enhance the understandings of
fields such as Internet topology and social network analysis previously only
researched by the network engineering and sociology communities. This new
interdisciplinary work has been very fruitful. However, there are gaps only
more recently being addressed. In particular, while the contribution of
physicists to the understandings of topology and community structure in
networks is substantial, often both physicists (and their counterparts in
sociology or network engineering) are often unaware of work on the similar
questions in other fields. In addition, similar theoretical understandings and
predictions for network dynamics such as those now being reached for topology
remain elusive and are still in the earliest stages. There have been good
papers written on dynamics by physicists but they have yet to formulate
results with the same generality or detail as the results on topology.
Just like the fist tentative measurements of Internet topology allowed the
field of networks to grow, Internet traffic dynamics have provided a similar
opportunity. Serious research on the macroscopic nature of Internet traffic
can be traced almost to its inception, however, only in about the last 15
years has the field come of age and begun to provide truly deep insights into
how communication over the largest technological edifice in human history
operates. Within this time the terms, “self-similarity”, “multifractal”, and
“critical phenomena” have emerged to try to refashion our ideas about the
Internet and how it behaves. This paper could just be a review of the work by
physicists and engineers on network traffic, however, my familiarity with the
research in the field has convinced me that for a full understanding of the
state of the art research in Internet traffic dynamics, separating the views
of engineers and physicists would make any analysis incomplete and inadequate.
An area for improvement in this research on network traffic is the increasing
collaboration and cross-citation of works from other fields. Though in the
study of networks there are notable exceptions, in general, physicists and
engineers studying the Internet conduct their own research projects, using
field specific methodologies, and publish in field specific journals with
little cross-citation of relevant results from the other disciplines. Indeed,
one can see from the average paper in physics journals such as Physical Review
E or Physica A or from engineering journals such as the IEEE or ACM series of
journals that many interrelated problems are being studied from totally
different perspectives. This has sometimes caused tension, especially between
physicists and network engineers, about the utility, details, or real-world
validation of physics based theories and the lack of generality and generally
applicable network principles among the engineering perspective.
In figure 1, I have tried to give a full diagram and summary of how these
different world views operate. If someone is trying to see which view is
completely “right” or “wrong” they are missing the point that the Internet is
in some part everything that both sides describe it as. It is an example of an
engineering system that is dependent on the precise nature of its protocols
and other workings to function. It is also a large-scale self-organized system
not far removed from those that physicists have studied in physical systems
for years. However, the research rarely reflects a full synthesis of both
views. Also, both perspectives are more correct on some points than on others.
For researchers interested in this area, or even those who feel themselves to
be seasoned veterans, one of the first papers I recommend is [102] which is a
short position paper released by the 2006 CAIDA Workshop on Internet Topology.
Attended by eminent physicists and engineers, this workshop clearly spelled
out the promise and peril of multidisciplinary research on Internet topology
and traffic.
In line with the increasing focus on network dynamics, and the reality that
many of such research projects involve the Internet, this review paper is
meant to familiarize mainly physicists and engineers with the major results of
each other and how they interrelate. For physicists, hopefully it will provide
more exact information on the workings of packet switching systems in the
Internet in order to allow us to better test our predictions against reality,
build more realistic models and simulations, validate models and theories with
real data such as traffic traces, and contribute to the study of network
dynamics through a more complete understanding of the dynamics of the
Internet. For network engineers, they can see the issues raised by statistical
mechanics approaches towards network features such as congestion and realize
that there can possibly be large-scale phenomena quite independent of detailed
technical specifications. Since the Internet probably has the largest readily
accessible and easily understandable archive of network traffic dynamics, it
could likely play a huge role in empirically validating theoretical ideas and
simulations of dynamics in networks.
This paper is organized in order to provide not only a review of current
research in the field but to provide a basic introduction in the workings of
the Internet for newcomers to the field. First, I will review the basic ideas
of Internet traffic including packets, definitions of flows and throughput,
and the basic protocols. While this may seem common knowledge, much work in
the field can only be understood if you have the correct definitions and
knowledge of the Internet basics. Therefore, this is provided to prevent
confusion and perhaps inform on less discussed topics. Next, we will discuss
the evolution and composition of Internet traffic as far as usage and
protocols are concerned and study the basic dynamics of packet flow including
packet size distribution and Internet flow characteristics. The meat of the
paper involves a detailed discussion of the self-similar nature of Internet
traffic and how it is defined and measured, the detailed workings and dynamics
of the TCP transport protocol, as well as the large body of work by physicists
on phase transition and critical phenomena models on packet switching
networks. These techniques will be reviewed and their possible promises and
current pitfalls addressed. Finally, several interesting and related phenomena
present in Internet traffic such as oscillations will be covered. Each idea is
given a firm grounding and a thorough introduction but there will be no
pretense that I can completely delve into all research on any of these ideas
in one paper. Self-similar traffic alone has already inspired several volumes
on even its most esoteric aspects. However, it is hoped this paper will allow
someone with a reasonably technical background and minimal familiarity with
the subject and research to quickly grasp the main themes and results that
have emerged from the research. Even for those that consider themselves
experts, there may be small insights or details that have been poorly covered
in most treatments and may add to their knowledge of the subject.
Figure 1: Network engineers and physicists often have diverging viewpoints on
similar Internet traffic phenomena due largely to their backgrounds and
training as well as the fundamental questions they ask. Here is a summary of
those perspectives. On the top and bottom are the background knowledge and
viewpoints of each side and in the middle are the problems they typically
tackle and how they ask the questions.
## 2 Packets, OSI Network Layers, and Key Terminology
This first section of the paper will be its longest and will explain from the
ground up the nature of Internet packet traffic and its key transport level
protocol, TCP. This information is essential to those attempting to understand
the Internet’s functions and allows one to both understand other work in
papers across many fields as well as develop more realistic models of the
Internet based off its actual parameters and not idealizations.
### 2.1 Packets and OSI
In 1969, the Internet (then ARPANET) was first established as a distributed
packet communications network that would reliably operate if some of its nodes
were destroyed in an enemy attack as well as facilitate communication between
computer centers in academia. Though the Internet has changed greatly up until
today, its packet switching mechanism and flexibility remain its key aspects.
The packet is the core unit of all Internet traffic. A packet is a discrete
bundle of data which is transmitted over the Internet containing a source and
destination address, routing instructions, data description, a checksum, and
data payload. Packet handling and traffic management are governed by a complex
set of rules and algorithms collectively defined as a protocol.
Different protocols are responsible for handling different aspects of traffic.
Though this may seem trivial, protocols heavily affect the nature of traffic
and models of traffic which may be completely valid for one protocol can be
completely invalid for another. Also, protocols are used in different
applications or tasks and these should inform any analysis of Internet traffic
or predictive models describing its behavior.
In addition, there are levels of tasks handled by certain protocols and not
others. These are broken out, traditionally into seven layers, by a model
known as the Open Systems Interconnection (OSI) model. These seven layers are
shown and described with examples in table 1. For analysis of packet data, the
application, transport, network, and data link layer are typically the most
relevant.
Layer | Number | Description | Example Protocols
---|---|---|---
Application | 7 | Network applications such as terminal emulation and file transfer | HTTP, DNS, SMTP
Presentation | 6 | Formatting of data and encryption | SSL
Session | 5 | Establishment and maintenance of sessions | TCP sessions
Transport | 4 | Provision of reliable and unreliable end-to-end delivery | TCP, UDP
Network | 3 | Packet delivery, including routing | IP
Data Link | 2 | Framing of units of information and error checking | Ethernet, ATM
Physical | 1 | Transmission of bits on the physical hardware | 10BASE-T, SONET, DSL
Table 1: Breakdown of 7 layer OSI model. Descriptions taken from [152]
The higher layers (higher number) always initiate a lower level protocol. For
example, for e-mail using the application protocol SMTP, SMTP starts a TCP
connection which itself uses IP packets to deliver data.
Even within the same layers though, protocols can function much differently.
By far the most well-known and widely used suite is the transport/network
protocol combination TCP/IP. Transmission Control Protocol(TCP), which manages
sessions between two interconnected computers, is a connection based protocol
which means it has various means of checking and guaranteeing delivery of all
packets. This is why it is widely used to transmit web pages using Hypertext
Transfer Protocol (HTTP), email with Simple Mail Transfer Protocol (SMTP), and
other widely used applications. TCP’s connectionless cousin is User Datagram
Protocol (UDP). UDP sends packets without bothering to confirm a connection or
receipt of packets. This can make it unreliable for delivery but much faster
and more useful for real-time applications like voice over IP (VoIP). TCP will
be covered in more detail and its differences elaborated later in the paper.
These differences cause TCP to react to feedback in its traffic patterns and
adjust its throughput based on these considerations.
### 2.2 Packet structure
Packets have two main parts: a data payload which contains specific data being
transmitted and overhead which contains instructions about packet destination,
routing, etc. Each level and protocol has a different amount of overhead as
shown in table 2. Overhead usually has both a fixed and variable portion.
However, for most transport and network layer protocols the fixed portion can
usually be considered the size of the entire header. When dealing with the
total size of packets and measuring throughput, one must be careful to specify
whether or not the packet size includes overhead. Also, at the data link
layer, there is a maximum frame size of 1500 bytes in most systems (minus data
link layer overhead). The effect of packet size and packet size distribution
will also be covered in more detail later in the paper.
Figure 2: Structure of a packet in this paper. Proportions based on a 50 byte UDP packet payload. Numbers are size of headers or payload in bytes. A is the Link Layer (i.e. Ethernet) header which contains MAC address source and destination and payload type, B is the Internet Protocol (IP) header, C is the transport layer TCP/UDP protocol header, D is the data payload and E is the Link Layer (i.e. Ethernet) CRC checksum hash to prevent accidental corruption of the frame. Protocol | Header Size
---|---
IP | 20 bytes for IPv4, 40 bytes for IPv6
TCP | Normally 20 bytes, can be up to 60 bytes
UDP | 8 bytes
Ethernet | 14 bytes for header and 4 bytes for checksum
Table 2: Packet header sizes for prominent protocols
### 2.3 Packet traffic characteristics
Larger than individual packets is the packet flow which can be statistically
described using many important measures. Probably the most widely known and
important are bandwidth, throughput, goodput, packet flow rate, flow, latency,
packet loss, and Round Trip Time (RTT).
Bandwidth - Bandwidth is the maximum possible throughput over a link.
Bandwidth, being an ideal, is almost never achieved under normal conditions
but provides a convenient benchmark to compare the capacity of data links.
Throughput - Throughput is the rate of packet transmission over a network
link, usually in megabits per second (Mbps) or kilobits per second (Kbps). It
is the most widely recognized measure of network data speed and essential in
understanding the performance of data traffic.
Goodput - Goodput is the measure of throughput excluding packet overhead. When
analyzing data, one must be careful to ascertain whether traffic data is
throughput or goodput. If it is goodput, the total throughput is actually
larger because you have to incorporate the average packet overhead in the
amount of data transferred. However, for both throughput or goodput, the
packet flow rate is the same.
Packet Flow Rate - The rate of packet flow over a network link. This differs
from throughput or goodput in only measuring the number of discrete packets
that travel over the link, regardless of their size. Throughput and packet
flow rate are related by the following equation
$T=s\lambda$ (1)
Where s is the average packet size including overhead, T is the average
throughput, and $\lambda$ is the average packet flow rate.
Session - A TCP communication dialog set up between two computers by first the
delivery of a SYN packet, coordination request and (SYN+ACK) packet, and
finally an ACK by the initiating party.
Flow - This flow must be carefully distinguished from the packet flow rate
mentioned above. A flow in Internet traffic is defined several ways but in
general is a connection between a source and destination which is continuously
transmitting data. Usually, this means a connection based protocol such as TCP
where a connection is made and data continuously transmitted until the
connection times out or a standard inter-packet arrival time is exceeded in
the “packet train” [92, 36]. Sometimes traffic instead of bytes is also
measured in terms of the number of flows. The distribution of flow sizes and
their properties will be covered later in the paper. Note when measuring flows
using traffic auditing software such as Netflow or ARGUS, one must make sure
to verify the exact method of flow measurement and make sure measurements
across different systems are comparable and whether or not they match any of
the theoretical definitions above.
Packet Loss - This is the percentage of all packets lost in transit. It is
usually measured as the percent difference between packets transmitted in a
packet flow and packets received on the other end from the same packet flow.
This affects all traffic and is caused by many causes that lead to the loss of
packets including congestion, physical connection problems, and other issues.
It has a large effect on TCP throughput.
Round Trip Time (RTT) - The statistical average time it takes a packet to
travel from a source to a destination and back. It is the most common measure
of latency on computer networks. It is closely related to throughput and along
with packet loss often used as a measure of link congestion.
### 2.4 Protocol traffic breakdown
Most prominent studies of Internet traffic have tended to be done over high
traffic transit networks or links linking smaller networks, such as those at a
university, to the overall Internet. Overall, TCP dominates all traffic with
about 95% or more of total bytes, 85-95% of all packets and 75-85% of all
flows using TCP as the transport protocol [184, 34, 104]. UDP comes second
representing about 5% or less of traffic with its main function being sending
DNS requests and communications. These results generally hold for the larger
Internet, however, in measuring traffic protocol representation one must be
careful to define whether a high traffic transit link, carrying traffic from
many sources and applications, is used versus a link on an edge network closer
to final users. Edge networks can have extremely particular traffic traces
when there is large use of UDP for VoIP traffic or other multimedia type
traffic.
TCP application traffic has generally evolved over time in three main eras
characterized by the dominant types of traffic influenced by available
applications and access speeds. In the Text Era (1969-1994) most TCP traffic
was driven by email, file transfers, and USENET newsgroups. In 1989, Cáceres
[26] at UC Berkeley characterized Internet traffic of being 80% TCP and 20%
UDP by packet and 90% TCP and 10% UDP by bytes. TCP traffic bytes were split
roughly evenly between SMTP for email and FTP for file transfer while UDP was
mostly DNS. An updated study by Cáceres and collaborators in 1991 [27]
monitored traffic at several universities again finding similar results. Once
again at UC Berkeley 83% of packets were TCP, 16% were UDP, and about 1% were
ICMP. UDP traffic was predominantly for DNS at 63% of its packets. TCP traffic
in terms of packets was 28% telenet, 16% rlogin (Unix host login utility), 12%
FTP, 12% SMTP, 12% NNTP (USENET), with the balance shared among other
protocols. FTP was the largest protocol in bytes at 36% of all bytes. These
dominant application level protocols were confirmed by Claffy and Polyzos as
well [35].
Next was the Graphics or Hyperlink Era (1994-early 2000). After CERN made the
World Wide Web free for any use in 1993, the graphics based web grew rapidly.
In 2004, Paxson [155] reported that in Internet traffic though FTP, SMTP, and
NNTP still held sway, HTTP was by far the fastest growing protocol growing
300-fold in traffic measured by connections in only two years and already
vying to be the 2nd most popular TCP application level protocol. By 1995, WWW
traffic had become the largest application level protocol with 21% of traffic
by packets compared to 14% for FTP, 8% for NNTP and 6% for SMTP [71]. By 1997,
Thompson, Miller, & Wilder [184] could report that HTTP dominated TCP traffic,
95% of all Internet traffic bytes at this point, with 75% of the overall
bytes, up to 70% of the overall packets, and 75% of the overall flows during
daytime hours. Its closest competitor, SMTP, was reduced to only 5% of packets
and 5% of all bytes. The Internet was now a popular, mainly web and graphics
based medium.
The current era is the Multimedia Era (early 2000-present). In this period,
sharing of multimedia through P2P file sharing applications and streaming
audio or video began to rival the web for dominance of Internet traffic. A
Sprint study on an IP backbone in early 2000 [70] reports that P2P was already
rivaling the web in terms of bytes transferred with at times P2P accounting
for 80% of all traffic. Streaming also accounted for as much as 26% of all
traffic as well. The web was still competitive, however, sometimes accounting
for 90% of all traffic. By 2004, however, Fomenkov, et. al. [68] could report
WWW traffic clearly peaked in late 1999/early 2000 and P2P had dominated
traffic growth ever since. A recent April 2008 traffic trace study [19] shows
the Web and P2P sharing 34% and 33% of total TCP/IP bytes respectively.
However, P2P only accounts for about 3% of all flows compared to the 40% of
all flows dominated by the web showing P2P flows are generally larger and more
likely to be “elephant flows”. Another earlier study by the same team [54]
gave similar results with Web and P2P (normal and encrypted) consisting of 41%
and 38% of bytes and 56% and 4% of flows respectively.
One caveat on methodologies is necessary with more recent traffic. Though it
is undisputable that multimedia and P2P type applications are carrying large
amounts of traffic, identifying their exact nature based on port numbers is
not always as exact as identifying HTTP traffic from port 80. There have been
several methodologies created for traffic identification that use statistical
measures in addition with TCP port numbers to determine traffic types. These
include CoralReef (from CAIDA), BLINC[95], WEKA a suite of machine learning
algorithms (from the University of Waikato, New Zealand), and graph based
methods[91]. Reviews of these methodologies evaluate their relative usage and
show though port numbers are still the largest factor, they are often not
completely reliable due to users or software intentionally or unintentionally
changing port numbers for various hardware and applications [96, 131, 100].
### 2.5 Topology
As mentioned before, topology is currently the most studied feature of the
Internet and other computer networks by physicists. Due to the wide range and
depth of research being done in this field, this paper will not present even a
cursory review of its main ideas and results. Dealing with Internet traffic,
however, it is important to balance the knowledgeable networking engineering
perspective with the more abstract methods used by physicists. For the now
widely familiar methods of physicists, the author recommends several
outstanding review papers [7, 142, 45, 20, 173, 166, 154]. However, anyone
measuring or analyzing network topology network must read network engineering
contributions and rebuttals to many of the now standard scale-free network
techniques and theories. This is usually poorly covered or not at all in the
physics community. In particular, the author recommends [199, 48, 118, 119, 9,
202, 86] which tackle the power-law distribution and “robust yet fragile”
nature of the Internet that physicists have theorized. A proposed alternative
to general preferential attachment growth models for the Internet is given in
[29, 119]. Here the authors present a growth model that optimizes growth based
on business and location considerations that match real Internet growth
processes.
Also in [103, 10, 121] the issues of data integrity in measuring the Internet
topology are described including the errors introduced by several common
methods such as traceroute and BGP router tables. Essential reading for anyone
looking at this area. Finally, as a guide before claiming anything as a power
law, the author urges these two excellent articles on tackling this
determination in a rigorous fashion [37, 128, 130].
### 2.6 Packet Sizes
#### 2.6.1 Distribution of packet sizes
Internet traffic, with its various protocols and traffic types, has many
widely varying packet sizes. However, there is an upper limit to packet size
and this is almost always determined at the data link layer. Various data link
communications schemes, such as Ethernet or ATM, impose an upper bound on the
size of transmitted packets through the hardware or operating system settings.
This upper bound packet size is often designated the Maximum Transmission Unit
(MTU) at the link layer.
In Ethernet, the current MTU on most systems is 1500 bytes. Packets at the
data link layer are often termed frames but the idea is the same. This 1500
bytes includes the payloads and headers of all lower level protocols but does
not include the Ethernet frame header and footer. Several studies on packet
size distributions have shown that packet size is in general a bimodal or
trimodal distribution with most packets being small (500 bytes or less) [99,
184, 46, 34, 104]. In addition, the distribution of packet sizes is not a
smooth long-tailed distribution in that some packet sizes can predominate due
to system defaults.
For example, [99, 184] describe that there are peaks in the frequency
distribution for packet sizes. In a traces of data over a day or longer on a
data link, they explain many reasons for the small packet size. First, for TCP
systems there is a protocol option for “MTU discovery” that tries to find the
MTU of the network in order to make packets as large as possible. If MTU
discovery isn’t implemented, TCP often defaults to an MTU of 552 or 576 bytes.
Also, nearly half of the packets are 40-44 bytes in length. These packets are
used by TCP in control communications such as SYN or ACK traffic to maintain
the connection between the source and destination systems. At 576 bytes [99]
the packet size increases linearly to 1500 bytes showing that packet sizes in
the intermediate region are relatively equally distributed. In general,
according to [184, 34, 104] about 50% of the packets are 40-44 bytes, 20% are
552 or 576 bytes, and 15% are 1500 bytes. Table 3 shows the distributions of
packet sizes from a traffic trace and fit well with the studies except the
absence of a strong peak in the 552 or 576 byte range.
Packet Size Range | ALL | TCP | UDP
---|---|---|---
0-19 | 0% | 0% | 0%
20-39 | 2% | 0% | 0%
40-79 | 59% | 69% | 19%
80-159 | 7% | 2% | 23%
160-319 | 3% | 1% | 15%
320-639 | 7% | 3% | 34%
640-1279 | 3% | 3% | 6%
1280-2559 | 18% | 22% | 4%
Table 3: Packet size distribution of a capture of 1 million packets in a 100
second trace from the MAWI [125] traffic trace archive from Samplepoint-B on
July 22, 2005.
Kushida [104], despite being one of the only papers that looks at packet size
distribution among UDP separately, clarifies that since 98.2% of the traffic
measured in the paper is TCP, the UDP contribution to overall IP and Internet
traffic packet size distribution should be considered negligible. Using a
different measurement for packet size distribution, that looks at the ratio of
packet size*number of packets versus the total traffic measured, Kushida finds
a series of peaks between 75 and 81 bytes and another large peak at 740 bytes.
None of these peaks are substantial, however, and no size of packet reaches
even 10% of the total. Since UDP has no connection based features such as TCP,
the reason for these peaks is not inherent in the protocol itself. UDP is
mainly used for Domain Name Server (DNS) and Simple Network Management
Protocol (SNMP; a network monitoring protocol) and applications related to
these functions drive the size of the UDP packets.
Finally, there is often an asymmetry in packet size for both directions in a
flow. For example, if a web page is being served to a PC, the PC will be
receiving large TCP packets with HTTP (WWW) data while it will only be sending
comparatively smaller packets as data requests back. Packet size also has
diurnal variations where it can be larger during daytime hours and in
international links, [99] showed that the average packet size on both
directions of the link oscillated out of phase by about 11 hours (2.9
radians).
Does packet size or MTU matter? Absolutely, in fact many network engineers
realize that average packet size and MTU are critical factors in determining
the overall maximum throughput in a network. Recalling equation 1, for a fixed
throughput, decreasing the packet size increases the packet flow rate.
Oftentimes, many believe that the key throttle in computer network throughput
is its stated bandwidth. In fact, bandwidth bottlenecks are rarely the
bottleneck on network performance. Computer network hardware typically has a
maximum packet flow rate it can effectively handle, afterwards packets begin
forming queues in the hardware buffer and congestion reduces throughput.
Smith, in [170] showed how on a normal Ethernet link between two computers,
the maximum throughput across varying packet sizes exhibited a transcritical
bifurcation.
In fact, one problem currently plaguing next generation high speed networks is
outdated, smaller MTUs on the systems of their users. Therefore, in order to
take advantage of the increasing bandwidth capabilities of the Internet, there
is a concerted push in some corners to raised the typical MTU above the normal
Ethernet 1500 bytes, up to 9000 bytes where possible, to allow more rapid
communication. There will need to be larger studies on network hardware such
as routers to understand how the MTU completely affects throughput and whether
the bifurcating behavior is present for saturated links in the Internet at
large.
### 2.7 Flow Size Structure and Distribution
#### 2.7.1 Definition and nature of flows
As mentioned in the definition of an Internet flow, a flow is defined as
continuous communication between a source and destination system. Flows are
typically described by one of two definitions: an identifiable clustering of
packets arriving at a link or by identifying characteristics such as the
source and destination addresses along with an identifying label such as a TCP
session ID or an IPv6 flow label.
For the first definition, the most widely used definition was given by Jain
and Routher in 1986 [92] while studying data on a token ring network at MIT.
While also noting that the interpacket arrival rate is neither a Poisson or
compound Poisson distribution, they defined individual flows as “packet
trains” where a packet train is defined as a sequence of packets whose
interarrival times are all less than a chosen maximum interarrival gap,
usually determined by system software and hardware configurations. If a packet
is received after a longer interval than the maximum gap, it is considered
part of a new flow. This brings up one important characteristic of Internet
flows: though they obviously have a time average, they are extremely bursty
and inhomogeneous compared to most other types of flows studied in physics.
For the second method, the first and still likely most widely used method of
identifying flows via address or label is using TCP packets. TCP flows start
with a SYN packet and end with a FIN packet. Therefore matching SYN and FIN
packets with source and destination IP addresses and session ID in the TCP
headers are often used to define flows. Another elaborate definition was
presented by Claffy et. al. [36] who represent a flow as active if there is
interpacket time less than a maximum value and distinguish flows by a group of
packets identified by aspects including source/destination pairs,
unidirectional nature (flows in only one direction), protocols used, and other
factors that may distinguish the packet destinations.
The new next generation Internet Protocol, IPv6, though not yet implemented
widely beyond a now defunct test network called 6Bone, has been designed with
a part of its header overhead reserved for a “flow label”. This flow label
would allow the traffic source to provide a unique identifier that would
clearly distinguish IP traffic flows. Besides improvements in routing and
traffic management, this will allow more accurate research as IPv6 is
implemented throughout the Internet.
#### 2.7.2 Distributions of flow characteristics
Early in the paper, it was mentioned that long-tail behavior is present in
Internet traffic to the same extent as it is in the topology. Flows are no
exception and several quantities used to describe flows have long-tail
distributions. In particular, the distributions of sizes in terms of data
transferred, duration in terms of length of the flow, and data rate of flows
have all been found to exhibit long-tail distributions.
These flows have been given certain names throughout the literature which are
summarized by Lan and Heidemann [111]. Flow sizes are divided into two
classes: “elephants” and “mice “where elephants are a small part of all flows
measured over a certain time but account for a large number of the bytes
transferred while many other flows account for proportionally smaller
components of the overall traffic (the mice). Elephant flows have been
described in detail in several papers [111, 56, 58, 110, 136, 147], in
particular in a paper by Mori et. al. describes a traffic trace where elephant
flows are only 4.7% of all flows measured but 41% of all traffic during the
period. Barthélemy et. al., [18] give a related result studying routers on the
French Renater scientific network. They conclusively find that a small number
of routers (a so-called ’spanning network’) submit the vast majority of data
on the network while the contribution of the other routers is exponentially
smaller.
Elephant flows, though agreed upon in principle, have been defined differently
in many papers. Estan [56] defined an elephant flow as a flow that accounts
for at least 1% of total traffic in a time period. Papagiannaki [147] uses
flow duration to classify elephant flows. Lan and Heidemann [111, 110] use a
statistical definition where a flow is considered an elephant flow if the
amount of data it transmits is at least equal to the mean plus three standard
deviations of flow size during a period. This is 152kB in their paper. This
final definition implicit assumes the scaling exponent among flow sizes,
$\alpha$ is at least 2, since the variance for the distribution is infinite if
$\alpha<2$.
In figure 3, the author has used data from the WIDE MAWI [125] traffic trace
archive which measured the daily traffic over a T-1 line between Japan and the
Western US to show the relative proportion of all traffic the top 10 flows
represented over time from 2001-2007. The upward tick in mid-2006 reflects the
upgrading of the data link speed from 100Mbps to 1Gbps. The % of all traffic
captured by the top 10 flows declines over time as the number of overall flows
per day increases and the top 10 occupy a declining share of the number of
flows.
Figure 3: Percent of data in all flows occupied by the top 10 flows over time.
From the WIDE MAWI traffic trace archive [125] using data from Samplepoint-B
from 12/31/2000 - 5/31/2007. Median daily flows total about 350,000
Research from Mori et. al. [136] also gives evidence that elephant flows not
only occupy disproportionate amounts of traffic but they are also more likely
than mice to be responsible for congestion in links.
The duration of flows has been classified with similar zoological flair. Most
flows have a relatively short duration while a small number of flows have a
comparatively very long duration. Brownlee and Claffy [24] analyzed duration
among Internet streams, which are individual IP sessions versus one way flows
of packets typifying flows. About 45% of streams were of a very short
duration, less than 2 seconds, and were termed “dragonflies”. Short streams
were defined has having a duration from 2s to 15 minutes and consisted of
another 53% of all flows. “Tortoises” were flows with a duration greater than
15 minutes and accounted for 1-2% of all streams but 50% of all bytes
transferred. The dragonfly/tortoise definition is simplified and extended to
flows in [111] where a dragonfly is a flow less than the mean flow duration
plus three standard deviations which is 12 minutes in their paper. They find
70% of all Internet flows are less than 10 seconds.
Lan and Heidemann [111] also introduce a new measure of flow, “cheetahs” and
“snails” to characterize the distribution of throughput in flows. Cheetahs are
flows with an average throughput greater than the mean plus three standard
deviations. Their dividing throughput is 101 kB/s in the paper. According to
their measurements, about 80% of Internet flows have a throughput of less than
10 kB/s.
These different types of measurements on flows are obviously not independent
and in fact are heavily correlated in several ways. Cheetahs tend to be high
throughput but small in size and short in duration Zhang [18] previously
showed a correlation between flow size and rate and Lan and Heidemann [111]
confirm this showing 95% of cheetah flows are dragonflies with a duration of
less than 1 second. 70% of cheetah flows are also smaller than 10 kB. Elephant
flows tend to be large in size and duration but low in throughput. Only 30% of
elephant flows in [111] are faster than 10 kB/s and 5% are faster than 100
kB/s. 50% of elephant flows lasted longer than 2 minutes and 20% of elephant
flows lasted at least 15 minutes.
Different flow types are also dominated by different types of traffic.
Elephants seem to be mostly web and P2P traffic while Tortoises are mostly
DNS. Cheetahs have by far mostly web and DNS traffic. This leads to the
question as to the causes of bursty flows (as differentiated from bursty
overall traffic described in the section on self-similarity). TCP mechanisms,
described in the next section, could cause the burstiness of some flows,
however, it could also be driven by the bursty data requests from user
applications such as interactive applications like MMORPGs.
The granularity and nuance in the characteristics of flows is an interesting
theoretical and practical challenge for those studying Internet dynamics. But
it still gets even better as our next section on the dynamics of TCP traffic
demonstrates.
Category | Large-size | Long-lived | Fast | Bursty
---|---|---|---|---
Elephant | Y | Y | N | N
Tortoise | N | Y | N | N
Cheetah | N | N | Y | Y
Table 4: Classification and description of flows from Lan and Heidemann [111]
## 3 Explanation and Mechanics of TCP
Now that we understand the basics of packet traffic, we can dig more deeply
into one of the key functions of the Internet: connection based packet
delivery through TCP. As stated earlier, TCP by far is the bulk of Internet
traffic. Therefore, any discussion on Internet traffic is by and large a
discussion of TCP/IP traffic. TCP is a connection based protocol and relies on
several programmed algorithms to manage and guarantee the delivery of packet
traffic. The fact is, however, TCP was developed when the Internet was
relatively small. Though it is still useful and efficient, the large-scale
macroscopic effects of its operation were not easily predictable and were only
measured or derived later.
Volumes of articles have been written on TCP behavior, possible algorithmic
improvements, and traffic management. TCP has several features including
buffering and congestion control that allow it to be one of the only Internet
protocols that uses feedback to adjust protocol performance. Nonlinear effects
combined with feedback have been well-known to produce complex systems
phenomena and TCP is no exception. In this section, TCP’s basic mechanisms
will be defined and explained and then linked with the most common theories of
network performance and congestion.
### 3.1 A short explanation of TCP
There are many good guides on TCP, but most information in this paper is taken
from an IBM guide [152]. TCP relies on several key features which are
necessary to ensure reliable and smooth delivery of packets between the source
and receiver. The TCP connection starts out with a “three-way handshake” which
consists of one SYN (synchronize) and two ACK (acknowledge) packets. TCP flows
are based on the concepts of windows and flow control. When packets are
transmitted they are given sequence numbers to determine the correct order of
data transmission. The source then waits to receive and ACK packets before
transmitting additional packets.
The number of packets a source can transmit before needing to receive at least
one ACK packet is the window. When a TCP connection is initiated and as it
continues, the receiver sends an ACK packet which lets the source know the
highest sequence number it is able to accept given buffer memory and system
constraints. The source then sends the number of packets to fit that window
and waits for an ACK. For every packet confirmed, an additional one is sent
and the window size is maintained. The window size can be changed by the
receiver in every ACK packet by varying the highest sequence number it can
receive so the window often varies over the course of a transmission. If an
ACK for a packet is not received within a timeout period, TCP considers the
unacknowledged packets lost and retransmits them.
Because of the possible need for retransmission, TCP must buffer all data that
has been sent but has not received an ACK. The size of this buffer at the
sender is usually calculated by the bandwidth delay product which is the
product of the link bandwidth and the RTT. Therefore, for high bandwidth links
or long RTT links, the buffer can become increasingly large and burdensome on
the operating system.
### 3.2 Congestion control
Figure 4: Illustration of the window size of a hypothetical TCP flow over
time. Each type of congestion control mechanism is illustrated in the figure.
ssthresh1, ssthresh2, and ssthresh3 are the first, second, and third slow
start thresholds over time respectively. ssthresh2 is 1/2 of the value where
the timeout congestion occurred and ssthresh3 is 1/2 the value where the
duplicate ACK congestion occurred. The second congestion is caused by
duplicate ACKs and initiates fast retransmit and fast recovery.
Congestion control is one of the most prominent and differentiating aspects of
TCP as compared to UDP and any other transport level protocol. Flow control
uses ACK feedback to coordinate smooth transmission between the sender and
receiver and congestion control uses feedback from the network throughput
environment to adjust the sender rate in order to not cause network
congestion. Thus Internet traffic, which is largely TCP/IP, behaves in part as
a massive closed loop feedback system between the transmission throughputs of
multiple senders which are modified by the measured traffic congestion
environment. This has doubtlessly led to much research in self-organization in
TCP traffic which will be discussed later.
The congestion control algorithm is not uniform across all TCP software
implementations and has various flavors named after resort cities in Nevada
including Vegas, Reno, and Tahoe. All implementations though essentially share
four general features: slow start, congestion avoidance, fast recovery (not in
Tahoe), and fast retransmit.
Slow start is used to address the inherent problem that regardless of what TCP
window the receiver advertises in its ACK packets, the network still may be so
slow or congested in points as to not be able to handle that many packets
transmitted over such a short time. Slow start handles this by overlaying
another window over the TCP window called the congestion window (cwnd). At
first, this window starts at one packet and tests to see if an ACK is
received. If so, the congestion window grows to two, waits for two ACKs, and
then grows to 4, increasing by powers of two at each successful step. This
occurs until the window is the maximum size advertised by the receiver.
Afterwards, the receiver’s window size can increase by one segment per ACK
successfully received. The sender will eventually use the smaller of the
congestion window or the window advertised by the receiver as the TCP window.
This insures that if the receiver’s window is too large for current network
performance, the congestion window will compensate. Slow start has a threshold
though which is the maximum size possible for any congestion window.
Congestion avoidance works in tandem with slow start. Congestion avoidance
assumes that any packet loss (packet loss is normally assumed to be much less
than 1%) signals network congestion. As stated earlier, this is not
necessarily true as hardware or network issues which cause lost or dropped
packets and aggressive “crowding out” by large flows of aggressive
connectionless protocols like UDP can also generate this effect. The
connection measures packet loss by a timeout or duplicate ACK packets. If
congestion is detected, congestion avoidance slows down the TCP connection by
setting the slow start threshold to one half of the current congestion window,
the so-called exponential backoff. If a timeout caused the congestion, the
congestion window is “reset” down to one packet and slow start repeats. Slow
start increases the window size to the new slow start threshold and then
congestion avoidance takes control of the congestion window size. Instead of
increasing the window size in an exponential manner, congestion avoidance
increases it in increments with every ACK received according to the following
equation
$\frac{\mbox{segment size}}{\mbox{congestion window size}}*\mbox{segment
size}$ (2)
where the segment size is the size in bytes of data TCP fits into each packet.
Therefore the congestion window increases linearly controlled by the
congestion window algorithm.
Fast retransmit is where TCP uses the number of duplicate ACKs received to
determine whether a packet was received out of order at the receiver or likely
dropped. If three or more duplicate ACKs are received it assumes the packet
was lost and retransmits it. This prevents TCP from having to wait the entire
timeout period before retransmitting. Since fast retransmit is based on the
assumption of a lost packet, congestion avoidance comes into play. However,
fast recovery takes over in the situation where fast retransmit is used and
allows the TCP window to not decrease all the way to one and restart using
slow start but by setting the threshold to one half of the congestion window
size and starting the congestion window size at the threshold size + 3 x
segment size. The congestion window then increases by one segment for each
additional ACK.
### 3.3 TCP macroscopic behavior
The intricacies of the operation of TCP have lead to much research
characterizing the protocol’s average or expected performance and its
influence on the overall traffic patterns of the Internet. One of the best-
known and widely reported results is a famous equation for the maximum
possible throughput for a TCP connection developed in early versions by Floyd
[65] and Lankshman & Madhow [109] and in a most widely known version by
Mathis, Semke, and Madhavi [123]. They explore the expected performance of TCP
against a background of random, but constant probability, packet loss given
the window resizing by the congestion avoidance backoff mechanism. Their
famous result (often called the SQRT model) is for the theoretical maximum
throughput for a TCP connection and is given by
$T\leq\frac{\mbox{MSS}}{\mbox{RTT}}\frac{C}{\sqrt{p}}$ (3)
where $MSS$ is the maximum segment size, typically defined by the operating
system for TCP and is usually 1460 bytes [160], $p$ is packet loss percentage,
and $C$ is a constant which varies based on assumptions of periodic or random
period loss and the handling of ACK by a congestion avoidance algorithm. Since
C is usually less than 1 the equation can be simplified to
$T<\frac{\mbox{MSS}}{\mbox{RTT}}\frac{1}{\sqrt{p}}$ (4)
This equation assumes packet loss is handled by congestion avoidance and
determined by receiving duplicate ACK packets, not packet loss via timeouts.
Though this is the most famous and widely used equation, a more accurate one,
especially when $p>0.02$ was introduced by Padhye, Firoiu, Towsley, and Kurose
[146]. This equation, based on a version of TCP Reno, also incorporates packet
loss due to timeouts which is more realistic for higher packet loss
situations. Their equation yields an approximation for throughput where
$T\approx\mbox{min}\left(\frac{W_{m}}{\mbox{RTT}}\mbox{,}\frac{1}{\mbox{RTT}\sqrt{\frac{2bp}{3}}+t_{0}\mbox{min}\left(1,3\sqrt{\frac{3bp}{8}}\right)p\left(1+32p^{2}\right)}\right)$
(5)
This equation is also known as the PFTK equation. Here $p$ is once again
packet loss, $W_{m}$ is the maximum window size advertised by the receiver,
$b$ is the number of packets acknowledged by each ACK (usually 2), and $t_{0}$
is the initial timeout value.
In [146] the authors compared the fit to their equation versus the SQRT
equation for real data and state PFTK fits better. A more complete analysis
and comparison was conducted by El Khayat, Geurts, and Leduc [53]. They note
both equations neglect slow start which makes them inappropriate for very
short TCP flows and the equations also neglect fast recovery. To test the
models they generated thousands of random networks with random graph
topologies where the number of nodes (10 - 600) was chosen at random and the
bandwidth (56 kbps - 100 Mbps) and the delay (0.1ms -500ms) were chosen
randomly for each link. They then tested the TCP throughput on these virtual
networks versus both equations and made comparisons using the mean squared
error, $R^{2}$, the over/under estimation ratio of average calculated
throughput to actual throughput, and an absolute ratio which takes the larger
of the first ratio or its inverse. For all metrics, the PFTK equation
performed better but was still a poor predictor of actual TCP performance
giving incorrect estimates roughly 70% of the time.
In addition to measuring the current macroscopic behavior of TCP, there are
many studies attempting to model the behavior of TCP flows under alternate
congestion management paradigms. The most prominent of these is TCP/RED
(Random Early Detection), first proposed by Sally Floyd [66]. TCP/RED actively
manages the router queue by increasing the probability of a router
deliberately dropping a packet as the queue buffer becomes increasingly full
in order to better manage congestion. There are several papers [159, 120, 127]
that state large-scale traffic patterns such as chaos or oscillations could be
induced but these are still mostly based on theory or simulation.
## 4 Packet Arrival Times - Self-Similarity, Long Range Dependence, and
Multifractals
### 4.1 Self-similar traffic and long range dependence
One of the most widely researched and discussed characteristics of Internet
data traffic among both the computer science and physics communities is the
self-similar nature of Internet packet arrival times. Interestingly enough,
the trajectory of research on this begins with a shattering of simplistic
preconceptions about network traffic similar to that of Faloutsos and Barabási
and Albert regarding Internet topology random graph models. Like pre-
Barabási/Albert theory assumed all telecommunications networks, including the
Internet, were random graphs, early research in Internet traffic regarded
packet arrival times as based on a Poisson (Erlang-1) or Erlang-k distribution
similar to that in telephone switching and call center traffic [201]. The
first cracks in this were a paper by Leland and Wilson [115] which showed
packet interarrival times that seemed to exhibit both diurnal fluctuations and
did not seem to adhere to a Poisson distribution. The second paper by Leland,
Wilson, Taqqu, and Willinger [116] thoroughly and convincingly debunked the
theory of Poisson arrival of packets in Internet traffic and using rigorous
statistics showed that Internet traffic had self-similar characteristics and
correlations over long time scales (long-range dependence or LRD). Like degree
distributions in scale-free topologies, the packet arrival per unit time
exhibited long-tail distributions where large bursts of traffic were not
isolated and extremely rare statistical coincidences but par for the course
over all time scales. The studies were based on four captures of data over
four years. The traffic traces, taken at the former AT&T Bellcore research
facility, varied from 20 to 48 hours in length and recorded the timestamps of
hundreds of millions of packets.
The authors are also the first to describe Internet traffic as having a
fractal character. This research has since been confirmed in a torrent of
papers which are too numerous to describe. Paxson and Floyd [157] confirmed
the failure of Poisson modeling in long-tail traffic behavior in Wide Area
Network (WAN) data in several protocols including TCP, FTP, and Telnet.
Crovella and Bestavros [41] demonstrated long-tail distributions in WWW
traffic including packet interarrival times, file download size distributions,
file download transmission time distributions, and URL request interarrival
times. Other papers have essentially confirmed in most cases that long-range
dependence is a key feature of Internet packet traffic.
### 4.2 Measuring self-similarity and long range dependence
There are several good review articles detailing the mathematical techniques
used to investigate self-similar processes in network traffic [149, 94, 199,
55, 162, 3, 198, 201]. Here we will cover the most prevalent and important
ones.
The simplest definition for self-similarity assumes that for a continuous time
process, $X(t)\mbox{ }t\geq 0$, for scaling the time by a factor $c_{1}$,
$X(t)=c_{1}^{-H}X(ct)$ (6)
where $H$ is the Hurst exponent and takes a value between 0 and 1 for self-
similar processes. For a self-similar process that exhibits long-range
dependence, $H$ is between 1/2 and 1. This definition, like most other for
self-similarity, implicitly assumes a stationary process.
The most accepted and widely used definitions are termed the so-called first-
order and second-order similarity. First order similarity is based on the
autocorrelation of the traffic trace. Assuming a traffic trace is defined as a
stationary stochastic process $X$ with a set of values at time steps t:
$X=(X_{t}:t=0,1,2...)$ (7)
and the autocorrelation function, $\rho(k)$, is defined as
$\rho(k)=\frac{E\left[(X_{t}-\mu)(X_{t+k}-\mu)\right]}{\sigma^{2}}$ (8)
Where $\mu$ is the mean and $\sigma^{2}$ is the variance of the traffic. The
self-similar behavior is manifested in that the behavior of the
autocorrelation function is not one which exponentially decays with time as
with a short range dependent time series but rather exhibits a power law
behavior
$\rho(k)\sim c_{2}k^{-\beta}\mbox{ }0<\beta<1$ (9)
Where $c_{2}$ is a positive constant and the approximation symbol indicates
this behavior is the asymptotic behavior of the system as
$k\rightarrow\infty$. Fitting a linear regression to an autocorrelation or
autocovariance plot should not be considered a rigorous or best practice
method of determining self-similarity and the Hurst exponent. There are
various other tools, with shortfalls as well, that are best used to make an
accurate determination.
_Second order similarity/aggregated variance analysis_
Second order similarity is defined as taking the original time series and
recreating it for different time “windows” $m$ where all time values in the
series in windows of length $m$ are averaged. For example, the new time steps
become $t=0,m,2m..,N/m$. Second order similarity, also known as aggregated
variance analysis, is formally defined as taking the new time series
$X^{(m)}_{k}=1/m(X_{km-m+1}+\ldots+X_{km})$ (10)
for all $m=1,2,3,\ldots$. The time series is called exactly self-similar if
the variance of $Var(X^{(m)})=\frac{\sigma^{2}}{m^{-\beta}}$ and
$\rho^{(m)}(k)=\rho(k)\mbox{ }k\geq 0$ (11)
For a normal independent and identically distributed time series the variance
would behave as $Var(X^{(m)})=\frac{\sigma^{2}}{m}$. With self-similarity it
decays much more slowly given the range of $0<\beta<1$ . The time series is
called asymptotically self-similar if the autocorrelation function of the new
time series for large k behaves as
$\rho^{(m)}(k)\rightarrow\rho(k)\mbox{ }m\rightarrow\infty$ (12)
For both definitions of self-similarity, the Hurst exponent $H$ can be derived
from the value of $\beta$ according to the equation $H=1-\beta/2$. This
confines the Hurst exponent to values of between 1/2 and 1 for a self-similar
system. Note $H$ = 1/2 exponent is identical to that of random Brownian motion
and $H$ = 1 reflects complete self-similarity. In most studies, $H$ is
estimated to be around 0.8 in most types of Internet traffic. The results from
the data trace analyzed by the author in figure 5 give a Hurst exponent of
0.81.
One must take care to differentiate two similar but not identical aspects of
Internet traffic: self-similarity, just defined above, and long-range
dependence. Long-range dependence is defined as a system where the
autocorrelation function behaves as
$\sum_{k}|\rho(k)|=\infty$ (13)
when $H>1/2$ for self-similar traffic long-range dependence is implied but in
other conditions you can have long-range dependence but not self-similarity as
long as equation 13 is satisfied. Long-range dependence is also called
persistence and is contrasted by short-range dependence (SRD) which manifests
in processes where $0<H<1/2$. LRD also depends on an assumption of
stationarity in traffic which is reasonable on timescales of minutes to hours
but is less useful over large timescales due to diurnal traffic variations and
long-term trends.
_R/S Statistic_
Again, we separate the time series into $m$ equal blocks of length $N/m$
except all values in each block are aggregated by simple summation. Define $n$
as $n=N/m$ and define the range $R(n)$ as the difference between the value of
the largest block and the smallest block. Define $S(n)$ as the standard
deviation of the values of the blocks. The ratio $R(n)/S(n)$ should scale with
$n$ such that
$E[R(n)/S(n)]\sim c_{3}n^{H}$ (14)
Note that one problem with both the R/S and other methods such as aggregated
variance is choosing the right range for the sizes of the blocks [38].
Choosing values of $m$ that are too small makes short term correlations
dominate, while a large $m$ has fewer blocks and gives a less accurate
estimate of $H$. One approach created to deal with this issue is wavelet
analysis of the logscale diagram which is covered in the next section on
wavelet methods.
_Periodogram_
An additional test for long-range dependence is the presence of $1/f$ noise in
the spectral density of the time signal at low frequencies. The exponent of
$1/f$ noise is related to $\beta$ as well where
$f(\lambda)=c\lambda^{-\gamma}$ where $c$ is a constant (unrelated to previous
ones), $\lambda$ is the frequency and $0<\gamma<1$ and $\gamma=1-\beta$.
Often, the spectral density, $I(\lambda)$ is estimated as
$I(\lambda)=\frac{1}{2\pi N}\left|\sum^{N}_{j=1}X_{f}e^{ij\lambda}\right|^{2}$
(15)
Whose log-log plot slope should be close to $1-2H$ near the origin.
_Scaling of Moments_
In [183], the authors use the fact that the moments scale with the length of
the time series to identify self-similarity. Define the absolute moment as
$\mu^{(m)}(q)=E|X^{(m)}|^{q}=E\left|\frac{1}{m}\sum^{m}_{i=1}X(i)\right|^{q}$
(16)
The absolute moment $\mu^{(m)}(q)$ scales as
$\log{\mu^{(m)}(q)}=\beta(q)\log{m}+C(q)$ (17)
Where $\beta(q)=q(H-1)$.
Table 5: Relationship among key exponents $H=1-\beta/2$
---
$\beta=2(1-H)$
$\gamma=1-\beta$
$\beta=1-\gamma$
$\gamma=2H-1$
$H=(\gamma+1)/2$
An excellent guide to measuring the Hurst parameter can be found in [38].
Though the Hurst exponent is well-defined mathematically, in practice all
measurements of it are only estimations and different techniques, software, or
noisy data sets can produce varying estimates. Even on artificially generated
data with a known Hurst exponent, the different techniques had divergent
measurements of the Hurst exponent and the R/S statistic performed poorly
underestimating $H$ on both generated and real data.
Many may realize that in all of this discussion of self-similarity and
fractals the fractal dimension has not been mentioned once. The omission is
purposeful and due to the convention that almost without exception, the Hurst
exponent is used as the measure of self-similarity in data traffic research.
In any case, the conversion is not difficult since the fractal dimension $D$
of the time series is related to the Hurst exponent by
$D=2-H$ (18)
Given equation 18 we can see that the typical fractal dimension of data
traffic is around 1.2.
_Nonstationary data methods_
As noted earlier, the previous techniques to measure self-similarity
implicitly assume a stationary signal, however, this is definitely not the
case in Internet traffic where on longer time scales, non-stationarity due to
periodicities such as daily usage patterns and growth in traffic over time
make the traffic data nonstationary.
A commonly used method of measuring long-range correlations and self-
similarity in nonstationary time series traffic is the use of detrended
fluctuation analysis (DFA), an approach adopted independently in several
papers [114, 174, 217]. DFA was first used to measure the long-range
correlations in non-coding regions of DNA and is often used to measure
correlations among fluctuations in physiological or financial time series. In
short, DFA is a modified RMS which calculated the deviation from a trend and
long-range correlation in a time series. To use DFA for a time series $X(t)$
of length $N$, first calculate the profile $y(t)$ given by
$y(t)=\sum^{t}_{i=1}[X(i)-\langle{X}\rangle]$ (19)
where
$\langle{X}\rangle=\frac{1}{N}\sum^{N}_{i=1}X(i)$ (20)
The next step involves separating the signal into $m$ equal sized non-
overlapping segments. In each segment, use least squares regression to find
the local linear trend $\tilde{y}_{t}$ in the segment and then calculate the
detrended profile of the signal, $y_{m}(t)$ where
$y_{m}(t)=y(t)-\tilde{y}_{t}$ (21)
and finally the detrended rms is calculated as
$F(m)=\sqrt{\frac{1}{N}\sum^{N}_{i=1}y_{m}(t)^{2}}.$ (22)
If the signal has long-range dependence from a $1/f$ spectrum, $F(m)$ will
scale with $m$ as
$F(m)\sim m^{\alpha}$ (23)
$\alpha$ is related to the $1/f$ exponent of the signal by $\gamma=2\alpha-1$,
which superficially makes it identical to the Hurst exponent. Give the RMS,
the $\alpha$ measured is a second order measurement of the power law scaling.
The Hurst exponent is most simply extracted by taking the mean value of
$\alpha$.
DFA is not the only method for analyzing nonstationary time series and as some
recent studies conclude, is neither always accurate nor optimum for analyzing
nonstationary trended processes. In [16] it is mathematically shown that for
trended processes DFA estimates for the Hurst exponent do not converge to an
accurate value and the wavelet method (next section) is recommended instead to
measure the Hurst exponent. Therefore, estimates of $H$ from papers using the
DFA method should be viewed with skepticism and tested against other methods.
### 4.3 Wavelet methods
The final methods usually used to measure self-similarity are wavelet methods
and are often the preferred method for nonstationary data. In [62, 2, 61, 78,
185], the logscale diagram method of interpreting Internet traffic is
described. A logscale diagram is created using discrete wavelet analysis of
the signal, where the signal, $X(t)$, is represented as filtered through a
wavelet defined given a timescale $j$ and time instant $k$ as
$\psi_{j,k}(t)=2^{-j/2}\psi(2^{-j}t-k)$ (24)
A typical wavelet used in the analyses of these discrete wavelets is the Haar
wavelet. Applying the wavelet transform the signal can be represented as
$X(t)=\sum_{k}c_{X}(j_{0},k)\phi_{j_{0},k}+\sum^{J}_{j\leq
j_{0}}\sum_{k}d_{X}(j,k)\psi_{j,k}(t)$ (25)
Where $c_{X}(j_{0},k)$ are called the scaling coefficients, $\phi$ is called
the scaling function, and $d_{X}(j,k)$ are called the wavelet coefficients.
Wavelet theory will not be covered in great detail here due to its complexity,
however, there are several useful guides [158, 143, 93] to the subject. Each
scale increment $j$ represents a scaling of the timescale of an order $2^{j}$
and $j$ is commonly termed the octave. In addition, in [2] it is shown for a
stationary, self-similar process that the expectation of the energy $E_{j}$
that lies around a given bandwidth $2^{-j}$ around the frequency
$2^{-j}\lambda_{0}$ where, $\lambda_{0}$ is the sampling frequency is
$E[E_{j}]=E\left[\frac{1}{N_{j}}\sum_{k}|d_{j,k}|^{2}\right]$ (26)
Where $d$ are the wavelet coefficients of octave $j$ and $N_{j}$ is the number
of wavelet coefficients in the octave $j$. Graphing the log of $E[E_{j}]$
versus the octave $j$, gives a logscale diagram, an example of which is the
bottom graph in figure 5. In addition, $E[E_{j}]$ is also related to the
sampling frequency and the Hurst exponent:
$E[E_{j}]=c|2^{-j}\lambda_{0}|^{1-2H}$ (27)
So the logarithm of the expected energy is directly proportional to the Hurst
exponent. In fact, monofractal behavior is indicated by a linear dependence of
$\log{E[E_{j}]}$ over multiple octaves. The different scaling regimes can be
seen in figure 5 by noting how the curve varies over the octaves 2 to 4 and 8
to 12. There are two important procedures when using logscale diagrams is to
always represent the confidence intervals (usually 95%) for the wavelet
coefficient values in each octave. Second, one must make sure to use a minimum
variance unbiased estimator instead of least squares regression fitting when
estimating $H$ from linear trends between at least three octaves. Appropriate
methods for calculating this estimator as well as its variance can be found in
[4, 5].
The second, and often considered more rigorous, method of looking at changing
self-similarity using wavelets is looking at the scaling of the partition
function for each moment of order $q$ over each octave where the partition
function is defined as
$S(q,j)=\sum_{k}|2^{-j/2}d_{X}(j,k)|^{q}$ (28)
The scaling behavior, besides being seen by graphing $log_{2}S(q,j)$ vs. $j$
is encapsulated using what is called the structure function
$\tau(q)=\lim_{j\to-\infty}\frac{\log{S(q,j)}}{j\log{2}}$ (29)
if the traffic is exactly self-similar with Hurst exponent, $H$, then for each
$q$, $\tau(q)=Hq-1$. When more than one scaling behavior is in the signal,
$\tau(q)$ is no longer linear, but concave and each scaling exponent
contributes to its value roughly according to its relative strength in the
signal at the relative timescale. For details see [62, 185]
There have been some theories of self-similar behavior in Internet traffic by
traffic engineers. This thesis was first broached and analyzed in [161]. In
short, it is believed that self-similarity is generated on long timescales by
user and application needs (see next section on the ON/OFF model) but on
shorter timescales the picture is much less clear. It is possible that a
different self-similarity affected by network or TCP congestion control
considerations is active in this region. Early papers tried to explain
multifractal properties of traffic by a process known as multiplicative
cascades or conservative cascades [62, 78, 76]. The cascade is mathematically
defined as a mass $M$ equally distributed over the interval $(0,1]$ where the
mass is broken up into two new masses, one with mass $p$ and the other with
mass $1-p$, where $p$ is a fraction of mass defined for the process, and these
two new pieces are broken up again according to the same process ad infinitum.
The multiplicative cascade model was rationalized in the relation to Internet
traffic by describing the encapsulating of flows into packets and the
fragmentation of these packets at the link layer as a conservative cascade
process where the total transmitted data is conserved but broken down into
many different packets. Since this process occurs over relatively short time
scales, it is given as additional evidence for the cause behind the different
scaling at shorter time scales.
Many questions about the multifractal paradigm, however, were raised in [186]
which openly criticized some of these claims and questioned whether
multifractal models are necessary and as proven as they purport to be. In
particular, Veitch, Hohn, & Abry, while analyzing some of the most common data
traces used in Internet traffic studies, raise questions about the rigor of
the statistical tests used such as logscale diagrams without confidence
intervals and the large size of these confidence intervals for some values of
the energy at higher octaves. They also raise the point that these tests rely
on an assumption of stationarity in Internet traffic which may not always be a
valid assumption, especially over longer timescales. In the end, they do not
completely rule out multifractals, however, they raise the point that current
statistical tools are not yet fully developed enough to give a definite answer
to existence of multifractals. Similar comments are made in [200] declaring
that multifractal patterns may exists but are not to be seen as an end in
themselves and any new model of multifractality must be matched with a
feasible mechanism.
Self-similarity and long-range dependence account for the “bursty” behavior of
Internet traffic at all time scales. Unlike telephone traffic, which is
Poisson and large spikes are rare deviations from a mean traffic level and
have an exponentially decreasing probability, burstiness in Internet traffic
at almost all-scales has a non-vanishing probability. This makes traffic
management schemes and infrastructure planning much more difficult from a
statistical standpoint. Sometimes, a scheme known as “small buffers, high
bandwidth” [149] is advanced to deal with bursty traffic to avoid trying to
create massive buffers to handle bursts of traffic. However, there is not yet
an easy answer to managing Internet traffic, especially one with practical
use.
|
---|---
|
|
Figure 5: A view of a 10,192 second trace of IPv6 6Bone experimental network
Internet traffic taken from the WIDE MAWI traffic trace archive Samplepoint-C
on July 22, 2005. The data was collected into 1s intervals. The following
figures show the packets/s of traffic, the distribution of packet arrivals,
the autocorrelation of the time series up to a lag of 1000, the $1/f$ noise
plot of the data trace, and the logscale diagram constructed from wavelet
coefficient data based on 100ms bins of packet arrivals and 95% confidence
intervals. The Hurst exponent was calculated with the statistical program R
with the _fractal_ package using the aggregated variance method estimating H =
0.81.
## 5 Theories on the causes of self-similar traffic
Once self-similarity was demonstrated in Internet traffic, the next logical
step was to look for the cause. Besides Internet topology models, this is
where most of the tension between the different fields has developed and
caused clashes of theories and methodology. Though some consider the question
controversial and unresolved, there are some models which currently have a
greater weight of evidence behind them as checked against real-world traffic.
There are three main theories which will be discussed at length throughout the
rest of the paper.
### 5.1 Application Layer Cause: Long-tail ON/OFF sources
First, there is the most widely known, and only empirically validated,
application layer theory which states that self-similar traffic is the cause
of the behavior or users. This was first elaborated in [196, 197]. This theory
models the traffic on the Internet as large number of ON/OFF sources with
identical duration distributions, an idea earlier broached by Mandelbrot
[122]. The ON/OFF sources, which reflect flows, are superimposed traffic
sources that alternate in ON and OFF periods according to a power-law
distribution. Though this is not usually explicitly mentioned, this model is
extremely close to modeling a large number of flows with long-tails dominated
by a few elephant flows found in actual traffic measurements. In [196, 197]
they give evidence both from theory and observation that many superimposed
ON/OFF sources behave to a limit as fractional Brownian motion and can account
for the self-similarity seen in overall Internet traffic. Specifically, the
Hurst exponent of the traffic of the superimposed is calculated by
$H=\frac{3-\mbox{min}(\alpha_{ON},\alpha_{OFF}))}{2}$ (30)
where $\alpha_{ON}$ and $\alpha_{OFF}$ are the power law exponents of the ON
and OFF distribution times respectively. ON/OFF flows with long-tail duration
are not merely a theoretical abstraction but have a basis in the access of
files over the Internet. Internet file sizes have been shown by several
studies [14, 47, 79, 129] to be at least long-tailed though there is dispute
over whether the distribution more closely fits a power law, lognormal, or
double Pareto.
Crovella & Bestavros and Crovella, Park & Kim [151, 41, 150] extend the model
to explain the self-similar nature of TCP and web traffic. They base their
studies on long-tailed file sizes with a power-law exponent of about 1.2
(implying a Hurst exponent of about 0.9 from equation 30). They also found
that in general, reducing the tail of the file length distribution by
increasing the power-law exponent, lowers the Hurst exponent as expected in
[196, 197]. This theory of ON/OFF sources, and its variants, has become the
most dominant explanation for self-similar network traffic in most network
engineering papers and has the largest theoretical and empirical backing. It
also is a frequently used model for simulating data traffic in Internet
simulations.
The second theory, discussed more in detail in the next section, considers
origins of the self-similar traffic at the transport layer. In particular, it
looks at possible effects the TCP congestion control algorithms may have on
network traffic given the feedback and collective behavior it can engender
among multiple traffic sources over the same path. The third main theory,
discussed in the section on phase transition models of traffic congestion is
that which connects self-similarity to critical phenomena in data traffic near
the transition point from free flow to congested traffic. This is the model
currently most favored amongst physicists and underlines many of the models of
data traffic that will be described later. In addition to describing these
models, the insights and flaws of each will be described in detail.
### 5.2 Transport Layer Cause: TCP Congestion Control
As discussed earlier, there has been interest in whether self-similar traffic
can find its causes in the congestion control of TCP rather than at the
application level. This has not been a generally accepted or verified cause or
contributor to self-similarity, at least on long time-scales. Veres and Boda
[187] first brought up the conjecture that since the assumption TCP
stochasticity or predictable periodicity itself in these equations is highly
flawed, TCP throughput then cannot be reduced to a closed form equation, and
that TCP instead exhibits deterministic and chaotic behavior. In addition,
most analyses look at TCP only at the single link level instead of treating it
as a network dependent entity given congestion control. They based their
papers on simulations that showed self-similarity, sensitivity to initial
conditions, strange attractors, and stable periodic orbits appeared. Fekete
and Vatay [60] also used simulations to show that the interaction of TCP with
buffers in routers can also cause chaotic behavior in the TCP flows. They
simulated the interaction of $N$ different TCP flows with a buffer that had
the capacity to hold a fixed number of packets. They show that the backoff
algorithm of TCP caused by lost packets can cause power-law behavior in packet
interarrival times and chaotic dynamics given the (buffer length)/($\\#$ of
TCP packets) ratio is below a critical value of 3.
Similarly, Hagà et. al. [82] used simulations to recreate self-similar traffic
and long-range dependence by the interactions of multiple TCP flows at a
buffer on a central router connecting three different hosts. They assume an
effective loss packet loss rate (real packet losses and RTT exceeding the
allowable timeout period) but have an infinite sized router buffer so there is
no real packet loss but a large effective packet loss due to timeouts. This
model can produce self-similar traffic with $H$ = 0.89 without any assumptions
of ON/OFF distributions or file sizes beyond a constant TCP flow size of 1000
packets. The only extra assumption is stochastic source and destination of TCP
flows among the three hosts.
These methods, particularly relying on simulations or analytical reasoning and
not actual Internet traffic trace methods, have been heavily criticized,
especially by those favoring an application layer explanation for self-
similarity. In [63, 64], a more comprehensive TCP model is developed
accounting for both the backoff phase and congestion avoidance phase after
slow start. Their argument is that TCP can possibly generate correlation
structure and possibly self-similarity but only short timescales (up to
1024xRTT for high packet loss) and not arbitrarily long time scales which
characterize most self-similar traffic. Their models shows that at low packet
loss rates the correlation structure is dominated by congestion avoidance
after slow start while the exponential backoff governs the correlation
structure at high loss rates. In [81], the authors focus on short TCP flows
and thus only model slow start and backoff. They again show self-similarity at
large enough packet loss rates but though the long-range dependence that is
also present would connote infinite variance, the TCP based self-similarity
only extends over certain short timescales. Thus the authors dub this self-
similarity pseudo self-similarity since its timescale is relatively limited.
Veres et. al. answer these criticisms in [188]. Here they concede that though
TCP’s congestion control may not by itself be the cause of LRD in Internet
traffic, they show through data, simulation, and mathematical arguments that
TCP’s congestion control suite may propagate self-similar traffic along its
path if it encounters a bottleneck that limits its send rate and has self-
similar traffic. Therefore even if TCP can’t create the full self-similar
effect, it may be responsible for propagating the self-similarity far beyond
the traffic it originated at.
In a variation of the above research, Sikdar and Vastola [169, 168] give a
model where self-similarity and long-range dependence emerge from the dynamics
of a single TCP flow instead of multiple flows. They model a single TCP flow
as the superposition of $W_{max}$ ON/OFF processes where $W_{max}$ is the
maximum window size advertised by the receiver. This is similar to the earlier
ON/OFF model but they show that for higher packet loss rates, a higher Hurst
exponent and more self-similar traffic is generated. In addition, there are
other papers detailing possible mechanisms by which TCP congestion control can
give rise to self-similarity in Internet traffic.
In the end, TCP congestion control has not emerged as a favored cause of self-
similar traffic. Perhaps it is a strong or even dominant factor of self-
similarity at relatively short timescales but is likely not the cause of the
pervasive self-similar traffic described in most papers. One final note not
connected to the idea of self-similarity is that TCP is one of the few ways
that traffic actively couples to the network topology via the dependence of
throughput on RTT. Though most latency in networks is likely caused by
congestion and other network conditions, given similar bandwidths and delays,
the average shortest path (number of hops) in topology can affect average RTT
as shown in [59]. In summary, TCP being the dominant protocol on the Internet
is one the main determinants of the traffic dynamics. However, TCP is a
complicated and feedback driven protocol whose actions can only be partially
estimated using analytical or stochastic models. The TCP protocol will
definitely hold promise in the future for those looking for more intricate
complex phenomena or pattern formation in Internet traffic dynamics.
## 6 Network Layer Cause: Theories of Phase Transitions and Critical
Phenomena in Networks
The most active area of work by physicists in the research of network dynamics
is a group of research which merges the new insights of Internet traffic
behavior with the mature and well-tested tools of statistical mechanics and
critical phenomena. Similar to papers written on vehicle traffic [98, 44, 33]
these papers have analyzed the onset of congestion in networks as a phase
transition from a free-flow to congested state determined by a critical
parameter. In fact, an explicit comparison was given in [74]. The papers, in
general, deal with three broad, though sometimes overlapping, themes regarding
the onset of congestion. First, are the papers that analyze the onset of
congestion as a function of the packet creation rate for various topologies
and also whether the self-similar structure of traffic can be reproduced in
these models. Second, are models primarily concerned with investigating the
rise of self-organized, emergent phenomena in networks in the critical state
and linking the studies of congestion with the study of self-organization in
general. Finally, are many papers who investigate how different routing
strategies can delay or affect the onset of congestion. The papers of the last
category often overlap with the first. Below I will describe some of the most
often quoted papers. A more comprehensive list and reference of papers is
given in table LABEL:routingtable for those wanting to delve into the topic in
more detail. In the papers described, the critical parameter is typically the
packet creation rate. This has different symbols depending on the paper but
here we will describe it as $\lambda$. Finally, in the next section I will
note many common, and unfortunately many times accurate, criticisms and
problems with these models.
Papers by physicists investigating congestion first concentrated on the onset
of congestion as a critical phenomenon and possible links between this and the
self-similar nature of Internet traffic. With few exceptions, these papers
focus on the link or network layer dynamics (IP) as the source of critical
phenomena in Internet traffic. One of the first papers to deal with a phase
transition model of Internet traffic was by Csabai in 1994 [43]. In this
paper, Csabai noted the presence of a $1/f$ power spectrum for the RTT times
for pings between two computers where the fitted slope is -1.15 (about an $H$
of 1.08). He also is among the first to compare Internet data traffic with
vehicle traffic [74]. It must be noted that the RTT from ICMP echoes is not
always equivalent to the RTT in TCP since many gateways give preferential
forwarding to TCP packets. Also, this power law spectrum based off of ICMP
echoes is different from the overall traffic whose self-similarity was
discussed earlier.
Takayasu, Takayasu, and Sato [181] followed up with a similar study where they
also note the $1/f$ distribution of RTT for ICMP pings between two computers
if there are many gateways on the route between them, likely because of
consecutive jamming due to filled buffers. For a short route, their echo
replies are distributed $1/f^{2}$ at low frequencies and as white noise at
higher frequencies ($f>10^{-4}$). They extend the analysis though to include a
theoretical derivation of the behavior of network traffic taking into account
a simple topology. They disregard loops and use the theoretical topology of a
Cayley tree where gateways are sites and cables are links. A contact process
(CP) is modeled where empty sites are considered jammed gateways and filled
sites (particles) are considered un-jammed gateways. A jammed gateway has a
probability $p$ of becoming un-jammed if it is adjacent to an un-jammed
gateway (particle reproduction) and an un-jammed will become jammed with an
independent probability $q$ (particle annihilation). An un-jammed gateway will
do neither with probability $1-p-q$. In analyzing the simulation, they assume
that the number of un-jammed gateways over time is equivalent to the
distribution of RTT. They derive a power-law result from the CP process which
shows the distribution of jammed sites over time follows a $t^{-\alpha}$ power
law distribution with time when a parameter $\delta=1-p/q$ equals 0 and that
this power law yields $1/f$ noise for the conditions $0\leq\alpha\leq 1$. A
comparison of ICMP echo RTTs to earthquake aftershocks is made by Abe and
Suzuki [1] who fit the RTT from pings in Internet traffic to a statistical
distribution that is similar to Omori’s Law which models the arrival of
aftershocks from an earthquake.
A similar hierarchical tree topology is used to investigate critical behavior
for data flow by Arenas, Diàz-Guilera, and Guimerá [11]. They derive a mean-
field theory solution for the critical packet creation density and also show
that most congestion occurs at the root of the tree and the first level of
branching. Power-law scaling of the total number of the packets in the system
is observed near the critical point $\lambda_{c}$.
Takayasu & Takayasu later expand on a theory of self-similarity in Internet
traffic as a critical phenomenon in [182, 180, 179] In [180] Takayasu,
Takayasu, and Fukuda describe what they believe is a phase transition in the
flow of overall Internet traffic. They separate data traffic into 500 s bins
and take autocorrelations of each bin comparing the correlation length in
seconds with the mean traffic density. The correlation length increases with
traffic flow density until a critical density $\lambda_{c}$ = 500 kbytes/sec
where the correlation length begins decreasing again. They associate this with
a second order phase transition in the flow where there is a transition from
free to congested flow. When they consider any flow above 300 kbytes/sec as
“congested” around the critical point they can show power law scaling of
lengths of congestion times confirming the critical nature of the phenomenon.
In [179], the same authors theorize that the critical nature of traffic
measured in Ethernet networks is due to the Ethernet collision detection
management algorithm (CSMA/CD) which employs an exponential backoff algorithm
on detection of an Ethernet frame collision that is qualitatively similar to
the congestion backoff mechanism described in TCP. They show that a binary
backoff algorithm can generate $1/f$ traffic distributions at the critical
point.
Most of the other prominent papers in the first category follow in the
tradition of the first Takayasu paper describing phase transitions using
modeled networks in simulations to infer a shift in dynamics above a certain
packet creation/flow threshold [28, 140, 213, 144, 72, 172, 203, 215, 80]. The
threshold can be changed either by topologies, which are more efficient with
small world networks such as random graphs or scale-free graphs than lattices
or Cayley trees, by changing the proportion of nodes that can generate (versus
only route) traffic, or by routing strategies such as in [144, 215]. In [72,
172, 203], mean field methods are used to calculate the onset of congestion
for a 2D lattice at
$\lambda_{c}=\frac{2}{pL}$ (31)
where $p$ is the proportion of nodes that can generate packets and $L$ is the
length in nodes of a side of the lattice. All of these papers purport to show
$1/f$ distributions of packet travel times which they link to self-similar
traffic. A paper looking at the problem from a different angle by Moreno,
Pastor-Satorras, Váquez, and Vespignani [135] approached the problem by
looking at what average traffic density in the overall network could lead to a
spread of congestion across all nodes and the collapse of the network. This is
a related viewpoint on the cascading router failures and percolation models
that have been studied on scale-free topologies [138, 134, 42] which links
cascading failures not just to topological sensitivity of certain hubs but
also the traffic levels in the network.
In the second category, are papers largely concerned not with the value of the
critical parameter but with emergent phenomena themselves. One of the earliest
papers hinting at this was a study by Barthélemy, Gondran, and Guichard [17].
Borrowing techniques from nuclear physics, they studied the eigenvalue
distribution and eigenvectors of the traffic correlation matrix of 26 routers
and 650 connections in the Renater computer network for two weeks of traffic
data. Their technique used random matrix theory to compare the eigenvalue
distribution of the correlation matrix of Renater traffic fluctuations to that
of a control random matrix. The traffic fluctuations in an interval $\tau$ in
the traffic between source $i$ and destination $j$ were defined as
$g_{ij}(t)=\log\left[\frac{F_{ij}(t+\tau)}{F_{ij}(t)}\right]$ (32)
and the correlation between connections $ij$ and $kl$ is defined as
$C_{(ij)(kl)}=\frac{\langle g_{ij}g_{kl}\rangle-\langle g_{ij}\rangle\langle
g_{kl}\rangle}{\sigma_{ij}\sigma_{kl}}$ (33)
They found that the largest eigenvalues were much larger than the largest
eigenvalues of a similar rank random traffic matrix whose flows have a mean of
0 and unit variance. Also, the largest values of the eigenvector for the
largest eigenvalue correspond to the most highly correlated routers. These
results all indicated spatiotemporal correlations among the routers in the
network that deviated from traffic defined purely by a stochastic process.
Among the most consistent researchers to address the emergent phenomena
question directly are Yuan and Mills [207, 208, 209, 211] who make a
persuasive case that emergent phenomena in networks could go beyond the simple
onset of congestion in simple network topologies and only treating packets at
the network (IP) layer. The main themes of their papers are measuring
spatiotemporal patterns that emerge in larger networks. The main features they
add lacking in many other models are size (more nodes), more realistic
topologies, as in [209], where their network includes four levels of hierarchy
in tree structure, and modeling of transport (TCP) level effects such as
congestion control [209]. In their first paper [207], they use a simplified
topology of a 2D cellular automata (CA) with all nodes as hosts and routers.
The state of a router on the CA is defined by the number of packets in its
queue and it “transitions” by passing off packets given the state of the
queues of the surrounding cells. The traffic sources can originate in any node
and are modeled as ON/OFF sources as in [196, 197]. Packets are routed via a
full routing table. In addition, they model the systems with three types of
congestion control algorithms: no congestion control, a congestion control
that stops transmitting above a threshold RTT per hop to the destination, and
a TCP-imitating congestion control that includes slow start and congestion
avoidance. Their main results use the TCP-imitating congestion and produce
power spectrums of the time series of the number of received packets at a
given node for various sample time lengths and network sizes. In general, they
find that increasingly longer sample times diminish the correlations and long-
range dependence measured in the power spectrums but increasing the network
size increases the correlations over both time and space. Comparing smaller
networks to similar sized subgraphs in larger networks shows that the
subgraphs exhibit stronger correlations and they deduce that large network
sizes can allow for wider coupling and self-organization. They also surmise
larger networks may be more predictable because congestion is stable over
longer time scales.
In subsequent papers, this idea is developed further. In [208], the authors do
an analysis by creating a weight vector for each node that is constructed from
the components of eigenvectors derived from the correlation matrix as in [17].
Yuan and Mills create a technique to analyze simulated networks of a larger
size. They define flow vectors, $x_{i}$, where $i$ ranges from 1 to N where N
is the number of nodes with N components each component $x_{ij}$ representing
the flow from node $i$ to node $j$ during a sample interval. They then create
a normalized flow vector by normalizing each element of each vector for the
entire sample time including all intervals where the normalized vector is
$f_{ij}=\frac{x_{ij}-\langle{x_{ij}}\rangle}{\sigma_{ij}}$ (34)
They analyze the eigenvector of the largest eigenvalue of the correlation
matrix among all normalized flow pairs over time and use the elements of the
subvectors of this eigenvector to create N $S$ vectors where $S_{ij}$ is the
relative contribution of node $i$ to node $j$ in terms of traffic
correlations. Performing simulated traffic on a 2D topology with ON/OFF
sources with Pareto distributed ON times and TCP congestion control they
observe complex fluctuation of the largest eigenvalue over times as well as
correlated traffic between certain nodes over time though they note their
largest eigenvalues tend to be smaller than those in [17]. They also raise the
point that during congested critical states, taking a sample of a few nodes
(or routers) may give a better and more overall cohesive picture of the entire
network if sampling all nodes is infeasible. This is mainly due to the
increased spatiotemporal correlations in congestion. Also, longer sampling
time windows tend to reduce the visibility of correlations in traffic.
In [209], they continue the same research based on the eigenvalue method but
using a four-tier (backbone router, subnet router, leaf router, and source
hosts) hierarchical network to model the actual AS-level and below topology of
the Internet. They also use the measures of spatiotemporal correlation to find
both hotspots and show that distributed denial of service (DDoS) attacks can
cause large-scale effects beyond the target router by disturbing traffic flows
in other correlated routers in the network. They suggest methods of analyzing
network-wide phenomena using small samples of nodes and possibly detecting
DDoS attacks by the signatures of large-scale perturbations in correlated
network traffic. In [211], they return to the 2D CA formalism but investigate
spatiotemporal dynamics using wavelets and logscale diagrams over varying
average packet creation rates, congestion control protocols, and average flow
durations. They looked for causes of LRD at the application level (file size
distribution), transport level (congestion control type) and network level
(varying the rate of ON/OFF sources and network size). They found that LRD
emerged on wide times scales with long-tail distributions of file sizes, an
increasingly large network size, or Pareto distributed ON/OFF source times but
only emerged on limited time scales when only the type of congestion control
was varied. Though they acknowledge the limits of their model they suggest
that most LRD emerges due to interactions in the network layer or possibly
file-sizes in the application layer. Like [63, 64, 81] they suggest TCP
congestion control plays only a limited part in the emergence of LRD. Yuan et.
al. [214] closely replicate the results of [208] except they compare
visualization of the largest eigenvalue over time with the information entropy
of the weight vectors. They find the eigenvalue more clearly shows the change
in correlation structure over time. There are also some very interesting
spatiotemporal plots of router congestion over time in [212] showing pattern
formation in the temporal congestion among routers in a 1D cellular automaton
model.
Other researchers, still following the theory that the onset of congestion
could be considered a critical phenomenon, began investigations on possible
new routing strategies that could help extend the tolerance of a network to
congestion. In short, all of the proposed routing strategies aim to be an
improvement over current state Internet routing where routers use a global
router table and shortest path metrics to route packets. In particular, these
papers show that the geodesic on the networks between two points defined
solely according to a graph shortest path are not always the best routing
paths in real traffic conditions. The newly proposed routing strategies tend
to explicitly take into account traffic and/or queue conditions at neighboring
routers or use different topological measures such as betweenness [51, 50] in
order to redefine the shortest path metrics and packet routing strategies.
Besides random walk routing, the most common routing heuristics are next
nearest neighbor (NNN) [6, 177, 178, 175, 176] based on delivery to the
destination if it is an adjacent node or either random walking or bias to
higher degree nodes otherwise (except a packet cannot travel to a node it has
just left). Typically it is shown that NNN is superior to random routing.
Preferential next nearest neighbor (PNNN) [206, 192, 87, 88] slightly alters
NNN by explicitly taking into account node degree in routing according to a
parameter $\alpha$ which creates a preference distribution for nodes of a
degree $k$ according to the relation $k^{\alpha}$, or a similar relation. In
[31] traffic congestion at nodes is also taken explicitly into account.
Typically, there is a purported value of $\alpha$ which minimizes average
packet travel time and tolerates a higher packet creation rate for the onset
of congestion.
Many of these papers, with some exceptions such as [72], do not show how these
new routing methodologies should compare against the current shortest-path
full routing table the Internet uses for routing. What would the value of a
random walk or walk based on the degree of an adjacent router add to network
routing infrastructure and performance? Or is it worse than the current system
(which most comparisons suggest)? The basic idea of the geodesic between
source and destination depending on traffic conditions is a very interesting
proposal but how could it be implemented in practice? These questions should
provide fertile ground for future research and cross-disciplinary
collaboration.
There have been some papers proposing feedback based routing in the network
engineering literature (for example [216]), however, they are not related to
similar research in physics and this line of research by physicists is often
not looked upon highly by the network engineering community as will be
discussed in the next section. One final note from equation 1 is that since
all of these models use the packet as the basic unit, the concept of the
relationship between data throughput and packet size shows that apart from
topology changes or new routing, as stated earlier, one easy way to reduce the
packet creation rate on a network is to increase the average packet size.
Since throughput is an important measure in the function of the Internet,
future measurements and experiments on packet creation and congestion should
thoroughly account for this.
Table 6: A general view of the statistical mechanics congestion and routing models discussed in the paper. Citation | Topology | Host/Router Distribution | Packet Creation Distribution | Routing Strategy
---|---|---|---|---
[11] | Cayley Tree | All nodes are both | uniform distribution; fixed probability | full routing table - shortest path
[182] | Cayley Tree | Hosts on perimeter | uniform distribution; fixed probability | full routing table - shortest path
[144] | 2D square lattice | Hosts on perimeter | Poisson process of rate $\lambda$ | 1) deterministic method where packets are routed to nodes who have received least packets, 2) probabilistic method where particles choose node biased against nodes having handled more packets
[28] | 2D square lattice | All nodes are both | uniform distribution; fixed probability | proximity of neighbors to destination, inverse “temperature” thermal agitation, and repulsion with sites with filled buffers
[140] | 2D square lattice | Top row nodes are hosts; bottom row destinations | uniform distribution; fixed probability | random walk
[213] | 2D square lattice | All nodes are both | uniform distribution; fixed probability | full routing table - shortest path
[72] | 2D square lattice + extra links | All nodes are both | Poisson process of rate $\lambda$ | using a routing table and lowest queue consideration in neighbors; use two types of routing tables: a full routing table with all paths and a partial table with paths only up to a distance m from each node
[171] | 2D square lattice | random nodes of probability $p$ are hosts | uniform distribution; fixed probability | full routing table - shortest path & congested node avoidance
[172] | 2D square lattice | random nodes of probability $p$ are hosts | uniform distribution; fixed probability | full routing table - shortest path; $\lambda$ moderated by congestion & congested node avoidance
[203] | 2D square lattice | All nodes are both | Poisson process and long range dependent distribution | full routing table - shortest path & congested node avoidance
[215] | random, regular, Cayley tree, scale-free | All nodes are both | uniform distribution; fixed probability | full routing table - shortest path, and biased against high degree or betweeness nodes
[80] | 1D chain, 2D lattice, Cayley tree | All nodes are both | uniform distribution; fixed probability | full routing table - shortest path
[135] | scale-free | All nodes are both | initial load created on edges by uniform distribution | N/A
[207] | 2D cellular automata | All nodes are both | Poisson duration ON/OFF sources | full routing table - shortest path
[208] | 2D square lattice | two tiers: one hosts, one routers | Pareto duration ON/OFF sources | full routing table - shortest path
[209] | four tier hierarchical network | tier four (lowest level) sources and receivers | Pareto duration ON/OFF sources | full routing table - shortest path; different forwarding capacities on each tier
[211] | 2D cellular automata | two tiers: one hosts, one routers | both Poisson and Pareto duration ON/OFF sources | full routing table - shortest path; different forwarding capacities on each tier
[214] | 2D square lattice | two tiers: one hosts, one routers | Pareto duration ON/OFF sources | full routing table - shortest path; different forwarding capacities on each tier
[212] | 1D chain | Fixed number of hosts at random positions on chain; inter-host spacing is buffer length | uniform distribution; fixed probability | right to left diffusion based off of max routing speeds and buffer sizes
[50] | scale-free | All nodes are both | uniform distribution; fixed probability; one time packet creation t=0 | global routing table; shortest path and congestion in neighbor nodes
[51] | scale-free based on Internet AS map | All nodes are both | uniform distribution; fixed probability | global routing table
[205] | 2D lattice, scale-free | All nodes are both | fixed number of packets per time step | global routing table; shortest path biased against high betweeness nodes
[177, 178] | scale-free, web like | All nodes are both | uniform distribution; fixed probability | NNN
[175, 176] | scale-free, web like, random grown tree | All nodes are both | fixed number of packets per time step | NNN, random walk, 1-2 distance awareness
[206] | scale-free | All nodes are both | fixed number of packets per time step | PNNN
[192] | scale-free | All nodes are both | fixed number of packets per time step | 1) routes to neighbors biased towards higher degree nodes; 2) also adjusted for congestion in neighbors
[87] | scale-free | All nodes are both | fixed number of packets per time step | PNNN
[88] | scale-free | All nodes are both | fixed number of packets per time step | PNNN
[31] | random, static scale-free, BA scale-free | All nodes are both | uniform distribution; fixed probability | full routing table - shortest path & congested node avoidance
[112, 113] | 2D square lattice + extra links | All nodes are both | Poisson process of rate $\lambda$ | full routing table - shortest path & congested node avoidance
## 7 Criticisms of Various Approaches to Self-Similarity
Though the physics literature on congestion and critical phenomena in networks
is becoming increasingly sophisticated and adept at “reproducing” self-similar
patterns seen in Internet traffic, there have been several valid criticisms,
particularly from the network engineering community, that the methodologies
may reproduce observations but do not take into account the actual workings of
the Internet in detail [67, 199, 48, 114, 103, 156, 10, 121, 8, 202]. In [67]
Floyd and Paxson, though not addressing physics approaches directly, note that
the while simulations are crucial to Internet research, the Internet is
extraordinarily complicated and difficult to accurately simulate, especially
on large scales. In particular, they point out three problems: the
increasingly unpredictable behavior of IP over increasingly diverse networks
and applications, the massive and continuously increasing size of the
Internet, and its penchant for changing in many drastic ways over time.
Heterogeneity is the rule not the exception and many activities such as
periodicities are often left out of simulations. The papers [199, 8, 48]
address the physics community more directly pointing out defects in the
theories of critical phenomena and the hub vulnerability of scale-free
networks respectively. In [199], Willinger et. al. describe evocative models,
which reproduce the observations using generic models, and explanatory models,
whose applicability are tested by experiment and measurement (what they term
“closing the loop”). They complain many models, from physicists and some
engineers, are evocative and ignore the research on the particulars of
Internet protocols, function, and traffic that could verify or refute their
model. These models instead act as a “black box” which try to come up with an
appealing model or theory to match self-similarity but do not provide any
prediction or verification using real world traffic trace data. For example,
the largest problem in many of the phase transition models is that they
demonstrate self-similar traffic only at critical loads while Internet traffic
measurements show self-similar traffic in both free-flow and congested regimes
and at all levels of flow. In [8], Alderson and Willinger further elaborate
that they believe models from statistical mechanics are not applicable to the
Internet which is designed by multiple economic and performance considerations
and not by simple rules of self-organization often present in models such as
preferential attachment.
Barabási-Albert preferential attachment model for the growth of scale-free
networks is criticized since many of the highly connected nodes in the
Internet are at the edges near final consumers rather than in the central
parts of the high speed AS network. This valid criticism though is partially
answered by disassortative mixing, which shows that in non-social networks
like the Internet, high degree nodes are more likely to connect with low-
degree nodes and rules out a core of highly connected, highly vulnerable hubs.
Many subsequent models of scale-free networks have taken this into account.
Finally, [48, 202] criticize the research in topology that says scale-free
networks are vulnerable to attack due to highly connected hubs, which they
once again say is fallacious because despite power-law degree distributions
the most highly connected hubs are often on the periphery of the Internet and
not along its crucial backbone. They blame much of the confusion on the
methodology in the widely cited paper by Faloutsos et. al [57]. He et. al [86]
partially rebut criticisms by studying relationships between customer-
providers and peers at Internet exchange points (IXPs). They find the customer
provider relationship closely fits a power law though the peer-peer
relationships show a Weibull distribution.
Lee in [114] points out both the aforementioned problems with the critical
phenomena models but also points out the ON/OFF model also has problems
because since an ON/OFF source has a long-tail duration time distribution, you
will have a finite probability of an ON/OFF source as long in duration as any
observation period you make. Lee also criticizes TCP models for not accounting
for similar self-similarity effects in UDP and other stateless protocols. In
[81, 63, 64, 188], the possibility is also raised that that TCP can effect the
traffic dynamics in a more market dynamic on shorter time scales (near the
order of multiple RTT) and this is an area of future investigation and debate.
In the author’s opinion, these criticisms of critical phenomena models are
extremely valid and that self-similarity only at critical loads and lack of
real world validation show many of the models in the previous section are
unrealistic or misleading. Given the current facts and validation against real
data, it seems that the current weight of the evidence for long timescale
self-similarity lies with the application layer explanation based on ON/OFF
sources. However, this does not mean the physics models based on statistical
mechanics are completely useless. Wide scale Internet traffic measurements are
nearly impossible currently and also large-scale theories of traffic are still
relatively undeveloped. Currently almost all real data is traffic traces over
one given link during a given time period which makes large scale Internet
studies very difficult. Large-scale congestion, traffic correlations, and
other complicated phenomena will probably draw useful lessons from physics
models of self-organization and long-range correlations though more realistic
models are absolutely necessary. The omission of TCP-like congestion control,
except in a few models, must be rectified. If these criticisms sound a bit
harsh, try to put yourself in the shoes of most network engineers who
understand the intricate processes in detail of how the Internet operates.
When shown a model of a 2D grid, no mention of congestion control, infinite
router buffers, and self-similar traffic only in congested conditions, their
incredulity is understandable. It is aggravated by the fact that almost none
of these papers try to match results with or analyze real traffic traces.
In defense of the efforts of physicists, however, I believe that physics
started out correctly choosing simplistic topologies and dynamics scenarios
that are both analytically tractable and amenable to rapid simulation.
However, though the earlier work of those like Takayasu began by looking at
traffic traces, this soon disappeared almost completely in favor of computer
simulation. Despite the obvious shortcomings of explaining self-similar
traffic, the demonstration that large-scale congestion, though on which
timescale is unknown, may be a theory the physics and engineering communities
should take note of for validation or refutation.
With the work in [17, 208, 209] showing large-scale correlations among router
traffic in both real and simulated data, can we really look at the Internet
dynamics from solely the viewpoint of a collection of single traffic traces?
The question is not if the Internet displays large-scale correlations and
self-organization well-known to complexity theory, but how these large-scale
effects play out and if realistic simulations with both realistic dynamics and
topology can predict effects that we have not yet observed or known how to
look for. Much more cross-disciplinary work is needed in this direction.
Though the analysis of the criticisms above seems like it tries to be even
handed and please everyone while solving nothing, the nature of the problem is
such that the issues regarding the core nature of Internet traffic cannot be
easily resolved. Willinger et. al. are right in that the loop must be closed
and just creating a simulation that outputs traffic with a Hurst exponent near
0.8 cannot be considered the final word in the “cause” of self-similarity in
Internet traffic. In addition, though it is difficult and near impossible,
large-scale and coordinated traces and models of a topologically and
dynamically correct Internet is the next logical step in modeling and studying
these phenomena.
## 8 Other Interesting Phenomena
### 8.1 Flows and fluctuations
Barabási and Argollo de Menezes from the physics community [12] proposed an
interesting result when they announced a relationship between the average
volume of the flow and its dispersion (standard deviation of traffic volume)
among nodes in a network. In particular, they found that accounting for all
nodes in a network, you find the average flux $\langle{f}\rangle$ and standard
deviation $\sigma$ per node are related by the scaling relationship
$\sigma\approx\langle{f}\rangle^{\alpha}$ (35)
Where $\alpha$ is near either 1 or 1/2 for two types of systems. The traffic
on nodes of a network of Internet routers and on/off state occurrence of
junctions in a microprocessor electronic network had scaling exponents of 1/2
while visitor traffic to a group of WWW pages, traffic at a group of highway
junctions, and water flow in different locations in a river network
demonstrated scaling exponents of 1. In two simulations, one based on random
walks on a scale-free network and the other by simulating shortest-path
traffic on a scale-free network, they were able to explain the scaling
exponent of 1/2 as being based on the channeling of traffic through select
nodes and arises from internal or endogenous network dynamics. The power
scaling exponent of 1 on the other hand is shown to be universal when the
amount of traffic is driven by external forces as well as endogenous dynamics
similar to an open system. They believe the power exponent of 1 is more
universal than since it results from the interplay of endogenous and exogenous
pressures. In a subsequent paper [13], they give a method of extracting the
endogenous and exogenous traffic and propose a metric, $\eta_{i}$, that
defines the predominance of external or internal influences on traffic
dynamics by the equation
$\eta_{i}=\frac{\sigma^{ext}_{i}}{\sigma^{int}_{i}}$ (36)
Where $\eta_{i}\gg 1$ indicates an externally driven system while $\eta_{i}\ll
1$ indicates a systems dominated by internal dynamics. $\eta_{i}$ can vary on
different time scales as [52] showed using trading records from the New York
Stock Exchange where internal dynamics were dominant on the scale of minutes
while external ones were dominant on the hours and days time scales. In [178],
a power law-scaling relationship was also found via an NNN routing simulation
on a scale-free network and scaling was demonstrated also exhibiting either an
exponent of 1/2 or 1.
The generality of the results and the universality of the classes proposed in
these papers has recently been disputed though. Duch and Arenas [49] perform
several measurements relating flow and fluctuations on data from the Abilene
Internet backbone and claim that $\alpha$ varies between 0.71 and 0.86 and not
1/2 as represented in the first papers. They also propose that this derivation
is due to the original papers disregarding congestion in networks and show
analytically that for short timescales of measurement, an $\alpha$ of 1/2 is a
trivial result but a false generality once the timescales are extended and
other parameters come into play. They conclude that there is a scaling
relationship but no universality classes as claimed. Meloni et. al. [126] go
even further and say that under certain conditions, power-law scaling between
flows and fluctuations should be abandoned. They conduct a simulation of a
random diffusion process on a network of packets measuring scaling as
influenced by the time window of measurements, the degrees of the nodes flows
and fluctuations are measured on, and the volume of packets in the network.
They produce an analytical result that explains power scaling behavior between
the two quantities only under the conditions where the noise fluctuations in
the system and/or the time window size are relatively small. Otherwise
$\alpha$ tends towards 1 and does not display power-law scaling. Also, they
show even in networks with power-law scaling, $\alpha$ can scale differently
at 1/2 for low-degree nodes or 1 for high degree nodes showing that within
networks there may be varying scaling depending on the degrees of the nodes.
Finally, Han et. al. in [83] measure $\alpha$ for the download rates of an
Econophysics web database and find an $\alpha$ varying from 0.6 to 0.89
depending on the length of the sample time window. They confirm the power law
scaling between flux and fluctuations but do not find any universal exponents.
From these results the research between flows and fluctuations in networks is
still in its earliest stages but holds out much promise for future progress.
### 8.2 Internet worm traffic & BGP storms
In [209], a simulation by Yuan and Mills was touched on that aimed to try to
predict part of the large-scale impact of a rapidly spreading Internet worm.
Recent increases in the amount and sophistication of malicious code released
on the Internet including the use of “zombie” computers for large-scale DDoS
attacks has demonstrated this is far from just a theoretical exercise. An
increasingly large literature base on the Internet traffic effects of
epidemics has arisen, particularly after the Code Red outbreak in 2001 (which
the author had the dubious honor of handling as a network security
administrator at the time). Again, to stay with the scope of the paper the
aspects of Internet worms discussed here will be tightly limited to effects on
traffic, both measured and predicted, and will not delve into the voluminous
theoretical work of epidemiology on scale-free networks or other topologies
[153] or much of the new literature with specialized epidemic models for
computer worms and the effect of topology on their spread [75, 204, 101, 133,
193]. Suffice to say, mathematical results indicate for an ideal epidemic
spreading on a scale-free topology, there is no minimum epidemic threshold and
that an infection of any size can theoretically spread through the entire
network.
The two most studied Internet worms have been the Code Red (start: July 19,
2001), Nimda (start: September 19, 2001), and SQL Slammer (start: January 25,
2003). The Slammer, though not holding a malicious payload, was the fastest
spreading worm in history [132]. What has often been found is that the worms
not only cause trouble for the computers they affect, they create large-scale
traffic patterns that can disrupt the normal behavior of entire networks.
Often, a worm spreads by exploiting a vulnerability in computers and will try
to infect random computers by testing an IP address at random or due to
certain rules. With potentially millions of computers sending out such probes
at once it is easy to see how normal traffic patterns can be seriously
disrupted.
In particular, worms have often been the culprit of what could be termed a
large-scale instability in the BGP routing system called a BGP update storm or
BGP storm. In a BGP storm, the normal level of BGP updates sent to update the
router table can rise by several orders of magnitude and sometimes disrupt
traffic [40, 164]. For example, in [40] they describe how during the Nimda
worm normal BGP update traffic of 400 advertisements per minute jumped to
10,000 advertisements per minute. This is not because the worms infected the
routers themselves but because the worms caused large packet flows which
overwhelmed the router memory and CPU limits and caused them to crash. These
router crashes caused frenzied reorienting of the Internet router topology.
BGP storms are interesting in that both traffic and topology is rapidly
changing. BGP storms may be an avenue for both physicists and engineers to
investigate the relationship between topology and traffic in a situation when
both are largely in flux.
Yuan and Mills expanded their work from [209] to a full paper [210] that looks
at spatiotemporal correlations between routers and hosts in several types of
large scale DDoS attacks. They find that DDoS attacks may cause traffic
variations at correlated routers and hosts besides just the target. Because of
these large traffic altering phenomena, certain spectral techniques have been
researched to identify DDoS attacks. Some of these are summarized in the next
section.
Figure 6: A view of a historical BGP storm caused by the onset of the Slammer
worm on January 25, 2003 similar to charts in [164]. The chart shows the
number of BGP updates divided by the standard deviation of the number of
updates over three days, January 24-26, 2003, based on 15 minute intervals.
Updates are from routers at the University of Oregon’s RouteViews project.
### 8.3 Traffic oscillations/periodicities
Earlier periodic behavior in Internet traffic was casually mentioned as
theoretical assumptions of TCP traffic. Also, one of the consequences the
self-similar nature of traffic is the $1/f$ spectral behavior of the traffic.
Beyond these, however, there are a plethora of traffic periodicities that
represent oscillations in traffic over periods of several scales of magnitude
from milliseconds to weeks. Many of these are well-defined and classified.
Their origin has two possible sources: first, software or transmission driven
periodicities which range on the time scale of milliseconds, seconds, or in
rare cases, hours. Second are user driven periodicities which range on the
time scale of days, weeks, and possibly longer. This new area of research has
been dubbed _network spectroscopy_[22] or _Internet spectroscopy_ and is
finding uses in applications such as identifying traffic sources via traffic
periodicity “fingerprints” to early detection of denial of service or other
hacker attacks by detecting anomalous oscillations in the traffic spectrum
similar to vibration analysis of faulty machinery.
The causes and periods of various known periodicities are summarized in figure
7. The values can have a general range of deviation so the periods are not
always exact, but are a good guide to the major periodicities. User traffic
driven periodicities were the first known and most easily recognized. The
first discovered and most well-known periodicity is the 24 hour diurnal cycle
and its harmonic of 12 hours. These cycles have been known for decades and
reported as early as 1980 and again in 1991 as well as in many subsequent
studies[165, 108, 148, 163, 139, 145, 69]. This obviously refers to the 24
hour work-day and its 12-hour second harmonic as well as activity from around
the globe. The other major periodicity from human behavior is the week with a
period of 7 days [148, 163, 25] and a second harmonic at 3.5 days and barely
perceptible third harmonic at 2.3 days. There are reports as well of seasonal
variations in traffic over months [84], but mostly these have not been firmly
characterized. Long period oscillations have been linked to possible causes of
congestion and other network behavior related to network monitoring [139,
145]. One note is that user traffic driven periodicities tend to appear in
protocols that are directly used by most end users. The periodicities appear
TCP/IP not UDP/IP and are mainly attributable to activity with the HTTP and
SMTP protocols. They also often do not appear in networks with low traffic or
research aims such as the now defunct 6Bone IPv6 test network.
The autonomous, non-user driven, periodicities operate mostly at timescales
many orders of magnitude smaller than user behavior. At the lowest period, and
correspondingly highest frequency, are the periodicities due to the throughput
of packet transmission at the link level. This has been termed the
“fundamental frequency” [84] of a link and can be deduced from the equation:
$f=\frac{T}{s}$ (37)
where $T$ is the average throughput of the link and $s$ is the average packet
size at the link level. A quick inspection reveals this equation is identical
to that for the flow rate given by equation 1. Indeed, the fundamental
frequency is the rate of packet emission across the link and is the highest
frequency periodicity possible. The theoretical maximum fundamental frequency
is given by
$f_{max}=\frac{B}{MTU}$ (38)
where B is the bandwidth of the link and the packet size is the MTU packet
size. Therefore for 1 Gigabit, 100 Mbps, and 10 Mbps Ethernet links with MTU
sizes of 1500 bytes, the theoretical maximum fundamental frequencies are 83.3
kHz, 8.3 kHz, and 833 Hz respectively.
The usual measured fundamental frequencies via power spectrum diagrams are
lower than the theoretical fundamental frequencies due to lower throughput.
The fundamental frequency also generally displays harmonics as well [84].
Broido, et. al. [23] believe there are thousands of periodic processes in the
Internet. Among other prominent recognized periodicities are BGP router table
updates sent every 30 seconds, SONET frames transmitted every 125$\mu$s, DNS
updates transmitted with periods of 75 minutes, 1 hour, and 24 hours due to
default settings in Windows 2000 and XP DNS software[22], and in TCP flows ACK
packets at a frequency of 1/RTT [23, 21, 32] with RTTs usually ranging from
10ms to 1s.
The main practical applications being researched for network spectroscopy are
inferring network path characteristics such as bandwidth, digital
fingerprinting of link transmissions, and detecting malicious attack traffic
by changes in the frequency domain of the transmission signal. In [97, 85],
the authors use analysis of the distribution of packet interarrival times to
infer congestion and bottlenecks on network paths upstream. In [32, 30, 90,
89, 117] various measures of packet arrival distributions, particularly in the
frequency domain, are being tested to recognize and analyze distributed denial
of service or other malicious attacks against computer networks. Inspecting
the frequency domain of a signal can also reveal the fingerprints of the
various link level technologies used along the route of the signal as is done
in [23, 39].
Figure 7: A rough breakdown of the major periodicities in Internet traffic
showing the responsible protocols and their period in seconds. The
periodicities span over 12 orders of magnitude and different protocol layers
tend to operate on different time scales.
### 8.4 Biological/ecological models and Internet traffic
Comparisons of the Internet to biological or ecological systems are legion and
range from the theoretically precise to philosophical speculations in both
popular fiction such as William Gibson’s Neuromancer and Masamune Shirow’s
Ghost in the Shell [77, 167] as well as in the opinions of some researchers
such as Vernor Vinge’s “Singularity”[189].
The focus here is on scientific papers which have used mathematical models,
biological or ecological, to describe functions of the Internet or compare
certain functions to physical systems. The growth of the Internet’s nodes in
terms of a birth/death process is covered in [154]. In [73] Fukuda, Nunes-
Amaral, and Stanley use several statistical analyses to show a striking
similarity between variations in daily Internet active connections in a data
trace and statistics on heartbeat intervals. By both separating both non-
stationary time series into stationary segments and using DFA, they show that
the magnitudes of activity for night time (non-congested) Internet connections
and healthy heartbeats are statistically very similar. Likewise day time
(congested) Internet connections and diseased heartbeat intervals are also
similar in their fluctuations. They propose that a general nonlinear systems
explanation underlies both systems and given that the heart rate is controlled
by the autonomic nervous system, understandings of Internet functions and
properties could be used to study the autonomic nervous system as well.
Several authors have also used ecological interaction models such as
mathematical models of competition and mutualism to study interaction between
web sites and search engines. In [124, 106, 107], competition and cooperation
between web sites are analyzed using the n-competitor Lotka-Volterra
differential equations from ecology. The steady state of “winner takes all” or
multiple participants is extracted from stability criterion and compared to
actual market competition. In [190], another model which includes a
cooperation effect is introduced to study the same dynamics. The interesting
analogy between search engines and websites as a mutualistic relationship is
introduced in [191]. The postulated mutualism is obligate for the search
engine and facultative for the websites. This is similar to the sea anemone
and hermit crab or mycorrhiza and plant mutualisms in nature. They show that
strong mutual support for web sites by search engines and vice versa offers
the best opportunity for long-term sustainability and growth.
The last paper covered in this section is a recent publication which draws
similarities between the energy use and scaling of information networks and
metabolic scaling phenomena such as Kleiber’s Law, the 3/4 power law scaling
of organism mass and metabolism [137]. Though the bulk of the paper is
comparing the circuitry density and area for electronic circuits
microprocessors, they derive, with a limited set of data points, a scaling
relationship between the total processing power of hosts on the Internet and
Internet backbone bandwidth with a scaling exponent of about 2/3. Future
research in this direction, especially if a valid scaling law relating
topology and dynamics is discovered, will surely be very fruitful.
Name | Data
---|---
CAIDA | Probably largest and most comprehensive repository of all types of Internet data and research. Hosted by UC San Diego
Crawdad | Collection of wireless data traces made at Dartmouth
Datapository | Collection of topology data, mostly based on BGP
Dimes | A project that seeks to more accurately measure Internet topology by applying traceroute and ping from a collection of different users with downloaded software around the world in a model similar to SETI@Home
MOME | Large collection of traffic traces conducted by European based researchers and collaborations
NLANR | Older traffic trace project; now mostly housed at CAIDA
ns-2 | Commonly used network traffic simulator in the network engineering community
PingER | Stanford project to monitor Ping response in IPv4 and IPv6 across the Internet
PREDICT | A collection of datasets of Internet topology and traffic tailored towards predicting and defending against cyberattacks. Must apply for data access
RouteViews | U of Oregon’s database on Internet routing tables and BGP data
tcpdump | Main program used to collect traffic for analysis; used in many packet sniffing programs
UMass Trace Repository | Collection of publicly available traffic traces conducted at the University of Massachusetts Amherst
WAND | Collection of traces done by the University of Waikato, New Zealand
WIDE MAWI | Japan’s WIDE Project traffic trace archive; data source of many graphs in this paper
Table 7: Common Internet traffic data sources & software.
## 9 Conclusion
With every passing year, research is making us more and more aware of the
complex dynamics and interplay of factors on the Internet. Though many may
haphazardly use the terms self-organization, emergence, or power law this
review has hopefully laid out the concrete facts about what is known clearly
about Internet traffic, what is less clear, and where many new paths can be
beaten. Unlike most systems which are amenable to constant analysis over long
time periods, the Internet is ever changing. What we understand today may not
completely apply several years from now. In addition, our knowledge of long
range correlations and dynamics among multiple sites and links is still in its
infancy. Large-scale congestion is the only possible large-scale property
which has been studied in any detail and it remains to be seen if it is the
only one that exists. There is much room for speculation on these matters
without being irresponsibly fanciful.
At the core, these issues are more than academic since the long-term
efficiency and stability will require us to understand the Internet and its
traffic well enough to optimize it for the ends of users. Advances in
understanding the Internet are also enabled and constrained about our
knowledge of nonlinear dynamics and complex systems in general. As more themes
and discoveries about these systems emerge, they will doubtlessly provide us
with more tools with which to investigate the Internet and uncover more of the
story behind its dynamics. Finally, as mentioned earlier, it is essential for
more cross-disciplinary cooperation to take place in order to accelerate our
understanding of Internet phenomena. The two groups have cooperated in some
areas and are hardly irreconcilable. Combining both toolkits can definitely
bring forth some more surprising and rewarding results.
Finally, though this paper has been heavy on esoteric technical aspects of the
Internet, we must not lose sight of the whole, as the poet Walt Whitman once
wonderfully wrote [195],
> When I heard the learn’d astronomer;
> When the proofs, the figures, were ranged in columns before me;
> When I was shown the charts and the diagrams, to add, divide, and measure
> them;
> When I, sitting, heard the astronomer, where he lectured with much applause
> in the
> lecture-room,
> How soon, unaccountable, I became tired and sick;
> Till rising and gliding out, I wander’d off by myself,
> In the mystical moist night-air, and from time to time,
> Look’d up in perfect silence at the stars.
>
For everything said about self-similarity, phase transitions, and related
matter we must never lose sight of the Internet as the wonderful invention it
has been in its cultural, economic, and technological aspects uniting those
from around the world. Even if in only a small part, this should animate and
encourage our research into the future.
## References
* [1] Abe, S. & Suzuki, N., Omori’s law in the Internet traffic, _Europhys. Lett._ 61 (2003), 852-855.
* [2] Abry, P. & Veitch, D., Wavelet Analysis of Long-Range Dependence Traffic, _IEEE Transactions on Information Theory_ , 44:1 (1998), pp.2-15.
* [3] Abry, P., Baraniuk, R., Flandrin, P., Riedi, R., & Veitch, D., Multiscale Network Traffic Analysis, Modeling, and Inference Using Wavelets, Multifractals, and Cascades, _IEEE Signal Processing Magazine_ , 19:3 (2002), 28-46.
* [4] Abry, P., Flandrin, P., Taqqu, M.S. & Veitch, D., Wavelets for the analysis, estimation, and synthesis of scaling data, In _Self-similar network traffic and performance evaluation_ edited by Park, K. & Willinger, W. (John Wiley & Sons, New York, 2000) pp. 39-88.
* [5] Abry, P., Flandrin, P., Taqqu, M.S. & Veitch, D. Self-similarity and long-range dependence through the wavelet lens, In _Theory and Applications of Long-Range Dependence_ edited by Doukan, P., Taqqu, M.S., Oppenheim, G. (Birkh auser, Boston, 2003) pp. 527-556.
* [6] Adamic, L., Lukose, R., Puniyani, A., & Huberman, B. Search in power-law networks, _Phys. Rev. E_ , 64 (2001), 046135.
* [7] Albert, R. & Barabási, A.L., Statistical mechanics of complex networks, _Rev. Mod. Phys._ , 74 (2002), 47-97.
* [8] Alderson, D.L. & Willinger, W., A contrasting look at self-organization in the Internet and next-generation communication networks, _IEEE Communications Magazine_ , 43:7, (2005) 94-100 .
* [9] Alderson, D.L., Li, L., Willinger, W., & Doyle, J. Understanding Internet topology: Principles, models, and validation, _IEEE/ACM Transactions on Networking_ , 13:6, (2005) 1205-1218.
* [10] Alderson, D.L., Chang, H., Roughan, M., Uhlig, S. & Willinger, W., The many facets of Internet topology and traffic, _Networks and Heterogenous Media_ , 1:4, (2006) 569-600.
* [11] Arenas, A., Diàz-Guilera, A. & Guimerá, R., Communication in Networks with Hierarchical Branching, _Phys. Rev. Lett._ , 86 (2001), 3196-3199.
* [12] Argollo de Menezes, M, & Barabási, A.L., Fluctuations in Network Dynamics, _Phys. Rev. Lett._ , 92 (2004), 028701.
* [13] Argollo de Menezes, M. & Barabási, A.L., Separating Internal and External Dynamics of Complex Systems, _Phys. Rev. Lett._ , 93 (2004), 068701.
* [14] Baker, M.G., Hartman, J.H., Kupfer, M.D., Shirriff, K.W., & Ousterhout, J.K., Measurements of a distributed file system , in _Proceedings of the 13th ACM Symposium on Operating System Principles_ , (Pacific Grove, CA ,1991), pp. 198-212.
* [15] Albert, R., Jeong, H., & Barabási, A.L., Internet diameter of the World-Wide Web, _Nature_ , 401 (1999), 130131 .
* [16] Bardet, J.M. & Kammoun, I., Asymptotic properties of the detrended fluctuation analysis of long-range-dependent processes, _IEEE Transactions on Information Theory_ , 54:5, (2008) 2041-2052.
* [17] Barthélemy, M., Gondran, B., Guichard, E., Large scale cross-correlations in Internet traffic, _Phys. Rev. E_ , 66 (2002), 056110.
* [18] Barthélemy, M., Gondran, B., & Guichard, E., Spatial structure of the internet traffic, _Physica A_ , 319:1 (2003), 633-642.
* [19] Basher, N., Mahanti, A., Williamson, C. & Arlitt, M., A comparative analysis of web and peer-to-peer traffic, _Proceedings of the 17th International Conference on the World Wide Web_ (Beijing, China, 2008), pp. 287-296.
* [20] Boccaletti, S., Latora, V., Moreno, Y., Chavez, M., & Hwang, D.U., Complex networks: Structure and dynamics, _Phys. Repp._ , 424:4-5 (2006), 175-308.
* [21] Broido, A., Invariance of Internet RTT spectrum, in _Proceedings of ISMA Conference_ , (San Diego, CA, 2002).
* [22] Broido, A., Nemeth, E., & Claffy, K.C., Spectroscopy of DNS Update Traffic, _ACM SIGMETRICS 2003_ , 31:1 (2003) 320-321.
* [23] Broido, A., R King, Nemeth, E., Claffy, K.C., Radon Spectroscopy of Packet Delay, in _Proceedings of the IEEE High-Speed Networking Workshop 2003_ , (San Diego, CA 2003).
* [24] Brownlee, N. & Claffy, K.C., Understanding Internet traffic streams: dragonflies and tortoises, _IEEE Communications Magazine_ , 40:10 (2002) 110-117.
* [25] Burgess, M., Haugerud, H., Straumsnes, S., & Reitan, T., Measuring system normality, _ACM Trans. on ComSys._ , 20:2 (2002) 125-160.
* [26] Cáceres, R., Measurements of wide-area Internet traffic University of California Berkeley Technical Report, CSD-89-550 (1989).
* [27] Cáceres, R., Danzig, P., Jamin, S., & Mitzel, D., Characteristics of wide-area TCP/IP conversations, _Proceedings of SIGCOMM ’91_ , 21:4 (Zurich, Switzerland, 1991), pp. 101-112.
* [28] Campos, I., Tarancón, A., Clérot, F., & Fernández, L.A., Thermal and repulsive traffic flow, _Phys. Rev. E_ , 52 (1995) 5946-5954.
* [29] Chang, H.S., Jamin, S., & Willinger, W., Internet connectivity at the AS-level: an optimization-driven modeling approach, _Proceedings of the ACM SIGCOMM workshop on Models, methods and tools for reproducible network research_ , (Karlsruhe, 2003) pp. 33-46.
* [30] Chen, Y. & Hwang, K., Collaborative detection and filtering of shrew DDoS attacks using spectral analysis, _Journal of Parallel and Distributed Computing,_ 66:9 (2006) 1137-1151.
* [31] Chen, Z.Y. & Wang, X.F., Effects of network structure and routing strategy on network capacity, _Phys. Rev. E_ , 73 (2006) 036107.
* [32] Cheng, C.M., Kung, H.T., & Tan, K.S., Use of spectral analysis in defense against DoS attacks, in _Proceedings of IEEE GLOBECOM ’02_ , 3 (Taipei, Taiwan, 2002) pp. 2143-2148.
* [33] Chowdhury, D., Santen, L., & Schadschneider, A., Statistical physics of vehicular traffic and some related systems, _Phys. Repp._ , 329:4-6 (2000) 199-329.
* [34] Claffy, K.C. & McCreary, S., Trends in Wide Area IP Traffic Patterns: A View from Ames Internet Exchange, _Proceedings of the 13th ITC Specialist Seminar on Internet Traffic Measurement and Modeling_ (Monterey, CA, 2000).
* [35] Claffy, K.C. & Polyzos, G.C., Traffic characteristics of the T1 NSFNET backbone, _Proceedings of INFOCOM ’93_ (San Francisco, CA, 1993) pp. 885-892.
* [36] Claffy, K.C., Braun, H.W., & Polyzos, G.C., A parameterizable methodology for Internet traffic flow profiling,, _IEEE Journal on Selected Areas in Communication_ , 13 (1995) 1481-1494.
* [37] Clauset, A., Shalizi, C.R., & Newman, M.E.J., Power-law distributions in empirical data, to appear _SIAM Review_ , preprint arXiv:0706.1062 (2009).
* [38] Clegg, R.G., A practical guide to measuring the Hurst parameter, _International Journal of Simulation: Systems, Science & Technology_, 7:2 (2006) 3-14.
* [39] Coates, M., Hero, A., Nowak, R. & Yu, B., Internet tomography, _IEEE Signal Processing Magazine_ , 19:3 (2002) 47-65.
* [40] Cowie, J., Ogielski, A.T., Premore, B.J., & Yuan, Y.G., Internet worms and global routing instabilities, in _Proceedings of SPIE_ , 4868 (San Jose, CA, 2002) pp. 195-199.
* [41] Crovella, M.E. & Bestavros, A., Self-similarity in World Wide Web traffic: evidence and possible causes , _IEEE ACM Transactions on Networking_ , 5:6 (1997) 835-846.
* [42] Crucitti, P., Latora, V., & Marchiori, M., Model for cascading failures in complex networks, _Phys. Rev. E_ , 69 (2004) 045104.
* [43] Csabai, I.J., 1/f noise in computer network traffic, _Phys. A: Math. Gen._ , 27 (1994) 417-421.
* [44] Daganzo, C.F., Cassidy, M.J. & Bertini, R.L. Causes and Effects of Phase Transitions in Highway Traffic University of California, Berkeley Institute of Transportation Studies Research Report UCB-ITS-RR-97-8, (1997).
* [45] Dorogovtsev, S.N. & Mendes, J.F.F., Evolution of networks, _Adv. in Phys._ , 51:4 (2002) 1079-1187.
* [46] Dovrolis, C., Ramanathan, P., & Moore, D., What do packet dispersion techniques measure?, _Proceedings of INFOCOM 2001_ , 2 (Anchorage, AL, 2001) 905-914.
* [47] Downey, A.B., The structural cause of file size distributions, in _Proceedings of the 9th IEEE/MASCOTS (Modeling, Analysis and Simulation of Computer and Telecommunication Systems)_(Cincinnati, OH, 2001) pp. 361-370.
* [48] Doyle, J.C., Alderson, D.L., Li, L., Low, S., Roughan, M., Shalunov, S., Tanaka, R., & Willinger, W., The ‘robust yet fragile’ nature of the Internet, _Proc. Natl. Acad. of Sci._ , 102:41 (2005) 14497-14502.
* [49] Duch, J. & Arenas, A., Scaling of Fluctuations in Traffic on Complex Networks, _Phys. Rev. Lett._ , 96 (2006) 218702.
* [50] Echenique, P., Gómez-Gardeñas, J., & Moreno, Y., Improved routing strategies for Internet traffic delivery, _Phys. Rev. E_ , 70 (2004) 056105.
* [51] Echenique, P., Gómez-Gardeñas, J., & Moreno, Y., Dynamics of jamming transitions in complex networks, _Europhys. Lett._ , 71 (2005) 325.
* [52] Eisler, Z., Kertész, J., Yook, S.H., & Barabási, A.L., Multiscaling and non-universality in fluctuations of driven complex systems, _Europhys. Lett._ , 69 (2005) 664-670.
* [53] El Khayat, I., Geurts, P., & Leduc, G., On the Accuracy of Analytical Models of TCP Throughput in _Networking Technologies, Services, and Protocols; Performance of Computer and Communication Networks; Mobile and Wireless Communications Systems_ , (Springer, Berlin, 2006) pp. 488.
* [54] Erman, J., Mahanti, A., Arlitt, M., & Williamson, C., Identifying and discriminating between web and peer-to-peer traffic in the network core, _Proceedings of the 16th International Conference on World Wide Web_ (Banff, Alberta, 2007) pp. 883-892.
* [55] Erramilli, A., Roughan, M., Veitch, D., & Willinger, W., Self-similar traffic and network dynamics, _Proc. of the IEEE_ , 90 (2002) 800-819.
* [56] Estan, C. & Varghese, G., New directions in traffic measurement and accounting: Focusing on the elephants, ignoring the mice, _ACM Trans. on ComSys._ , 21:3 (2003) 270-313.
* [57] Faloutsos, M., Faloutsos, P., & Faloutsos, C., On power-law relationships of the Internet topology _ACM SIGCOMM ComComm. Rev._ , 29 (1999) 251-262.
* [58] Fang, W. & Peterson, L., Inter-AS traffic patterns and their implications, _Proceedings of IEEE GLOBECOM ’99_ 3 (Rio de Janeiro, Brazil, 1999) pp. 1859-1868.
* [59] Fei, A.G., Pei, G.Y., Liu, R., & Zhang, L.X., Measurements on Delay and Hop-Count of the Internet, in _Proceedings of IEEE GLOBECOM ’98-Internet Mini-Conf._ , (Sydney, Australia, 1998).
* [60] Fekete, A. & Vattay, G., Self-similarity in bottleneck buffers, in the _Proceedings of IEEE GLOBECOM ’01_ , 3 (San Antonio, TX, 2001) pp.1867-1871.
* [61] Feldmann, A., Gilbert, A.C., Willinger, W., & TG Kurtz, The changing nature of network traffic: scaling phenomena, _ACM SIGCOMM ComComm. Rev._ , 28:2 (1998) 5-29.
* [62] Feldmann, A., Gilbert, A.C., & Willinger, W., Data networks as cascades: investigating the multifractal nature of Internet WAN traffic, _ACM SIGCOMM ComComm. Rev._ , 28:4 (1998) 42-55.
* [63] Figueiredo, D., Liu, B., Misra, V., & Towsley, D., On the autocorrelation structure of TCP traffic, _Computer Networks_ , 40:3 (2002) 339-361.
* [64] Figueiredo, D., Liu, B., Feldmann, A., Misra, V., Towsley, D., & Willinger, W., On TCP and self-similar traffic, _Performance Evaluation_ , 61:2-3 (2005) 129-141.
* [65] Floyd, S., Connections with multiple congested gateways in packet-switched networks part 1: one-way traffic, _ACM SIGCOMM ComComm. Rev._ , 21:5 (1991) 30-47.
* [66] Floyd, S. & Jacobson, V., Random Early Detection gateways for Congestion Avoidance, _IEEE/ACM Transactions on Networking_ , 1: 4 (1993) 397-413.
* [67] Floyd, S., & Paxson, V., Difficulties in simulating the internet, _IEEE/ACM Transactions on Networking_ , 9:4 (2001) 392-403.
* [68] Fomenkov, M., Keys, K., Moore, D., & Claffy, K.C., Longitudinal study of Internet traffic in 1998-2003, _Proceedings of the Winter International Symposium on Information and Communication Technologies_ (Dublin, Ireland, 2004) pp. 1-6.
* [69] Fowler, H.J. & Leland, W.E., Local area network characteristics, with implications for broadband network congestion management, _IEEE Journal on Selected Areas in Communications_ , 9:7 (1991) 1139-1149.
* [70] Fraleigh, C., Moon, S., Lyles, B., Cotton, C., Khan, M., Moll, D., Rockell, R., Seely, T., & Diot, C., Packet-level traffic measurements from the Sprint IP backbone, _IEEE Network_ , 17:6 (2003) 6-16.
* [71] Frazer, K.D., NSFNET: A Partnership for High-Speed Networking, Final Report 1987-1995, Merit Network, Inc. (1995).
* [72] Fukś, H. & Lawniczak, A., Performance of data networks with random links, _Mathematics and Computers in Simulation_ , 51:1 (1999) 101-117.
* [73] Fukuda, K., Numes Amaral, L.A., & Stanley, H.E., Similarities between communication dynamics in the Internet and the autonomic nervous system, _Europhys. Lett._ , 62 (2003) 189-195.
* [74] Gábor, S. & Csabai, I.J., The analogies of highway and computer network traffic, _Physica A_ , 307:3-4 (2002) 516-526.
* [75] Ganesh, A., Massoulie, L., Towsley, D., The effect of network topology on the spread of epidemics, in _Proceedings of INFOCOM 2005. 24th Annual Joint Conference of the IEEE Computer and Communications Societies_ , 2 (Miami, 2005) 1455-1466.
* [76] Gao, J.B. & Rubin, I., Multiplicative multifractal modeling of long-range-dependent network traffic, _Intl. J. of Comm. Sys._ , 14:8 (2001) 783-801.
* [77] Gibson, W., _Neuromancer_ (Ace Books, New York, 1984).
* [78] Gilbert, A.C., Willinger, W., & Feldmann, A. Scaling analysis of conservative cascades, with applications to network traffic, _IEEE Transactions on Information Theory_ , 45:3 (1999) 971-991.
* [79] Gong, W.B., Liu, Y., Misra, V., & Towsley, D., On the tails of Web file size distributions, in _Proceedings of the 39th Allerton Conference on Communication, Control, and Computing_ , (Monticello, Illinois, 2001).
* [80] Guimerà, R., Arenas, A., Díaz-Guilera, A., & Giralt, F., Dynamical properties of model communication networks, _Phys. Rev. E_ , 66 (2002) 026704.
* [81] Guo, L., Crovella, M., & Matta, I., How does TCP generate pseudo self-similarity, in _Proceedings of the 9th IEEE/MASCOTS (Modeling, Analysis and Simulation of Computer and Telecommunication Systems)_ (Cincinnati, OH, 2001) pp. 215-223.
* [82] Hága, P., Pollner, P., Simon, G., Csabai, I.J., & Vattay, G., Self-generated self-similar traffic, _Nonlinear Phenomena in Complex Systems_ , 6:4 (2003) 814-823.
* [83] Han, D.D., Liu, J.G., & Ma, Y.G., Fluctuation of the Download Network, _Chin. Phys. Lett._ , 25 (2008) 765-768.
* [84] He, X., Papadopoulos, C., Heidemann, J., & Hussain, A., Spectral Characteristics of Saturated Links, University of Southern California Technical Report, USC-CSD-TR-827 (2004).
* [85] He, X., Papadopoulos, C., Heidemann, J., Mitra, U., Riaz, U., & Hussain, A., Spectral Analysis of Bottleneck Traffic, University of Southern California Technical Report, USC/CS Technical Report 05-853 (2005).
* [86] He,Y., Siganos, G., Faloutsos, M., & Krishnamurthy, S., “A Systematic Framework for Unearthing the Missing Links: Measurements and Impact” in _4th USENIX Symposium on Networked Systems Design & Implementation_, (Cambridge, MA, 2007), pp. 187-200.
* [87] Hu, M.B., Wang, W.X., Jiang, R., Wu, Q.S., & Wu, Y.H., Phase transition and hysteresis in scale-free network traffic, _Phys. Rev. E_ , 75 (2007) 036102.
* [88] Hu, M.B.,, Wang, W.X., Jiang, R., Wu, Q.S., & Wu, Y.H., The effect of bandwidth in scale-free network traffic, _Europhys. Lett._ , 79 (2007) 14003.
* [89] Hussain, A., Heidemann, J., & Papadopolous, C., A framework for classifying denial of service attacks, in _Proceedings of the 2003 Conference on Applications, Technologies, Architectures, and Protocols for Computer Communications_ (Karlsruhe, Germany, 2003) pp. 99-110.
* [90] Hussain, A., Heidemann, J., & Papadopolous, C., Identification of Repeated Attacks Using Network Traffic Forensics, USC/ISI Technical Report ISI-TR-2003-577b (2004).
* [91] Iliofotou, M. et. al., Graph-based P2P traffic classification at the Internet backbone, _Proceedings of the 28th IEEE international conference on Computer Communications Workshops_ (Rio de Janeiro, 2009) pp. 37-42.
* [92] Jain, R. & Routhier, S., Packet trains: measurements and a new model for computer network traffic, _IEEE Journal on Selected Areas in Communication_ , 4:6 (1986) 986-995.
* [93] Kaiser, G., _A Friendly Guide to Wavelets_ , (Springer, Berlin, 1994).
* [94] Karagiannis, T., Molle, M., & Faloutsos, M., Long-Range Dependence: Ten Years of Internet Traffic Modeling, _IEEE Internet Computing_ , 8:5 (2004) 57-64.
* [95] Karagiannis, T., Papagiannaki, K., & Faloutsos, M., BLINC: multilevel traffic classification in the dark,_ACM/ SIGCOMM ComComm Review_ , 35:4 (2004) 229-240
* [96] Karagiannis, T., Broido, A., Faloutsos, M. & Claffy, K.C., Transport layer identification of P2P traffic, _Proceedings of the 4th ACM SIGCOMM conference on Internet measurement_ (Taromina, 2004) pp. 121-134.
* [97] Katabi, D. & Blake, C., Inferring Congestion Sharing and Path Characteristics from Packet Interarrival Times, MIT Technical Report, MIT-LCSTR-828 (2001).
* [98] Kerner, B.S., & Rehborn, H., Experimental Properties of Phase Transitions in Traffic Flow, _Phys. Rev. Lett._ , 79 (1997) 4030-4033.
* [99] Khalil, K.M., Luc, K.Q., & Wilson, D.V., LAN traffic analysis and workload characterization, _Proceedings of the 15th Conference on Local Computer Networks_ (Minneapolis, MN, 1990) pp. 112-122.
* [100] Kim, H. et. al., Internet traffic classification demystified: myths, caveats, and the best practices, _Proceedings of the 2008 ACM CoNEXT Conference_ , (Madrid, 2008) Article 11.
* [101] Kim, J.H., Radhakrishnan, S., Dhall, S.K., Measurement and analysis of worm propagation on Internet network topology, in _Proceedings of the 13th International Conference on Computer Communications and Networks_ (Chicago, 2004) pp. 495-500.
* [102] Krioukov, D., Chung, F., Claffy, K.C., Fomenkov, M., Vespignani, A., & Willinger, W., The Workshop on Internet Topology (WIT) Report, _ACM SIGCOMM ComComm. Rev_ , 37:1 69-73 (2007).
* [103] Krishnamurthy, B. & Willinger, W., What are our standards for validation of measurement-based networking research?, _Performance Evaluation Review_ , 36:2, 64-69 (2008).
* [104] Kushida, T., An empirical study of the characteristics of Internet traffic, _ComComm._ , 22:17 (1999) 1607-1618.
* [105] Lévy-Véhel, J. & Riedi, R., Fractional Brownian motion and data traffic modelling, in _Fractals in Engineering: New Trends in Theory and Applications_ , edited by J Lévy-Véhel & E Lutton, (Springer-Verlag, Berlin, 1997) pp. 185-202.
* [106] López, L. & Sanjuán, M., Defining strategies to win in the Internet market, _Physica A_ , 301:4 (2001) 512-534.
* [107] López, L., Almendral, J.A., & Sanjuán, M., Complex networks and the WWW market, _Physica A_ , 324:4 (2003) 754-758.
* [108] Lakhina, A., Papagiannaki, K., Crovella, M.E., Diot, C., Kolaczyk, E., & Taft, N., Structural analysis of network traffic flows, _Performance Evaluation_ Review, 32:1 (2004) 61-72.
* [109] Lakshman, T.V. & Madhow, U., The performance of TCP/IP for networks with high bandwidth-delay products and random loss, _IEEE/ACM Transactions on Networks_ , 5:3 (1997) 336-350.
* [110] Lan, K. & Heidemann, J., On the correlation of Internet flow characteristic, University of Southern California Information Sciences Institute Technical Report, ISI-TR-574, USC/ISI, (2003).
* [111] Lan, K. & Heidemann, J., A measurement study of correlations of Internet flow characteristics, _Computer Networks_ , 50:1 (2006) 46-62.
* [112] Lawniczak, A.T. & Tang, X.W., Network traffic behaviour near phase transition point, _Eur. Phys. J. B_ , 50:2 (2006) 231-236.
* [113] Lawniczak, A.T. & Tang, X.W., Packet Traffic Dynamics Near Onset of Congestion in Data Communication Network Model, _Acta Physica Polonica B_ , 37:5 (2006) 1579:1604.
* [114] Lee, C.Y., Higher-order correlations in data network traffic, _J. Korean Phys. Soc._ , 45:6 (2004) 1664-1670.
* [115] Leland, W.E., & Wilson, D.V., High time-resolution measurement and analysis of LAN traffic: Implications for LAN interconnection, _Proceedings IEEE lNFOCOM ’91_ (Bal Harbour, Florida, 1991) pp. 1360-66.
* [116] Leland, W.E., Taqqu, M.S., Willinger, W., & Wilson, D.V., On the self-similar nature of Ethernet traffic (extended version) _IEEE/ACM Transactions on Networking_ , 2:1 (1994) 1-15.
* [117] Li, L. & Lee, G., DDoS Attack Detection and Wavelets, _Telecom. Sys._ , 28:4 (2005) 435-451.
* [118] Li, L., Alderson, D.L., Willinger, W., & Doyle, J., A first principles approach to understanding the Internet’s router-level topology, _ACM SIGCOMM ComComm. Rev_ , 34:4, 3-14 (2004).
* [119] Li, L., DL Alderson, Willinger, W., & Doyle, J., Towards a theory of scale-free graphs: Definition, properties, and implications, _Internet Mathematics_ , 2:4, 431-523 (2006).
* [120] Low, S.H., Paganini, F., Wang, J.T., & Doyle, J.C., Linear stability of TCP/RED and a scalable control, _Computer Networks_ , 43:5 (2003) 633-647.
* [121] Mahadevan, P., Krioukov, D., Fomenkov, M., Huffaker, B., Dimitropolous, X., Claffy, K.C., & Vahdat, A., Lessons from three views of the Internet topology preprint arXiv:cs/0508033 (2005).
* [122] Mandelbrot, B.B., Long-run linearity, locally Gaussian processes, H-spectra and infinite variances, _Intl. Econ. Rev._ , 10 (1969) 82-111.
* [123] Mathis, M., Semke, J., Madhavi, J., & Ott, T., The Macroscopic Behavior of the TCP Congestion Avoidance Algorithm, _ACM SIGCOMM ComComm. Rev._ , 27:3 (1997) 67-82.
* [124] Maurer, S.M. & Huberman, B.A., Competitive dynamics of web sites, _Journal of Economic Dynamics and Control_ , 27:12 (2003) 2195-2206.
* [125] MAWI, Widely Integrated Distributed Environment (WIDE) Project, Kanagawa, Japan, MAWI Working Group Traffic Archive (WIDE) Backbone traffic traces) http://mawi.wide.ad.jp/mawi/
* [126] Meloni, S., Goméz-Gardeñas, J., Latora, V., & Moreno, Y., Scaling Breakdown in Flow Fluctuations on Complex Networks, _Phys. Rev. Lett._ , 100 (2008) 208701
* [127] Misra, V., Gong, W.B., & Towsley, D., Fluid-based analysis of a network of AQM routers supporting TCP flows with an application to RED, _ACM/SIGCOMM ComComm Review_ , 30:4 (2000) 151-160.
* [128] Mitzenmacher, M., A brief history of generative models for power law and lognormal distributions, _Internet Mathematics_ , 1:2, 226-251 (2004).
* [129] Mitzenmacher, M., Dynamic models for file sizes and double Pareto distributions, _Internet Mathematics_ , 1:3, 305-333 (2004).
* [130] Mitzenmacher, M., Editorial: The future of power law research , _Internet Mathematics_ , 2:4, 525-534 (2005).
* [131] Moore, A. & Papagiannaki, K., Toward the accurate identification of network applications, in _Passive and Active Network Measurement_ edited by Dovrolis, C. (Springer, Berlin, 2005) pp. 41-54.
* [132] Moore, D., Paxson, V., Savage, S., Shannon, C., Staniford, S., & Weaver, N., Inside the Slammer Worm, _IEEE Security & Privacy_, 1:4 (2003) 33-39.
* [133] Moore,D., Shannon, C., Voelker, G., & Savage, S., Internet Quarantine: Requirements for Containing Self-Propagating Code, in _Proceedings of INFOCOM 2003_ , 3 (San Francisco ,2003) pp.1901-1910.
* [134] Moreno, Y., Gomez, J.B., & Pacheco, A.F., Instability of scale-free networks under node-breaking avalanches, _Europhys. Lett._ , 58 (2002) 630-636.
* [135] Moreno, Y., Pastor-Satorras, R., Váquez, A., & Vespignani, A., Critical load and congestion instabilities in scale-free networks, _Europhys. Lett._ , 62 (2003) 292-298.
* [136] Mori, T., Kawahara, R., Naito, S., & Goto, S., On the Characteristics of Internet Traffic Variability: Spikes and Elephants, _Proceedings of the 2004 International Symposium on Applications and the Internet_ (Tokyo, Japan, 2004) pp. 99-106.
* [137] Moses, M.E., Forrest, S., Davis, A.L., Lodder, M.A., & Brown, J.H., Scaling theory for information networks, _J. Royal Soc. Interface_ , 5:29, (2008) 1469-1480.
* [138] Motter, A.E. & Lai, Y.C., Cascade-based attacks on complex networks, _Phys. Rev. E_ , 66 (2002) 065102.
* [139] Mukherjee, A., On The Dynamics and Significance of Low Frequency Components of Internet Load, University of Pennsylvania Technical Reports, MS-CIS-92-83 (1992).
* [140] Mukherjee, G. & Manna, S., Phase transition in a directed traffic flow network, _Phys. Rev. E_ , 71 (2005) 066108.
* [141] Newman, M.E.J., Scientific collaboration networks. I. Network construction and fundamental results, _Phys. Rev. E_ , 64 (2001) 016131.
* [142] Newman, M.E.J., The structure and function of complex networks, _SIAM Review_ , 45 (2003) 167-256.
* [143] Nievergelt, Y., _Wavelets Made Easy_ , (Springer, Berlin, 1999).
* [144] Ohira, T. & Sawatari, R., Phase transition in a computer network traffic model, _Phys. Rev. E_ , 58 (1998) 193-195.
* [145] Owezarski, P., & Larrieu, N., Internet Traffic Characterization - An Analysis of Traffic Oscillations, in _High Speed Networks and Multimedia Communications_ edited by MM Freire, P Lorenz, & M Lee, (Springer, Berlin, 2004) pp. 96.
* [146] Padhye, J., Firoiu, V., Towsley, D., & Kurose, J., Modeling TCP Reno performance: a simple model and its empirical validation, _IEEE/ACM Transactions on Networks_ , 8:2 (2000) 133-145.
* [147] Papagiannaki, K., Taft, N., Bhattacharyya, S., Thiran, P., Salamatian, K. & Diot, C., A pragmatic definition of elephants in internet backbone traffic, _Proceedings of the 2nd ACM SIGCOMM Workshop on Internet Measurement_ (Marseille, France, 2002) pp. 175-176.
* [148] Papagiannaki, K., Taft, N., Zhang, Z., & Diot, C., Long-term forecasting of Internet backbone traffic, _IEEE Transactions on Neural Networks_ , 16:5 (2005) 1110-1124.
* [149] Park, K. & Willinger, W., Self-similar network traffic: an overview, in _Self-Similar Network Traffic and Performance Evaluation_ , edited by Park, K. & Willinger, W. (John Wiley & Sons, New York, 2000) pp. 1.
* [150] Park, K., Kim, G., & Crovella, M.E., The protocol stack and its modulating effect on self-similar traffic, in _Self-Similar Network Traffic and Performance Evaluation_ , edited by Park, K. & Willinger, W., (John Wiley & Sons, New York, 2000) pp. 349.
* [151] Park, K., Kim, G., & Crovella, M.E., On the relationship between file sizes, transport protocols, and self-similar network traffic, in _Proceedings of the Fourth International Conference on Network Protocols (ICNP’96)_ (Columbus, OH, 1996) pp. 171-180.
* [152] Parziale, L., Britt, D.T., Davis, C., Forrester, J., Liu, W., Matthews, C., & Rosselot, N., _TCP/IP Tutorial and Technical Overview_ , IBM: Internal Technical and Support Organization (2006).
* [153] Pastor-Satorras, R. & Vespignani, A., Epidemic spreading in scale-free networks, _Phys. Rev. Lett._ , 86 (2001) 3200-3203.
* [154] Pastor-Satorras, R. & Vespignani, A., _Evolution and Structure of the Internet: A Statistical Physics Approach_ , (Cambridge University Press, New York, 2004).
* [155] Paxson, V., Growth trends in wide-area TCP connections, _IEEE Network_ , 8:4 (1994) 8-17.
* [156] Paxson, V., Strategies for sound Internet measurement in Proceedings of the 4th ACM SIGCOMM conference on Internet measurement, (Sicily, Italy, 2004) pp. 263-271.
* [157] Paxson, V., & Floyd, S., Wide area traffic: the failure of Poisson modeling, _IEEE/ACM Transactions on Networking_ , 3:3 (1995) 226-244.
* [158] Percival, D.B. & Walden, A.T., _Wavelet Methods for Time Series Analysis_ , (Cambridge University Press, New York, 2000).
* [159] Ranjan, P., Abed, E.H., & La, R.J., Nonlinear instabilities in TCP-RED, _IEEE/ACM Transactions on Networking_ , 12:6 (2004) 1079-1092.
* [160] Stanford SLAC PingER (Ping end-to-end reporting) Project www-iepm.slac.stanford.edu/pinger/
* [161] Riedi, R. & Lévy-Véhel, J., TCP traffic is multifractal: A numerical study, preprint (1997).
* [162] Roberts, J.W., Traffic theory and the Internet, _IEEE Communications Magazine_ , 39:1 (2001) 94-99.
* [163] Roughan, M., Greenberg, A., Kalmanek, C., Rumsewicz, M., Yates, J., & Zhang, Y., Experience in measuring backbone traffic variability: models, metrics, measurements and meaning, in _Proceedings of the 2nd ACM SIGCOMM Workshop on Internet Measurement_ (Marseille, France, 2002) pp. 91-92.
* [164] Roughan, M., Li, J., Bush, R., Mao, Z.Q., & Griffin, T., Is BGP Update Storm a Sign of Trouble: Observing the Internet Control and Data Planes During Internet Worms, in _Proceedings of IEEE SPECTS’06_ (Calgary, Alberta, 2006).
* [165] Shoch, J.F., & Hupp, J.A., Measured performance of an Ethernet local network, _Communications of the ACM_ , 23:12 (1980) 711-721.
* [166] Shan, X.M., Wang, L., Ren, Y., Yuan, J., & Song, Y.H., Advances in the research of Internet complexity, _Journal of Beijing University of Posts and Telecommunications_ , 29:1 (2006) 1-8 (in Chinese).
* [167] Shirow, M., _The Ghost In The Shell (Kokaku Kidotai)_ , (Kodansha, Tokyo, 1991).
* [168] Sikdar, B. & Vastola, K., The effect of TCP on the self-similarity of network traffic, in _Proceedings of the Conference of Information Science Systems_ , John Hopkins University, (Baltimore, MD, 2001).
* [169] Sikdar, B. & Vastola, K., On the contribution of TCP to the self-similarity of network traffic in _Evolutionary Trends of the Internet_ , (Springer, Berlin, 2001) pp. 596.
* [170] Smith, R.D., Data traffic dynamics and saturation on a single link, _International Journal of Computer Systems Science & Engineering_, 3:1, 11-16 (2009).
* [171] Solé, R. & Valverde, S., Information transfer and phase transitions in a model of internet traffic, _Physica A_ , 289:4 (2001) 595-605.
* [172] Solé, R. & Valverde, S., Self-organized critical traffic in parallel computer networks, _Physica A_ , 312:4 (2002) 636-648.
* [173] Strogatz, S.H., Exploring complex networks, _Nature_ , 410 (2001) 268-276.
* [174] Tadaki, S., Long-Term Power-Law Fluctuation in Internet Traffic, J. Phys. Soc. Japan, 76 (2007) 044001.
* [175] Tadić, B. & Thurner, S., Information super-diffusion on structured networks, _Physica A_ , 332 (2004) 566-584.
* [176] Tadić, B. & Thurner, S., Search and topology aspects in transport on scale-free networks, _Physica A_ , 346:2 (2005) 183-190.
* [177] Tadić, B., Thurner, S., & Rodgers, G.J., Traffic on complex networks: Towards understanding global statistical properties from microscopic density fluctuations, _Phys. Rev. E_ , 69 (2004) 036102.
* [178] Tadić, B., Thurner, S., & Rodgers, G.J., Transport on complex networks: flow, jamming and optimization, _International Journal of Bifurcation and Chaos_ , 17:7 (2007) 2363-2385.
* [179] Takayasu, M., Takayasu, H., & Fukuda, K., Origin of critical behavior in Ethernet traffic, _Physica A_ , 287:2 (2000) 289-301.
* [180] Takayasu, M., Takayasu, H., & Fukuda, K., Dynamic phase transition observed in the Internet traffic flow, _Physica A_ , 277:2 (2000) 248-255.
* [181] Takayasu, M., Takayasu, H., & Sato, T., Critical behaviors and 1/f noise in information traffic, _Physica A_ , 233:3 (1996) 824-834.
* [182] Takayasu, M., Takayasu, H., & Tretyakov, A.Y., Phase transition in a computer network model, _Physica A_ , 253:1 (1998) 315-322.
* [183] Taqqu, M.S., Teverovsky, V., & Willinger, W., Is network trac self-similar or multifractal, _Fractals_ , 5 (1997) 63-73.
* [184] Thompson, K., Miller, G.J., & Wilder, R., Wide-area Internet traffic patterns and characteristics, _IEEE Network_ , 11:6 (1997) 10-23.
* [185] Uhlig, S., Non-stationarity and high-order scaling in TCP flow arrivals: a methodological analysis, _ACM Comm. Rev._ , 34:2 (2004) 9-24.
* [186] Veitch, D., Hohn, N., & Abry, P., Multifractality in TCP/IP traffic: the case against, _Computer Networks_ , 48:3 (2005) 293-313.
* [187] Veres, A. & Boda, M., The chaotic nature of TCP congestion control in the _Proceedings of IEEE INFOCOM 2000_ , 3 (Tel-Aviv, Brazil, 2000) pp. 1715-1723.
* [188] Veres, A., Kenesi, Z., Molnár, S., & Vattay, G., On the propagation of long-range dependence in the Internet, in _ACM SIGCOMM ComComm. Rev._ , 30:4 (2000) 243-254.
* [189] Vinge, V., Signs of the Singularity, _IEEE Spectrum_ , 45:6 (2008) 76-82.
* [190] Wang, Y.S. & Wu, H., _Physica A_ , Dynamics of a cooperation-competition model for the WWW market, 339:4 (2004) 609-620.
* [191] Wang, Y.S. & Wu, H., _Physica A_ , Dynamics of a macroscopic model characterizing mutualism of search engines and web sites, 363:2 (2006) 537-550.
* [192] Wang, W.X., Yin, C.Y., Yan, G., & Wang, B.H., Integrating local static and dynamic information for routing traffic, _Phys. Rev. E_ , 74 (2006) 016101.
* [193] Wang, Y., Chakrabarti, D., Wang, C.X., Faloutsos, C., Epidemic Spreading in Real Networks: An Eigenvalue Viewpoint, in _Proceedings of the 22nd International Symposium on Reliable Distributed Systems (SRDS’03)_ (Florence, 2003) pp. 25-44.
* [194] Watts, D.J. & Strogatz, S.H., Collective dynamics of ‘small-world’ networks, _Nature_ 393 (1998) 440442
* [195] Whitman, W., When I heard the Learn’d Astronomer, _Leaves of Grass_ , (Bantam, New York, 1983).
* [196] Willinger, W., Taqqu, M.S., Sherman, R., & Wilson, D.V., Self-similarity through high-variability: statistical analysis of Ethernet LAN traffic at the source level, _ACM SIGCOMM ComComm. Rev._ , 25:4 (1995) 100-113.
* [197] Willinger, W., Taqqu, M.S., Sherman, R., & Wilson, D.V., Self-Similarity through High-Variability: Statistical Analysis of Ethernet LAN Traffic at the Source Level, _IEEE/ACM Transactions on Networking_ , 5:1 (1997) 71-86.
* [198] Willinger, W., Paxson, V., & Taqqu, M.S., Self-similarity and heavy tails: Structural modeling of network traffic, in _A Practical Guide To Heavy Tails: Statistical Techniques and Applications_ edited by RF Adler, RE. Feldman, & MS. Taqqu, (Birkhauser, Boston, 1998), pp. 27-53.
* [199] Willinger, W., Govindan, R., Jamin, S., Paxson, V., & Shenker, S., Scaling phenomena in the Internet: Critically examining criticality, _Proc. Natl. Acad. of Sci._ , 99 (2002) 2573-2580.
* [200] Willinger, W., Paxson, V., Riedi, R., & Taqqu, M.S., Long-range dependence and data network traffic in Theory and Applications of Long-range Dependence edited by Doukhan, P., Oppenheim, G., & Taqqu, M.S. (Birkhauser, Boston, 2003) pp. 373-408.
* [201] Willinger, W., & Paxson, V., Where Mathematics Meets the Internet, _Notices of the AMS_ , 45:8 (1998) 961-970.
* [202] Willinger, W., Alderson, D., & Doyle, J.C., “Mathematics and the Internet: A Source of Enormous Confusion and Great Potential”, _Notices of the AMS_ , 56:5, p. 586-599.
* [203] Woolf, M., Arrowsmith, D.K., Mondragón, R.J., & Pitts, J.M., Optimization and phase transitions in a chaotic model of data traffic, _Phys. Rev. E_ , 66 (2002) 046106.
* [204] Yan, G. et. al., Epidemic spread in weighted scale-free networks, _Chinese Physics Letters_ , 22:2, (2005) 510-513.
* [205] Yan, G., Zhou, T., Hu, B., Fu, Z.Q., & Wang, B.H., Efficient routing on complex networks, _Phys. Rev. E_ , 73 (2006) 046108.
* [206] Yin, C.Y., Wang, B.H., Wang, W.X., Yan, G., & Yang, H.J., Traffic dynamics based on an efficient routing strategy on scale free networks, _Eur. Phys. J. B_ , 49:2 (2006) 205-211.
* [207] Yuan, J. & Mills, K.J., Exploring Collective Dynamics in Communication Networks, _J. Res. Natl. Inst. Stand. Technol._ , 107:2 (2002) 179-191.
* [208] Yuan, J. & Mills, K.J., A cross-correlation-based method for spatial temporal traffic analysis, _Performance Evaluation_ , 61:2 (2005) 163-180.
* [209] Yuan, J. & Mills, K.J., Macroscopic Dynamics in Large-Scale Data Networks, in _Complex Dynamics in Communication Networks_ , (Springer, Berlin, 2005) pp. 191-211.
* [210] Yuan, J. & Mills, K.J., Monitoring the macroscopic effect of DDoS flooding attacks, _IEEE Transactions on Dependable and Secure Computing_ , 2:4 (2005) 324-355.
* [211] Yuan, J. & Mills, K.J., Simulating Timescale Dynamics of Network Traffic Using Homogeneous Modeling, _J. Res. Natl. Inst. Stand. Technol._ , 111:3 (2006) 227-242.
* [212] Yuan, J., Yong, R., & Shan, X.M., Analysis of a type of computer network cellular automata model, _Acta Physica Sinica_ , 49:3 (2000) 398-402 (in Chinese).
* [213] Yuan, J., Ren, Y., Liu, F., & Shan, X.M., Phase transition and collective correlation behavior in the complex computer network, _Acta Physica Sinica_ , 50:7 (2001) , 1221-1225 (in Chinese).
* [214] Yuan, J., Wang, J., Xu, Z.X., & Li, B., Time-dependent collective behavior in a computer network model, _Physica A_ , 368:1 (2006) 294-304.
* [215] Zhao, L., Lai, Y.C., Park, K.H., & Ye, N., Onset of traffic congestion in complex networks, _Phys. Rev. E_ , 71 (2005) 026125.
* [216] Zhu, D.P., Gritter, M., & Cheriton, D.R., Feedback based routing, _ACM SIGCOMM ComComm. Rev._ , 33:1 (2003) 71-76.
* [217] Zhu, X.Y., Liu, Z.H., & Tang, M., Detrended Fluctuation Analysis of Traffic Data, _Chin. Phys. Lett._ , 24 (2007) 2142-2145.
| arxiv-papers | 2008-07-22T02:49:08 | 2024-09-04T02:48:56.922285 | {
"license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/",
"authors": "Reginald D. Smith",
"submitter": "Reginald Smith",
"url": "https://arxiv.org/abs/0807.3374"
} |
Subsets and Splits